CFB VS Ultralinear
ee
I'll like to try a SE and PP amp using Cathode Feedback.
Can someone give specs for a CFB transformer for the EL34?
What are the ideal load resistance and winding ratios?
EE
John Stewart
Good Luck John L Stewart
Chris Merren
"ee" <ee1000@*not*hotmail.com> wrote in message
news:9URn8.5$8f5....@news0.telusplanet.net...
I will give you a HINT:
Whatever plate load you decide on...(this is "primarily" a function of
screen voltage)...lets say 6K for giggles with 2500 turns on the primary..
Well..when you add a cathode winding you DO NOT change the "total" primary
turns of 2500... you simply distribute a portion of the 2500 turns to the
cathode.. For example: 2000 turns on the plate 500 turns for the
cathode...the total turns will still remain in SERIES...since the tube will
be between the two winding in series....in this arangement watch for the
instantaneous heater to cathode potential..
Regards
Chris
ee
design.
I am rather curious as to proper impedance matching when it comes to the
CFB design.
There is quite a lengthy thread on pentode VS triode mode.
I'd like to consider another mode.
Would a 1:1 cathode to anode ratio make it into a MacIntosh Unity
Transformer for SE?
EE
Patrick Turner
ee wrote:
> I hope no one else take it to mean they should have all the fun in the amp
> design.
Fun with amp design is not limited to the chosen few...
>
>
> I am rather curious as to proper impedance matching when it comes to the
> CFB design.
The load seen by the tube regardless of UL, CFB use,
UL + CFB, pentode, triode, McIntosh "unity coupling",
or plain cathode follower,
will be the same in each case.
But the gain, distortion, and Rout will vary quite a lot.
Plain UL reduces THD dramatically and only marginally increases
the drive needed to the output grids, whilst
incurring only a slight loss in power from the pure beam tetrode/pentode mode.
CFB, as used in the Quad II is just as effective.
as CFB is increased, the need for linear wide bandwidth
higher drive voltages is also increased, and this offsets the
some of the gains made by using CFB.
>
>
> There is quite a lengthy thread on pentode VS triode mode.
> I'd like to consider another mode.
> Would a 1:1 cathode to anode ratio make it into a MacIntosh Unity
> Transformer for SE?
Yes it would indeed.
And the McIntosh amps need about 140 vrms applied to the
output tube grids to make them clip.
The distortion in the driver amp is then substantial,
which is reduced by global loop NFB in the McIntosh.
But the mac has a very very low Rout,
and considering it runs in class B for most of the time,
people say it sounds OK.
Patrick Turner.
>
>
> EE
Denis
>
The CFB circuit simply allows one to obtain the ultimate performance
from a pentode or ray tetrode.
With CFB one can easely choose the optimal Eg2, which is always
dfferent from Ep, and this allows one to reduce the odd harmonics to
even lower value than that of the triodes. The CFB stage is usually
more load-tolerable than any other output configuration. This can be
an important issue, for instance, when driving the electrostatics.
Finally, the CFB circuit can have the lowest possible output impedance
and usually eliminates the need for global NFB.
The CFB is also interesting option with high-mu triodes.
> I'll like to try a SE and PP amp using Cathode Feedback.
> Can someone give specs for a CFB transformer for the EL34?
>
If you are serious with this ppoject, I can design and make the output
transformers.
A SE EL34 CFB amp seems to be very interesting, because it will be
directly comparable with popular SE 300B amps by the power range.
Therefore it would be exceptionally interesting to compare the sound.
Best regards,
Denis N. Afanassyev www.cortmi.com.ua/omak
Denis
> I hope no one else take it to mean they should have all the fun in the amp
> design.
>
> I am rather curious as to proper impedance matching when it comes to the
> CFB design.
>
> There is quite a lengthy thread on pentode VS triode mode.
> I'd like to consider another mode.
You are quite right
> Would a 1:1 cathode to anode ratio make it into a MacIntosh Unity
> Transformer for SE?
>
Certanly not. The McIntosh Unity coupling was a wrong idea for either
PP or SE.
It is enough to use from 12.5% to 16% of the total primary winding as
the CFB section to get it all. The EL34 will probably be happy with
just 12.5%.n this case you'll get a power stage equivalent to a triode
with Mu=8 and Ri<1kOhm.
The requred driver voltage will be below 50V peak, which will be
easely provided ba half of 6SN7 tube. By the way, 300B requires 70V
peak, so any front end designed for 300B will do this job with even
lower distortions.
In contrary, the McIntosh stage would require four times this voltage,
and there would be too much local feedback then needed.
The turn ratio depends on loudspeaker load, and could be from 25:1 to
40:1 for the best results with EL34.
Regards,
Denis
Patrick Turner
Denis wrote:
> ee <ee1000@*not*hotmail.com> wrote in message news:<%sVn8.216$Kl.1...@news1.telusplanet.net>...
> > I hope no one else take it to mean they should have all the fun in the amp
> > design.
> >
> > I am rather curious as to proper impedance matching when it comes to the
> > CFB design.
> >
> > There is quite a lengthy thread on pentode VS triode mode.
> > I'd like to consider another mode.
>
> You are quite right
>
> > Would a 1:1 cathode to anode ratio make it into a MacIntosh Unity
> > Transformer for SE?
> >
>
> Certanly not. The McIntosh Unity coupling was a wrong idea for either
> PP or SE.
But the gentleman asked if using unity coupling, ie, equal
plate and cathode windings would make his idea like a McIntosh.
Well, it would.
Whether the Mac idea is the wrong way to build a tube amp is debatable.
There are a lot of happy mac owners out there.
However I agree with Denis that the large drive voltage required for the Mac
amp sort of makes a good idea difficult to achieve.
McIntosh got around this problem with bootstrapping.
It isn't my cuppa tea.
But it seems to work, as the THD for the mac isn't too bad.
The use of unity coupling in the mac involved
bifilar wound transformers, which mean leakage between cathode
and anode windings is negligible.
But in designs with only 12.5 % of CFB, the leakage has to be very seriously
considered, and its an art to wind good OPTs with CFB.
Patrick Turner.
Ian Iveson
la Macintosh), UL, UL with separate screen windings, etc. Some of
those trannies were designed with EL34 in mind.
Alternatively, you might grab a copy of the book published by the
designer of those trannies (Menno van der Veen, "Modern High End
Valve Amplifiers"). There you will find most of the maths.
One point worth considering is that CFB allows you to get pentode
power, but with triode-like output impedance and distortion. As has
been pointed out, it is also relatively load-tolerant.
My CFB trannies were wound by Sowter, who did an excellent job. You
need to be careful to consider the leakage and capacitance between
anode and cathode. Anyone who winds a CFB or UC tranny needs to
know exactly what they are doing. Especially with EI trannies.
cheers, Ian
"ee" <ee1000@*not*hotmail.com> wrote in message
news:9URn8.5$8f5....@news0.telusplanet.net...
John Byrns
<cer...@engineer.com> wrote:
> I will give you a HINT:
> Whatever plate load you decide on...(this is "primarily" a function of
> screen voltage)...lets say 6K for giggles with 2500 turns on the primary..
> Well..when you add a cathode winding you DO NOT change the "total" primary
> turns of 2500... you simply distribute a portion of the 2500 turns to the
> cathode.. For example: 2000 turns on the plate 500 turns for the
> cathode...the total turns will still remain in SERIES...since the tube will
> be between the two winding in series....in this arangement watch for the
> instantaneous heater to cathode potential..
Actually the correct load impedance for CFB is somewhat higher than you
would calculate from the static screen Voltage due to the fact that the
signal Voltage on the cathode subtracts from the effective screen Voltage
as the current through the tube increases. CFB as usually implemented
with pentodes includes a dash of UL.
Regards,
John Byrns
Surf my web pages at, http://www.enteract.com/~jbyrns/index.html
John Stewart
Denis wrote:
> ee <ee1000@*not*hotmail.com> wrote in message news:<%sVn8.216$Kl.1...@news1.telusplanet.net>...
> > I hope no one else take it to mean they should have all the fun in the amp
> > design.
> >
> > I am rather curious as to proper impedance matching when it comes to the
> > CFB design.
> >
> > There is quite a lengthy thread on pentode VS triode mode.
> > I'd like to consider another mode.
>
> You are quite right
>
> > Would a 1:1 cathode to anode ratio make it into a MacIntosh Unity
> > Transformer for SE?
> >
>
> Certanly not. The McIntosh Unity coupling was a wrong idea for either
> PP or SE.
Sounds like Unity to me. Equal winding in cathode & anode = unity,
I thought. I built one & published in AudioXpress last year. It was used
as a high powered line stage. It was posted on one of the binaries
a while back. If anyone is interested I will post it again.
>
>
> It is enough to use from 12.5% to 16% of the total primary winding as
> the CFB section to get it all.
You have just described the Quad circuit. Cheers John
Shiva
with unity trans. configuration. I was thinking of taking the center tap
from a push-pull tranny, and unsoldering it so i have two equal(ish)
windings... Any problems you could foresee? (other than a good chance of
destroying a tranny, but, hey, I've accumulated loads of them...) Thanks
for the (possible) reply.
"John Stewart" <jh.st...@sympatico.ca> wrote in message
news:3CA08305...@sympatico.ca...
Ian Iveson
your side, but leakage inductance is not. Could end up with a nice
oscillator.
McIntosh reckoned that the quality factor of the OPT must be at
least 72,000. The QF is given by Primary Inductance / Primary
Leakage Inductance.
One good thing is that the tranny sees a low impedance at its
primary, so you don't need so many turns to get the bandwidth.
cheers, Ian
"Shiva" <help...@666.com> wrote in message
news:a7qao9$n4ut1$1...@ID-45739.news.dfncis.de...
John Byrns
Aside from the leakage inductance problem that others are addressing, it
sounds like you are talking about using a modified push pull transformer
for a CFB SE amp, and PP transformers generally aren't designed to handle
the unbalanced DC of a SE output stage.
Regards,
John Byrns
In article <a7qao9$n4ut1$1...@ID-45739.news.dfncis.de>, "Shiva"
<help...@666.com> wrote:
ee
It would be something worthwhile if common EL34 can be made into 300B
sound.
Now with EL34 where is the grid 2 and suppressor screen connected to?
Ground and not cathode I would guess.
The Quad uses a tetrode so that doesn't tell.
Now if grid 2 is connected to UL tap, is that still a CFB?
EE
John Stewart
alt.binaries.pictures.radio
& alt.binaries.radio.pictures that you can have some
fun with if you are successful in getting all those wires
out before destroying the transformer.
Have fun!! John L Stewart
Ian Iveson
Feeling a little foolish...sometimes the wood gets lost in the
trees!
cheers, Ian
"John Byrns" <jby...@enteract.com> wrote in message
news:jbyrns-2603...@216-80-74-144.d.enteract.com...
Ian Iveson
anode.
You won't get a 300B sound, either, unless you drive the output
stage with a 300B. UC pretty much wipes out the character of the
valves in the output stage. It gives maximum power and minimum
distortion, even in class B.
Hence the tricky drive requirement: no distortion, no gain, just
power.
If you combine UL and CFB, you get what some have called "super
triode". If you cross couple the screens, each to the UL tap on the
other side, you get what some have called "super pentode". Both of
these terms are claimed by van der Veen, but he would not be able to
defend the copyright IMO. Both are easy to drive too, compared to
UC.
What has been largely omitted so far is the point that these are
ways of extracting more power from the output valves. Twice as
much, pretty much, for CFB configurations above, and more for UC I
think.
cheers, Ian
Shiva
running 6V6's (or el84's) single-ended through *tiny* transformers. Sure,
the things weren't Hi-Fi, but... If core saturation was as big a problem as
it is made out to be, those things should have sounded awful, and more and
more so as time went on, as the core got permanently magnetized. And yet...
Kind'o sounded OK, and, with a few basic mods, looked OK on the scope (hey,
I use one of those power amps to pipe in the evil MP3's into my workshop,
just a junker with a better power supply & different drive circuit, and I
can can't complain...). Now, if the cores in those things weighed less than
a carton of smokes, and the core of the transformers I'm planning on using
weigh up to 30 times as much, wouldn't the saturation be negligible? The
core design seems the same... Or were those tiny things gapped? What that
makes single-ended Hi-Fi trannys so different? also, are toroidal trannys
used in Hi Fi? Some music amps use them, and it sure makes them lighter
Would seem to make sense... efficient & not leaky. Thanks for the reply,
btw.
"Ian Iveson" <ianives...@virgin.net> wrote in message
news:RN7o8.102$pA6....@news6-win.server.ntlworld.com...
Shiva
triode-strapped 6V6 push-pull guit amp "project" (I'm hoping that it's the
choke in the power supply, have just finished the thing, played it, and,
whoops, all kinds of nastiness ~1Mhz shows up when I really dime it out with
the output tubes in... And I don't even like the sound... Too "squashed",
and compresses too much. I could kill the oscillation by jumping the driver
plates with a .1n cap, but that's cheating, and the thing's already
lackluster sounding...) and do some necessary evil (fix some amps before 4pm
tomorrow) I'll jump on it.
-dimnews:3CA0F810...@sympatico.ca...
Fred Nachbaur
gradual phenomenon. Those consumer audio units you're referring to were
not, as you point out, "hi-fi", and 10% THD was considered entirely
acceptable. Another point is that, even though the *tube* may have been
capable of 10 watts, I'd hazard that in most listening situations they'd
rarely be run higher than one or two watts.
That being said, I'm sure that the OPT was the limiting factor in these
units, and that by simply replacing the transformer a vast sonic
improvement could be achieved.
Fred N
Shiva wrote:
--
+---------------------------------------+
| Music: http://www.netidea.com/~fredn/ |
| Vacuum Tube projects & other stuff: |
| http://www.fna.muohio.edu/dogstar/ |
+---------------------------------------+
Tim
> units, and that by simply replacing the transformer a vast sonic
> improvement could be achieved.
Afraid I'll have to mention the single 125 ohm resistor for all 4
6V6s in that stereo 6V6 PP Magnavox I have.. or that .0033uF cap
that's coupling the first half of the 12AX7 to the one 6V6...
But at least it has a choke in the PS.
But yes, those OPTs look a little small for even 5W...
Tim
--
"WOOHOO! Who would've guessed reading and writing would pay off!"
- Homer Simpson
Tim
> triode-strapped 6V6 push-pull guit amp "project" (I'm hoping that it's the
> choke in the power supply, have just finished the thing, played it, and,
> whoops, all kinds of nastiness ~1Mhz shows up when I really dime it out with
> the output tubes in... And I don't even like the sound... Too "squashed",
> and compresses too much. I could kill the oscillation by jumping the driver
> plates with a .1n cap, but that's cheating, and the thing's already
> lackluster sounding...) and do some necessary evil (fix some amps before 4pm
> tomorrow) I'll jump on it.
If you've been paying attention to ABSE, you've seen my stereo 6V6 SE..
well, with the 12AU7 output in, I used to have weird behavior in the
one channel (in the underside pic it's closest to the PSU, down the
middle), it used to get bad distortion. After tossing in the resistors
on the grids, it's all smooth with everything in.
I'd toss in some grid stoppers - the resistors must be close to the
socket, so I hear.
It's unlikely to be the power supply IMHO. If you absolutely must
suspect it, put a few ceramic and/or film caps across B+ to shunt
the RF.
John Daniel
>Ok, more silly questions: Think back to those *lovely* console stereos
>running 6V6's (or el84's) single-ended through *tiny* transformers. Sure,
>the things weren't Hi-Fi, but... If core saturation was as big a problem as
>it is made out to be, those things should have sounded awful, and more and
>more so as time went on, as the core got permanently magnetized. And yet...
>Kind'o sounded OK, and, with a few basic mods, looked OK on the scope (hey,
>I use one of those power amps to pipe in the evil MP3's into my workshop,
>just a junker with a better power supply & different drive circuit, and I
>can can't complain...). Now, if the cores in those things weighed less than
>a carton of smokes, and the core of the transformers I'm planning on using
>weigh up to 30 times as much, wouldn't the saturation be negligible? The
>core design seems the same... Or were those tiny things gapped? What that
>makes single-ended Hi-Fi trannys so different? also, are toroidal trannys
>used in Hi Fi? Some music amps use them, and it sure makes them lighter
>Would seem to make sense... efficient & not leaky. Thanks for the reply,
>btw.
Yes they did sound surprisingly good.
Why?
I think it was because at LF's the phase turns around, and the
feedback becomes selective, giving a "boomer" effect.
Using a pentode for ouput probably adds to this effect.
Seem to remember a passage from RDH4 mentioning "eliptical load
lines".
JD
Denis
> The winding needs to be bifilar. Capacitance, within reason, is on
> your side, but leakage inductance is not. Could end up with a nice
> oscillator.
>
The bifilar windings are of bad side as to transformer reliability,
and it is really unnecessary to get the best performance of the CFB
stage.
For example, my CFB OPTs have neither bifilar windings nor unity ratio
between the cathode and anode sections.
The optimal CFB percentage for most pentodes is from 12.5% to 16%,
except 6L6 and 6V6, which will improve up to 25% of CFB.
All these circuits consider the screen grids are fed by the constant
B+, which is usually significantly lower than the Ep.
This means a bit of negative feedback also applied to the screen
grid, as in the UL circuit. This little amount of UL feeback is just
what is doctor prescribed.
From other hand, the UC circuit always needs pure pentode operation.
The leakage inductance between the small catode sections and the rest
of secondary is not an issue with proper winding sectioning.
The resulting performance with such output stage is truly amazing: the
ouput impedance is low and constant up to 100 kHz, the power bandwidth
is above 100 kHz(-3dB), and the low-level frequency responce is still
flat up to 200kHz( the upper limit of my LF oscillator). The OPT does
not saturate at full power even below 20 Hz, and at this limit no
extra distortions are measured.
The frequency responce and distortion do not change significantly for
the loars from open circuit to 2 Ohms.
The filament-to-cathode voltage is also OK.
Above data apply to my push-pull CFB amplifiers with EL509 tubes.
I think the SE CFB one with EL34 can be made similarly good.
> McIntosh reckoned that the quality factor of the OPT must be at
> least 72,000. The QF is given by Primary Inductance / Primary
> Leakage Inductance.
>
This QF is easely made above 100000.
> One good thing is that the tranny sees a low impedance at its
> primary, so you don't need so many turns to get the bandwidth.
>
Another matter is core saturation. The SE OPTs do not saturate so
rapidly as PP ones, and they feature a stable inductance, but this
inductance also must be high enough not to eat out all available plate
current.
Generally speaking, the design of good OPT is not a black magic, but a
complex enginnring art. It is always necessary to find certain
compromise between the opposites, and this compromise can be both
superb and mediocre.
Regards,
Denis.
Patrick Turner
Ian Iveson wrote:
> The winding needs to be bifilar. Capacitance, within reason, is on
> your side, but leakage inductance is not. Could end up with a nice
> oscillator.
The LL between the two half primaries of an OPT
may not be so large a figure that it would be a problem,
unless the LL was high between the whole primary and the
secondary speaker coils.
The LL between half P sections is reduced to low levels with
the mutual coupling between the P and S sections.
The more interleaving, the less LL where ever you measure it.
I have tried using a garden variety OPT in this manner, and there were
no dramas,
and all that is required is the usual zobel damping networks
judiciously connected to load down the filter effect which is inherent
in all
amps with OPTs and/or input Ts.
If a PP tranny is going to be used for SE unity coupling,
like half a McIntosh circuit, there would be a need for the core
to have an air gap, to stop the tranny from saturating at too high
a low frequency, from the effect of the unbalanced DC.
I just wouldn't do any of this unless I could gap the tranny by
removing all the E&I lams, and rebuilding with
non interleaved lams, and just had the right sized gap
between butted E and Is.
>
> McIntosh reckoned that the quality factor of the OPT must be at
> least 72,000. The QF is given by Primary Inductance / Primary
> Leakage Inductance.
But this 72,000 thing is the ratio of the Lp to the LL.
The LL is the amount of series inductance measure anode to anode
in the anode circuit.
Suppose the spec of a tranny says it is 13 mH ( which is that of the
Williamson design ).
Then the PP output stage at HF is equivalent to the Ra-a,
which is about 3.2 kOhms for two KT66 in triode, plus
RL, about 10 kOhms, and this LL all in series powered from a
low Rout gene of Mu x Vg at its output.
So we have a network of R + L + R of a total R of 13.2 kOhms,
and 13 mH. The current in this network will fall to 0.7 times the
LF value at some HF, in this case when XL = 13.2 k,
which is at 161 kHz.
The LL effect is explained at page 212, RDH4.
>
>
> One good thing is that the tranny sees a low impedance at its
> primary, so you don't need so many turns to get the bandwidth.
There is a lot of local NFB, so hence the bandwidth.
Patrick Turner.
Patrick Turner
ee wrote:
> Thanks to all fine gentlemen for really useful information and offers.
>
> It would be something worthwhile if common EL34 can be made into 300B
> sound.
>
> Now with EL34 where is the grid 2 and suppressor screen connected to?
> Ground and not cathode I would guess.
Connect G3 suppressor to cathode.
>
>
> The Quad uses a tetrode so that doesn't tell.
>
> Now if grid 2 is connected to UL tap, is that still a CFB?
Yes, but if you have unity anode and cathode windings, just take the
screen to the B+,
and this will be still CFB, with UL.
Patrick Turner.
Patrick Turner
Shiva wrote:
> Ok, more silly questions: Think back to those *lovely* console stereos
> running 6V6's (or el84's) single-ended through *tiny* transformers. Sure,
> the things weren't Hi-Fi, but... If core saturation was as big a problem as
> it is made out to be, those things should have sounded awful, and more and
> more so as time went on, as the core got permanently magnetized.
Core material for all transformers is chosen because it can never be
permanently magnetised.
> And yet...
> Kind'o sounded OK, and, with a few basic mods, looked OK on the scope (hey,
> I use one of those power amps to pipe in the evil MP3's into my workshop,
> just a junker with a better power supply & different drive circuit, and I
> can can't complain...). Now, if the cores in those things weighed less than
> a carton of smokes, and the core of the transformers I'm planning on using
> weigh up to 30 times as much, wouldn't the saturation be negligible?
This would be true, if all you had was a huge OPT and an EL84.
> The
> core design seems the same... Or were those tiny things gapped?
Yess, those tiny OPTs for SE operation are all gapped.
> What that
> makes single-ended Hi-Fi trannys so different?
The iron needs to be gapped to stop it severely magnetising
in one direction from the idle DC current.
Without a gap, the THD would be horrendous.
> also, are toroidal trannys
> used in Hi Fi? Some music amps use them, and it sure makes them lighter
> Would seem to make sense... efficient & not leaky. Thanks for the reply,
> btw.
Yes, toroidal OPTs are used as audio trannies.
I don't use then due to the complexity of the tap arrangements I use, and the
tendency of the core to saturate very abruptly, since it is
a high mu material.
Very hard for the DIY person to wind.
Excellent results can be had with E&I lams, or C-cores.
Patrick Turner.
Patrick Turner
Shiva wrote:
> Hey, thanks... As soon as I deal with the RF (ish) oscillation in my
> triode-strapped 6V6 push-pull guit amp "project" (I'm hoping that it's the
> choke in the power supply, have just finished the thing, played it, and,
> whoops, all kinds of nastiness ~1Mhz shows up when I really dime it out with
> the output tubes in... And I don't even like the sound... Too "squashed",
> and compresses too much. I could kill the oscillation by jumping the driver
> plates with a .1n cap, but that's cheating, and the thing's already
> lackluster sounding...) and do some necessary evil (fix some amps before 4pm
> tomorrow) I'll jump on it.
What can I say, without talking about a return to square one....
Patrick Turner.
Ian Iveson
> > least 72,000. The QF is given by Primary Inductance / Primary
> > Leakage Inductance.
>
> But this 72,000 thing is the ratio of the Lp to the LL.
Yes, Patrick.
Er...Thanks
regards, Ian
John Stewart
Here is an excerpt from an article I recently did that explains what is
happening.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
An important addition to this amp are three grid stopper resistors. This tube
has very high transconductance &
will likely oscillate on it’s own if given a chance. I knew I had problems when
my FM receiver was interfered with
while the amplifier was running. I observed spectrum up to 150 Mhz. Grid stopper
resistors lower the Q
(Quality Factor) of the circuit according to the formula
Q = (1/R) * SQR ( L/C )
As little as 100 ohms in a conductor interferes with the RF to the extent that
it is eliminated. The audio is left intact.
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
I typically use 1K to 10 K in the G1 leads. No more than 100 ohms in the G2
lead.
They should be 1/2 watt (least inductance, not wirewound) mounted as close as
possible to the tube socket.
Hope that helps so you can use your time to try the other trick circuit.
Good Luck John L Stewart
Fred Nachbaur
Tim wrote:
>>That being said, I'm sure that the OPT was the limiting factor in these
>>units, and that by simply replacing the transformer a vast sonic
>>improvement could be achieved.
>>
>
> Afraid I'll have to mention the single 125 ohm resistor for all 4
> 6V6s in that stereo 6V6 PP Magnavox I have..
That's actually not so bad, shared cathode resistors is a way to get
cathode biasing without bothering with bypass caps. I've never myself
seen one shared between the two channels though... but hey, why not?
Saves the manufacturer $.13 :)
> or that .0033uF cap
> that's coupling the first half of the 12AX7 to the one 6V6...
Yeah... I'm sure you could find other things to improve as well, such as
screen supplies in many of these kinds of units.
> But at least it has a choke in the PS.
Ooh! Deluxe! :)
Fred N
Shiva
and it's a lousy cap (the pitfalls of using scavenged parts - gumdrops seem
to hold their own, but the oil-filled... And I thought I'd be so *HiFi*...
Since I prob'ly will be hangin around here, let me quickly
explain/introduce myself, since it (should) make my posts/questions a bit
clearer. I'm a vintage (mostly) guitar amp tech, who's gotten sick of aping
existing circuits (mostly fender/ marsh, both of which sound great, but are
little more than rip-offs from tube data "typical application" schematics).
Always was an electronics/computer/auto(racing) geek, but always had a
shallow engineering background in electronics, doing more prototyping than
CAD/paper design. My way of designing relied on a frequency generator, a
scope, crates and crates of salvaged parts, tube data on the 'net, and
substitution boxes, so the finished circuits wound up resembling the
original schematics only in concept.
One fine day I decided to blow fifty bucks on Morgan Jones's "Valve
Amplifiers" book, at which point I realized just how little I knew, and my
free time vanished... Downloaded some "demo" spice-e apps (still looking
for one which is (basically) free, has tube models, does ac/dc
analysis/simulations/plots/, and is easy to use (read: mouse everything
other than values, able to import schematics, flashes "YOU'RE AN IDIOT!!!"
when I blunder)), and, unfortunately for my customers, started getting into
Hi Fi design. Also started blowing my money on audio test gear, and
modifying (read: op amp op amp op amp) junkshop RF test gear into audio test
gear. I've built many guitar amps /effects before, and, at this point,
built one stereo power amp (gutted trashfind amp down to sockets &
transformers, and built a trivial single-ended 6V6 power amp for my shop
(running into a pair of Alesis studio monitors, which I got innstead of $$$
for work (unfortunately, players are usually broke, so I usually work out
out "swap" deals, or wind up giving their gear to them with an "as soon as
you get the $$$, drop by". Only been ripped off twice in my *life*.) Now
I've got two matching decent-quality OT's (from Ampegs), and a *huge*
chassis (3ft by 1.5 ft), & a beefy PT (from a Leslie - the thing sits on a
twin tube rects, and is the *heaviest* tranny I've seen... Haven't measured
the HT amperage, but must be impressive). The OT's are potted, and I'm
thinking of <lunacy> of potting the PT </lunacy> - I think it will be a bit
better than simply having bells with an exposed core... Anyway, planning on
making a 2-channel, 2 6L6 a side (or even <gasp> 6550's, power amp, PP with
<gasp again> screen bias, and <lunacy>variable feedback</lunacy>. I seem to
remember old HiFi amps having a "damping" knob - was that what it was?
(speaker seeing different reflective (is that the right word) impedance, and
acting loose as a goose when it's high, and constipated and uptight when
it's low???) I don't do vinyl anymore - my last turntable was a Rec-o-Cut
with skateboard-wheel drive... And, i'd guess, a 1/4horse motor, so all I
need from a preamp is the EQ section... Anyhow, I can get 600V from the PT
straight, and, hey, there's always a voltage doubler, so any *reasonably
priced* tube suggestions are welcome. Oh, never heard an EL34 audio amp I
like (keep thinking - "Dynakit!"), and I *like* 6l6's (haven't heard a HiFi
amp with 6550's, they sound nasty in anything but weirded-out Ampegs, but,
hey, they handle goofy voltages (way beyond their spec sheet, considering
what the plates *actually* get hit with in an overdriven PP amp), so, if I
use them conservatively, maybe I'll even buy some speakers to handle the
power...
Back to the cellar with me - I *do* have to fix other people's stuff, or
starve...
-dim
"Tim" <tmor...@charter.net> wrote in message
news:ua2nvkl...@corp.supernews.com...
Ian Iveson
I knew nothing at all until I read Menno van der Veen's "Modern High
End Valve Amplifiers". Having read it forwards and backwards
several times, I got Morgan's book and realised how little he deals
with output transformers. They make a good combination. I came
here just to check that people really do these things...
That's what we are best at I think. Some people here know lots and
some don't know so much, but we generally make an effort and there
are regular claims that progress is being made! Private knowledge
is pretty unconvincing, I find. I also believe that audio is an
area where social knowledge is particularly important.
I use spice. Can't really say how easy it would be for someone who
already knows about the subject. The package I use is an old
version of Proteus, cobbled together from demos with different
disablements. I have full functionality except that I can't save
test set-ups or graphs. To ask for a totally free,
fully-functional, point-and-click system with decent libraries that
you can add to is a bit like challenging the fundamental operation
of capitalism, I'm afraid. I tried quite a few packages before I
settled on Proteus. You must expect some messing about to get a
collection of valve models, and do your own transformers.
Around about now, someone is going to suggest that you beg, steal or
borrow a copy of RDH4. Beware that this is not easy reading. It is
complete, up to the time, which is its great value. For me, I use
it to fill in the bits that more readable books leave out (it is
good for really complicated graphs too, and variable names invented
in some bygone age).
Finally, I arranged a few pages a while ago when I got lost with
screens during the evolution of my present amps (E88CC
input/concertina + 4*6CH6 with CFB and +/- UL). You can still see
where spice led me at
http://freespace.virgin.net/ianiveson.home/circuit.htm Notice the
ringing, incidentally, on the square wave test with feedback. This
is commonly dealt with by using a judicious (?) zobel network across
the primary, or secondary, or between screen and anode. In reality,
I seem to have dealt with it by specifying a bifilar wound
primary/cathode winding. The added capacitance of this winding is
not present in the simulation as it appears on the pages.
So what's wrong with your concertina, then?
Don't go away...
best wishes, Ian
"Shiva" <help...@666.com> wrote in message
news:a7t035$o30ro$1...@ID-45739.news.dfncis.de...
Patrick Turner
Ian Iveson wrote:
> Welcome, Dim (is that your name?)
>
> I knew nothing at all until I read Menno van der Veen's "Modern High
> End Valve Amplifiers". Having read it forwards and backwards
> several times, I got Morgan's book and realised how little he deals
> with output transformers. They make a good combination. I came
> here just to check that people really do these things...
So where did van de Veen and Jones get their basic info from?
I have some of their material in my library, but I got the info
from which these modernists draw their designs from,
and so I am not copying them, but working from first principles.
These guys are good value, and I recommend them,
but also I recommend getting hold of the older books on tube audio.
>
>
> That's what we are best at I think. Some people here know lots and
> some don't know so much, but we generally make an effort and there
> are regular claims that progress is being made! Private knowledge
> is pretty unconvincing, I find. I also believe that audio is an
> area where social knowledge is particularly important.
>
> I use spice. Can't really say how easy it would be for someone who
> already knows about the subject. The package I use is an old
> version of Proteus, cobbled together from demos with different
> disablements. I have full functionality except that I can't save
> test set-ups or graphs. To ask for a totally free,
> fully-functional, point-and-click system with decent libraries that
> you can add to is a bit like challenging the fundamental operation
> of capitalism, I'm afraid. I tried quite a few packages before I
> settled on Proteus. You must expect some messing about to get a
> collection of valve models, and do your own transformers.
First the brain is used, and if you have time for Spice, go to it,
but for me it would spoil the use of my brain.
I like to work things out, not have somebody do it for me.
Maybe my method leads to better hands-on learning and experience.
I build handmade amps for a living.
I have to work it out and do it, I cannot afford to explore the
possibilities for
a year whilst building an amp.
The hobbyist can do so.
I am glad I learned the old way.
>
> Around about now, someone is going to suggest that you beg, steal or
> borrow a copy of RDH4. Beware that this is not easy reading. It is
> complete, up to the time, which is its great value.
The laws of tube use are valid up to right now, and
all I can say is that if you can digest and apply what is in the RDH4,
then you might hear music for your troubles.
Parts of it are a bit dated, like tone controls.
But if you do master this Book, then all else you read becomes far
easier.
It is not a book for folks with short attention spans, and expect
knowledge to
leap easily from page to brain, in some kind of easy instant wisdom fix.
> For me, I use
> it to fill in the bits that more readable books leave out (it is
> good for really complicated graphs too, and variable names invented
> in some bygone age).
>
> Finally, I arranged a few pages a while ago when I got lost with
> screens during the evolution of my present amps (E88CC
> input/concertina + 4*6CH6 with CFB and +/- UL). You can still see
> where spice led me at
> http://freespace.virgin.net/ianiveson.home/circuit.htm Notice the
> ringing, incidentally, on the square wave test with feedback. This
> is commonly dealt with by using a judicious (?) zobel network across
> the primary, or secondary, or between screen and anode. In reality,
> I seem to have dealt with it by specifying a bifilar wound
> primary/cathode winding. The added capacitance of this winding is
> not present in the simulation as it appears on the pages.
You could try a pair of Zobel networks across each plate half primary
on the OPT using about 2.2 k plus 0.0022 uF.
This may provide a resistive load to the output tubes, at F above 50
kHz, and help stabilise
any amp which uses feedback.
Also zobel networks from the plate circuit of an input tube
to ground can help with oscillation or square wave overshoot.
The principles are used at
http://www.turneraudio.com.au/70-200monotubeamp.gif
Patrick Turner.
Shiva
collect my 200 dollars? Actually, I've found the enemy, and it wasn't us...
It was bad couplin' cap. I have this huge collection of Lili caps I don't
want to use in customer's amps (they're "salvaged from a tube HiFi preamp),
so I use them in my hobbyshoppin' projects. Well, one was marginal, and
lived in a nasty spot (driving the concertina phase splitter - I really
*did* have a reason not to drive it straight from the previous plate - I was
too lazy to get the DC on the previous plate low enough, and, 'sides, the
thing's all anti-HiFi anyway - the concertina is sharing the heater (in the
same envelope) as it's driving tube, and it's a high Mu tube, so there's got
to be some glitches... 'sides, this circuit is a glorified distortion box
anyway ( a guitar amp which wants to fit into a Champ-sized box *with a 10"
speaker*, so the layout is a bit less than... spacious...), so, other than
*try* to be good, I can't do much - not much room, so the tube/transformer
placement is dictated more by available space than correct layout practice
(I'll post a picture of the silly thing - you'll be *appalled*). Anyway,
the thing's as simple as things could get - 12AT7 (don't cringe, now...) 1/2
used as the first stage, 1/2 as a concertina (yes, I *am* that clueless),
driving 6gu7 (each side driving output tubes from the plates), driving a
pair of cathode-biased 6V6's into a neat-o ancient Jefferson output tranny.
No global feedback - it sounds neat in guitar amps when the speakers have
lengthy conversations with the output tranny, instead of seeing next to a
dead short... Anyhow, I've resolved the RF bit, and *still* hated the
sound. Nice clean, but the overdriven sound sounded more like a compressor
than anything with any, oh, you know, *those* things... The idiotic thing
I did was not really having enough drive voltage to drive the 6v6's as
triodes, and still be able tooverdrive them as much as I wanted. So - back
to pentode connection. Now it sounds great, but... It's basically the same
stuff I was going to stop aping - a PP cathode-biased configuration <yawn>.
"Patrick Turner" <in...@turneraudio.com.au> wrote in message
news:3CA1C055...@turneraudio.com.au...
Patrick Turner
Shiva wrote:
> Hey - not the dreaded square one... I don't even get to pass "go" and
> collect my 200 dollars?
From the following post of yours, I don't think you needed too
much guidance; you seemed to have sorted out many little problems,
and you seem to be having fun with tubes...
I was a bit lost for words, late last night.....
Patrick Turner.
Denis
> Ian Iveson wrote:
>
> > The winding needs to be bifilar. Capacitance, within reason, is on
> > your side, but leakage inductance is not. Could end up with a nice
> > oscillator.
>
> The LL between the two half primaries of an OPT
> may not be so large a figure that it would be a problem,
> unless the LL was high between the whole primary and the
> secondary speaker coils.
> The LL between half P sections is reduced to low levels with
> the mutual coupling between the P and S sections.
> The more interleaving, the less LL where ever you measure it.
I would rather say "optimal interleaving", because of another
important factor, parasitic capacitance.
The bandwidth is much influenced also by the winding geometry, eash
physical dimension of a transformer coil is an independent parameter,
and not just core and window cross-sections. If the coil dimensions
are significantly different from the optimal ones, the desired
bandwidth simply cannot be obtained with any interleaving.
For example, I typically use 7 or 8 sections of the secondary, and the
full power bandwidth reaches 180 kHz. The LL referred to the secondary
is below 3uH(microhenries).
The SS amplifiers often need similar or even larger inductance in
series with output to stop oscillations.
> Suppose the spec of a tranny says it is 13 mH ( which is that of the
> Williamson design ).
This is rather a mediocre OPT.
The good one intended for two KT66s or 807s shouls have at least five
times lower LL.
Generally speaking, the good OPT will not require use of Zobel
networks, will allow one to leave the amplifier without load even at a
presence of large input signal, and will provide unconditional
stability with any load.
Regards,
Denis.
ee
I would use Spice because it just takes a min. to change a component and
analyze it whereas rebuilding or reordering a transformer can take weeks.
I also have a breadboard to mix and match parts and determine how the sound
changes. What may be tube sound is some non-linearity and this is
emphasized in SET with little or no feedback.
Spice let's me experiment without any capital investment.
Both methods are valid.
EE
Marie Antoinette
That's a good boy. Get it all out of your system before next Friday. You
might also want to check out alt.rec.forgotten.cabinet.doors, as well as
alt.rec.I'mHome. And tubes rule.
Patrick Turner
Denis wrote:
> Patrick Turner <in...@turneraudio.com.au> wrote in message news:<3CA1BB00...@turneraudio.com.au>...
> > Ian Iveson wrote:
> >
> > > The winding needs to be bifilar. Capacitance, within reason, is on
> > > your side, but leakage inductance is not. Could end up with a nice
> > > oscillator.
> >
> > The LL between the two half primaries of an OPT
> > may not be so large a figure that it would be a problem,
> > unless the LL was high between the whole primary and the
> > secondary speaker coils.
> > The LL between half P sections is reduced to low levels with
> > the mutual coupling between the P and S sections.
> > The more interleaving, the less LL where ever you measure it.
>
> I would rather say "optimal interleaving", because of another
> important factor, parasitic capacitance.
>
> The bandwidth is much influenced also by the winding geometry, eash
> physical dimension of a transformer coil is an independent parameter,
> and not just core and window cross-sections. If the coil dimensions
> are significantly different from the optimal ones, the desired
> bandwidth simply cannot be obtained with any interleaving.
>
> For example, I typically use 7 or 8 sections of the secondary, and the
> full power bandwidth reaches 180 kHz. The LL referred to the secondary
> is below 3uH(microhenries).
But what effect would the shunt capacitance then have, due to all these sections?
I suggest you post a design in full for your OPT, like I have done,
at my web site, so we can see what is needed to get the responses claimed.
>
> The SS amplifiers often need similar or even larger inductance in
> series with output to stop oscillations.
>
> > Suppose the spec of a tranny says it is 13 mH ( which is that of the
> > Williamson design ).
>
> This is rather a mediocre OPT.
Not really, 13 mH is sufficiently low LL
to suit a 10 k a-a RL,
If you had an amp with its RL = 4 k a-a, then
the LL would need to be 2.5 times less, for the same response.
> The good one intended for two KT66s or 807s shouls have at least five
> times lower LL.
It depends on load Denis, and mode of operation.
> Generally speaking, the good OPT will not require use of Zobel
> networks, will allow one to leave the amplifier without load even at a
> presence of large input signal, and will provide unconditional
> stability with any load.
I agree, but Zobels across the OPT increase the margin
of stability. One can do without them if the tranny is that good.
But at least a Zobel with say 0.1 uF and 10 ohms is not
a bad idea across the output secondary, and exactly what is
used needs to be tailored for the OPT, and to be the minimum values
which least affect the desired bandwidth.
The resonances one gets wth OPT at above 50 Khz
sometimes need to be damped, by having the Zobel
become essentially a resistive load by about 80 kHz.
The R then prevents the high Q, and rapid phase shift
associated with a resonance, near Fo.
The Zobel networks also reduce HF gain but without the penalty
of extra phase shift at above 120 kHz, which is often about the F
of oscillations, should they occur.
Patrick Turner.
>
>
> Regards,
>
> Denis.
Patrick Turner
ee wrote:
> What are some other old books to recommend besides RH4?
Any book on radio/electronics by Terman.
Tremain's Audio Encyclopedia is good.
Take a look in the university, or tech college library near you,
if there is one, the older the college, the better, as there's
more likelyhood of the existance of an old reference archive,
full of hard cover electronics text books, used by teachers
and students, in the 50s and 60s.
> I would use Spice because it just takes a min. to change a component and
> analyze it whereas rebuilding or reordering a transformer can take weeks.
I agree, Spice would be convenient.
But nothing quite equals reality.
So mucking around with old amps used as a testbed for ideas
is a good way to get to understand things.
I began wionding all my own transformers.
>
> I also have a breadboard to mix and match parts and determine how the sound
> changes. What may be tube sound is some non-linearity and this is
> emphasized in SET with little or no feedback.
>
> Spice let's me experiment without any capital investment.
But how do you get to hear what each modification
made with Spice sounds like?
> Both methods are valid.
Up to a point, yes, they are, and for some things,
Spice would be essential, such as complex analog filter circuits, that
without a lot of knowledge, and time for maths, it would be hopeless
to try to work them out, just let someone else do the work for you,
which is what happens with simulation.
But for the simulation to be accurate, the data fed in to the
program must be correct, and with an amp with an OPT,
it may be difficult to know what the exact R and C and L
components are in the circuit.
I get by very well without any simulations,
I just try to think of everything.
Patrick Turner.
Denis
> What are some other old books to recommend besides RH4?
>
Some valuable references on CFB are collected at
http://homepages.enterprise.net/icedragon/new/distribu.htm
Although the basic principles of transfomer designs were known efen
before RDH4, the progress in understanding them was considerable
during the past years.
Indeed, Menno Van der Veen has made some mistakes too.
> I would use Spice because it just takes a min. to change a component and
> analyze it whereas rebuilding or reordering a transformer can take weeks.
>
Be aware of Spice! There is no accurate nonlinear model of a tube,
either triode or pentode. The existing models are limited in
applications, for instance, even the best known triode model
(Duncanamps) is relatively good only for quite narrow range of
resistive loadlines. The most difficult issue here is the nonlinear
internal feedback, which is inherent for any triode.
Perhaps, the pentode coud be simpler to simulate.
Now I'm working on more realistic tube models, and I use MathCAD, and
not Spice.
The Spice may be helpful in analysing the passive components, like the
transformers providing your equivalent circuit is correct.
> I also have a breadboard to mix and match parts and determine how the sound
> changes. What may be tube sound is some non-linearity and this is
> emphasized in SET with little or no feedback.
Remember, any triode has nonlinear internal feedback, which makes it
the voltage amplifier. The return ratio of this NFB is beta=1/Mu. The
nonlinearity of this feedback is evident as this Mu depends on Vg and
Vp. When the triode operates with Rl>>Ri, the resulting nonlinearity
is almost completely defined by this nonlinear NFB. Fortunately, the
most important nonlinear product will be 2nd harmonic.
The pentode has this NFB almost completely suppressed due to action of
the screen grid. The CFB circuit introduces an external and perfectly
linear local NFB reducing distortions far better. The pentodes have
another possibility to lower the distortions before applying the NFB,
the screen current.
The screen grid eats some portion from the total cathode current, and
influences the harmonic spectrum of the plate one. From other hand,
Ig2 depends on Eg2, and one can tailor the transfer curves of a
pentode by adjusting the screen voltage. Generally it is possible to
lower the open loop distortions of a pentode by factor of two, or even
more, in respect to those of a triode.
Note: The open loop distortions of a triode are solely defined by the
Gm nonlinearity, and they always are as high as 15% for 100% of class
A current swing.
Regards,
Denis
Patrick Turner
Denis wrote:
> ee <ee1000@*not*hotmail.com> wrote in message news:<nLIo8.6327$8f5.1...@news0.telusplanet.net>...
> > What are some other old books to recommend besides RH4?
> >
>
> Some valuable references on CFB are collected at
>
> http://homepages.enterprise.net/icedragon/new/distribu.htm
>
> Although the basic principles of transfomer designs were known efen
> before RDH4, the progress in understanding them was considerable
> during the past years.
> Indeed, Menno Van der Veen has made some mistakes too.
>
> > I would use Spice because it just takes a min. to change a component and
> > analyze it whereas rebuilding or reordering a transformer can take weeks.
> >
>
> Be aware of Spice! There is no accurate nonlinear model of a tube,
> either triode or pentode. The existing models are limited in
> applications, for instance, even the best known triode model
> (Duncanamps) is relatively good only for quite narrow range of
> resistive loadlines. The most difficult issue here is the nonlinear
> internal feedback, which is inherent for any triode.
But one never needs to take this aspect of a triode into account
when trying to set up a circuit.
You just set up the circuit, and everybody knows
its no use worrying about the self regulation of a triode, because
that is the result of the way theu made the tube,
and you cannot change or alter the insides of a tube.
So simple modelling excluding the idea
behind the NFB in a triode will work just fine.
>
> Perhaps, the pentode coud be simpler to simulate.
>
> Now I'm working on more realistic tube models, and I use MathCAD, and
> not Spice.
>
> The Spice may be helpful in analysing the passive components, like the
> transformers providing your equivalent circuit is correct.
Yes, and that's the hard part, knowing what that
equivalent circuit actually is.
> > I also have a breadboard to mix and match parts and determine how the sound
> > changes. What may be tube sound is some non-linearity and this is
> > emphasized in SET with little or no feedback.
>
> Remember, any triode has nonlinear internal feedback, which makes it
> the voltage amplifier. The return ratio of this NFB is beta=1/Mu. The
> nonlinearity of this feedback is evident as this Mu depends on Vg and
> Vp. When the triode operates with Rl>>Ri, the resulting nonlinearity
> is almost completely defined by this nonlinear NFB. Fortunately, the
> most important nonlinear product will be 2nd harmonic.
Well, not many folks in the group will follow you
about a "return ratio = 1/mu."
For working out triode behavoiur, there is NO NEED
to take into account what Denis is trying to say.
And as RL gets larger and larger,
until the load is a constant current source,
the linearity of the triode is at its maximum.
And most triodes with loads which are many times the Ra,
or approaching a CCS in value, are very linear.
On page 511,
The RDH4 has a test between a 6SJ7 connected as a triode
or as a pentode and the THD for either is shown on a graph.
In this, the pentode has less THD up to 30 vrms output,
where the two tubes have equal THD,then the
triode is much better by 60 vrms.
But the graph is a bit strange because the triode THD
is shown increasing along a straight line, which
most triodes do not do.
All my experiments with triodes show the THD
increasing along a curved line as shown by the pentode line on
page 511, figure 12.16A.
It depends what sort of triode or pentode is used, and how
it is used, to get the least THD for any given application.
> The pentode has this NFB almost completely suppressed due to action of
> the screen grid. The CFB circuit introduces an external and perfectly
> linear local NFB reducing distortions far better. The pentodes have
> another possibility to lower the distortions before applying the NFB,
> the screen current.
In all my experiments with CFB in output stages,
if one simply applies enough CFB to reduce the effective plate
resistance down to that of the same tube connected as a triode,
then the triode is often just as linear as the pentode.
If it isn't, its because the triode has only more 2H THD than the pentode with CFB,
and this pentode will always have more odd order THD, reduced by half,
by CFB, but still there.
Meanwhile the triode has very low amounts of odd order THD....
The real reward for CFB is that a pentode with CFB can
still produce its full power, and the price paid is gain,
ie, the input voltage required is higher,
but not much worse than a triode.
But if the OPT and driver amp have wide bandwidth,
and low phase shift, there is nothing wrong with
simple plate loaded pentode stages with just
one good loop of global NFB applied.
>
>
> The screen grid eats some portion from the total cathode current, and
> influences the harmonic spectrum of the plate one. From other hand,
> Ig2 depends on Eg2, and one can tailor the transfer curves of a
> pentode by adjusting the screen voltage. Generally it is possible to
> lower the open loop distortions of a pentode by factor of two, or even
> more, in respect to those of a triode.
It all depends on how much CFB is used.
I like using some UL screen taps as well as CFB,
I found this improved the spectrum content of THD,
but did not reduce the power very much.
>
> Note: The open loop distortions of a triode are solely defined by the
> Gm nonlinearity, and they always are as high as 15% for 100% of class
> A current swing.
Wrong, Denis, wrong.
Triodes set up properly in PP class A
output stages will give only 1% at a dB below clip.
The alarmist figures which you quote above
are simply irrelevant to most folks.
In SET, most amps are set up for a load which yeilds
5% at clipping.
The only way you are going to get 15% THD
with a triode at near clipping, or grid current,
or both, ie full available power,
is if the load value is far too low.
I have had many customers very happy with
two KT88 in triode to supply their audio needs.
The two KT88 will provide better
sound than a pair of EL34 in pentode, and perhaps better than UL,
but not much.
The use of KT88/6550 as triodes instead of pentodes in old amps which
may have had pentodes or beam tetrodes is an easy fix.
Whether CFB is better by a mile is purely speculative.
Patrick Turner.
>
>
> Regards,
>
> Denis
Ian Iveson
> either triode or pentode. The existing models are limited in
> applications, for instance, even the best known triode model
> (Duncanamps)
Known by who? How do they qualify for "best known model"?
> is relatively good only for quite narrow range of
> resistive loadlines. The most difficult issue here is the
nonlinear
> internal feedback, which is inherent for any triode.
> Perhaps, the pentode coud be simpler to simulate.
They all have cathodes. That is the hard part. Pentode introduces
more complications.
> Now I'm working on more realistic tube models, and I use MathCAD,
and
> not Spice.
Would you care to tell us what the functions are, Denis. Perhaps
you could also explain why they cannot be implemented in Spice?
> Indeed, Menno Van der Veen has made some mistakes too.
Perhaps you could give an example?
I would also be interested in how you manage to get "perfectly
linear" CFB, and how you reduce primary leakage to a few uH.
cheers, Ian
"Denis" <denis_af...@mail.ru> wrote in message
news:32f1783d.02032...@posting.google.com...
> ee <ee1000@*not*hotmail.com> wrote in message
news:<nLIo8.6327$8f5.1...@news0.telusplanet.net>...
> > What are some other old books to recommend besides RH4?
> >
>
> Some valuable references on CFB are collected at
>
> http://homepages.enterprise.net/icedragon/new/distribu.htm
>
> Although the basic principles of transfomer designs were known
efen
> before RDH4, the progress in understanding them was considerable
> during the past years.
>
>
>
> > I would use Spice because it just takes a min. to change a
component and
> > analyze it whereas rebuilding or reordering a transformer can
take weeks.
> >
>
Patrick Turner
September 1952 Wireless World
article about "Amplifiers and Superlatives ",
by none less than D.T.N. Williamson, and P.J. Walker.
For those not familiar, Williamson was the guy who got famous
for the amp named in his honour,
and Walker got famous with the Quad range of products.
The two blokes were obviously not enemies, and it appears they
co-operated on the article topic.
They would have been paid well by WW, and the article
gave them the chance to say nice things about
the operation of triodes, ultralinear, and cathode feedback
used in the output stages of pentode and tetrode amps.
At the end of the article which describes the salient aspects of each topology,
there is a table for class A operation of various topologies, using KT66,
and we assume about B+ = 320 volts, :-
Triodes,
efficiency = 27 %,
power = 10 watts,
THD at just under grid current = 1%,
RL / Rout = between 2 and 4, so let's say 3.
( This means a damping factor of 3,
which isn't bad for a circuit with no external loop of feedback )
Beam Tetrodes,
efficiency = 38 %,
power = 14 watts,
THD = 2%,
RL/Rout, 0.05 to 0.1, so let's say 0.075,
( This means a very poor damping factor,
and the amp without NFB is a just a current source,
with negligible signal voltage regulation.)
Standard UL, optimum screen taps at 40% of anode voltage,
efficiency = 36%,
power = 13.5 watts,
THD = 1.5%,
RL / Rout = 0.5 to 1.0, so let's say 0.7.
( So with screen UL, Rout is nearly equal to RL,
and damping factor is about unity.)
Tetrodes with Quad arrangement,
efficiency = 36%,
power = 13.5 watts,
THD = 0.5%,
RL / Rout = 2.0, no other figures given.
( So the dampoing factor is less than that of triodes,
or in other words, the Rout is higher than triodes.)
The article does not state exactly what the % of CFB
is used in the "Quad arrangement".
But prior to the table, they say the test used an
amount of CFB which gave similar operating
conditions as the case for 40% UL plate coil taps.
This would mean that the OPT primary would be divided
into sections giving a plate winding with 78 vrms appearing at each anode,
and a cathode winding with voltage of say 52 appearing at each
cathode..
The % of CFB would be 40%, of the total windings.
The screens are taken to the B+.
The relative voltages at grid, screen, plate and cathode
would then be identical to the plain 40% screen UL arrangement.
The huge difference with CFB is that the distortion voltage
is not only applied to between the cathode and screen, as in the case of plain UL,
but also to between the grid and cathode.
But something is really wrong here in the way the guys wrote the article.
If there is indeed 40% of the turns in the cathode winding,
output stage needs a drive voltage to the grids which would be very high.
If 13.5 watts was produced into say 5 kOhms a-a,
then 260 vrms is across the load of 5 kOhms,
and anode to cathode volts in each tube would be
130 vrms, so 78 volts is at the anode, and 52 at the cathode,
so if the gain of the tube is say 10, then 13 volts
is needed between cathode and grid, and the
total voltage needed at each of the output grids will be 52 plus 13,
or 65 vrms.
Unfortunately, none of these operating conditions was made
completely clear, but the great benefits of CFB
were touted as being very good, because the drive voltage required for
CFB was not much worse than triode, but as all can see,
it is a lot worse, as about 30 vrms are required for each triode grid drive.
In the Quad II amp, nothing like the amount of CFB is used
as is suggested by the sloppy article from these two guys in 1952.
Only about 10% of CFB is used in the Quad amp, not 40%.
I would humbly suggest that Walker was clever to promote
his idea about the "Quad arrangement", but fell short
of actually giving the real thing to the public in his Quad II
design.
The article goes on to give the maths for calculating the
gain and output resistance of the various ways of connecting output tubes.
Unfortunately, where they were needed to make sure all would be clear,
nothing is much clearer than plain mud.
The equations ( containing up to 26 items for Quad Rout )
contain an item 'm', and it really isn't clear what it is, or how to
calculate it so it can be placed into their equations.
Even I found it impossible to fathom the math, basically schoolboy stuff,
but complex enough to confound 99% of readers at the time.
There are other analomies, but a more profound way to confuse
nearly all readers would be hard to find.
The general concept was explained OK, but basically,
Walker seemed bent on advertising Quad, rather than being
plain honest about CFB, without any mention of his brandname.
My own deep and meaningful experiences
with CFB in output stages led me to
building 90 watt amps with 4 x 6550 tubes,
with 12.5% of CFB and 25% taps for the screens,
so the tube operating conditions are exactly the same as
37.5% UL .
Whilst in class A :-
In beam pentode, the 6550 has about 12,000 Oms of Ra.
Plain 37.5% of UL reduces this to about 4,000 Ohms,
and does little to reduce the max power, but reduces gain only
about 30%.
With the CFB windings added, the Ra approaches that of triode,
and the power stays the same, but instead of 35 vrms needed
at each of the the input grids of the 6550, some 70 vrms is needed.
The THD went lower than that of triode at 1 / 2 the power.
Loads used were that which would normally be used for triodes.
The lowest THD was achieved with the UL taps, as well as the CFB.
The THD was measured at the output using the same drive amp.
I found that raising the CFB to 25%, the THD rose ever so slightly,
because the additional drive amp THD was greater than the
reduction in THD in the output stage, due to higher CFB%.
As CFB is increased the amount of global feedback
is reduced, to keep the total amount of NFB, both global and
local CFB in the output stage to about 15 dB.
So as the global NFB is reduced, and drive voltage increased,
the distortion in the drive amp is not reduced as well as when there is
zero CFB, and the NFB is all global, as it is in a Williamson.
I am not going to give you all the maths for the various
ways of calculating gain and Rout, because
it just isn't needed, and I don't have all the maths
for combined CFB and UL.
One can still build something really good,
and tweak it one way or the other, depending
on what you observe yourselves when you accurately measure your amps,
and then carefully consider the results very skeptically over a cuppa.
There would not be any need for simulation, just do the hard yakka,
and you will be then sure of what is going on, not what might go on.
At least Walker and Williamson agreed at the beginning of the article
there were more ways
to build decent amps than just one, and all could lead
to excellent amp performance, depending on the implementation of an idea,
which needed to be valid at the outset.
Pentode / Tetrode amps needed more global NFB than others to
achieve the low Rout and THD of triodes, although there was more power.
This meant a decent OPT, which makers were reluctant to wind,
as makers were allergic to costs, and many companies then
had the grandfather of Mr Bean Kownter working for them, who prefered to spend
company profits on Rolls Royces, than transformer iron and wire.
CFB in the OPT offered the chance to make the margin of stability
greater, since not as much NFB was being applied around all those stages.
But what this translates to is the opportunity to reduce costs by
cheapening up the design buy making all the parts lighter,
and having hummy power supplies, and relying on NFB
to tidy up the mess.
The Quad II amps look, and feel very chic, and are very neatly wired up,
and engineered, but are very poorly designed amps.
It did not matter at that time because 99%
of Quad's market were rich folks who would only use a civilised half watt.
May the Easter bunny be bringing you bundles of eggs and stuff....
May your OPTs be heavy, make you reluctant to move,
and deliver you good music.
Patrick Turner.
Patrick Turner
Ian Iveson wrote:
>
> I would also be interested in how you manage to get "perfectly
> linear" CFB, and how you reduce primary leakage to a few uH.
The "perfectly linear" supply of NFB to the cathodes
is due to the use of the output transformer, which would conribute
little to the THD
if it is well made.
Most NFB is transfered from one part of the circuit to another via
resistors
or some non distortion producing network.
But Denis maintains the internal triode feedback is non linearly
generated.....
So what, folks still love 'em.
Most LL quantities for OPTs are quoted as a L figure seen by the
output tubes in series with the total tube load, ie,
normally the anode to anode load.
So if we had an OPT with a 10 k to 16 ohm ratio,
then say 13 mH of LL would be seen in series with that 10 k load,
but if we measured at the secondary coil, that 13 mH is transformed down
by the impedance ratio of the OPT, which in this case would be 625 : 1,
so LL at the output would be 20 uH, and this would result
in a pole with the 16 ohm load at 127 kHz.
> cheers, Ian
Patrick Turner.
Shiva
inherently rude, and *definitely* overloaded wit 5% work (one needs to eat)
and 95% hobbyshopping (just got a 331A distortion analyzer (through this
NG), and spent most of the day trying to eliminate ground loops/calibrate
everything so that it *at least* jives with the rest of the gear (my freq.
generators were calibrated by using the *~1000hz* probe-matching tap on my
scope... I thought it would be a crystal-controlled oscillator, but, I
guess, it's not (465 Tek)) Error wasn't important before, but with that
distortion analyzer, it's nice to be able to just set the knobs in the right
position, and only need to fine-tune, rather than remembering that 95 on the
freq. generators & scope means 100 on the distortion rig. Anyhow, I'm
goin' to try to work a bit more instead of hobbyshopping , and get a decent
freq. counter/ cap meter (I'm using a home-made bridge for that now, with a
+-5% ref. cap) The new multitesters are so cheap now, and pretty
convenient... for 200 bucks I can get one which I really lust after, freq.
counter, cap bridge down to something like 10pf, nice big display, and looks
beatable (I sometimes *step* on my tools...)
Anyhow, got to get back to work - promised that a bunch of stuff will be
done by 4pm today, so I got to get to it. Oh - that project I had that was
supposed to be so different & weird & had ringing/rf problems? Degenerated
to a PP pretty-much pentode run-of the mill guit amp. Sounds neat, but so
do they all... I've got a pair of matching Ampeg output trannys, so I'm
making a fairly generic PP stereo amp, and see if I could get 6550's or
KT88's to sound reasonable, and, if not, go to 6L6s.Oh, why does everyone
*love* 6sl7's? I mean, I like the look of octals, but it's so much easier
going with 12**7 family, 'coz then you have a whole slew of tubes to play
with, without rewiring the socket. Oh, and (sorry, guitar background) AU's
and AY's are great-sounding tubes, and AX's are still the "cooking"
double-triode as far as I'm concerned. What's so good about6SL 7's thatt
they show up in so many 'modern" amps?
Oh, and I've spent half an hourstudying your circuit (the 4 el84 one, with
individual bias & (if I remember correctly) ac adjustments on all of them.
I'd love to see the *real live* graphs on the thing. Do you *like* the
sound? What are you driving it with? After the initial shock of seing the
cathode hardwired to the screen grid (it was the tube symbol you had to use
(internal connection to wrong pin), I'm not in a position to discuss the
virtues or shortcomings of your design. Everything *looked* correct, but
I'd have to hear the thing and put it on the bench to know for sure. Oh, I
use Design Works Lite 4 for drawing schematics, and, other than a few common
tubes, it has generic pentodes in it's library, in case you need a nice,
intuitive schematic drawing (no analysis) appy.
Thanks to everybody who's spent time replying to me - if anyone is looking
for (pulls or US-Mil surplus) power trannys, I might be able to help out/
swap for other stuff...
-dim.
"Ian Iveson" <ianives...@virgin.net> wrote in message
news:svZo8.10620$Ko1.1...@news6-win.server.ntlworld.com...
Shiva
doing something miss manners doesn't approve of.
A question which will make everyone gasp with shock and disdain: What
*IS* RDH4? something like radio handbook 4th edition? Sorry - clueless. I
used to be all proud of myself for having RCA RT book...
-dim
"John Daniel" <NOSP...@dhaen.co.uk> wrote in message
> <snip>
Shiva
am willing to bone up on tranny design, but what are the costs? Is it
possible to get the E & I iron in small quantities? At what cost? I used
to restore vintage racecars, so I had a machine shop, and still have access
to one, so I could build a winding rig, but things like design & suppliers
are a mystery to me (well, I mean, I've certainly destroyed/ripped apart
many trannys as a kid, and, even as I write this, I have an output
transformer from a PP pentode 6L6 -> 2 ohms tranny which I'm dissecting
(very much easier than I thought, splitting the center tap is trivial, but I
also found a bunch of primary taps which were not connected to anything, and
the secondary consists of 3 separate windings wired in parallel - oh, the
possibilities...
Or should I get that thought 9homemade trannys) out of my head, and do
something constructive, like play with matches in the traffic?
-dimnews:3CA1BFFC...@turneraudio.com.au...
Fred Nachbaur
For the most part, people here don't seem to care much about the form of
postings, the main criterion being understandability. We do sometimes
get someone throwing a hissy fit about "top-posting twits," but that
usually fizzles out quickly. I try to post in a manner that follows
previous history in a thread; sometimes I'm a top-posting twit, other
times a bottom-posting twit, and yet other times a replies-inline-twit.
RDH4 is "The Bible", aka Radio Designer's Handbook, 4th edition as you
surmised. Don't feel too bad, I don't have a copy anymore either (a
casualty of a move years ago) and haven't yet replaced it. My favourite
books are also the RCA RT and TT manuals, and more and more I find
myself referring to online resources.
Cheers,
Fred N
Shiva wrote:
Ian Iveson
qualms about that. Easy. But I don't think he did.
Even a well made transformer does not supply "perfectly linear" CFB,
by virtue of the fact that it cannot provide "perfectly linear"
*anything*. I would also take some convincing that "perfectly
linear" CFB is what is actually required. I don't even know what
"perfectly linear" feedback means. With respect to what, the
voltage output signal at the anode?
You see, in the basic debate here about CFB pentode versus triode,
even if we accept the marginally-acceptable notion of inherent
feedback of which Denis speaks, we would tend to come down on the
side of whichever arrangement delivered the most appropriate
feedback by the most direct and speedy route. In this context,
"perfectly linear" CFB sounds like a claim to have found the Holy
Grail. If it is perfect, then the triode can't beat it.
But of course it is not perfect. What we end up doing, as always,
is comparing one imperfection with another. The best amplifiers,
IMO, achieve greatness by combining a number of imperfections in a
way that sounds best. I think we all know this. This means that
extolling the virtues of one *stage* at the expense of alternatives
is daft.
For example. I use CFB and UL very successfully. I was drawn to
CFB and the use of parallel output valves because I wanted no
overall nfb, even though my speakers are not easy. OK, so I get a
wonderful output stage. But since I have no overall nfb, then I
must pay particular attention to how I reduce distortion before the
signal gets to there, and this is made a bit harder by the fact that
I need a bigger signal.
I also think, now I'm in the mood, that claims concerning an
"optimum" proportion of CFB are silly. There is a fair amount of
pseudo-analytical claptrap circulating here.
That's why I asked Denis.
cheers, Ian
"Patrick Turner" <in...@turneraudio.com.au> wrote in message
news:3CA47591...@turneraudio.com.au...
zonked
news:a8245f$p0n4i$1...@ID-45739.news.dfncis.de:
> Ok, what's the etiquette here? top posting or not?
Experience people tend to post relevent comments after each point so it
reads like a conversation (like this)
>Anyhow, sorry if
> i'm doing something miss manners doesn't approve of.
> A question which will make everyone gasp with shock and disdain:
> What
> *IS* RDH4? something like radio handbook 4th edition? Sorry -
> clueless. I used to be all proud of myself for having RCA RT book...
Radiotron Designers Handbook 4th edition. Many say it's the 'bible' for
valve lovers...
>
Tim
That's what I hear...
> I'd love to be able to do that, and
> am willing to bone up on tranny design, but what are the costs?
Not much if you have the junk.
> so I had a machine shop, and still have access to one
Geez, everyone but ME...
> Or should I get that thought 9homemade trannys) out of my head, and do
> something constructive, like play with matches in the traffic?
No, you just have to be very patient... If you wind by hand, you
might get between 1 to 2 seconds per turn (based on my experience),
so a primary section - say, 500T - you'll take a few hours solid
of hand-stressing turning for a transformer. Doesn't sound like
much but trust me you have to be patient...
I once tried winding a small power transformer, but the primary wire
was a little too thick, so the thing doesn't quite work right.
Shiva
kind'o like NJ7P's site, http://www.nj7p.org/, a very clean & friendly front
end for a very complete database. (the actual tube database addy (it's a
CGI) is http://www.nj7p.org/cgi-bin/tube?tube=.
-dim
"Fred Nachbaur" <fnac...@netscape.net> wrote in message
news:3CA49DC2...@netscape.net...
<snip> ...and more and more I find
> myself referring to online resources.
>
> Cheers,
> Fred N
Fred Nachbaur
> At this point, even if I have hardcopy, I go straight for the computer - i
> kind'o like NJ7P's site, http://www.nj7p.org/, a very clean & friendly front
> end for a very complete database. (the actual tube database addy (it's a
> CGI) is http://www.nj7p.org/cgi-bin/tube?tube=.
> -dim
Yes. I like Bill's site very much. Another great resource is Duncan
Munro's http://www.duncanamps.com . His TDSL (Tube Data Sheet Locator)
is absolutely terrific, especially due to the built-in hyperlinks to
online data sheets. (It's even more useful for people on dial-up modems,
for most tube types the built-in data and ratings tabs are enough to get
you going.)
Fred N
Patrick Turner
Shiva wrote:
> Ok, what's the etiquette here? top posting or not? Anyhow, sorry if i'm
> doing something miss manners doesn't approve of.
> A question which will make everyone gasp with shock and disdain: What
> *IS* RDH4? something like radio handbook 4th edition?
It is the Radiotron Designer's Handbook, 4th edition,
from the 1950s,
and is regarded as a sort of New Testament Tube Bible.
For serious understanding, its contents should be studied,
and the many references, which are listed, but which
now would be hard to find.
Patrick Turner.
Patrick Turner
Shiva wrote:
> Patrick, do you wind your own trannys?
Yes, I couldn't get what I wanted, so I built the lathe,
and started winding. It was the best move I could have made.
> I'd love to be able to do that, and
> am willing to bone up on tranny design, but what are the costs?
First you need a winding lathe, then start with chokes,
to become use to the process, and make the inevitable mistakes
on non critical gear, and get used to the wire tangling and breaking,
and things trying to stop you.
I could have trained to be an insurance lawyer, I'd have made millions.
I build darn tube gear instead, and you ask me about costs?
Its a funny question.
Real devotion never counts the cost.
> Is it
> possible to get the E & I iron in small quantities?
Maybe from where you are.
I can here in Oz. and small wire quantities too.
> At what cost?
Top grade E&I lams here cost about US $4.50 per Kg.
Wire is about the same.
But your own time to make bobbins, terminate properly, design,
varnishing or waxing, all come into the picture.
> I used
> to restore vintage racecars, so I had a machine shop, and still have access
> to one, so I could build a winding rig, but things like design & suppliers
> are a mystery to me (well, I mean, I've certainly destroyed/ripped apart
> many trannys as a kid, and, even as I write this, I have an output
> transformer from a PP pentode 6L6 -> 2 ohms tranny which I'm dissecting
> (very much easier than I thought, splitting the center tap is trivial, but I
> also found a bunch of primary taps which were not connected to anything, and
> the secondary consists of 3 separate windings wired in parallel - oh, the
> possibilities...
I built my lathe using a stout timber frame, and a motor in a box,
and a fan belt pulley system driving a 12 mm shaft to hold the bobbin,
and some lightbulbs to further control the motor speed.
Part of an odometer with a wheel sprung loded onto the drive shaft
counts the turns, in both directions.
I think it took a few days to make the lathe.
>
> Or should I get that thought 9homemade trannys) out of my head, and do
> something constructive, like play with matches in the traffic?
It depends what you want to end up doing.
There is always Hammond, but if you want perfection,
then DIY is all you can do cheaply,
but you gotta get that design way better than Hammond.
And this takes Knowledge, and Practice, and Devotion,
and Passion......
If I want a Hammond, the air fare is huge, and sea fare takes too long,
so I just wind my own.
I tried to get firms intersested in my designs, but they
all vomit in my lap when I show them the details.
That's why I know nobody will steal my designs
I have at my web pages, its all to hard to do.
But once upon a time, even more complex designs like
Williamsons was just routine winding work, and the
practised skilled tranny winder could do 3 or 4 a day,
with all carefully layered windings.
Patrick Turner.
Patrick Turner
Ian Iveson wrote:
> If Denis meant a few uH measured at the secondary, then I have no
> qualms about that. Easy. But I don't think he did.
He did spell out the LL being at the secondary, and compared it
to the typical SS amp which has a few uH to deliberately
decouple the amp from the load at HF.
He was quite correct about that.
>
>
> Even a well made transformer does not supply "perfectly linear" CFB,
> by virtue of the fact that it cannot provide "perfectly linear"
> *anything*. I would also take some convincing that "perfectly
> linear" CFB is what is actually required. I don't even know what
> "perfectly linear" feedback means. With respect to what, the
> voltage output signal at the anode?
The THD generated in the OPT will get fed back, and this
will get corrected to some extent depending on the gain and
beta, the fraction of the total output fed back.
So even if the feedback delivery vehicle has some contributary
THD, it gets corrected.
Resistor networks tend to have less linearity problems, and no
parasitic reactances either.
>
> You see, in the basic debate here about CFB pentode versus triode,
> even if we accept the marginally-acceptable notion of inherent
> feedback of which Denis speaks, we would tend to come down on the
> side of whichever arrangement delivered the most appropriate
> feedback by the most direct and speedy route. In this context,
> "perfectly linear" CFB sounds like a claim to have found the Holy
> Grail. If it is perfect, then the triode can't beat it.
Well, nothing is perfect, so Triode amps will always have their place
in a complicated world, and for forgetful tranny winders, who forget
to put screen taps on the OPT.
>
> But of course it is not perfect. What we end up doing, as always,
> is comparing one imperfection with another. The best amplifiers,
> IMO, achieve greatness by combining a number of imperfections in a
> way that sounds best.
Yes, imperfections are inevitable.
Maybe the best amps combine the best amounts of positive things,
such as decent transformers, tubes, simple effective design,
rather than some deliberate consideration of a combination of negatives,
which are what imperfections are.
As long as the good outweighs the bad,
but again, personal preference comes into it.....
> I think we all know this. This means that
> extolling the virtues of one *stage* at the expense of alternatives
> is daft.
>
> For example. I use CFB and UL very successfully. I was drawn to
> CFB and the use of parallel output valves because I wanted no
> overall nfb, even though my speakers are not easy. OK, so I get a
> wonderful output stage. But since I have no overall nfb, then I
> must pay particular attention to how I reduce distortion before the
> signal gets to there, and this is made a bit harder by the fact that
> I need a bigger signal.
>
> I also think, now I'm in the mood, that claims concerning an
> "optimum" proportion of CFB are silly. There is a fair amount of
> pseudo-analytical claptrap circulating here.
You could be right about the claptrap.
I have heard and built splendid amps with Zero CFB.
Even darn triode types.
But CFB is an interesting thing to optimise as an exercise.
Its not about being always right, and winning, its
all about taking part, and partaking of the wonderment of the
simple electronics, and enjoying some music.
I have some EL509 tubes, and one distant day when I get time,
I might see if I can duplicate Denis's views on CFB using this tube with
CFB.
It does seem like a very suitable tube for this.
I have been right through all this business of CFB, years ago,but
as long as the driver amp and OPT are quite blame free,
there is no NEED for CFB, but it is an interesting thing
to get working right.
Patrick Turner.
Ian Iveson
> to the typical SS amp which has a few uH to deliberately
> decouple the amp from the load at HF.
> He was quite correct about that.
That's true. The post actually implies a LL at primary of 2.5mH at
most. I would still be interested to hear how Denis does this. Or
from Denis at all, really.
We seem to agree on the rest. My CFB UL amps are nice. Many other
designs are nice too.
regards, Ian
"Patrick Turner" <in...@turneraudio.com.au> wrote in message
news:3CA58D3D...@turneraudio.com.au...
Patrick Turner
Ian Iveson wrote:
> > He did spell out the LL being at the secondary, and compared it
> > to the typical SS amp which has a few uH to deliberately
> > decouple the amp from the load at HF.
> > He was quite correct about that.
>
> That's true. The post actually implies a LL at primary of 2.5mH at
> most. I would still be interested to hear how Denis does this. Or
> from Denis at all, really.
I am sure Denis would go on and on and on about CFB
for a long time with just a couple wind ups from the wind up key.
But how does one get LL down to 2.5 mH at the primary?
Well, Williamson had 4400 pri turns, and 5 x 4 interleaved sections.
LL was about 13 mH
But if the turns were halved, and the core size and turn length
and geometry was kept constant, then the F saturation at full power
would move
from about 10 Hz to 20 Hz, and the LL would be reduced
by the square root of the primary turn reduction,
and as turns are x 1/2, LL is reduced to 1/4 of what it
was at 4,400 turns, so LL would be about 3.25 mH.
With a core of twice the Williamson cross sectional centre leg area,
ie, 44 x 64 mm, we could reduce turns by half again, to 1,100,
and still get Fsat down to 20 Hz.
But because the turn length is longer, the LL rises,
although the less turns reduce it, as well as the greater winding
traverse width
of the winding using the larger core, because with a larger core size,
the window length would be greater.
So to get better Fs, and get better power, turns need to be raised,
and at about 2,000 turns on a 44 x 64 mm large core, with 4 x 5
interleaving,
we get LL consistantly around the les than 5 mH figure,
which allows us to use a 5 kOhm load, and get wide bandwidth.
The larger core and fewer turns allows lower winding losses.
It will allow at least 50 watts, four times Williamsons amp.
I am currently winding two OPTs with a 110 mm stack of 52 mm toungue
GOSS E&I lams with max mu of 17,000, and
the tranny is for a 350 watt amp, using 12 x 6550 UL.
The optimum load is 1.4 k to 6 ohms.
With 1,030 primary turns and 5 x 6 interleavings, I can get LL
to just under 1 mH, and Fsat down to less than 15 Hz, at full power,
and the max primary L is still very, very high.
In fact the iron Mu is so high for this wonderful aussie
product I sourced that I may have to strip the whole first OPT down and
place the E&I into the bobbin in bunches of 4 at a time, so inserting
a slight effective gap, to better control saturation at LF.
Winding losses are less than 5%.
So although huge OPTs like what I am winding have been said to be hard
to wind,
and difficult to get low LL, it is not true that it is impossible.
The Ra of the PPP 12 x 6550 will be about 800 ohms a-a,
so capacitance effects will not be a problem because
I am using the same 0.7 mm distance between P and S
windings as I would use in a two tube tranny, where UL Ra-a
is about 5 kOhms.
But the low LL and the Cshunt will still resonate at some HF,
not sure until I see how it goes in tests.
It is important to get both LL and shunt C to low values
in OPTs, so that low phase shift at HF is possible.
> We seem to agree on the rest. My CFB UL amps are nice. Many other
> designs are nice too.
I agree, the implementation of the idea is crucial to design success,
in puristic terms, but for marketting success,
it pays to throw out perfectionist ideas, and
just use con artist tricks.
There was a design in WW in 1957, which had 10 x KT88, UL,
and it was rated for 400 watts. Pity that the cut off of the large OPT
was rated for 75 Hz, which says something about turns used.
The LL was only 0.75 mH, Lp only 4.5 H, which seems very low.
One sees these old designs, and they always leave something out,
like extended bass response.
Nevertheless, for use at a stadium PA system it would have been OK,
and not bad for breadboard construction, a large peice of 20 mm
plywood was used for a chassis.
> regards, Ian
>
Patrick Turner.
Denis
The intenal structure of a triode cannot be modified, and the internal
feedback is here.
There is just only way to design a good amplifier with a triode: To
understand clearly and quantitatively the physics of its action.
Of course there is published information on the dynamic performance of
the tubes, for instance, almost complete data are available for the
dynamic operation of 300B. Some pentodes too. Why a lot of people make
the same mistakes?
> > Perhaps, the pentode coud be simpler to simulate.
> >
> > Now I'm working on more realistic tube models, and I use MathCAD, and
> > not Spice.
> >
> > The Spice may be helpful in analysing the passive components, like the
> > transformers providing your equivalent circuit is correct.
>
> Yes, and that's the hard part, knowing what that
> equivalent circuit actually is.
>
Particularly, this is difficult for the output transformers. There are
several equivalent circuits with different placement of equivalent
capacitances.
I'm using an individual approach to each transformer I design.
>
> For working out triode behavoiur, there is NO NEED
> to take into account what Denis is trying to say.
>
Indeed, this is STRICTLY NEEDED to consider.
There are operating conditions, wehere the internal fedback clearly
shows its teeth. For instance, the operation of a triode power stage
with loudspeaker load at the presence of an input signal with high
crest factor(the music one).
> And as RL gets larger and larger,
> until the load is a constant current source,
> the linearity of the triode is at its maximum.
> And most triodes with loads which are many times the Ra,
> or approaching a CCS in value, are very linear.
>
They would be far better, if the internal feedback would be perfectly
linear, i.e., like the CFB.
> On page 511,
> The RDH4 has a test between a 6SJ7 connected as a triode
> or as a pentode and the THD for either is shown on a graph.
> In this, the pentode has less THD up to 30 vrms output,
> where the two tubes have equal THD,then the
> triode is much better by 60 vrms.
>
> But the graph is a bit strange because the triode THD
> is shown increasing along a straight line, which
> most triodes do not do.
The straight line increase of THD with output level is a clear
indication of @H as a prevalent distortion product. Any triode has
such behaviour of THD except the clipping region.
The pentode in the comparable operating conditions has sufficiently
lower 2H, but often somewhat higher 3H, etc. Therefore the THD vs.
output voltage curve will be like something between the straight line
and square parabola.
>
> It depends what sort of triode or pentode is used, and how
> it is used, to get the least THD for any given application.
>
Yes , there are certain differences between the performance of
particular tubes, but there are fundamental physical principles, which
are universal for all tubes.
>
> In all my experiments with CFB in output stages,
> if one simply applies enough CFB to reduce the effective plate
> resistance down to that of the same tube connected as a triode,
> then the triode is often just as linear as the pentode.
> If it isn't, its because the triode has only more 2H THD than the pentode with CFB,
> and this pentode will always have more odd order THD, reduced by half,
> by CFB, but still there.
> Meanwhile the triode has very low amounts of odd order THD....
Indeed, the pentode can have almost zero 3H.
Even when comparing the simplest power stages, the SE class A ones,
the pentode with CFB can always work better(in optimal mode) than the
same pentode in triode connection, to say nothing about UL.
Moreover, that kosher 300B has nothing to say against the good CFB
pentode SE stage with comparable power. I do not discuss here the
efficiency.
>
> But if the OPT and driver amp have wide bandwidth,
> and low phase shift, there is nothing wrong with
> simple plate loaded pentode stages with just
> one good loop of global NFB applied.
>
This is probably the simplest way to get a good sound for little
money. I adopt this approach too. One of my amps (OMAK AMBASSADOR)
uses EL84s in plain pentode PP, and the Concertina phase splitter, and
operates in pure class A up to 2 Ohm load. Meanwhile, the OPT has 150
kHz bandwidth.
> > The screen grid eats some portion from the total cathode current, and
> > influences the harmonic spectrum of the plate one. From other hand,
> > Ig2 depends on Eg2, and one can tailor the transfer curves of a
> > pentode by adjusting the screen voltage. Generally it is possible to
> > lower the open loop distortions of a pentode by factor of two, or even
> > more, in respect to those of a triode.
>
> It all depends on how much CFB is used.
> I like using some UL screen taps as well as CFB,
> I found this improved the spectrum content of THD,
> but did not reduce the power very much.
>
I mention here the open-loop distortions, i.e., those before
application of CFB or another feedback. This is exceptionally
important, for instance, at transient reproduction, when the dynamic
overload situation will take place.
> >
> > Note: The open loop distortions of a triode are solely defined by the
> > Gm nonlinearity, and they always are as high as 15% for 100% of class
> > A current swing.
>
> Wrong, Denis, wrong.
> Triodes set up properly in PP class A
> output stages will give only 1% at a dB below clip.
> The alarmist figures which you quote above
> are simply irrelevant to most folks.
> In SET, most amps are set up for a load which yeilds
> 5% at clipping.
Wrong, Patrick, wrong!!!
When I say: the open-loop THD for a triode, I suppose the internal NFB
is removed. When it is a SE triode stage at full class A current
swing, the respective distortion components are distributed the
following way: 14.4% of 2H; 1.6% of 3H, 0.6% of 4H, totally 14.5%.
These are the ideal figures, any real triode has at least a bit higher
figures.
When a triode sees some finite load, an internal NFB takes place and
lowers these distortion. For instance, when a 300B sees a 2.5kOhm load
(a standard mode, Ep=350V, Ip=80mA), the typical THD =6.5%, but the
amount of internal NFB is 1+Rl/Ri=1+2500/700=4.57, and one would
expect THD=3.17%. The rest is determined by the nonlinearity of
internal NFB (Mu=F(Vg,Vp).
Regards,
Denis.
Denis
> > Be aware of Spice! There is no accurate nonlinear model of a tube,
> > either triode or pentode. The existing models are limited in
> > applications, for instance, even the best known triode model
> > (Duncanamps)
>
> Known by who? How do they qualify for "best known model"?
I mention the Generic Triode Model by Duncanamps.
At least, this model gives rasonable predictions for a resistive load
operation(for instance, for 6SN7).
The full approximation of a triode is a problem of great complexity.
It is essential to represent a triode as a current amplifier with a
loop of voltage NFB. This latter is a greatest problem, as this NFB is
nonlinear.
>
> They all have cathodes. That is the hard part. Pentode introduces
> more complications.
>
The only complication for a pentode is the current redistribution
between the plate and the screen grid. Indeed, this mechanism can be
relatively easy to simulate using such tools of nonlinear math as
cubic spline approximation. The internal feedbacks, which are present
in the pentodes too, have a little effect and could be omitted as far
as the screen grid is kept at constant potential in respect to
cathode. The things again become much complicated, if one applies some
feedback signal to the screen grid, i.e., in the UL configuration.
> > Now I'm working on more realistic tube models, and I use MathCAD,
> and
> > not Spice.
>
> Would you care to tell us what the functions are, Denis. Perhaps
> you could also explain why they cannot be implemented in Spice?
>
I use MathCAD simply because it is my main instrument for all math
work. This software also contains a powerful set of nonlinear
procedures, especially cubic splines. All analysis is clear and
transparent from the math side, and I can alter my model at every
point and see what happens.
I don't exclude the possibility to create such model for Spice too.
> > Indeed, Menno Van der Veen has made some mistakes too.
>
> Perhaps you could give an example?
>
The steady-state operation into resistive load has nothing common with
a real operation of a power stage with loudspeaker load. Such
parameters as Q factor and Tuning factor are of little use.
Instead, the performance of an OPT over a whole range of loads, from
open circuit to shorted output is to be considered.
The design principles for a perfect OPT must take into account all
parasitic parameters, like LL and various capacitances as the
independent factors.
The choise of toroidal cores is strange too.
First, it is much simpler to reduce the LL to the desired value by
using the conventional coils. The EI- or double-C cores allow one to
make the huge cross-section thus reducing the number of turns and the
overall winding thickness. By the way, the leakage inductances are
proportional to number of turns squared.
Second, it is essential to have an airgap in both SE and PP OPTs.
The toroids also can be airgapped, but all their magic will go.
The stary fields are not the majour issue for well-designed double-c
or EI-cored transformer.
I use exstensively the double-C cores, because of their evident
advantages: optimal dimensions, easely inserted airgap, and high
magnetic properties.
Finally the double-C output transformers are much more compact than
the toroidal ones with similar performance.
> I would also be interested in how you manage to get "perfectly
> linear" CFB, and how you reduce primary leakage to a few uH.
The perfectly linear CFB means its transfer factor, beta is solely
determined by the ratio of turns. In contrary, this factor of the
internal NFB of a triode is a nonlinear function of Vg and Vp, and
there are several physical mechanisms, which are responcible for this
nonlinearity.
I meant the total LL referred to the SECONDARY. If referred to the
whole primary including the CFB sections attached, it is 1200uH=1.2mH,
or if referred to the half of the primary together with its CFB
section, it is 300uH.
The turn ratio of this transformer is 20:1(whole primary to secondary)
or 10:1(half primary to secondary).
This is quite low. In fact, the most important factors determining the
HF response of my transformers are the capacitances.
Regards,
Denis.
Shiva
nice to see the <turning read with shame> "typical application" data & plate
curves... For some reason, he decided not to include NJ7P's site when i hit
the "links" button. Nice for pinouts of unknown tubes, though... And it's
free! (well, there's the annoying add at the bottom, but I just didn't have
the heart or motivation to fire up a debugger and kill it.)news:3CA53B66...@netscape.net...
Shiva
(mostly to my guru/mentor/the guy who helped me when I first started an amp
repair business. He gave me parts at cost and advice & hanging around
spewing vintage amp trivia for free). Sam's manuals that i found in a
library dumpster (we've got a great library here, there are no books (you
won't even find Dante or Dostoyevsky, you won't find any Sam's or Chilton's
(auto equivalent to Sam's)), everything has been replaced by coffee table
books & "book of the month" dealys.... And it's been closed for over a
year, since the town is spending 3.5 mil on "renovation" (this doesn't mean
books, BTW). I go to a much poorer town's library, which has a *great*
collection of everything, from dated techno books to everything by Burgess,
obscurities like Huissman and Compte de Lautreamont (hey, I can't
spell...)).</self pitying reminiscence>news:3CA583B7...@turneraudio.com.au...
Bob Groschen
> If I want a Hammond, the air fare is huge, and sea fare takes too long,
> so I just wind my own.
> I tried to get firms intersested in my designs, but they
> all vomit in my lap when I show them the details.
Well now *that* would certainly give you a case of
"smelly crotch".... :-)
> But once upon a time, even more complex designs like
> Williamsons was just routine winding work, and the
> practised skilled tranny winder could do 3 or 4 a day,
> with all carefully layered windings.
Ah, for the good old days.... :-)
--
Best Regards,
Bob Groschen
Tim
Let's have a toast of home-made non-Bourbon then! ;)
Patrick Turner
Denis wrote:
The most difficult issue here is the nonlinear
> > > internal feedback, which is inherent for any triode.
> >
> > But one never needs to take this aspect of a triode into account
> > when trying to set up a circuit.
> > You just set up the circuit, and everybody knows
> > its no use worrying about the self regulation of a triode, because
> > that is the result of the way theu made the tube,
> > and you cannot change or alter the insides of a tube.
> > So simple modelling excluding the idea
> > behind the NFB in a triode will work just fine.
> >
>
> The intenal structure of a triode cannot be modified, and the internal
> feedback is here.
> There is just only way to design a good amplifier with a triode: To
> understand clearly and quantitatively the physics of its action.
> Of course there is published information on the dynamic performance of
> the tubes, for instance, almost complete data are available for the
> dynamic operation of 300B. Some pentodes too. Why a lot of people make
> the same mistakes?
They try to get much power, and low THD, which is nearly impossible
with SET, with zero feedback loops.
If they ran the tubes cooler, and used two instead of one,
the THD is low, and so is Rout.
And maybe the 300B is then reliable....
>
>
> > > Perhaps, the pentode coud be simpler to simulate.
> > >
> > > Now I'm working on more realistic tube models, and I use MathCAD, and
> > > not Spice.
> > >
> > > The Spice may be helpful in analysing the passive components, like the
> > > transformers providing your equivalent circuit is correct.
> >
> > Yes, and that's the hard part, knowing what that
> > equivalent circuit actually is.
> >
>
> Particularly, this is difficult for the output transformers. There are
> several equivalent circuits with different placement of equivalent
> capacitances.
> I'm using an individual approach to each transformer I design.
This is the only way.
Sure, design carefully, but at the end of the day,
you get what you wind.
>
>
> >
> > For working out triode behavoiur, there is NO NEED
> > to take into account what Denis is trying to say.
> >
>
> Indeed, this is STRICTLY NEEDED to consider.
>
> There are operating conditions, wehere the internal fedback clearly
> shows its teeth. For instance, the operation of a triode power stage
> with loudspeaker load at the presence of an input signal with high
> crest factor(the music one).
We should agree to disagree.
I don't think you like power triodes much.
Many folk just do like them
Crest Factor?
Where is the reference to CF in RDH4??
>
>
> > And as RL gets larger and larger,
> > until the load is a constant current source,
> > the linearity of the triode is at its maximum.
> > And most triodes with loads which are many times the Ra,
> > or approaching a CCS in value, are very linear.
> >
> They would be far better, if the internal feedback would be perfectly
> linear, i.e., like the CFB.
Not necessarily.
Let's agree to disagree, shall we.
>
>
> When I say: the open-loop THD for a triode, I suppose the internal NFB
> is removed.
But the self regulating action within a triode cannot ever be disconnected.
It comes with the tube when you buy it, so what on earth is the relevance
of considering the removal of triode feedback when you cannot.
> When it is a SE triode stage at full class A current
> swing, the respective distortion components are distributed the
> following way: 14.4% of 2H; 1.6% of 3H, 0.6% of 4H, totally 14.5%.
> These are the ideal figures, any real triode has at least a bit higher
> figures.
> When a triode sees some finite load, an internal NFB takes place and
> lowers these distortion. For instance, when a 300B sees a 2.5kOhm load
> (a standard mode, Ep=350V, Ip=80mA), the typical THD =6.5%, but the
> amount of internal NFB is 1+Rl/Ri=1+2500/700=4.57, and one would
> expect THD=3.17%. The rest is determined by the nonlinearity of
> internal NFB (Mu=F(Vg,Vp).
So what, many folks prefer 300B sound, which they say is better
than anything else, and no amount of argument with measurements
and figures will change their minds.
And in any case why use only 2.5 k load for a 300B?
I would prefer 5 kOhms.
I have built SE amps, and I know some people like them
in preference to PP despite the fact the THD is 10 times
worse than PP, right across the power range.
Don't ask me why, but it happens.
Patrick Turner.
>
>
> Regards,
>
> Denis.
John Byrns
<in...@turneraudio.com.au> wrote:
> So what, many folks prefer 300B sound, which they say is better
> than anything else, and no amount of argument with measurements
> and figures will change their minds.
A little double blind testing should settle that issue once and for all.
> And in any case why use only 2.5 k load for a 300B?
> I would prefer 5 kOhms.
Watch it Patrick, that statement has been the cause of mega flame wars
here in "RAT" in the past. In fact I think it was at the root of the
largest ever flame war here. I guess that was before your time. At any
rate it is not a good idea to be advocating high impedance loads for
300B's here. On the other hand with luck you may escape serious flaming,
as all the religious faithful may have left the group.
Regards,
John Byrns
Surf my web pages at, http://www.enteract.com/~jbyrns/index.html
Patrick Turner
John Byrns wrote:
> In article <3CA6CBE6...@turneraudio.com.au>, Patrick Turner
> <in...@turneraudio.com.au> wrote:
>
> > So what, many folks prefer 300B sound, which they say is better
> > than anything else, and no amount of argument with measurements
> > and figures will change their minds.
>
> A little double blind testing should settle that issue once and for all.
If only double blind testing would really change anything.
You can prove there are no differences between some amps, and still
folks like a certain type, and maybe its all down to subconscious
psycho acoustic perceptions.
Who am I to argue.
Two years I demonstrated a standard UL PP amp with two x 6550 tubes per
channel
to the Audiophile Society of New South Wales, and at the meeting a guy had a
PP amp with 4 x triodes in each channel, but only 6 dB of global NFB
wheras mine had 16 dB.
Mine seemed to have more pizazz, and the triode amp
seemed to be toppy and harsh.
But levels were elevated, in a room where 30 of us listened.
Maybe at low levels in a more intimate listening session, the triode amp
may have fared better.
I have found this to be the case with PP triode amps,
with just 2 x KT88 / 6550.
But when we played some music through my SEUL amps with
half the power, using 13E1, they fell about, and they could tell there
was some remarkable sound going on.
I produce many different types of amps from time to time,
including SS. I enjoy music, and the engineering.
If there are differences, they may be just differences,
not necessarily meaning any one amp is better....
>
>
> > And in any case why use only 2.5 k load for a 300B?
> > I would prefer 5 kOhms.
>
> Watch it Patrick, that statement has been the cause of mega flame wars
> here in "RAT" in the past. In fact I think it was at the root of the
> largest ever flame war here. I guess that was before your time. At any
> rate it is not a good idea to be advocating high impedance loads for
> 300B's here. On the other hand with luck you may escape serious flaming,
> as all the religious faithful may have left the group.
Ah well, I don't want to create a silly war over the size of the RL for a
triode.
I have found, in my experience, that all triode amps like a high B+,
and highish ohmic load, either PP or SE, and I don't expect a single living
soul
to embrace that philosophy, but that's where I am.
The THD and Rout is lower if you do it this way.
But the OPT needs to have more turns, and then more interleaving,
and I can see where some folks would get upset about
me asserting that present transformers often used for SET amp kits
are not much good....
I am not a huge fan of the 300B.
I don't much like DHTs.
But, a lot of people do like them, and I have heard remarkably
good sound from these old tubes.
But I do know of a case where a guy has used 3 quads since 1995
in a PP pair of monobloc amps which means a heck of a lot of money spent.
I was asked yesterday to quote on OPT replacements, as one had developed
shorted turns.
Some guys go yachting, or racing cars, its all just there thing, eh.
The yachties don't winge too much when they tear
a $20,000 spinnaker to bits in a big blow,
just buy a new one.
I am building a pair of 350 watt amps at present, and it would be very
easy to simply use the speaker secondary for the source
of CFB to the 12 x 6550 output tubes.
About 45 vrms is made at the secondary, and enough for
a non trivial amount of CFB.
But NO. Too much trouble for doubtful sonic returns.
Audio Research do this, it doesn't mean I have to.
I have made it possible to use 20% of CFB,
by having one of the five primary section available for CFB if wanted.
Then the speaker windings are free to remain totally isolated from the
tube circuit, and with one side grounded.
And its easy to adjust the 12 secondary windings for matching
from 15 to 6 to 1.25 ohms with no unused sections of windings,
or increased losses.
I tested the first 14 Kg OPT today for these amps, and with
the 5P and 6S sections, I got 300 kHz of bandwidth,
with saturation at 23 Hz at full power.
It was only ever intended that 200 watts be produced,
but I like to allow for a bit extra, so in real usage,
saturation will never be a problem.
This is good enough result to simply use a normal loop
of NFB with standard 40% UL, and get
no undesirable delays in the application of NFB.
The total amp, including driver amp and output tubes, and OPT
will have an open loop bandwidth of about 80 Khz.
Then with some step networks, I will judiciously reduce
ultrasonic frequency gain, and end up with
very good stabilty and bandwidth,
and that magical clarity that tubes can provide.
I will not be relying on NFB to reduce much phase shift.
I am sure there are other ways to go about making
350 watt amps, but I will do it my way.
I have learned from the group over the last 12 months,
and talking with you guys has
led to my own broader understanding of the techniques required for
simple effective circuits to make blameless sound.
I also do double blind testing here whenever I build new amps.
I have a simple switch box to change the speakers
from one pair of amps to another.
I have 2 sets of identical outputs on my preamp to do this,
so the levels can be set previously to listening to exactly the same level.
Then I call in a friend or two, and ask them to work the switch
and try to choose which amp sounds "better",
or different, while they dunno which direction
of switch is for which pair of amps.
Sometimes it is impossible to hear any differences.
But I can always fool any listener, by adjusting the level
of one pair of amps to be 1 dB louder;
they will always pick that pair as being "more detailed"
Then I reverse the procedure, and see if they pick it, and sure enough,
the other amps are found to be better.
Then you know there isn't much difference.
Some amps don't need this treatment, they just sound terrible,
many cheaper SS amps are in this brigade.
And it isn't just crossover distortion.
I have altererd the bias on some class B SS amps,
while listening to music, from between real class B, where the devices
have maybe 1 mA of idle current to heavy class A bias,
and I have heard no change in sound, still bad as ever.
Generally it is all the other circuitry which seems to dry out the sound,
and leave it edgy, apart from the output stage, with all its NFB.
But some SS amps are very good I have found.
Patrick Turner http://www.turneraudio.com.au
BFoelsch
brought over one of his creations to demonstrate. It was an all-triode all
fixed bias design, with 2A3s and 56s. NO feedback or degeneration or
anything. Not even a cathode resistor. We listened to it and it sounded like
an amplifier, then we compared it to some of my McIntosh stuff, and my
friend immediately commented on the "harshness" of the Mac.
To make a long story short, the McIntosh was flat from 20-20kHz and dead
quiet. His homemade design was down 6db at 50 Hz, had a gentle peak at
around 2kHz and then rolled off down to -10dB at 10 kHz. Above 10 kHz it
fell off very rapidly, and there was nothing left at 15kHz. It also had a
moderate helping of hiss and hum. We won't even touch on THD and IMD,
although it wasn't TOO bad up to about 1 watt.
He designed his amp to suit his taste, rather than any standard of technical
perfection.
In the 1970s, I think it was Bob Carver who went around making one of his
amps sound like any brand you specified, by tweaking the characteristics
which he had so well researched.
There is nothing wrong with designing a product to suit your own taste. Just
be aware that taste is not science. I suspect that a lot of the "sweetness"
attributed to certain designs could be more economically realized with a
handful of resistors and capacitors.
"Patrick Turner" <in...@turneraudio.com.au> wrote in message
news:3CA726EB...@turneraudio.com.au...
Henry Pasternack
> here in "RAT" in the past. In fact I think it was at the root of the
> largest ever flame war here. I guess that was before your time. At any
> rate it is not a good idea to be advocating high impedance loads for
> 300B's here. On the other hand with luck you may escape serious flaming,
> as all the religious faithful may have left the group.
From the archives, February, 1998:
In article <34D51F...@indigo.ie>, an...@indigo.ie wrote:
[André's tirade about Henry snipped.]
I must say, when I hear André pontificating on the size of Henry's ego, it
is a serious case of the pot calling the kettle black. In fact after
having had a number of Mr Jute's writings paraded in front of me over the
last couple of months, I would have to say that in comparison with André,
Henry has no ego at all. It appears to me that Mr. Jute's ego is fully
the equal of the sum total of the ego's of all the other regular posters
here on r.a.t who are accused from time to time of having outsized ego's.
Audio systems have both a touchy feely aspect, and a scientific side to
their design. André seems not to have learned the lesson yet, that a
pleasing audio system requires both elements to achieve a successful
result. Both are essential, and neither overrides the other. From
reading his writings, it is clear that André comes to audio maters from
the touchy feely side of things. Like many people who put their entire
stock in the touchy feely approach, André seems to feel that allows him to
define the scientific side of audio design as he sees fit, without concern
for the facts.
It is interesting that André puts Henry down for supposedly presenting
formulas from the RDH. It is less clear if André is also trying to put
the RDH down. While the RDH has become somewhat of a cliché in audiophile
circles, the information it provides is generally quite accurate. It is
not at all clear to me that Henry got his formulas from the RDH. Anyone
who is competent with math, and starts from first principal's is going to
derive the same formula's that are in the RDH. It's not rocket science,
and is really quite elementary. I would be suspicious of anyone who came
up with a formula that is different than the one in the RDH, I would give
their work close scrutiny. Most of this stuff (formulas) has one right
answer, that has been known for a long time, and since the RDH has mostly
right answers, I would expect any formula relating to matters discussed in
this newsgroup to be identical to one from the RDH. Just because Henry's
formula is the same as the one in the RDH does not mean he got it from
that source. Most of these formulas can be derived from first principals
in 5 minutes, with a pencil and a sheet of paper.
Another way that André seems to like to insult people is by mentioning the
"Joenet" in connection with the person being insulted. Could someone tell
me what this "Joenet" is, and why its mention seems to be an insult?
It is amazing the amount of foolishness that is taken as fact in this
newsgroup. And it is even more amazing the level of the personal attack
against Henry, that André felt it necessary to go to when his error's were
pointed out by Henry. If André is correct, then why are the personal
attacks necessary? Why not mount a technical defense?
Regards,
John Byrns
Patrick Turner
BFoelsch wrote:
> Some time ago a fellow I knew was piddling around with amplifiers and
> brought over one of his creations to demonstrate. It was an all-triode all
> fixed bias design, with 2A3s and 56s. NO feedback or degeneration or
> anything. Not even a cathode resistor. We listened to it and it sounded like
> an amplifier, then we compared it to some of my McIntosh stuff, and my
> friend immediately commented on the "harshness" of the Mac.
>
> To make a long story short, the McIntosh was flat from 20-20kHz and dead
> quiet. His homemade design was down 6db at 50 Hz, had a gentle peak at
> around 2kHz and then rolled off down to -10dB at 10 kHz. Above 10 kHz it
> fell off very rapidly, and there was nothing left at 15kHz. It also had a
> moderate helping of hiss and hum. We won't even touch on THD and IMD,
> although it wasn't TOO bad up to about 1 watt.
>
> He designed his amp to suit his taste, rather than any standard of technical
> perfection.
>
> In the 1970s, I think it was Bob Carver who went around making one of his
> amps sound like any brand you specified, by tweaking the characteristics
> which he had so well researched.
>
> There is nothing wrong with designing a product to suit your own taste. Just
> be aware that taste is not science. I suspect that a lot of the "sweetness"
> attributed to certain designs could be more economically realized with a
> handful of resistors and capacitors.
Sweetness and deftness and precision that you hear with triode amps
at low power is very beguiling.
My AM radio has a triode output stage, and although
it has NFB, it clearly sounds different to any transistor amp.
I like it a lot.
Its my source of news in the kitchen,
but it really comes alive with rythym and blues.
Sure, the radio I built has been done to get a 10 Khz bandwidth,
and no more, and it lacks the full range to 22 kHz.
It is still listenable, and far better than some radios with 2 kHz of bandwidth,
for AM, and this is common, even tube ones.
And it plainly conveys the fact that some programs have been
networked, and pre-recorded, sent here and there in some complex
process before they reach my ears, and damn modern technology,
for the resulting distortion and compression is all to plainly
audible.
But I also like the seductive ease which my high powered PP UL amps
deal with any music, at any level, and yet remain "tubey",
which I think is simply the music getting through properly,
because when I go to live concerts, mainly of classical music,
I hear that warmth without a single amp anyplace around..
Sound gear is a bit like wine, there are the reds, and then the whites, and the
ports, sherries, and you name it, some are better than others.....
some we like, and others don't.
Patrick Turner.
Patrick Turner
Lord preserve me from the Merchants of Bull Twang.
Surely we are currently not under seige by the tiresome ego challenged
personalities, we could expect to have civil discussions
about the way we connect our electrodes, for quite some forseeable time,
could we not?
Patrick Turner.
Henry Pasternack
> Could somebody define what is "touchy feely audio design methodology"?
>
> Lord preserve me from the Merchants of Bull Twang.
>
> Surely we are currently not under seige by the tiresome ego challenged
> personalities, we could expect to have civil discussions
> about the way we connect our electrodes, for quite some forseeable time,
> could we not?
>
> Patrick Turner.
Hmmm. I thought if John felt the need to troll, it would be best to answer
him with his own words. An elegant solution, in my opinion. As for the
content of his early and recent postings, I have no comment.
-Henry
ee
Would the 5x4x1 sections : primary/sec/cathode winding be adequate?
Crest factor=peak/RMS of a signal.
EE
Choky
just piss off with under water brain damaging blablabla.........
--
Choky
Prodanovic Aleksandar
choky*remove*@eunet.yu
YU
remove *remove* to reply!!
John Stewart
BFoelsch wrote:
> Some time ago a fellow I knew was piddling around with amplifiers and
> brought over one of his creations to demonstrate. It was an all-triode all
> fixed bias design, with 2A3s and 56s. NO feedback or degeneration or
> anything. Not even a cathode resistor. We listened to it and it sounded like
> an amplifier, then we compared it to some of my McIntosh stuff, and my
> friend immediately commented on the "harshness" of the Mac.
>
> To make a long story short, the McIntosh was flat from 20-20kHz and dead
> quiet. His homemade design was down 6db at 50 Hz, had a gentle peak at
> around 2kHz and then rolled off down to -10dB at 10 kHz. Above 10 kHz it
> fell off very rapidly, and there was nothing left at 15kHz. It also had a
> moderate helping of hiss and hum. We won't even touch on THD and IMD,
> although it wasn't TOO bad up to about 1 watt.
Many years ago your friend's somewhat sick triode amp performance
was referred to as the "bar room sound". For the most part it didn't offend
anyone & after a few brews no one cared.
Cheers John L Stewart
BFoelsch
would guess those sets had a big boomy peak at around 100 Hz and rolled off
starting at about 3 kHz.
You are correct, that was indeed a fashion, particularly in inexpensive
consoles built in the 1950s. My parents had an early stereo Magnavox that
was a beaut. 1 35W4, 2 35C5 and a 12A?7. ( I think it was a 12AV7, but I am
not sure any more.)
The circuit was a hoot. It was stereo, but had 3 speakers; left, right and
center. It had 3 output transformers, left, right and center. The center
channel transformer was center-tapped, center tap to B+, and the two legs
provided the B+ to the left and right OPTs. This gave you an arrangement
where each 35C5 was loaded by its own OPT plus half of the center channel
OPT, which by the way had about .1 mfd connected across the primary. To
prevent the LF signals from cancelling in the PP winding of the center OPT,
the right channel was reverse phased at the crystal cartridge, and again at
the right speaker. Result was that the center channel was L+R up to 350 Hz
or so. Plenty of good bar room BOOM. Console had no baffling and no back.
Sound had no top, no bottom, but plenty of BOOM.
"John Stewart" <jh.st...@sympatico.ca> wrote in message
news:3CA75936...@sympatico.ca...
John Byrns
<he...@ginger.sri.com> wrote:
> "Patrick Turner" <in...@turneraudio.com.au> wrote in message
> news:3CA73FD4...@turneraudio.com.au...
> > Could somebody define what is "touchy feely audio design methodology"?
> >
> > Lord preserve me from the Merchants of Bull Twang.
> >
> > Surely we are currently not under seige by the tiresome ego challenged
> > personalities, we could expect to have civil discussions
> > about the way we connect our electrodes, for quite some forseeable time,
> > could we not?
> >
> > Patrick Turner.
Hi Henry,
It's good to know that a friend of the old gang is keeping an eye on
things here, to maintain religious purity as it were. I thought Patrick's
comments re 300B load lines were refreshing, given that I don't think he
was involved with the rec.audio.tubes newsgroup in the days of the "Wars",
and in any case is geographically remote from both Philadelphia, and
Cork. His was an interesting data point re the earlier discussion of Bob
C's Bubba 300 amp vs. Andre's sublime 300B amp.
> Hmmm. I thought if John felt the need to troll, it would be best to answer
> him with his own words. An elegant solution, in my opinion. As for the
> content of his early and recent postings, I have no comment.
It's hardly an elegant solution at all, its just plain disingenuous, but I
suppose that is what we should expect from someone who was trying so hard
to be admitted to the "gang of thugs". What you neglect to note, and what
really heated up the Magnequest Wars was that by fall I had fully realized
what the "gang" was up to, and switched to Andre's side, renouncing all
the negative things I had said about him previously. Actually I am amazed
at the insight into a number of matters that is displayed by that quote of
mine. While I initially joined the battle on the Magnequest side, before
I even knew what the "joenet" was, and later renounced my earlier negative
statements about Andre, I still would say that 95% of what I said was
right on, the only inaccuracies being some of the statements made about
specific personalities. About the only personality related statement that
I made, that I would say was actually true, was that Andre had a big ego,
but a lot of people have that, and I clearly wrongly characterized you by
saying that you didn't have a big ego. It was of course very
disheartening to see you, who had always stuck up for what was technically
correct, bend the technical truth on several occasions in order to "flame
Andre at any cost". Without going back in the archives, I believe I made
that post shortly after Mike himself had blasted Andre for having posted a
few kind words about Lundahl transformers, and before I realized who Andre
actually was. I had not previously noticed that the Magnequest partisans,
with the exception of Mike himself, had only been posting in
rec.audio.tubes for three or four months, and that the bulk of their
posting activity was directed at flaming any post of Andre's. They
claimed at the time that they were doing this because Andre had said
negative things about Magnequest, and he needed to be driven off the
internet, but a review of the archives will show that prior to Mike's
blasting Andre's Lundahl post, Andre had said nothing negative about
Magnequest in rec.audio.tubes. If Andre had said anything negative about
Magnequest at all, prior to that time, it was on the so called "joenet",
for which there was no publicly available archive. There was no good
reason for you and the others to make it your goal to run Andre out of
rec.audio.tubes, when he had not previously, before being ruthlessly
attacked, said anything negative about Magnequest here.
It is interesting to note that Mike, all of the gang, and various hangers
on, such as yourself, all cleared out of rec.audio.tubes once the stated
goal of driving Andre off the internet had been accomplished. You are the
only one we ever hear from, and you seem to mainly appear when someone
like Patrick inadvertently says something that goes against the party
line, and tends to support what Andre was saying.
Anyway, the issue of Bob C's low load theory of triodes vs. Andre's high
impedance load theory is an interesting one, and will probably never be
fully aired here, given the religious baggage it carries.
Henry Pasternack
> [Deleted]
As always, you portray a warped view of reality.
I'm not interested in debating your peculiar interpretation of ancient
history. All I ask is that you stop making provocative comments, like
this bit about the "religious faithful". Knowing your (IMHO unjustified
and unjustifiable) opinion of me, when I see comments like that, I take
them as insults. I'd rather not have the aggravation.
-Henry
John Byrns
Hi Henry,
If you didn't want the "aggravation", why did you join this thread in the
first place? Why are you insulted by my reference to the "religious
faithful", given that you don't even qualify as a true believer, and in
anycase I didn't even mention your name in the post where I spoke of the
"religious faithful". Your only connection with the "religious faithful"
seems to be your need to interject yourself into the discussion as some
sort protector of the "religious faithful". Why do you seem to have an
uncontrollable need to defend the purity of the religion when I point out
that someone like Patrick has made a statement contradicting one of the
fundamental beliefs of the "religious faithful"?
You have a good technical head on your shoulders, if you want to
contribute to the discussion why don't you address the issue that Patrick
raised when he said, "And in any case why use only 2.5 k load for a 300B?
I would prefer 5 kOhms", and forget your urge to be a religious protector?
Regards,
John Byrns
In article <a885k2$obl$1...@bob.news.rcn.net>, "Henry Pasternack"
<he...@ginger.sri.com> wrote:
Henry Pasternack
> If you didn't want the "aggravation", why did you join this thread in the
> first place? Why are you insulted by my reference to the "religious
> faithful", given that you don't even qualify as a true believer, and in
> anycase I didn't even mention your name in the post where I spoke of the
> "religious faithful". Your only connection with the "religious faithful"
> seems to be your need to interject yourself into the discussion as some
> sort protector of the "religious faithful". Why do you seem to have an
> uncontrollable need to defend the purity of the religion when I point out
> that someone like Patrick has made a statement contradicting one of the
> fundamental beliefs of the "religious faithful"?
>
> You have a good technical head on your shoulders, if you want to
> contribute to the discussion why don't you address the issue that Patrick
> raised when he said, "And in any case why use only 2.5 k load for a 300B?
> I would prefer 5 kOhms", and forget your urge to be a religious protector?
Optimal loading of a SE 300B is a subject that's been beaten to death more
times than I care to remember. I can't for the life of me imagine what I
might have to say about it that couldn't be found in the archives. In fact,
the same could be said about most topics of discussion on this newsgroup.
With respect to the so-called religious faithful, I don't believe there is
such a thing. This is just a simplistic characterization, like the "Gang
of Thugs" moniker, that you use to cast some people you don't like in a
bad light. I'm not loyal to anyone and I have no need to defend anyone.
The only thing I'm interested in defending is truth. I don't care if it's
the truth about how a circuit works, or the truth about how a personality
disordered lunatic came to rec.audio.tubes five years ago and turned one
of the group's most respectable members into a sycophantic zombie. And
I'm not describing Michael LaFevre and myself, so spare me the retort.
I deeply regret the Jute fiasco, and I feel deeply soiled for having been
part of it. Nevertheless, while the destruction of Andre's reputation was
inevitable and self-initiated, the battles served a purpose. People, most
notably you, forget the harm he caused. All those bitter arguments caused
Andre to reveal his true nature. As a result, whatever good will he had
created for himself was thoroughly and justifiably demolished. The internet
is a better and safer place because of it.
My own reputation seems to be somewhat the worse for wear, but I haven't
really tried hard to keep it up, have I? That's partly on purpose, and
so be it. Maintaining a reputation takes effort. It's not a priority for
me at the moment. Lurking on this newsgroup suits my purpose. I have
contributed thousands of pages of useful information to this group over
the years. I owe it nothing, and it gives me nothing in return. Fair's
fair.
I responded to you, not to Patrick. I have no argument with Patrick. He
is one of the most sensible people on this group, technically astute in
spite of not being a whiz with math. I like Patrick even if he doesn't
appreciate me. No, my reason for replying was simply because you irritated
me. The Jute fiasco leaves a bad taste in my mouth. You brought it up,
and for no good reason. I responded. Don't feign indignation.
Your loyalty to Andre Jute is irrational. If you were as critical of Andre
as you are of me, you would have to give up your sycophantic devotion to
him. This annoys me. I value order and sanity. Your presence on this
newsgroup disrupts my sense of things being in their place. It's not just
your affinity for Andre, but also your peculiar inability to be reached in
technical arguments. I am keenly aware that nobody but you and I has the
least interest in this disagreement. If the world doesn't end before then,
you and I may still be arguing about it thirty years from now. I hope not,
but I can see it happening.
There is only one solution, John, and that is for you to discover and deal
with the thing that is bothering you about Bob and Ron and Frank and
Michael.
Then you have to take off the blinders and come to terms with the sad
reality
of Andre Jute. When you do that, you and I will be in a position to talk.
In the meantime, I strongly urge you to behave yourself and avoid those
stupid little trolls. I and the newsgroup will thank you.
And finally, for anybody who objects to this posting, well, blow me.
Good night.
-Henry
Patrick Turner
The bobbin layout shown at
http://turneraudio.com.au/OutputTrans.htm
has 5P x 4S sections, but it can be
easily altered to make use of CFB.
There are a total of 16 layers of primary, so if we devoted two to
CFB, this would give 12.5% of CFB.
But if we used a low THD driver amp, we could devote
4 layers to CFB, and then have an effective 25% of CFB.
But which layers would we use out of the 16 for CFB?
If we wanted 12.5 %, then,
if we number the Primary layers starting from the bottom of the bobbin,
and work our way up,
then no7, and layer no10
would be best for the CFB.
Extra insulation between no7 and no8 and
between no9 and no10 would have to be used to
allow for the B+ voltage difference. Ends of these windings
would have to be taken out to terminations.
The reasor for no7, and no10, is that they are located between
the primary plate winding an the secondary,
so as much of the output message as possible
is sent to the output tube cathodes.
Should we want 25 %,
then maybe simply layers 7,8,9,10 could simply be devoted to
CFB.
Alternatively, layers 3, 7 10, and14 could be devoted to CFB,
but shunt capacitance will slightly affect the outcome.
It is important to use the layers for CFB which are
on the same side of the plate windings, for
each output tube, so that when the amp is used class AB,
coupling between the CFB winding of a given tube
is coupled closely to the plate winding of the same tube.
So it is important to consider the phasing of your
windings carefully.
ee wrote:
I have not used this term, crest factor,
it seems to not have had much significance to me.
But we all know peak voltage = 1.414 x the measured RMS voltage
of our voltmeters.
>
>
> EE
Patrick Turner
John Stewart wrote:
> BFoelsch wrote:
>
> > Some time ago a fellow I knew was piddling around with amplifiers and
> > brought over one of his creations to demonstrate. It was an all-triode all
> > fixed bias design, with 2A3s and 56s. NO feedback or degeneration or
> > anything. Not even a cathode resistor. We listened to it and it sounded like
> > an amplifier, then we compared it to some of my McIntosh stuff, and my
> > friend immediately commented on the "harshness" of the Mac.
> >
> > To make a long story short, the McIntosh was flat from 20-20kHz and dead
> > quiet. His homemade design was down 6db at 50 Hz, had a gentle peak at
> > around 2kHz and then rolled off down to -10dB at 10 kHz. Above 10 kHz it
> > fell off very rapidly, and there was nothing left at 15kHz. It also had a
> > moderate helping of hiss and hum. We won't even touch on THD and IMD,
> > although it wasn't TOO bad up to about 1 watt.
>
> Many years ago your friend's somewhat sick triode amp performance
> was referred to as the "bar room sound". For the most part it didn't offend
> anyone & after a few brews no one cared.
> Cheers John L Stewart
But had the tube been a pentode amp, similarly configured,
maybe no amount of Bourbon would improve the sound.
>
Patrick Turner.
Patrick Turner
John Byrns wrote:
A lot of what you say was before my time.
I am beginning to see how criticism of a commercial product
here on RAT could land someone in deep dodo,
because there are folks with a commercial interest who would
of cours hate any negative critique, since it would reduce sales.
The same sort of sensitivities lie with criticisms of
real estate developers, who rip down historic houses, and
erect garbage, and want everyone to wax lyrical about
their creations to generate sales.
I saw it happen in a posh area of Sydney where my old mum lives.
The developers sued people who wrote letters to the newspaper,
complaining about the greed and appalling style of greedy
types who seemed would sell their mother to make a buck.
They held sway for a year or two, but a nearly all green
member Municipal Council was then voted in, and for once we
had some sense applied to town planning, and no
more garbage foisted onto us by "the gang".
The council members who were in bed with the developers were out
on their arse, and good riddance.
The goings on here at RAT are not quite the same, there are no elections
about who is going to moderate any posts, so its up to us to do so.
I dunno If I helped Andre leave the group, it was not my first
intended wish when I saw his postings.
What got me was the level of unadulterated anger,
and tirades directed at us all for seemingly no reason,
rather than steadfastly sticking to the point of
all this, and that is to keep the craft alive, and facilitate
the spread of tube knowledge, with as little personal fuss as possible.
I have had to wear some criticisms about methods
of connecting electrodes of vacuum tubes,
but I won't write off somebody who disagrees with me,
but if they maintain an opposite position, I hope they
would save me from myself by proving they make a bit of sense.
I used to maintain there was no feedback in triodes,
until the obvious hit me in the eye so well, that I have to admit, triodes do
have inherent
self regualtion of their output voltage.
OK, so what?
Well, not much we can do to manipulate the properties of a
given triode once it has been made.
So, we can use whatever model we chose for its operation and
get by with considering it a low impedance device,
( yet to replicated in the world of SS )
I don't mind if folks want to use low value loads for their triode amps.
Normally this involves a low B+, and a high idle current.
All I would ask of them is that they try things with a high B+,
and a low idle current.
Maybe they might like the sound better.
Thre are a range of operating points for any output tube which can be chosen
to suit the load at hand, which may well be dictated by the OPT
available.
Good amps are built around the OPT, IMHO.
Especially for DIY types, where people get old OPTs, and have to
build around them, as they are not exactly alterable,
and are like the innards of a tube.
Patrick Turner.
Fred Nachbaur
Henry Pasternack wrote:
[...snip]
> ... I have
> contributed thousands of pages of useful information to this group over
> the years. I owe it nothing, and it gives me nothing in return. Fair's
> fair.
Indeed. And I just want you to know that I, for one, have valued your
contributions - and those of John Byrn. It truly saddens me that things
went as badly as they did, and hope sincerely that we can once again
enter into lively and informative discussions without fearing massive
flame wars laced with ad hominems and other unpleasantries.
> [... more snip]
>
> In the meantime, I strongly urge you to behave yourself and avoid those
> stupid little trolls. I and the newsgroup will thank you.
Yes, agreed. We already have one resident troll, and IMO that's one too
many.
> And finally, for anybody who objects to this posting, well, blow me.
No objection here. I think you've stated your point accurately and
without acrimony.
> Good night.
>
> -Henry
Good night, Henry, and John, and everyone else. Just before I read this
last post in the series, there was a tune playing on my DIY tube hi-fi,
by one L. van Beethoven based on a little poem by Schiller. It goes
something like this:
"O Freunde, nicht diese Töne! Sondern, lasst uns angenehmere anstimmen,
und Freudenvollere!"
(translation: O friends! Not these sounds! Instead, let us strike up
more pleasant ones, and more filled with joy!)
Cheers,
Denis
>
> But the self regulating action within a triode cannot ever be disconnected.
> It comes with the tube when you buy it, so what on earth is the relevance
> of considering the removal of triode feedback when you cannot.
>
Indeed, this self-regulating action, or internal feedback is often
disconnected during the operation of a triode power stage at dynamic
overload conditions. These conditions can be caused, for instance, by
the back-EMF generated by a loudspeaker. Sometimes the loudspeaker
wants to draw more current than an amplifier can provide at this
instant output voltage. The triode falls to regulate as all available
current is supplied to the load. Because the low Ri is a result of the
action of the internal NFB, the tride suddenly becomes a high-Ri
device, like a pentode, and a loudspeaker becomes undamped at this
instance. This situation is more likely to occur in an SE stage,
because the PP triode stages have more current headroom, and they are
always symetrical.
Also there are more situations when the NFB does not work. This
phenomen is worth consideration in any amplifier. Generally speaking,
to get teh best sound, one must make the amplifier as linear as
possible before applying any NFB. In this respect the pentodes are
superior.
> And in any case why use only 2.5 k load for a 300B?
> I would prefer 5 kOhms.
>
I would too. A lot of pilots may say: the best two-seater plane is a
four-seater without two rear seats.
> I have built SE amps, and I know some people like them
> in preference to PP despite the fact the THD is 10 times
> worse than PP, right across the power range.
> Don't ask me why, but it happens.
>
It is intersting to me how a 300B amp, but without those distortions,
will sound?
Particularly, when a standard, and even 2.5kOhm loaded 300B amp works
with some sensitive loudspeaker, perhaps >100dB, it operates in very
linear region for most of time(below 100mW), and all distortion it
produces is just a bit of 2nd harmonic.
Indeed, it is possible to outperform it, but one needs to solve a
number of major engineering problems, the transformers first.
Regards,
Denis.
Denis
> > He did spell out the LL being at the secondary, and compared it
> > to the typical SS amp which has a few uH to deliberately
> > decouple the amp from the load at HF.
> > He was quite correct about that.
>
> That's true. The post actually implies a LL at primary of 2.5mH at
> most. I would still be interested to hear how Denis does this. Or
> from Denis at all, really.
>
My transformers are even better. They have just 1.2mH of primary LL,
and the turn ratio is from 20:1 to 23:1 (full primary to secondary) so
the impedance ratio is from 400:1 to 529:1. The small primary LL is
achieved by making use of several factors, such as reduced number of
turns thanks to a large and high-quality double-c core, optimal
interleaving technique, and, finally, taking into account the geometry
factor.
The more important issue is the parasitic capacitance, or, to be more
correct, the system of capacitances. The complex winding structure
means the appearance of a number of capacitances at the different
places of the detailed eqivalent circuit. These capacitances can do
the twofold job: they may cause some unwanted resonances and
additional phase shift, or they may actually improve the frequency
response making it more stable at different loads. I've managed to
achieve the latter.
In fact, it is possible to extend the bandwidth of my OPTs above 300
kHz by making minor adjustments, but this bandwidth will be achieved
only for a particular(8Ohm)resistive load. I've preferred 180 kHz, but
for any load, and without resonance peaks.
More detail: the action of CFB extends the power stage bandwidth even
further, pehaps by an octave, because nearly half of the total LL is
enclosed into the local feedback loop, but the undistorted power
begins to decrease since 100 kHz for obvious reason. I've used also
just 3 db of global NFB, and the open-loop responce has -3dB at 70kHz,
this point is naturally defined by the first gain stage, which is made
of high-mu triode. The final response extends again to those 100 kHz,
and it is now the same for low and high levels. The circuit has great
stability margin and does not require the compensation networks
anywhere.
Regards,
Denis.
Patrick Turner
Denis wrote:
> Patrick Turner <in...@turneraudio.com.au> wrote in message news:<3CA6CBE6...@turneraudio.com.au>...
> >
> > But the self regulating action within a triode cannot ever be disconnected.
> > It comes with the tube when you buy it, so what on earth is the relevance
> > of considering the removal of triode feedback when you cannot.
> >
>
> Indeed, this self-regulating action, or internal feedback is often
> disconnected during the operation of a triode power stage at dynamic
> overload conditions. These conditions can be caused, for instance, by
> the back-EMF generated by a loudspeaker. Sometimes the loudspeaker
> wants to draw more current than an amplifier can provide at this
> instant output voltage. The triode falls to regulate as all available
> current is supplied to the load. Because the low Ri is a result of the
> action of the internal NFB, the tride suddenly becomes a high-Ri
> device, like a pentode, and a loudspeaker becomes undamped at this
> instance. This situation is more likely to occur in an SE stage,
> because the PP triode stages have more current headroom, and they are
> always symetrical.
> Also there are more situations when the NFB does not work. This
> phenomen is worth consideration in any amplifier. Generally speaking,
> to get teh best sound, one must make the amplifier as linear as
> possible before applying any NFB. In this respect the pentodes are
> superior.
First, hardly anyone I know runs their triode amps anywhere near
dynamic overload, ie clipping, and from my experience, the back emf issue
does not seem to be any problem, as observerved when a square wave is run
into speaker at LF.
A triode amp will cope just as well as any other.
Second, whatever upsets a triode amp will upset any other amp, pentode,
bjt, mosfet, etc, depending on how things are set up.
Second, pentodes are notoriously non linear without feedback,
and act just like a wild current source.
Sure, make the pentode amp as linear as possible without any feedback, including
CFB, but you still got a long way to go to get them to perform.
>
> > And in any case why use only 2.5 k load for a 300B?
> > I would prefer 5 kOhms.
> >
>
> I would too. A lot of pilots may say: the best two-seater plane is a
> four-seater without two rear seats.
>
> > I have built SE amps, and I know some people like them
> > in preference to PP despite the fact the THD is 10 times
> > worse than PP, right across the power range.
> > Don't ask me why, but it happens.
> >
>
> It is intersting to me how a 300B amp, but without those distortions,
> will sound?
But that is easy to find out about.
Just build two, set them up to power 3 ohms each,
and then bridge them.
Exit nearly all 2H THD, and then all you have is a little 3H.
I have never done this, and I havn't a clue if it sounds well.
> Particularly, when a standard, and even 2.5kOhm loaded 300B amp works
> with some sensitive loudspeaker, perhaps >100dB, it operates in very
> linear region for most of time(below 100mW), and all distortion it
> produces is just a bit of 2nd harmonic.
>
> Indeed, it is possible to outperform it, but one needs to solve a
> number of major engineering problems, the transformers first.
Maybe, but simple is sufficient for many people.
And some folks say the 300B sound can never be "outperformed".
Others will say 300B amps are gutless wonders.
It depends on speakers, the size of the house, listening levels,
lotsa things.....
Patrick Turner.
>
>
> Regards,
>
> Denis.
Patrick Turner
Denis wrote:
> "Ian Iveson" <ianives...@virgin.net> wrote in message news:<Gqgp8.17133$Ko1.3...@news6-win.server.ntlworld.com>...
> > > He did spell out the LL being at the secondary, and compared it
> > > to the typical SS amp which has a few uH to deliberately
> > > decouple the amp from the load at HF.
> > > He was quite correct about that.
> >
> > That's true. The post actually implies a LL at primary of 2.5mH at
> > most. I would still be interested to hear how Denis does this. Or
> > from Denis at all, really.
> >
>
> My transformers are even better. They have just 1.2mH of primary LL,
> and the turn ratio is from 20:1 to 23:1 (full primary to secondary) so
> the impedance ratio is from 400:1 to 529:1. The small primary LL is
> achieved by making use of several factors, such as reduced number of
> turns thanks to a large and high-quality double-c core, optimal
> interleaving technique, and, finally, taking into account the geometry
> factor.
Pardon me for asking, but could you post the full winding details of your
OPT, so we could all appreciate the way you achieve the wondrous figures
you claim.
I have my details at my website at
http://turneraudio.com.au/OutputTrans.htm
So perhaps we could all better analyse
your designs.
Of course it is necessary to have interleaving, and low LL and low capacitance.
But what is the core cross sectional area, at what anode to anode voltage
does the transformer saturate, and at what frequency?
If you just give the complete wind up details for the bobbin, we can work it all
out about your transformers.
>
> The more important issue is the parasitic capacitance, or, to be more
> correct, the system of capacitances. The complex winding structure
> means the appearance of a number of capacitances at the different
> places of the detailed eqivalent circuit. These capacitances can do
> the twofold job: they may cause some unwanted resonances and
> additional phase shift, or they may actually improve the frequency
> response making it more stable at different loads. I've managed to
> achieve the latter.
I agree about capacitances, and the necessity to keep them to a minimum.
And I agree, as you say, the complex quantities of LL and Cshunts
in a complex arrangement is extremely difficult
to predict, but of little concern, as long as the detrimental effects
of resonances are well up from the audio band, where they can be
minimised by careful critical damping.
The greater the number of interleavings is, the more complex things get,
and the OPT above 50 kHz resembles a stop band filter with
multiple undamped resonances.
> In fact, it is possible to extend the bandwidth of my OPTs above 300
> kHz by making minor adjustments, but this bandwidth will be achieved
> only for a particular(8Ohm)resistive load. I've preferred 180 kHz, but
> for any load, and without resonance peaks.
> More detail: the action of CFB extends the power stage bandwidth even
> further, pehaps by an octave, because nearly half of the total LL is
> enclosed into the local feedback loop, but the undistorted power
> begins to decrease since 100 kHz for obvious reason.
In other words, the CFB reduces the phase shift and attenuation caused by
the leakage L.
That is normal feedback action.
> I've used also
> just 3 db of global NFB, and the open-loop responce has -3dB at 70kHz,
> this point is naturally defined by the first gain stage, which is made
> of high-mu triode. The final response extends again to those 100 kHz,
> and it is now the same for low and high levels. The circuit has great
> stability margin and does not require the compensation networks
> anywhere.
That's fair enough, it sounds like the total amount of NFB ,
ie, CFB and loop FB is not very high, and hence
stability would be easy to achieve.
I like to have the open loop phase shift in the tube section of any amp
to be less than that of the OPT.
But I also like to step the extreme HF bandwidth down,
so that when the HF step circuits are included,
the open loop bandwidth is not 100 kHz, because if it is, and we apply
any sort of feedback, some loads may cause oscillation, such as
capacitive loads.
But at 150 kHz, the phase shift between V1 and V2 is reduced,
because the network allows the
miller effect of V2 to be shunted by a low resistance
When NFB is added, the response is then
levelled so -3dB is 65 kHz, and phase shift reduced to negligible amounts, and
by the time one gets to 65 kHz, the amp gain has been lowered,
where there are no audio signals, and the amount of applied feedback
has been reduced, so stability is easier.
The networks for achieving this are at
http://turneraudio.com.au/50-50StereoampKT88.gif
If you take a look at the set up between the power amp input stage,
and the LTP driver, you can see what I mean.
Williamson did the same in 1947....
And it stops many amps from oscillation, if the values chosen are just right,
and no more or less, depending on the quality of the OPT.
There are suitable principles for networks in the last pages of the
1957 GE book, 17 amp circuits from 5 to 1,100 watts.
But what of the LF performance?
Patrick Turner.
>
>
> Regards,
>
> Denis.
Ian Iveson
> I used to maintain there was no feedback in triodes,
> until the obvious hit me in the eye so well, that I have to admit,
triodes do
> have inherent
> self regualtion of their output voltage.
You were rather less wrong then than now.
cheers, Ian
John Stewart
an earth shaking design. If you think of it the quick roll off in the base
must have been a pretty good rumble filter for the low budget turntable.
The roll off just above a duck's quack would have removed the nasty
Intermod. Products & "dissonant" Harmonics.
The slight rise at 2 khz could have been useful as a "brilliance" effect.
Perhaps we have all missed the boat here!!!!!!!!!!!!!
John L Stewart
Denis
> What is an appropiate geometry for a CFB tranny?
>
> Would the 5x4x1 sections : primary/sec/cathode winding be adequate?
>
Four sections of secondary is not enough, the leakage inductance will
be too high.
The Golden Middle is 7 or 8 single-layer sections, all in parallel.
Use only one- or two-layered sections in the primary.
The CFB section is actually the part of the primary, it is better to
place it somewhere in the middle of the structure.
It is better to use somewhat thicker insulation between all layers.
This helps reducing the capacitances.
For EL34 12.5% of CFB will be OK, perhaps it worth incerasing up to
16%, but no more(Ep=350V). The optimal Eg2 is to be figured out by
means of experiment, it could be easy if one installs a voltage
regulator for this purpose.
The optimal geometry also means another thing: the optimal coil
dimensions.
The minimal LL for a given transformer coil is achieved whan its
dimensions are quite definite. If the core window has other
dimensions, there will be even impossible to achieve the desired
performance even with increased number of sections. The parasitic
coapacitances will increase too thus stopping the further HF
exstension. Moreover, these capacitances have additional importance
for SE design: one should bear in mind the limited plate current. For
a single EL34 it is important to keep total C<1000pF. This will
provide the full-power bandwidth of at least 50kHz, and no
considerable increase in THD at 20 kHz.
The precise coil dimensioning depends on various factors. The good
results are likely to be achieved with the window aspect ratio of 3:1
(height to width).
Regards,
Denis.
Ian Iveson
Ah, C-cores!
You confirm most of my findings, although your transformers are
better in the respects you relate than my EI examples. I wonder how
necessary this is? Although I haven't got much test equipment, it
seems my cfb+ul amps are good for 10-80kHz (-1dB), and show little
distortion throughout this range at full power.
I am very interested in your "detailed equivalent circuit".
Coincidentally, I was building a new model yesterday, before an
application crash wiped out my efforts. I can figure out where the
capacitances are, but measuring them and confirming my analysis is
problematic. I would be very grateful for any assistance in this.
I find that I need no corrective networks, mostly because of the
capacitance between anode and cathode windings. That seems to have
been a good move (thanks in part to Brian Sowter!).
I wonder if you can shed any light on the issue of differential
phase distortion? That has been a residual concern for some time.
I have been told not to worry about it, but nonetheless I would be
interested in your view, and particularly in any measurements you
have done. It seems to me to fit somewhere in the CFB pentode
versus triode debate.
Oh, and I would still like to know where van Veen's errors are!
best wishes, Ian
"Denis" <denis_af...@mail.ru> wrote in message
news:32f1783d.02040...@posting.google.com...
Patrick Turner
Ian Iveson wrote:
Yeah, funny how there has been this reaction
to my changed perceptions.
Zero NFB triode amps still exist, except that by these words, we mean
no loops of external feedback are applied.
What goes on inside a triode is mainly its own private business,
and its all done very much more elegantly than
men with wires, iron, resistors, and compensation capacitors.
The mystical part about a triode is why the plate voltage
changes when there is NO current change, when
a CCS load is used for an anode load.
It then sure shows they are voltage devices, and quite linear.
But on the other hand, pentodes have wild amounts of gain
when there is a CCS load used, and very non linear.
And they need a load to keep them linear, and respectable.
Maybe Madame Neggy Feedback, with her lash, too.
They are current devices.
Patrick Turner.
>
>
> cheers, Ian
Patrick Turner
Denis wrote:
> ee <ee1000@*not*hotmail.com> wrote in message news:<fkIp8.1091$hS3....@news0.telusplanet.net>...
> > What is an appropiate geometry for a CFB tranny?
> >
> > Would the 5x4x1 sections : primary/sec/cathode winding be adequate?
> >
>
> Four sections of secondary is not enough, the leakage inductance will
> be too high.
Not necessarily, it depends on the overall size of the transformer, and load used,
and performance desired.
A small size OPT for a SET with say EL84 may not need more than
2P x 3S sections.
Then the amount of NFB used changes interleaving
requirements.
With nearly no NFB, extremely low LL
with a tube amp is not necessary,
if we can achieve good enough bandwidth .
But I do like wide bandwidth anyway,
and the designs I have will be adequate for most folks.
> The Golden Middle is 7 or 8 single-layer sections, all in parallel.
> Use only one- or two-layered sections in the primary.
Then shunt C is too high.....and precious winding space in the
window is taken up with insulation, which must be good to
allow use of high voltages.
>
> The CFB section is actually the part of the primary, it is better to
> place it somewhere in the middle of the structure.
>
> It is better to use somewhat thicker insulation between all layers.
> This helps reducing the capacitances.
The self capacitance between layers of primaries is negligible.
If the OPT has say 5 sections of P, each with 2 to 4 layers,
then the self capacitance of each section might be
100 pF, But the sections are in series, screened off from
each other between S windings, and so self C is in series,
and across a whole OPT, will be only 50pF, much much less
than the P to S shunt C.
I use thin insulation between layers of primaries, but much thicker between
P and S windings.
>
>
> For EL34 12.5% of CFB will be OK, perhaps it worth incerasing up to
> 16%, but no more(Ep=350V). The optimal Eg2 is to be figured out by
> means of experiment, it could be easy if one installs a voltage
> regulator for this purpose.
Unless you have an Ep far above the voltage rating for Eg2,
I would suggest the screens Eg2 can stay the same as Ep,
which is how they designed the tube.
Musical Reference use 700 volts for Ep, and 350 for Eg2,
and some CFB, and 14 k a-a, and they get 36 watts
from 2 x EL84. That's what I measured.
>
>
> The optimal geometry also means another thing: the optimal coil
> dimensions.
> The minimal LL for a given transformer coil is achieved whan its
> dimensions are quite definite. If the core window has other
> dimensions, there will be even impossible to achieve the desired
> performance even with increased number of sections. The parasitic
> coapacitances will increase too thus stopping the further HF
> exstension. Moreover, these capacitances have additional importance
> for SE design: one should bear in mind the limited plate current. For
> a single EL34 it is important to keep total C<1000pF. This will
> provide the full-power bandwidth of at least 50kHz, and no
> considerable increase in THD at 20 kHz.
Agreed, perhaps you would like to post a typical wind up detail
to show what you really mean.
> The precise coil dimensioning depends on various factors. The good
> results are likely to be achieved with the window aspect ratio of 3:1
> (height to width).
Very nearly all standard pattern laminations, and C-cores
have windows with 3 to 1 dimension ratios.
Most E&I lams are wasteless pattern,
so that the Is are cut from the two slots of two Es
back to back.
So it hard to find anything else but 3 to1 ratio windows.
6 to1 ratios could be used, by butting Es together, and running
lots of Is across, and the LL would be reduced
because the traverse width of the winding is increased,
hence half the interleaving could be used, but Cshunt rises,
and magnetic length increases, so best give this idea a miss.
I getv good results where the stack height
is up to 2.5 times the toungue width.
But we have only scratched the surface of what makes a decent OPT.
I am sure a picture would speak a thousand words,
so we need your posted full design details for full appraisal.
Patrick Turner.
>
>
> Regards,
>
> Denis.
Denis
> I have just been reading my copy of the
> September 1952 Wireless World
> article about "Amplifiers and Superlatives ",
> by none less than D.T.N. Williamson, and P.J. Walker.
>
> For those not familiar, Williamson was the guy who got famous
> for the amp named in his honour,
> and Walker got famous with the Quad range of products.
> The two blokes were obviously not enemies, and it appears they
> co-operated on the article topic.
> They would have been paid well by WW, and the article
> gave them the chance to say nice things about
> the operation of triodes, ultralinear, and cathode feedback
> used in the output stages of pentode and tetrode amps.
>
> At the end of the article which describes the salient aspects of each topology,
> there is a table for class A operation of various topologies, using KT66,
> and we assume about B+ = 320 volts, :-
>
> Triodes,
> efficiency = 27 %,
> power = 10 watts,
> THD at just under grid current = 1%,
> RL / Rout = between 2 and 4, so let's say 3.
> ( This means a damping factor of 3,
> which isn't bad for a circuit with no external loop of feedback )
>
>
>
> Beam Tetrodes,
> efficiency = 38 %,
> power = 14 watts,
> THD = 2%,
> RL/Rout, 0.05 to 0.1, so let's say 0.075,
> ( This means a very poor damping factor,
> and the amp without NFB is a just a current source,
> with negligible signal voltage regulation.)
>
>
> Standard UL, optimum screen taps at 40% of anode voltage,
> efficiency = 36%,
> power = 13.5 watts,
> THD = 1.5%,
> RL / Rout = 0.5 to 1.0, so let's say 0.7.
> ( So with screen UL, Rout is nearly equal to RL,
> and damping factor is about unity.)
>
> Tetrodes with Quad arrangement,
> efficiency = 36%,
> power = 13.5 watts,
> THD = 0.5%,
> RL / Rout = 2.0, no other figures given.
> ( So the dampoing factor is less than that of triodes,
> or in other words, the Rout is higher than triodes.)
>
> The article does not state exactly what the % of CFB
> is used in the "Quad arrangement".
> But prior to the table, they say the test used an
> amount of CFB which gave similar operating
> conditions as the case for 40% UL plate coil taps.
> This would mean that the OPT primary would be divided
> into sections giving a plate winding with 78 vrms appearing at each anode,
> and a cathode winding with voltage of say 52 appearing at each
> cathode..
> The % of CFB would be 40%, of the total windings.
> The screens are taken to the B+.
> The relative voltages at grid, screen, plate and cathode
> would then be identical to the plain 40% screen UL arrangement.
>
> The huge difference with CFB is that the distortion voltage
> is not only applied to between the cathode and screen, as in the case of plain UL,
> but also to between the grid and cathode.
>
> But something is really wrong here in the way the guys wrote the article.
>
> If there is indeed 40% of the turns in the cathode winding,
> output stage needs a drive voltage to the grids which would be very high.
But I’ve verified that a pair of KT66 will give exactly such
performance with those 10% of CFB as used in the Quad II amplifier!
For instance, the Gm for KT66 will be about 6mA/V. The CFB percentage
of 10% means that we have the infinite load voltage gain Mu=10, and
Ri=Mu/Gm=1.7kOhm. The push-pull series Ri is twice this figure, i.e.,
3.4kOhm, and the specified plate-to-plate load of Quad II is 6.6kOhm,
roughly twice the Ri! The 0.5% THD is also a realistic figure for a
pair of KT66s in such arrangement. The required driving voltage
between the grids is 72 V, or just 36 Vpeak for each grid.
It is a pity, the original QUAD phase splitter with those two tiny
EF86s is not enough linear for this job. Perhaps, the results would be
better if a Williamson front end was used instead that original QUAD
circuit. From other hand, to avoid the use of global NFB, the CFB
percentage with these tubes should be at least 20%. The required
driver voltage will be approx. 50Vpeak, and this will be within the
abilities of such driver tube as 6SN7.
Regards,
Denis
Patrick Turner
Ian Iveson wrote:
> Thanks, Denis.
>
> Ah, C-cores!
>
> You confirm most of my findings, although your transformers are
> better in the respects you relate than my EI examples. I wonder how
> necessary this is? Although I haven't got much test equipment, it
> seems my cfb+ul amps are good for 10-80kHz (-1dB), and show little
> distortion throughout this range at full power.
>
> I am very interested in your "detailed equivalent circuit".
> Coincidentally, I was building a new model yesterday, before an
> application crash wiped out my efforts. I can figure out where the
> capacitances are, but measuring them and confirming my analysis is
> problematic. I would be very grateful for any assistance in this.
>
> I find that I need no corrective networks, mostly because of the
> capacitance between anode and cathode windings. That seems to have
> been a good move (thanks in part to Brian Sowter!).
>
> I wonder if you can shed any light on the issue of differential
> phase distortion? That has been a residual concern for some time.
> I have been told not to worry about it, but nonetheless I would be
> interested in your view, and particularly in any measurements you
> have done. It seems to me to fit somewhere in the CFB pentode
> versus triode debate.
>
> Oh, and I would still like to know where van Veen's errors are!
>
> best wishes, Ian
I would to like to know what van Veens errors are too.
Don't you worry about limitations due to the use of
E&I lams, in preference to C-cores.
I just wound some whoppa sized OPTs for 350 watts,
and bandwidth is 300 kHz, when set up with the
correct tube source resistance and secondary load of
5 ohms.
Only slight +/- 1 dB undulations in the response occurred
between the first resonance at 84 kHz, and to 250 kHz.
Only 6S sections and 5P sections, and altogether,
16 layers of neatly layer wound wires.
I used Grain Oriented Silicon Steel, E&I laminations,
0.35 mm thick.
The LF distortion is many times less than
basic NON Oriented Silicon Steel.
Just today, I tested a small sample and compared the GOSS
to the NOSS lams, under similar conditions,
and the GOSS is far superior, in every way to the NOSS.
But I have made good amps with NOSS material.
Just used enough, and with plenty turns.
The HF losses with GOSS are quite low.
The mu of the material I have is over 10,000,
at 50 Hz, with little voltage excitation.
The maximum Lp is over 275 H,
but Lp still remains quite high down to a low a-a signal LF signal
which is a favourable ratio to the load of 1,400 ohms.
Williamson had Lp max of 600 H, which was a lot, compared to the
10 k a-a load, but at low level signals a-a,
Lp is far lower at 100H, due to the iron
having less mu than when a high voltage exists across
the winding at LF.
The Williamson OPT gave excellent sound, but hardly any makers
gave any more than lip service to his design ideology,
if old Leaks and Quads are any guide.
Iron is iron, and very basically non linear stuff it is too,
so decent iron is a must, and a low source resistance,
and a healthy relationship between Lp minimum to the RL.
I quite like C-cores, but they are not always available,
and the Mu of these is not always so wonderful,
I have seen batches from the same maker
where Mu varied from 4,500 to 11,000.
Those figures are acceptable though.
To know exactly what Denis thinks makes up a decent OPT,
We need him to post an easy to read diagram of the wind up
of the bobbin, with all sizes and numbers shown,
and with all core material details.
Patrick Turner.
Patrick Turner
Denis wrote:
What you are saying about operating voltage conditions
is about right.
But I can't remember what I measured the Rout to be
when I last did a pair of amps up.
The THD was sure more than 0.5 %.
> It is a pity, the original QUAD phase splitter with those two tiny
> EF86s is not enough linear for this job. Perhaps, the results would be
> better if a Williamson front end was used instead that original QUAD
> circuit. From other hand, to avoid the use of global NFB, the CFB
> percentage with these tubes should be at least 20%. The required
> driver voltage will be approx. 50Vpeak, and this will be within the
> abilities of such driver tube as 6SN7.
I agree about triode drivers. The LTP
plus one input triode is simpler than Williamson.
Then if the LTP had a CT choke in its plate circuit,
its THD would be reduced 3 times, and definately no need for
loop feedback, and even more CFB could be used.
Have you got a Time Machine to take us back to 1949,
and a suitcase full of funds to build a factory?
Even a suitcase full of today's Roubles would go far.....
With hindsight, I am sure either of us could build far better
amps than Mr Walker.
But the Electrostatic speakers would be a far harder challenge.
That's where Walker was the genius, speakers!
Patrick Turner.
>
>
> Regards,
>
> Denis