Re: Preamp
Patrick Turner
> I've been toying with the idea of building a preamp with bandaxall
> tone controls.
>
> I've made an orphaned (not on menu) web page showing one idea.
>
> http://flipperhome.dyndns.org/6DT8%20Preamp
Hi Flipper,
Not a bad type of preamp design you have there.
I have to admit I have never considered a bootstrapped cascode stage
for hi gain approaching µ squared where µ is the amp factor of one of
the triodes.
If you didn't wish to use 4 triodes for your preamp then you could use
a CF triode input stage as you have, say 1/2 a 12AU7.
Then use a triode-pentode such as a 6U8A and use the pentode section
for the gain stage to replace the two triode cascode stage. The
pentode can be fed its Ia = 4mA through 50k from the 300V supply to
give Ea = 100V.
Open loop gain would be about 200 with 50k if ther gm of the pentode =
4mA/V.
The 6U8A triode section could be used as the CF output buffer as you
have with its divider to get the closed loop gain down to about 4X
from 200X, and thus THD is reduced by a factor = 1/50 And so is most
of the noise of the pentode.
So far, you would have two triodes and one pentode.
>
> The idea started with using a (super linear) pentode-triode pair in
> mind but I began wondering about pentode noise and, so, have done a
> (bootstrapped) triode cascode.
I am not so sure the super linear "pentode-triode pair" is all that
linear but then because you have so much shunt NFB the THD will always
become low, especially at low signal levels below a volt.
Maybe you could consider a µ follower stage where you have the 6U8A
pentode gain tube on the bottom with 10k from its anode to triode
section cathode, with the triode anode taken to =300V. The triode grid
has 1M to + 180V. There is a 0.47uF cap from pentode anode to triode
grid.
The pentode then sees a very high RL value of around 260k, and maximum
open loop gain at LF will be maybe 700x.
This gain will roll off a bit early since the stray shunt C will have
attenuationg effect since the anode circuit resistance is so damn
high. But because you have a lot of shunt FB with the baxandall
network, the final BW will be very wide.
The divider network which drives the FB to the Baxandall network is
powered from the 6U8A triode cathode.
The noise of your amp will never be dominated by the presence of a
pentode.
Your placement of the gain control before the this amp will mean the
gain control resistance will more likely cause much more nois than the
pentode will. The noise of the Baxandall resistances might also be a
worry but they are surrounded by a NFB network which tends to reduce
the noise.
In a normal situation where this preamp was used as a gain stage with
tone control between a power amp and CD player, you would have 1.4Vrms
max coming from the CD player and you might attenuate this down to
20mV at the gain control.
Your stage would then boost the 20mV plus any noise to about 85mV, and
if the power amp had voltage gain of say 20x you would have 1.7Vrms at
the output to give 0.5Watts to a 6 ohm speaker for average listening
levels that won't cause a divorce.
Obviously, if there was just 1mV of noise at the pre-amp output then
you have 20mV at the speaker, and that's intolerable !
The distortion you might generate under normal circumstances of use of
you preamp is almost unmeasurable.
Your amp may be capable of 30Vrms max at 0.2%, so that at 85mV the
noise may be 0.0006%, and well below the noise levels.
So it just does not matter whether you have a triode cascode pair or
real pentode.
In the bad old days many tone control stages did use a *passive*
Baxandall network but with logarathmic pots of high value, 1M was
typical, and the network was driven off the anode of 1/2 a 12AU7 with
a lousy 2mA of Ia.
So only 1 triode was used to do what you are doing, and the insertion
loss of the Baxandall was -20dB so that if the 12AU7 had gain = 12, or
about 22dB, there was a slight amount of overall gain. The gain
control and a damn balance control network was always placed before
such a single triode tone control stage. Often there was a another
12AU7 gain stage to boost typical line levels of 200mV up to 2.4V
before the gain/balance network.
Usually the passive networks with high resistances and log pots meant
the response in the flat position was anything but flat and there were
large unwanted differences between any two channels. Noise was always
a problem.
Nobody used Idc for filaments. The distortion of the passive
minimalist input stages of 1950's lowest common denominator consumer
audio gear was still quite low though, but certainly lower than what
was going on in the radio detectors and audio power amps.
As a result of such primitive bean conter designed amp stages of 1955,
tone and balance adjustments earned themselves a terrible reputation
for ruining good music amoung audiophiles who wanted only the best
sound.
The use of the Baxandall network in a shunt FB loop using linear pots
was a huge step forward for tone control. It was **SO MUCH** easier to
get right. Nearly all such tone stages were driven off a previous amp
stage with low Ra or a CF buffer with Rout < 10k. Again, 1/2 a 12AU7
was a common drive tube which easily gave 22dB gain and Rout was 10k.
The tone stage could be another 1/2 12AU7 with its closed loop gain
reduced to just below unity due to the Baxandall tone R&C networks.
Rout of this stage was low enough to drive anything, less than 2k, so
no need for any output CF.
This set up used the same two triodes as the earlier stages with
passive networks but it was quieter and had less THD.
My answer to preamp requirements can be seen at
http://www.turneraudio.com.au/preamp-10tube-integrated-2006.htm
Scroll down to SHEET-3, about 2/3 down the page.
I rarely ever need to use a tone control stage. So I can bypass mine
when I don't need it.
I rarely ever need any voltage gain between any signal source and
power amp, so the gain stage can be bypassed.
Usually I have the input signal applied to the gain control pot and
its output passes through a cathode follower to reduce the signal
output resistance to less than 1k to over come any stray C of cabling
and power amp input C.
In a preamp I built long ago I had switched inputs to a volume control
pot followed by a 1/2 6CG7 common cathode gain stage with Rdc loading
then direct coupled to 1/2 6CG7 cathode follower output buffer. Gain
was about 17x.
I once measured such a simple line level preamp under normal op
conditions where about 0.1Vrms as being produced to drive a power amp.
THD was always under 0.01%, and inaudible. Noise was low enough.
Later I decided I preferred the sound of a µ-follower gain stage, or
normal common cathode triode gain stages with a solid state CCS to get
Idc to the anode. Reductions of THD due to the CCS were typically
-10dB.
I have not tried the µ-follower with a pentode. I aim to do this soon
maybe using a 6U8A because its triode and pentode seem so naturally
suitable and one may have a nice 4mA or more for Ia. Pentodes give
much more gain than anyone wants or needs, but that allows one to use
a shunt FB network between the follower output and the input so that
ensures excellent gain matching of two channels and extremely low
output resistance. In order to keep the FB network easy to drive at
its input, R1 might be 47k, with R2 = 470k which does not adversely
load the follower troide. Thus open loop gain of maybe 700 is reduced
to 10. Open loop THD of maybe 3% at 10Vrms is reduced to 0.043%, so at
0.1Vrms output THD = 0.00043%, or unmeasuarble.
I find 20dB of preamp gain is way too much. 12dB is usually plenty, so
the FB network used might be 47k + 220k approximately..
The 47k FB network Rin with 100k biasing R makes Rin = 33k and it is
easily driven by any source. So if a CD player is fed in with 1.4V and
gain = 4 you get 5.6V at about 0.02% and this feeds the volume control
and noise is extremely low.
If only 0.1V is needed for the power amp, the volume control setting
is very low at about -54dB. Obviously using the gain stage is
pointless unless the source signal is the old fashioned 200mV -- or if
the power amp and speakers are horrendously insensitive, or if the
source is from an MC phono amp and there is a low output cartridge.
The only time I seem to use a tone stage is for boosting treble on
poor old recordings or on radio sometimes.
And I NEVER need more than +/-6dB of boost or cut although my tone
stage does provide slightly more than +/- 6dB.
I long gave up the idea of having +/- 20dB maximum cut of boost to
both bass and treble.
The troubles with tone controls is that they often don't adress the
perceived deficiency of bass or treble levels in recordings or in
following amps or speakers. it is because the frequency deficiencies
just never follow curves which can be corrected by the curves of boost
or cut in the tone control. Not even multi band graphic equalisers can
adress music deficiencies because people would be only guessing about
what's wrong. And with rooms one might need to make adjustments but
its useless unless its all done automatically with a sampling
computer, microphone and test signals. There have been several
products on the market which do this. One should never have to do such
Eq, because the Eq involves many opamps and introduced phase shifts.
>
> Distortion is lower in the bandaxall feedback so the input cathode
> follower dominates.
If the input follower was a 1/2 12AT7, the gain reduction is about
1/45, so if THD open loop was 0.1% at 1.4V from a CD player then
closed loop THD of the CF = 0.1% / 45 = 0.0022%, and about
unmeasurable, and definately inaudible. Cathode followers when
properly used and set up may be extremely sonically neutral amp
stages. I quite like 12AU7 with CCS cathode sinks as CF stages.
>
> I'm curious what people think.
>
> I've not used pentodes in preamps so would feedback take care of the
> noise?
The Quad-22 control unit had ONE lone EF86 as the phono amp around
which was a shunt FB loop to give RIAA eq.
Leak and many other preamps of the 1950s also used EF86 which was
regarded as quiet enough. None were as quiet as many alternative
triodes though. The main reason why anyone used an EF86 in 1955 was
because you could get open loop gain of 200 which was equal to a pair
of cascaded halves of a 12AU7 or 6SN7 and the EF86 circuitry used
maybe 1 damn resistor less. Such savings allowed CEOs to buy a better
brand of car.
Leak made their power amps all sensitive to 0.1Vrms input for
clipping. So if the amp made 30W to 8 ohms the closed loop gain = 155.
But radio sources etc of 1955 were often only 0.1V output. If one
wanted an SNR = -60dB considered good in 1955, then noise had to be <
0.1mV and usually it simply wasn't, and systems of 1955 were
notoriously noisy, considering the sensitivity of speakers from that
era. What the hell, sources often had SNR = -45dB.
These days people expect SNR 20dB better than in 1955. But if your
input signal from a CD player or other source >1V then getting the SNR
wanted is dead easy. I still have junk brought to me for a fix because
its noisy. Its not uncommon for preamps and DA converters etc to have
1mV of buzz in their outputs and the customers I have with sensitive
speakers find the noise intolerable because or typical power amp gain
= 20x.
Williamson's amp of 1947 meeded 2Vrms for 16W into 16 ohms, so his amp
had gain = 8x.
This means if a preamp had 1mV of noise, you get 8mV at the speakers.
Speakers of 1947 were often 93dB/W/M sensitive so that 8mV was quite
audible. Today's speakers are all mainly 88dB/W/M so 8mV is barely
audible from 4 metres away it it was just mains hum with few
harmonics. I have a client with Horning speakers which have a claimed
sensitivity of over 95dB/W/M so he needs all his gear to be dead
quiet.
I have had to rewire a few of his amps which have had triodes
throughout. Pentode noise isn't a huge problem to overcome considering
just how many of them were routinely used in many samples of
professional audio and scientific gear of the 1950s and 60s.
Just build your pentode things then measure and tweak until the noise
goes away. Maybe you may find some pentodes noiser than others and
theory says the higher the gm the lower the noise so pentodes like
6SJ7 and following derivatives like EF86 may never be as quiet with a
lousy 0.7mA of ia as an EF80/6BX6 with 7mA. Gm of a pentode is often
very variable as the vertical spacing of the Ra lines on the data
curves suggest. A favourite pentode of mine is the 6SH7 which has this
sweet zone in the middle of its set of curves resulting in wide
voltage swing at quite low THD despite the fact its a pentode. All
pentodes have a reputation for their alarming levels of distortion
with even number plus odd number harmonics. Not all are perfectly
awful.
When you look at the pentode Ra curves for varying values of Eg1 bias,
you see that for most RL values there will be the same THD for the
same current change even if the load value varies between say 5k and
50k.
This is a different situation to a triode suited to the same app. As
the load goes higher in the triode the THD rapidly falls as the
internal open loop gain rises and there is increasing correction of
THD by means of the internal electrostatic NFB network which acts very
much like a shunt NFB network with external resistors around a
pentode. The triode is without a screen to prevent obstruction of its
internal NFB. It seems folly to use a pentode which has a screen to
prevent its internal FB operating and then have to use an external
network orf resistors to restore the FB the screen obliterates.
But if you arrange the pentode to have a very high RL value then for a
given voltage swing you reduce the Ia change and then the THD becomes
very difficult to predict by graphical methods because Ra lines are
almost horizontal and so are the load lines. So the only real way to
find out about is to set a pentode up and have a direct coupled CF
buffer so you can measure the anode output without the probes/meter/
CRO affecting the measurement.
If gain =700, you only need 0.05Vin for 35V output, so careful noise
free low impedance sources and adjustments are essential. If you had a
pentode circuit with such gain without any shunt FB then the screen
noise could be a problem.
This noise appears at the anode like any other noise such as grid
input noise or shunt FB network resistors. It is reduced with shunt
NFB when used but never ever completely reduced, but methinks if you
have a gain reduction of 1/70 because of applied loop FB then the
noise will be low enough. Try it anyway, and tell us what you find.
You might find that with a 6BX6 with 7mA of Ia and load of say 300k
that the open loop gain is well over 1,000x.
Who knows what the THD might be, despite the tiny amount of Ia and IRL
change. Its usually dissapointingly high.
But the 6BX6 with its csreen connected to its anode becomes quite a
fine triode with gain of maybe 50 with a load of 300k and that is over
20 times lower than the pentode with 300k so whatever the pentode THD
may have been.
With a voltage swing of say 100Vrms into 300k the 6BX6 THD might be
2%, but its a huge improvement over what the THD was in pentode mode
with 100V into 300k.
So this leads us to use a 6BX6 as a triode in your tone control amp so
that you end up reducing triode gain of say 50 down to 4, so THD at
10V is also reduced from say 1% to 0.08%, and at 0.1V you'd get
0.0008% which of course should be quite acceptable to anyone.
You may also find that were you to use a 6U8A with pentode section
triode strapped and with triode section used as the upper triode of a
µ-follower, then THD and general performance would be as good as or
better than if the pentode were allowed to operate as such. The
pentode connection always means you'll need a screen bypass cap and a
dropping resistor from the +300V or elsewhere, and screen current can
be 30% of Ia and it has to be paid for.
Patrick Turner.
Raymond Koonce
Not to be anal, but the term is baxandall, named for P. J. Baxandall,
not bandaxall. See this link:
http://www.schmarder.com/radios/tech/tone.htm for some details.
Best regards,
Raymond
flipper wrote:
> I've been toying with the idea of building a preamp with bandaxall
> tone controls.
>
> I've made an orphaned (not on menu) web page showing one idea.
>
> http://flipperhome.dyndns.org/6DT8%20Preamp
>
> The idea started with using a (super linear) pentode-triode pair in
> mind but I began wondering about pentode noise and, so, have done a
> (bootstrapped) triode cascode.
>
> Distortion is lower in the bandaxall feedback so the input cathode
> follower dominates.
>
Patrick Turner
> On Tue, 23 Mar 2010 04:53:45 -0700 (PDT), Patrick Turner
>
Snip a lot of good stuff for brevity, ( SALOGSFB )
> >As a result of such primitive bean conter designed amp stages of 1955,
>
> You and your damn 'bean counter' myth. No bean counter ever 'designed'
> an amp and it's pointless to design a Ferrari work car, a marvel of
> engineering and wonderful though it may be, when 99.9% of the workers
> can't afford to buy it.
I think I will firmly declare to agree to disagree on matters
appertaining to Bean Counters.
Just think for a minute that had you been a bright young stary eyed
young engineer in an the Average Good Audio Company of America, ie,
the AGACA, in 1955. You submit your design to the CEO for next year's
model, then several things would have happened at least. The CEO who
was technically incompetent would have had Bean Counter decide on cost
feasability, and you'd have been told to drop 2 triodes out of your
design. The Sales Director or what is now called a marketting
specialist would have weighed up whether it was better to go for
something more elaborate than competitors and then promote its better
specs and raise the price, or not better, and just agree with Bean
Counter. So just what happens and what is sold and what value is
enshrined in a product or what garbalogical value is dumped on what
could otherwise been so much better is a very vague set of ideas,
company goings on and bullshit.
If the design could be promoted and more Gee Whiz than the
competition, and there was available capital to invest in a better
case to dress the preamp up all nice'n pretty, then the whole exercize
might be profitable, perhaps, depending on consumer sentiment that
year and whether there was a recession going on or not or whether the
chief of finance had the OK to spend more money on upgrading the
company product profile.
I look at the Quad-II and its control amp and I can only smile. There
is a heck of a lot of variable functionality built in
to suit ppl who wanted a wide variety of sources available from radio,
disc and tape for 1960.
The actual circuits used are dependant on a myriad of switching to get
the best from a smallest possible number of tubes.
The EF86 phono stage is NOT the world's best phono stage but it did
give many folks a sound in the lounge rather like Morris cars gave ppl
a ride from A to B. When we compare Quad22 phono stages to what can be
done with say 3 triodes in a phono amp, say a couple of cascaded 12AX7
halves plus a cathode follower to power the output and series global
NFB loop, then the Quad is primitive, like a Model T Ford.
So why didn't Quad just do it better?
Why didn't Quad fill the two nine pin input tube sockets with a
parallled 12AX7/12AT7 input tube followed by a 6CG7 long tailed pair
instead of their 2 x EF86 with some positive FB as well as the NFB?
Why was the Quad-II OPT such a toy of the thing? I could do on and on.
Why did the Dynaco ST70 kits contain such a huge abundance of hardly
anything decent within?
Profit.
All about profit, and paring back costs of production to the absolute
minimum.
Quad even bought hundreds of gallons of grey paint which the Royal
Navy had left over from WW2. Grey is a colour which appeals to the
British People. Brilliant marketting strategy. Dear old Peter Walker
probably bought a Jaguar instead of a Morris.
Anyway, while judging other products of the 1950s, Peter W deserved
his Jag IMHO. And IMHO, he sure earned a gong for his ESL57 !
Anyway, to be sure its of no benenfit to anyone that I argue about
costs and what Well Known Good Blokes did in 1955. Baxandall was one
them of course, and at least he gave the world better tone control
compared to what had been done before he came along.
>
> The people you ran into were not 'bean counters', they were cost
> control managers, and construction contracting is a whole different
> world than product design.
>
> >tone and balance adjustments earned themselves a terrible reputation
> >for ruining good music amoung audiophiles who wanted only the best
> >sound.
>
> Then they shouldn't be buying lowest cost consumer commodity
> equipment.
>
> See, that's the advantage of free enterprise. You get to chose from a
> myriad of pin head designers in competing companies for what best
> suits your needs, taste, and budget instead of being stuck with
> whatever horse designed by committed government socialist planners
> mandate. Not to mention the pin head designers in competing companies
> know you're going to be the ultimate judge so they, naturally, try to
> predict and cater to it, some better than others, where socialist
> planners couldn't care less what dumb, ignorant, you think because
> they're mandating it anyway and if you don't like it then that just
> proves how dumb and ignorant you are.
Well indeed.
I wonder what nice quality ridden preamps were being made by Soviet
Socialist Russia in 1955?
Maybe not many while the stary eyed idealist engineers were being
cooled down in a Siberian Education Centre
where they were forced to design rocket guidance systems to help blow
up capitalism.
I am socialist, and I plan my social centralistically from my kitchen
where I sit and think, but hardly anyone turns up to the party or just
have a bonk if she's good looking.
I don't really care much about what goes on in AGACA.
Do you read that DILBERT comic strip in newpapers? I do, and it tells
me all I know about what happens in Big Companies, and in quite a few
Small Ones.
I work alone as a sole trader and have no need to employ Bean Counter
so that he too can earn a living while removing quality from gear made
under my name, not his.
SALOGSFB....
>
> Yeah. For this one I thought I'd try no SS. Well, okay, I put in some
> diodes.
One does not have to have any SS for CCS, but they are virtually
passive blameless enterties in a tube circuit.
Diodes for clamping and limiting are just fine where they normally
spend their life conducting microscopic and negligible currents.
Another way to do cascode with boostrapped top triode anode is to
simply set up a nornal 2 triode µ-follower and ground the bottom
triode grid and have a third triode underneath to make the cascode.
The top triode acts like a follower and bootstraps the load for the
grounded grid triode underneath it.
The cathode input resistance of the GG triode = load / gain so Rki is
fairly high so the very bottom triode creates considerable useful
gain. So if say 6DJ8 are used, total gain might be 30 x 15 = 450.
Hopefully one may find the totem pole arrangement to be stable. One
needs about 100Vdc across each triode though and such things cannot
make huge VO so their use is limited to preamps.
> >I have not tried the µ-follower with a pentode. I aim to do this soon
> >maybe using a 6U8A because its triode and pentode seem so naturally
> >suitable and one may have a nice 4mA or more for Ia. Pentodes give
> >much more gain than anyone wants or needs, but that allows one to use
> >a shunt FB network between the follower output and the input so that
> >ensures excellent gain matching of two channels and extremely low
> >output resistance. In order to keep the FB network easy to drive at
> >its input, R1 might be 47k, with R2 = 470k which does not adversely
> >load the follower troide. Thus open loop gain of maybe 700 is reduced
> >to 10. Open loop THD of maybe 3% at 10Vrms is reduced to 0.043%, so at
> >0.1Vrms output THD = 0.00043%, or unmeasuarble.
>
> >I find 20dB of preamp gain is way too much. 12dB is usually plenty, so
> >the FB network used might be 47k + 220k approximately..
>
> I agree. I figured 9dB was enough.
>
> >The 47k FB network Rin with 100k biasing R makes Rin = 33k and it is
> >easily driven by any source. So if a CD player is fed in with 1.4V and
> >gain = 4 you get 5.6V at about 0.02% and this feeds the volume control
> >and noise is extremely low.
>
> I haven't really decided on how to handle the vol-bal controls yet.
>
> >If only 0.1V is needed for the power amp, the volume control setting
> >is very low at about -54dB. Obviously using the gain stage is
> >pointless unless the source signal is the old fashioned 200mV
>
> Where do you get "old fashioned 200mV" from?
200mV was the stated output of many sources before CD came along.
MM cartridges often had a stated output of 5mV, 1kHz ( Shure V15 ).
OK, so you then designed your phono stage to make 200mV at 1kHz, and
also at 50Hz and 20kHz.
SALOGSFB,
>
> >The Quad-22 control unit had ONE lone EF86 as the phono amp around
> >which was a shunt FB loop to give RIAA eq.
>
> Yeah, I looked at that thing and darn near went blind trying to follow
> all those switches but I'll be damned if I can find the RIAA network
> for channel 1.
Quad systems were purchased new by mainly Rich People. It was in the
days when Rich People often meant Bright People with over average
intelligence. They had degrees from unis, and were doctors or lawyers
and they could figure out how to use the Quad systems they bought.
Nowdays nearly everyone is dumb, but still able to buy almost
anything. They complain bitterly about the complexity and
incompatibility and stupidity in the stuff they buy. It all really
started with a vengeance when ordinary people were supposed to be able
to get a video tape player to be preset to record something on TV
while they were down at the pub.
Oh how times ahve changed......
I know lots a people with an integrated system allowing satellite TV
and CD, DVD, and even LP, plus casstte tape, and surround sound. I've
been asked to come out to their houses to get it all working because
they have spent days just totally bamboozled.
I now have a set top box for TV and I need two remore controllers and
to surf stations I don't have a display of just the 6 same networked
programmes available from 90 choices. Surfing has been made like
riding a bicyle with square wheels.
So media use in 2010 is +40dB worse than figuring out Walker's buttons
from 1955.
SALOGSFB,
>
> Is there a reason why you're talking about triode strapped pentodes
> rather than simply using a triode to begin with?>Patrick Turner.
Well, some pentodes when strapped as a triode make a triode that is
not available as a pure real triode.
Take the EF80/6BX6 for example. Examine the Ra, gm and µ, and what
else would you use if you wanted a pure triode?
Maybe a paralleled 12AT7.
I have many pentodes laying around.
I have used a 6EJ7 strapped as a triode in cascode with a 2SK369 for
my phono preamp. The load is about 20k and the gain = gm of the 2SK369
x RL, ie, 0.04A/V x 20,000 ohms = 800.
I could use 12AT7, 6DJ8, 12AU7, or triodes 6BX6, or just about any
darn tube which is happy with Ia = 5mA and get the same nice result.
Tube craft is not about following what commercial makers have foisted
upon us often at a huge price. We are at liberty to experiment with
simplicity or complexity as we wish.
Patrick Turner.