MC PHONO PREAMP
RealInfo
Unknown
A bipolar stage would be a better idea, as it intrinsically has a
lower input impedance than a FET. But generally speaking unfiltered
active stages followed by RIAA stages have poorer signal headroom than
those implemented together. This will not end up as a very coherent
design if you do it this way.
BTW, could you please switch your news reader to plain text instead of
HTML - this is a text only group, and the line length gets screwed
rather.
d
John Byrns
"RealInfo" <therig...@yahoo.com> wrote:
You could always use a 1:10 step-up transformer to boost the MC signal
up to the level of an MM cartridge, that's what they did back in the
days when the god of triodes ruled the universe,
--
Regards,
John Byrns
Surf my web pages at, http://fmamradios.com/
mebrat...@yahoo.com
> In article <gmu6m1$6v...@localhost.localdomain>,
>
> "RealInfo" <therighti...@yahoo.com> wrote:
> > Hi all
>
> > I tried to find a tube based MC phono preamp with google but did not find,
> > only MM preamps I found many and good .
>
> > I guess there is a basic thrmionic problem with preamps tubes that makes hard
> > to amplify MC cartdige output which is about 10 times
> > weaker the the MM cartrige in general .
>
> > So my question is : will it be a good idea to precede a tube MM preamp with a
> > simple Jfet 1 stage X10 amp (with suitable load resistor at it's input )
> > so it will cpmpensate the amplification that the MM preams lacks regardinng
> > MC cartridges ?
>
> You could always use a 1:10 step-up transformer to boost the MC signal
> up to the level of an MM cartridge, that's what they did back in the
> days when the god of triodes ruled the universe,
>
A MC cart (the low level kind) is quite useless without first quality
transformers to feed them into.
Patrick Turner
First Quality transformers can be found and they are usually small and
not difficult to make with well shielded coils that are prone to hum if
not shielded adequately.
The MC is a signal source with output impedance below 20 ohms and easily
drives a 1:10 step up tranny with impedance ratio of 600 ohms to 60k.
The amount of primary inductance can be lower than for 600 ohms because
the cartridge source impedance is low.
However, transformers are more expensive than j-fets such as the 2SK147
which was made by Hitachi, maybe some are still available. An exact
equivalent is the 2SK369, which is available from many suppliers, Mouser
perhaps, and it has 400mW Pd rating instead of 600mW rating of the '147.
Its gm = 40mA/V at Id = 5mA, and you can examine the characteristics
from the online data sheets for the device. Distortion is a lot higher
than a triode for the same output voltage but if used to make a gain
stage with gain = 10, the output voltage is only 2mV for for 0.2mV input
and at such low Vo the Dn is less than 0.01%, especially if there is an
unbypassed Rs = 12 to 25 ohms for local current FB.
I bought 2SK369 for less than usd $1.00 each. So I got 20, and they all
were matched for gm within 2%, and if I blow one, OK, I solder a new one
in without costing me a lot. I have not blown one yet, and I don't even
have any clamping diodes.
For gain of say 20 not counting the effect of the current FB the drain
load need only be 500 ohms.
Conditions would be to use a +12V supply, very well regulated and
bypassed to 0V, then have Id = 5mA with 2.5V across the 500 ohm load for
Ed = 10.5V. Then you have about 15 ohms for Rs, and ES will be about
0.075V. The gate should have 47k biasing it to 0V with an extra RCA
input jack to take loading resistors for the cartridge, often 470 ohms
is common with perhaps 0.1uF or less across the 470 ohms to shunt any
stray RF. Such inputs are fragile and a diode clamp is a good idea so
that input voltage cannot exceed +/-0.7V especially if you touch the
input and have a stray static electricty discharge to the gate.
Such high transconductance j-fets have 20dB less noise than an tube, and
its a nicer type of noise which will later be more effectively reduced
with the following RIAA filter after stage 2 of your amp.
I found there wasn't any need to use anything except single ended
unbalanced j-fet inputs. But you can always used a balanced input and
balanced following amp for less THD, but because its twice the parts
list and twice the complexity, why bother, when the typical THD outcome
is so low, and SNR is so good.
Allen Wright has a lot to say about using j-fets in cascode with triodes
which IMHO is far better than having the separate pre-preamp before a
normal MM cart preamp. I also have a lot to say about using cascode
circuits, They do take some skill to make them work without oscillating
as perhaps 100MHz which is very easy if you have any stray unwanted
loops with wiring too long or you have not bypassed well enough even for
RF frequencies.
See my pages at
http://www.turneraudio.com.au/preamp-10tube-integrated-2006.html
and
http://www.turneraudio.com.au/preamp-rocket-phono-2005.html
Patrick Turner.
RealInfo
Your works are real great and real state of the art !!
Just what I am llokinf for .
EC
"Patrick Turner" <in...@turneraudio.com.au> ???
??????:4993ED9B...@turneraudio.com.au...
Paul Crawford
> Allen Wright has a lot to say about using j-fets in cascode with triodes
> which IMHO is far better than having the separate pre-preamp before a
> normal MM cart preamp. I also have a lot to say about using cascode
> circuits, They do take some skill to make them work without oscillating
> as perhaps 100MHz which is very easy if you have any stray unwanted
> loops with wiring too long or you have not bypassed well enough even for
> RF frequencies.
>
> See my pages at
>
> http://www.turneraudio.com.au/preamp-10tube-integrated-2006.html
>
> and
>
> http://www.turneraudio.com.au/preamp-rocket-phono-2005.html
I too am interested in designing a MC preamp at some point, in the past I
have tended to use paralleled ECC88 to get low noise for MM but of course
MC is 20dB more problematic. However, you web pages don't seem to be
working!
Regards,
Paul
Patrick Turner
Yes, the site isn't fully operational after a move to the new ISP.
I am working on restoring links right now.
Patience is a virtue, Anyone wanting a one stop shop for audio ideas
should have downloaded my whole site and put it on a CD and they'd not
have to rely on surfing the site remotely all the time. Its only 23MB in
total.
Paralleling 6DJ8 ECC88 etc is a hard way to get low noise, and I bet you
never get what the theory books say you ought to get. Equivalent input
noise resistance for a triode = 2.5 / Gm where Gm is in amps per volt,
so if gm was say 0.0012A/V for 1/2 a 12AX7, EINR = 2.5 / 0.0012 = approx
2,000 ohms and the noise 20Hz to 20kHz for a 2k resistance is about 2uV
at room temp - in theory. so if the signal is 2mV, then SNR = -60dB but
2mV is a high level signal quoted for an MM cart and average signals
might be a lot less to SNR isn't so great. Obviously with MC with a
typical 0.4mV signal for high level the SNR isn't good enough at all.
Tubes tend to have large amounts of LF noise in their spectrum of noise
so hence the bassy rumble you hear with MC using a 12AX7 input tube. IE,
the noise profile isn't a flat profile.
To reduce noise by 20dB with tubes you'd need to reduce the EINR by
1/100, because any noise reduction noise is proportional to the square
root of the resistance reduction, so if R reduces by 0.5, then noise
reduces by 0.7, or if you have 6 x 1/2 12AX7, EINR = 2,000/6 = 333 ohms,
so noise reduces by 1/square root of 6 = 1/2.44 = 0.4, which is only
-8dB.
In practice I have found ECC88/6DJ8 to rarely be any less noisy than a
well chosen 12AX7, despite the gm of the 88 being up to ten times higher
than 12AX7. So 1/2 an ECC88 should have about 1/3 of the noise of 1/2
12AX7, -10dB, and with 6 x 1/2 ECC88, noise should be -18dB below 1/2 a
12AX7, but this is only if you run 60mA total at idle in the 3 x ECC88
tubes to get the high gm and EINR = 33 ohms. If only one of the EC888
isn't quiet, the result isn't wonderful. So you spend all week searching
for quiet samples of tubes and just when you get a crew together, they
don't all stay quiet. Managing a team of 8 female rowers would be
easier.
Now using just one single 2SK369 at only 5mA of idle, you get noise at
-20dB below a 12AX7, and reliably so, and noise spectra is less bassy so
the noise should sound like a high pitched hiss with gain turned up
+20dB above normal use levels and not at all noticeable at normal
volumes. The noise of an unmodulated vinyl track will then be much
greater than the amplifier noise.
J-fet noise theoretically = 0.7 / gm. 2SK369 at only 5mA has gm =
0.04A/V so EINR = 0.7/0.04 = 17.5 ohms.
The noise of 17.5 ohms = 0.2uV approx. The j-fet noise formula is from
the UK amateur radio handbook and the triode formula from RDH4.
Mosfets are not much good despite their higher gm than a j-fet. Its
because mosfets have "popcorn noise", ie, pops and sploshes like tubes.
Maybe a better tube for MC is the 645pi russian twin triode. The gm per
two parallel halves appraches that of the 2sk369, but you need quite
high Ia.
Patrick Turner.
Paul Crawford
> Yes, the site isn't fully operational after a move to the new ISP.
>
> I am working on restoring links right now.
>
Seemed to work when I looked today :)
> Paralleling 6DJ8 ECC88 etc is a hard way to get low noise, and I bet you
> never get what the theory books say you ought to get. Equivalent input
> noise resistance for a triode = 2.5 / Gm where Gm is in amps per volt,
> so if gm was say 0.0012A/V for 1/2 a 12AX7, EINR = 2.5 / 0.0012 = approx
> 2,000 ohms and the noise 20Hz to 20kHz for a 2k resistance is about 2uV
> at room temp - in theory. so if the signal is 2mV, then SNR = -60dB but
> 2mV is a high level signal quoted for an MM cart and average signals
> might be a lot less to SNR isn't so great. Obviously with MC with a
>
I was just using the two halves of a ECC88 and came to much the same
conclusion, it was just usable with a 'good' tube for MM input and
really not a match (sadly) for a decent op-amp.
> Now using just one single 2SK369 at only 5mA of idle, you get noise at
> -20dB below a 12AX7, and reliably so, and noise spectra is less bassy so
> the noise should sound like a high pitched hiss with gain turned up
> +20dB above normal use levels and not at all noticeable at normal
> volumes. The noise of an unmodulated vinyl track will then be much
> greater than the amplifier noise.
>
> J-fet noise theoretically = 0.7 / gm. 2SK369 at only 5mA has gm =
> 0.04A/V so EINR = 0.7/0.04 = 17.5 ohms.
> The noise of 17.5 ohms = 0.2uV approx. The j-fet noise formula is from
> the UK amateur radio handbook and the triode formula from RDH4.
>
I was looking at this:
www.linearsystems.com/datasheets/LSK170.pdf
as a possible choice. Given the need to buy a lot (at least, from
someone) then paralleling them, each with a gate stopper, makes some
sense prior to a high dynamic range valve system..
> Mosfets are not much good despite their higher gm than a j-fet. Its
> because mosfets have "popcorn noise", ie, pops and sploshes like tubes.
>
The reason they are unloved for RF oscillators - too much 1/f noise :(
> Maybe a better tube for MC is the 645pi russian twin triode. The gm per
> two parallel halves appraches that of the 2sk369, but you need quite
> high Ia.
>
I did wonder if the planned new UK made vales from here:
http://www.techtubevalves.com/index.php
Might include an E88CC alternative. They claim to have higher emission
cathodes, and I guess the cloud of surplus electrons might help reduce
noise if they did, not sure if it would be enough for the very low
frequency lumpiness though..
Paul
bigwig
Does anyone know what is going on with these "new valves". I read
about them a few months ago and they seem to be based on CRT
technology. Are they still going ahead with it, or was it just a flash
in the pan?.
Matt.
Patrick Turner
That has about 1/2 the gm of the 2SK369...but a good noise figure.
>
> as a possible choice. Given the need to buy a lot (at least, from
> someone) then paralleling them, each with a gate stopper, makes some
> sense prior to a high dynamic range valve system..
Paralleling j-fets is very doable; using 4 will reduce noise -6dB, ie,
halve noise voltage.
Gain into a given load will 4 x more, or if you want, RL can be lower
than for one j-fet.
> > Mosfets are not much good despite their higher gm than a j-fet. Its
> > because mosfets have "popcorn noise", ie, pops and sploshes like tubes.
> >
>
> The reason they are unloved for RF oscillators - too much 1/f noise :(
> > Maybe a better tube for MC is the 645pi russian twin triode. The gm per
> > two parallel halves appraches that of the 2sk369, but you need quite
> > high Ia.
> >
> I did wonder if the planned new UK made vales from here:
>
> http://www.techtubevalves.com/index.php
>
> Might include an E88CC alternative. They claim to have higher emission
> cathodes, and I guess the cloud of surplus electrons might help reduce
> noise if they did, not sure if it would be enough for the very low
> frequency lumpiness though..
I emailed this crew for samples i'd test independantly for them.
But so far they are wanting to make only equivalents of 3 staple tubes,
the 12AX7, 12AT7, and 12AU7.
The gm of such tubes is mainly determined by the distance between grid
and cathode and the fine wire coil pitch; the problem for tube makers
who wanted to make high gm tubes was that normal helical grids close to
a cathode were hard to produce.
So they turned to making tubes with the frame grids as used in the 6DJ8
(i think?) and definately in quite a few pentodes such as 6EJ7.
Now if ever anyone wanted to have a high gm input and hence low EINR,
one could have say 4 x 6EJ7 in triode and get 50mA/V and
thus theoretically EINR = 50 ohms, and hence should be OK for MC. Now
the µ of the 6EJ7 = about 60 in triode, so such an input stage would
have good gain with a CCS load.
I use a 6EJ7 in a 10 tube pre myself but cathode driven by iddy-biddy
2SK369; see sheet 2 about 1/2 way down the page...
http://www.turneraudio.com.au/preamp-10tube-integrated-2006.htm
Never try to cathode drive a high gm pentode set to run as a pentode.
You'll get a fabulously high gain but with terrible instability and bad
microphonics that make the slightest sound come out like St Mary's
cathedral on sunday morning.
Even the rather simple circuit I used was prone to RF oscillations at
well above anything my 20MHz CRO can see.
Once built, without adequate ceramic cap bypassing at points needing to
be bypassed to 0V and with leads no longer than 12mm if possible,
the circuit oscillated at HF. I could tell something was wrong when it
was hooked up to a following power amp and speaker and I'd get a slight
click when touching various points on the 0V rails around the tube and
fet. The screw driver acts like an antenna, and loads the RF differently
at different points causing a slight shift in dc conditions of the set
up so hence the click you hear. Its a sure sign you have RF
oscillations. Searching with a CRO proble may stop the oscillation as
soon as the circuit becomes connected to the probe capacitance so a CRO
might miss seeing the oscillations.
At say 100MHz, 10pF is a low impedance! With no loops in the 0V rail and
small size of area for the circuit and with short leads and with ceramic
caps as well as the other electros and plastic caps you then get zero RF
at HF.
To be sure I bypassed the dc heater filaments carefully to 0V near the
tube and used RF chokes between fils and PSU.
In a cascode circuit the fils are beside a driven cathode electrode
which is not tied to 0V.
Patrick Turner.
>
> Paul
Paul Crawford
> Does anyone know what is going on with these "new valves". I read about
> them a few months ago and they seem to be based on CRT technology. Are
> they still going ahead with it, or was it just a flash in the pan?.
> Matt.
I emailed them recently and they said:
"The E813CC valve (12AX7 / ECC83 equivalent) will be available for
purchase April this year and is a direct replacement in amplifiers for
the ECC83 currently used from China, Russia, Czech Republic, etc."
also
"The price: 1 E813CC valve shipped to a UK address will costs £36.26, 2 =
£66.26, 4 = £126.26
The E812CC is scheduled to be available by June and will retail at the
same pricing."
So looking forward to seeing them.
Paul
Ian Bell
I have yet to come across any work that verifies the old 2.5/gm
equivalent input noise resistance for a tube. I have yet to see better
than 2uV from a single tube.
Cheers
Ian Bell
> On Thu, 05 Mar 2009 17:40:05 -0800, bigwig wrote:
>> Does anyone know what is going on with these "new valves". I read about
>> them a few months ago and they seem to be based on CRT technology. Are
>> they still going ahead with it, or was it just a flash in the pan?.
>> Matt.
>
> I emailed them recently and they said:
>
> "The E813CC valve (12AX7 / ECC83 equivalent) will be available for
> purchase April this year and is a direct replacement in amplifiers for
> the ECC83 currently used from China, Russia, Czech Republic, etc."
>
> also
>
> "The price: 1 E813CC valve shipped to a UK address will costs £36.26, 2 =
> £66.26, 4 = £126.26
>
Bit pricey IMHO. Wonder if they will survive the recession?
Cheers
ian
bigwig
>
> - Show quoted text -
That is pricey. Still people pay more for NOS, like you say Ian I
wonder if they will survive.
Nick Gorham
Have you tried things like E810F, D3A and C3G ?
Not saying they weould be better, but they are high Gm, and claim low
noise, I use a e810f myself at the front of a phono stage.
--
Nick
Ian Bell
No but I have tried lots of types with gms in the 1 to 4mA/V range and
the best I get is 2uV in a 20K bandwidth.
>
> Not saying they weould be better, but they are high Gm, and claim low
> noise, I use a e810f myself at the front of a phono stage.
>
I must ry a higger gm one just to see if there really is any difference.
Cheers
Ian
Patrick Turner
The prices seem rather expensive.
36 quid or british pounds is about aud $85.00, and way above an EH twin
triode made in russia.
A 12AX7 is only a 12AX7 and can only be that and nothing else.
But Gordon Brown has instructed the British Mint to begin printing a
shirt & trouser load of British Quids to be distributed to well chosen
folks to use because so many seem to have run out of quids and are
skint.
Now the Brooks bicycle saddle company wants aud $300 for a leather and
steel basic bike saddle when all the rest in the shops here are mainly
below $100. So devaluing of the British Quid can't happen fast enough
IMHO and its an over valued currency. I use this odd example to
illustrate something but recently I have had a lot of trouble with 3
Brooks saddles which have failed badly in various ways. The sales
brochures indicate they are trouble free and offer the world's most
comfortable ride and that you'll only ever have to buy one for a
lifetime of cycling on a bicycle, but lemme tellya, it ain't like that
at all.
So because of overvalued BP, we get absurdly high tube prices.
Now the levels of specialist technical abilities in some british
companies is stupendously higher and better than anywhere else.
But before we ever grant anyone status of being the best, it must be
proven to be that.
Part of the stupendousness they have must translate into a competitively
priced product and methinks right now that the market for niche market
tube gear has about plunged to very low levels. At present the
manufacture of tubes generally in China and Russia is or will be
challenged and if the Russians and Chinese are forced to close shop and
dismantle factories, then those tube factories just won't be anymore
because when times come good again *maybe* nobody will tool up to make
tubes with new made machinery like the old which may be melted down to
make container loads of cheap Chinese consumer products. Its mostly
ancient gear that's being still used to make gear and in some really old
factories, often with an ageing but highly skilled workforce.
But anyone going into tube making with all new tooling may well be able
to use the benefits of modern computer controlled robotics to make the
tools required and produce thousands of tubes per week without hardly
hiring anyone and exposing them to toxic processes. Robots work 24/7 and
don't require holiday pay and sick pay.
Therefore tubes should be much cheaper than now, and not the other way
around.
But wherever a market can be monopolised or gauged by western nation
companies, then that's what they do. Something made in Russia for $3 at
the factory is sold to a westerner for $30. And something sold for $30
is usually an octal beam tetrode output tube such as a 6L6GC, so a small
twin triode should be a lot cheaper.
Greed ain't over yet, and many wannabes are planning how to destroy some
good old industries to be able to muscle in when times get better.
I have a supply of decent tubes I won't sell to anyone unless they buy
an amp built around them. I have enough stocks until i retire sometime
within the next 15 years, and enough for my own use for 15 more years if
I live that long and retain my hearing and my marbles.
Mike at New Sensor has given me a list of nice cheaper US $ prices, but
trouble is the Aud has gone down from usd $0.90 last year to $O.63 now
so the prices for US tube here have actually risen a bit. Mike is
becoming an "old bugger" and what happens when he retires? I should have
bought more tubes last year but then right now hardly anyone wants a
tube amp. Ageing baby boomers have been the main buyers of tube gear for
the last 15 years, and by using part of their high disposable income.
Maybe they feel sick as they watch their superanuation nest eggs drop in
value by 50% or more, and so all possible luxury spending will be banned
at least by their wives.
I wish the Brits the best of British Luck in making tubes in Britain.
They'll sure need luck to sell any.
Just how anyone could make a new 6CG7 better than the way Seimans did in
Germany in 1960 or how they were made in Oz in 1960 will remain a
mystery IMHO.
Someone might do it. Then they need to proove it. Tubes need to be
tested for thousands of hours to see how they last, and so you need a
year to test a tube for 8,760 hours. Such tests were going on all the
time while Mullard was king.
Not so easy.
Patrick Turner.
Patrick Turner
Ian Bell wrote:
>
> Patrick Turner wrote:
> >
> > Paul Crawford wrote:
snip
>
> I have yet to come across any work that verifies the old 2.5/gm
> equivalent input noise resistance for a tube. I have yet to see better
> than 2uV from a single tube.
The EINR isn't the only source of tube noise. There is shot noise at the
anode. And the noise spectra has more amplitude at low F.
But your'e right, not all tubes measure less than 2uV of calculated
input noise. I have a few with 0.8uV though, rather rare.....
One has to measure the noise after an amp raises the signal at the anode
of a tube under test x 1,000 times then divide the noise figure by a
total of [1,000 x tube gain]. We assume the noise at the anode of the
tested tube is mainly all due to EINR. Its difficult to measure it any
other way.
The noise of the tester amp with gain of 1,000 will not significantly
contribute to noise measurements.
Patrick Turner.
>
> Cheers
Patrick Turner
Have you ended up with a better SNR?
If minimum noise was say 1uV, then it doesn't matter what the following
gain is due to high gm.
For a -60dB SNR, input signal must be no lower than 1mV.
But I recall a 2SK369 has typically 0.14uV of input noise; somewhat
difficult to measure.
For a -60dB SNR required for vinyl to get below the vinyl noise the
input signal may be not less than 0.14mV.
Now my Denon MC gives a rated 0.35mV, so the expected SNR should be
-68dB.
And I seem to get it.
Patrick Turner.
>
> --
> Nick
Ian Bell
>
> Ian Bell wrote:
>> Patrick Turner wrote:
>>> Paul Crawford wrote:
> snip
>> I have yet to come across any work that verifies the old 2.5/gm
>> equivalent input noise resistance for a tube. I have yet to see better
>> than 2uV from a single tube.
>
> The EINR isn't the only source of tube noise. There is shot noise at the
> anode. And the noise spectra has more amplitude at low F.
>
Er, 2.5/gm IS the formula for shot noise (see RDH4 pages 936 & 937).
Other sources of noise are generally considered to be much smaller. The
2.5/gm seems to come from a single article written in 1940 and RDH4 only
tables values calculated on this basis rather than actual measurements.
> But your'e right, not all tubes measure less than 2uV of calculated
> input noise. I have a few with 0.8uV though, rather rare.....
>
Which is very interesting because you would expect a tube with a gm of
2.5mA/V to have a equivalent noise resistance of 1000 ohms which
produces a noise voltage of about 0.6uV. What was the gm of the types
that measures 0.8uV.??
> One has to measure the noise after an amp raises the signal at the anode
> of a tube under test x 1,000 times then divide the noise figure by a
> total of [1,000 x tube gain]. We assume the noise at the anode of the
> tested tube is mainly all due to EINR. Its difficult to measure it any
> other way.
Depends what you measure it with. My Ferrograph test set has a 1mV range
so a 40dB amp would make its full scale 10uV. Even a tube stage with
0.5uV EIN and a mere 20X gain will produce 10uV of noise at the output.
Cheers
Ian
Nick Gorham
>
>>
>>Have you tried things like E810F, D3A and C3G ?
>>
>>Not saying they weould be better, but they are high Gm, and claim low
>>noise, I use a e810f myself at the front of a phono stage.
>
>
> Have you ended up with a better SNR?
>
I would have to know better than whet to answer that, but I certainly
have ended up with what I believe is a good S/N ratio.
This is a post I made to Dave Slagles forum about the noise measurments
I made of the thing. Hopefully they make some sense.
> OK, I don't know if these are any interest , but I wanted to get some idea of the S/N ratio I am getting out of the E810F Pentode front end in my LCR phono. This is a Denon 103 (0.3mv) into lundahl TX's wired for 1:20. The first graph is the background noise, with the phono stage off. It does show I have some hum to sort out in the wireing to the computer, but as I am more interested in noise, it shouldn't be a problem. All the samples were done in 24bit at 96k.
>
> http://www.lurcher.org/nick/audio/lcr/background.jpg
>
> The next one is playing a 1khz +6db (thats what the test record said) test tone. Hopefully that will give some reference to the noise floor.
>
> http://www.lurcher.org/nick/audio/lcr/+6db-1k-source.jpg
>
> And finally, this is the phono stage with its inputs disconnected, so this should be the noise generated by the phono stage.
>
> http://www.lurcher.org/nick/audio/lcr/no-input.jpg
>
> Hopefully someone who has a clue (not me) can use this to say what the S/N ratio is. But it seems to me to indicate that the pentode isn't doing too much in the way of damage on the noise front.[
--
Nick
Paul Crawford
> Bit pricey IMHO. Wonder if they will survive the recession?
Who knows, you can say the same about most companies just now :(
Still, I am wondering if a dozen or two tubes would be a better
investment than my current business account (0.1% interest) or recent
pension schemes (-30% 'interest')?
Like decent wine, you can always enjoy it even if it is not paying off...
Paul
Ian Bell
> Patrick Turner wrote:
>>
>
>>>
>>> Have you tried things like E810F, D3A and C3G ?
>>>
>>> Not saying they weould be better, but they are high Gm, and claim low
>>> noise, I use a e810f myself at the front of a phono stage.
>>
>>
>> Have you ended up with a better SNR?
>>
>
> I would have to know better than whet to answer that, but I certainly
> have ended up with what I believe is a good S/N ratio.
>
> This is a post I made to Dave Slagles forum about the noise measurments
> I made of the thing. Hopefully they make some sense.
>
>
>> OK, I don't know if these are any interest , but I wanted to get some
>> idea of the S/N ratio I am getting out of the E810F Pentode front end
>> in my LCR phono. This is a Denon 103 (0.3mv) into lundahl TX's wired
>> for 1:20. The first graph is the background noise, with the phono
>> stage off. It does show I have some hum to sort out in the wireing to
>> the computer, but as I am more interested in noise, it shouldn't be a
>> problem. All the samples were done in 24bit at 96k.
>>
>> http://www.lurcher.org/nick/audio/lcr/background.jpg
>>
>> The next one is playing a 1khz +6db (thats what the test record said)
>> test tone. Hopefully that will give some reference to the noise floor.
>>
Since this is an FFT we need to convert the apparent noise floor - which
looks about -123dB to me, to total noise in a 20KHz bandwidth by adding
the magic 43dB (20 log square root of 20000) which gives the noise floor
as -80dB. This really represents the resolution of the instrumentation.
>> http://www.lurcher.org/nick/audio/lcr/+6db-1k-source.jpg
>>
This seems to have a peak at about 300Hz and -10dB. if this signal is
really 6dB above nominal then the nominal is -16dB on the scale given.
>> And finally, this is the phono stage with its inputs disconnected, so
>> this should be the noise generated by the phono stage.
>>
>> http://www.lurcher.org/nick/audio/lcr/no-input.jpg
>>
This shows some 1/f noise as expected and the lf end is about 10dB up on
the HF end. I would guess the mean is about -110dB which, by applying
the magic 43dB figure means the 20KHz BW noise is about -67dB.
Since the test tone was at -16dB that means the S/N ratio is:
-16 - -67dB = 51dB
which seems a little low to me. If that tone was really -6dB and not
+6dB then the S/N would be a more respectable 63dB.
Cheers
Ian
Paul Crawford
> halve noise voltage.
> Gain into a given load will 4 x more, or if you want, RL can be lower
> than for one j-fet.
Another advantage is less overall variability. Given the small size, etc,
there is little harm in using a few and it takes care of the worst noise/
microphony aspects.
> I emailed this crew for samples i'd test independantly for them.
>
> But so far they are wanting to make only equivalents of 3 staple tubes,
> the 12AX7, 12AT7, and 12AU7.
>
> The gm of such tubes is mainly determined by the distance between grid
> and cathode and the fine wire coil pitch; the problem for tube makers
> who wanted to make high gm tubes was that normal helical grids close to
> a cathode were hard to produce.
These days, given the lack of significant cost considerations on the rest
of the circuit (compared to the total) it seems sensible to go for a tube
with easy supply, and make up the difference in the design (e.g. JFET
input, extra stage of amplification, regulated supply).
> oscillations. Searching with a CRO proble may stop the oscillation as
> soon as the circuit becomes connected to the probe capacitance so a CRO
> might miss seeing the oscillations.
You need a spectrum analyser and a small insulated loop probe. Jolly easy
to find hooting then!
I normally always have a grid/gate stopper resistor for this reason. OK
it adds a small amount of noise, but not much compared to the active
device itself. Also if you have N paralleled inputs, the stoppers are
paralleled as well, so their contribution goes down with SQRT(N) as well.
Paul
Nick Gorham
Ok, I will take your word for that, but given that if I record at 16bit,
I get the trace at about -90dB, is that suggesting that a CD can only
manage a S/N of 47dB?
>
>>> http://www.lurcher.org/nick/audio/lcr/+6db-1k-source.jpg
>>>
>
>
> This seems to have a peak at about 300Hz and -10dB. if this signal is
> really 6dB above nominal then the nominal is -16dB on the scale given.
>
Yes, its my typo (again) it was a 300hz test.
--
Nick
Ian Bell
No, definitely not. I would have expected the FFT noise floor to be just
as low with a 16 bit input. The actual noise floor, with the input
shorted, is limited by the noise in the analogue circuits before the A/D
converter and I would expect them to achieve in the region of -80dB
which should show up as around -123dB on the FFT. Which program did you use?
Cheers
ian
Nick Gorham
>
>
> No, definitely not. I would have expected the FFT noise floor to be just
> as low with a 16 bit input. The actual noise floor, with the input
> shorted, is limited by the noise in the analogue circuits before the A/D
> converter and I would expect them to achieve in the region of -80dB
> which should show up as around -123dB on the FFT. Which program did you
> use?
>
>
> Cheers
>
> ian
>
>
SpectraPlus
--
Nick
Ian Bell
I just had a look at its manual on line. There is an option in the
spectrum display set up to show 'Total Power Meter' which is a bar that
appears on the right hand side of the plot - see page 31 of the user
manual. If you enable that option it will effectively do the magic 43dB
thing for you. Try it again with the input shorted and see what you get.
By the way, that is one expensive program!
Cheers
Ian
Patrick Turner
Ian Bell wrote:
>
> Patrick Turner wrote:
> >
> > Ian Bell wrote:
> >> Patrick Turner wrote:
> >>> Paul Crawford wrote:
> > snip
> >> I have yet to come across any work that verifies the old 2.5/gm
> >> equivalent input noise resistance for a tube. I have yet to see better
> >> than 2uV from a single tube.
> >
> > The EINR isn't the only source of tube noise. There is shot noise at the
> > anode. And the noise spectra has more amplitude at low F.
> >
>
> Er, 2.5/gm IS the formula for shot noise (see RDH4 pages 936 & 937).
> Other sources of noise are generally considered to be much smaller. The
> 2.5/gm seems to come from a single article written in 1940 and RDH4 only
> tables values calculated on this basis rather than actual measurements.
I thought the RDH4 said it my way. I'll have to re-read it again.
Anyway, its the only formula it seems for triode noise. Pentode noise is
higher because of "partition noise" caused by the screen.
The British amateur handbook quotes 0.7/gm for j-fets. I have found some
co-relation.
> > But your'e right, not all tubes measure less than 2uV of calculated
> > input noise. I have a few with 0.8uV though, rather rare.....
> >
>
> Which is very interesting because you would expect a tube with a gm of
> 2.5mA/V to have a equivalent noise resistance of 1000 ohms which
> produces a noise voltage of about 0.6uV. What was the gm of the types
> that measures 0.8uV.??
Some 12AX7 seemed to measure down to 1uV. gm varies, and is only say
2.5mA/V for say 1/2 6CG7 when Ia = 10mA and Ra = 7.7k.
Gm then is 2.6mA/V, but at 3 mA have found some samples 1uV when Ra is
much higher hence gm is much lower.
So what you get theoretically is only pie in the sky and most days the
tubes are more noisy than the data and calcs suggest they ought to be.
gm does obviously have some bearing on the noise.
>
> > One has to measure the noise after an amp raises the signal at the anode
> > of a tube under test x 1,000 times then divide the noise figure by a
> > total of [1,000 x tube gain]. We assume the noise at the anode of the
> > tested tube is mainly all due to EINR. Its difficult to measure it any
> > other way.
>
> Depends what you measure it with. My Ferrograph test set has a 1mV range
> so a 40dB amp would make its full scale 10uV. Even a tube stage with
> 0.5uV EIN and a mere 20X gain will produce 10uV of noise at the output.
Hang on, if you had 0.1uV and 40dB of gain, ie, 100x, then you'd have
10uV after the gain.
There is 1000uV per mV so the 10uV would be not a huge meter swing, only
1/100 of the linear scale.
But if your meter can tell you that you have 10uV then fine, but perhaps
its better to amplify the 0.1uV x 1,000 and get 0.1mV, and then you'd
see a swing of a meter needle of 10% of the scale on a 1mV range.
When testing 1/2 12AX7 with shorted grid via a short lead, and if you
have 2uV of EIN, then expect 120uV at the anode if the AX7 gain is 60x.
In my case the following amp has gain of 1,500, so this becomes 180mV,
and a 1V scale is OK. The input noise of the amp used to lift the signal
to measure it
is maybe 3uV or less and so the noise at the amp input is the square
root of ( 120uV squared + 3uV squared ) and if you work this out then
the following amp noise makes virtually no difference One could have its
noise at 10uV and the 12AX7 noise would swamp the noise of the following
amp. My amp I use is a 2 stage tube type. One could use an opamp.
My noise measures are all with dc on the heaters.
With ac on the heaters expect much more hum often due to leakage current
between cathode and heaters.
All my preamps always use dc heaters to avoid cathode hum.
A bad tube will have intermittent "sploshy" noise whose amplitude varies
between good and bloody terrible, and so you have to bin thse tubes
unless where you use it doesn't matter, maybe a second stage in a guitar
amp.
In the old days the recording studios may have had quite a number of
12AX7 in consoles etc and a tech would have to swap out tubes to keep
the noise low at service time. Of course the EF86 became immensely
popular because you could get more gain than with a triode and then use
NFB to help control noise. RDH4 has a few typical ideas used in studios
and broadcast stations. One is to use a pentode with its anode taken to
0V and then use the screen as the anode with a lowish "anode" current.
Good screening, but I assume the EIN is still related to the 2.5/gm, and
gm of this sort of triode connected pentode is rather low with typical
types, although perhaps with 6EJ7 whichy is a high gm tube, or an E280F,
then maybe they are quieter than normal full triode connection. I recall
I tried the grounded anode idea but found magnetically induced hum to be
a bother with EF86.
I have not read any reports on the Net or in old books where anyone
successfully managed to get tube input noise lower than has been
mentioned as typical in RDH4.
Nobody bothers to really examine the issue of tube input noise very much
in these times of opamps and fets and so forth.
But I found that afer several years of business I hade acquired a
collection of preamp tubes, mainly pulls from garbage gear, and I began
to wonder what was actually good amoung the pile. Some had to be good if
the garbage gear had had its tube replaced just before it was junked.
And I found some tubes were, low EIN, not very microphonic, nice bright
looking getter, hmm, maybe very close to NOS, so I went through the
piles and soon discovered what I could put in a "good" pile, ie, less
than 3uV for AX7, AU7, AT7 CG7, and what would be N, for noisy with the
average per section written on the tube, and M, for microphonic, ie,
well above a good sample. Some were quiet but M. Some are marked N,M,
meaning a bit noisy and microphonic but usable in a power amp second
stage for good while. Quite a few were just trash, more than 10uV and
micro, so out they went to the bin. Some had terrible hum.
Usually if I supply someone with an amp with tubes I select on the
criteria I have outlined, they never have noise problems.
Patrick Turner.
Patrick Turner
Well, suppose some bottles in your cellar remain drinkable. Some don't,
and turn to vinegar like many investments you've made. As you become
poorer, getting drunk to ease the pain would seem to become more
expensive, which needs you need another drink godamn it......and another
to cope with the doctor's report on your liver....
May as well light up a cigarette while your'e laying in the cellar.
Patrick Turner.
Ian Bell
I am operating 6CG7 at 5mA where gm is just a shade over 2mA/V which
gives a calculated effective input noise resistance of 1250 ohms which
is about 0.6uV. The best 6CG7 I have measured so far gives just over 3uV.
> So what you get theoretically is only pie in the sky and most days the
> tubes are more noisy than the data and calcs suggest they ought to be.
>
Too bloody right.
> gm does obviously have some bearing on the noise.
>
Pass. I guess one would need to measure tubes with large differences in
gm to be sure about that.
>>> One has to measure the noise after an amp raises the signal at the anode
>>> of a tube under test x 1,000 times then divide the noise figure by a
>>> total of [1,000 x tube gain]. We assume the noise at the anode of the
>>> tested tube is mainly all due to EINR. Its difficult to measure it any
>>> other way.
>> Depends what you measure it with. My Ferrograph test set has a 1mV range
>> so a 40dB amp would make its full scale 10uV. Even a tube stage with
>> 0.5uV EIN and a mere 20X gain will produce 10uV of noise at the output.
>
> Hang on, if you had 0.1uV and 40dB of gain, ie, 100x, then you'd have
> 10uV after the gain.
>
But you don't have 0.1uV, even at the input. Say we have a perfect tube
with gm =2mA/V at the operating point and a gain of 20. Its input noise
will be 0.6uV so the noise you measure at its output will be 12uV.
> There is 1000uV per mV so the 10uV would be not a huge meter swing, only
> 1/100 of the linear scale.
Indeed, that's why I am suggesting a 40dB amp before it.
> But if your meter can tell you that you have 10uV then fine, but perhaps
> its better to amplify the 0.1uV x 1,000 and get 0.1mV, and then you'd
> see a swing of a meter needle of 10% of the scale on a 1mV range.
>
But it's not 0.1uV, its 12uV so you anly need 40dB of gain.
> When testing 1/2 12AX7 with shorted grid via a short lead, and if you
> have 2uV of EIN, then expect 120uV at the anode if the AX7 gain is 60x.
> In my case the following amp has gain of 1,500, so this becomes 180mV,
> and a 1V scale is OK.
Yes, but my Ferrograph has a 1mV scale so I don't need anywhere near as
much extra gain.
The input noise of the amp used to lift the signal
> to measure it
> is maybe 3uV or less and so the noise at the amp input is the square
> root of ( 120uV squared + 3uV squared ) and if you work this out then
> the following amp noise makes virtually no difference One could have its
> noise at 10uV and the 12AX7 noise would swamp the noise of the following
> amp. My amp I use is a 2 stage tube type. One could use an opamp.
>
> My noise measures are all with dc on the heaters.
>
Agreed
> With ac on the heaters expect much more hum often due to leakage current
> between cathode and heaters.
Yes, been there, done that, pulled my hair out and bought the T-shirt.
> All my preamps always use dc heaters to avoid cathode hum.
>
> A bad tube will have intermittent "sploshy" noise whose amplitude varies
> between good and bloody terrible, and so you have to bin thse tubes
> unless where you use it doesn't matter, maybe a second stage in a guitar
> amp.
>
Selection for low noise is definitely to be recommended for the first stage.
> In the old days the recording studios may have had quite a number of
> 12AX7 in consoles etc and a tech would have to swap out tubes to keep
> the noise low at service time. Of course the EF86 became immensely
> popular because you could get more gain than with a triode and then use
> NFB to help control noise. RDH4 has a few typical ideas used in studios
> and broadcast stations. One is to use a pentode with its anode taken to
> 0V and then use the screen as the anode with a lowish "anode" current.
Yes, I am well aware of that one. I even managed to get a hold of the
article the RDH4 refers to. The best noise I have got so far was from a
6AU6 pentode wired as a triode as per this article.
> Good screening, but I assume the EIN is still related to the 2.5/gm, and
> gm of this sort of triode connected pentode is rather low with typical
> types,
I have measured gm between 3 and 4mA/V for a 6AU6 in triode mode.
although perhaps with 6EJ7 whichy is a high gm tube, or an E280F,
> then maybe they are quieter than normal full triode connection. I recall
> I tried the grounded anode idea but found magnetically induced hum to be
> a bother with EF86.
>
> I have not read any reports on the Net or in old books where anyone
> successfully managed to get tube input noise lower than has been
> mentioned as typical in RDH4.
>
Me neither. I have an article by Burgess from the Wireless Engineer
February 1945 which quotes the 2.5/gm derived by Harris and says it
'shows close agreement with measured values'. Now I think Harris and I
know for certain the Burgess is talking about radio receivers and for
them the lowest noise stage needs to be the rf amp or mixer. Now maybe
2.5/gm applies at rf frequencies but it looks to me like it does not
apply for audio. Indeed, Burgess references Harris' paper which is
entitled 'Fluctuations in space charge limited currents at MODERATELY
HIGH FREQUENCIES' (my capitals) which does no sound like audio to me.
> Nobody bothers to really examine the issue of tube input noise very much
> in these times of opamps and fets and so forth.
>
But it is something WE should examine as tube users.
> But I found that afer several years of business I hade acquired a
> collection of preamp tubes, mainly pulls from garbage gear, and I began
> to wonder what was actually good amoung the pile. Some had to be good if
> the garbage gear had had its tube replaced just before it was junked.
> And I found some tubes were, low EIN, not very microphonic, nice bright
> looking getter, hmm, maybe very close to NOS, so I went through the
> piles and soon discovered what I could put in a "good" pile, ie, less
> than 3uV for AX7, AU7, AT7 CG7, and what would be N, for noisy with the
> average per section written on the tube, and M, for microphonic, ie,
> well above a good sample. Some were quiet but M. Some are marked N,M,
> meaning a bit noisy and microphonic but usable in a power amp second
> stage for good while. Quite a few were just trash, more than 10uV and
> micro, so out they went to the bin. Some had terrible hum.
>
Can you remember what proportion of the total had low noise and
microphony?/
Cheers
ian
Patrick Turner
Ian Bell wrote:
>
> Patrick Turner wrote:
> >
> > Ian Bell wrote:
> >> Patrick Turner wrote:
> >>> Ian Bell wrote:
snip
> I am operating 6CG7 at 5mA where gm is just a shade over 2mA/V which
> gives a calculated effective input noise resistance of 1250 ohms which
> is about 0.6uV. The best 6CG7 I have measured so far gives just over 3uV.
>
> > So what you get theoretically is only pie in the sky and most days the
> > tubes are more noisy than the data and calcs suggest they ought to be.
> >
>
> Too bloody right.
>
> > gm does obviously have some bearing on the noise.
> >
>
> Pass. I guess one would need to measure tubes with large differences in
> gm to be sure about that.
You can see what the gm is for a triode by looking at the typical Ia Ea
curves showing Ra for various Eg1 bias voltages.
Draw a vertical line through a wanted Ea for the tube and plot where Eg1
voltages intersect this line.
For say 6CG7/6SN7 at 10mA of Ia and Ea = 150V, gm is fairly constant
about a small Ia change of +/- 1mA/
But if you have Ia at 3mA, and Ia changes +/- 2mA, then gm changes a
lot.
The change in gm relates to a cube root of something squared.
With pentodes, the gm is more easily seen with a vertical load line, ie,
load that is a short circuit.
> >>> One has to measure the noise after an amp raises the signal at the anode
> >>> of a tube under test x 1,000 times then divide the noise figure by a
> >>> total of [1,000 x tube gain]. We assume the noise at the anode of the
> >>> tested tube is mainly all due to EINR. Its difficult to measure it any
> >>> other way.
> >> Depends what you measure it with. My Ferrograph test set has a 1mV range
> >> so a 40dB amp would make its full scale 10uV. Even a tube stage with
> >> 0.5uV EIN and a mere 20X gain will produce 10uV of noise at the output.
> >
> > Hang on, if you had 0.1uV and 40dB of gain, ie, 100x, then you'd have
> > 10uV after the gain.
> >
>
> But you don't have 0.1uV, even at the input. Say we have a perfect tube
> with gm =2mA/V at the operating point and a gain of 20. Its input noise
> will be 0.6uV so the noise you measure at its output will be 12uV.
OK
>
> > There is 1000uV per mV so the 10uV would be not a huge meter swing, only
> > 1/100 of the linear scale.
>
> Indeed, that's why I am suggesting a 40dB amp before it.
OK
>
> > But if your meter can tell you that you have 10uV then fine, but perhaps
> > its better to amplify the 0.1uV x 1,000 and get 0.1mV, and then you'd
> > see a swing of a meter needle of 10% of the scale on a 1mV range.
> >
>
> But it's not 0.1uV, its 12uV so you anly need 40dB of gain.
12uV x 100 = 1.2mV. If you have a 1mV or 3mV range then fine.
My mV meter I made long ago is only 10mV for the lowest range.
>
> > When testing 1/2 12AX7 with shorted grid via a short lead, and if you
> > have 2uV of EIN, then expect 120uV at the anode if the AX7 gain is 60x.
> > In my case the following amp has gain of 1,500, so this becomes 180mV,
> > and a 1V scale is OK.
>
> Yes, but my Ferrograph has a 1mV scale so I don't need anywhere near as
> much extra gain.
Indeed...
>
> The input noise of the amp used to lift the signal
> > to measure it
> > is maybe 3uV or less and so the noise at the amp input is the square
> > root of ( 120uV squared + 3uV squared ) and if you work this out then
> > the following amp noise makes virtually no difference One could have its
> > noise at 10uV and the 12AX7 noise would swamp the noise of the following
> > amp. My amp I use is a 2 stage tube type. One could use an opamp.
> >
> > My noise measures are all with dc on the heaters.
> >
>
> Agreed
>
> > With ac on the heaters expect much more hum often due to leakage current
> > between cathode and heaters.
>
> Yes, been there, done that, pulled my hair out and bought the T-shirt.
>
> > All my preamps always use dc heaters to avoid cathode hum.
> >
> > A bad tube will have intermittent "sploshy" noise whose amplitude varies
> > between good and bloody terrible, and so you have to bin thse tubes
> > unless where you use it doesn't matter, maybe a second stage in a guitar
> > amp.
> >
>
> Selection for low noise is definitely to be recommended for the first stage.
That was one of the techie's job when he visited a studio or radio
station for regular maintenance, or when a tube became noisy.
>
> > In the old days the recording studios may have had quite a number of
> > 12AX7 in consoles etc and a tech would have to swap out tubes to keep
> > the noise low at service time. Of course the EF86 became immensely
> > popular because you could get more gain than with a triode and then use
> > NFB to help control noise. RDH4 has a few typical ideas used in studios
> > and broadcast stations. One is to use a pentode with its anode taken to
> > 0V and then use the screen as the anode with a lowish "anode" current.
>
> Yes, I am well aware of that one. I even managed to get a hold of the
> article the RDH4 refers to. The best noise I have got so far was from a
> 6AU6 pentode wired as a triode as per this article.
Maybe the shot noise is less because the metal area the electrons have
to land on is much smaller than the much larger area of the official
anode.
>
> > Good screening, but I assume the EIN is still related to the 2.5/gm, and
> > gm of this sort of triode connected pentode is rather low with typical
> > types,
>
> I have measured gm between 3 and 4mA/V for a 6AU6 in triode mode.
So it should be because that's the pentode gm.
When screen and anode are connected to a fixed B+, and a 10 ohm anode
current sensing R connected, the gm can easily be measured by biasing
the g1 for a given idle Ia and then altering the bias +/- with a signal
and measuring the Ia change. Because Ea and Eg2 doesn't change in the
test, triode and pentode gm should be the same.
>
> although perhaps with 6EJ7 whichy is a high gm tube, or an E280F,
> > then maybe they are quieter than normal full triode connection. I recall
> > I tried the grounded anode idea but found magnetically induced hum to be
> > a bother with EF86.
> >
> > I have not read any reports on the Net or in old books where anyone
> > successfully managed to get tube input noise lower than has been
> > mentioned as typical in RDH4.
> >
>
> Me neither. I have an article by Burgess from the Wireless Engineer
> February 1945 which quotes the 2.5/gm derived by Harris and says it
> 'shows close agreement with measured values'. Now I think Harris and I
> know for certain the Burgess is talking about radio receivers and for
> them the lowest noise stage needs to be the rf amp or mixer.
Yes, indeed, especially when you are trying to sift out a low level
signal from all the RF noise picked up and the noise of the receiver
itself.
But with AM sets RF amps or mixer noise doesn't matter much for
broadcast stations.
> Now maybe
> 2.5/gm applies at rf frequencies but it looks to me like it does not
> apply for audio.
Why would it not apply?
And wouldn't frequency vary things? as F rises, Rin drops, due to input
C.
> Indeed, Burgess references Harris' paper which is
> entitled 'Fluctuations in space charge limited currents at MODERATELY
> HIGH FREQUENCIES' (my capitals) which does no sound like audio to me.
>
> > Nobody bothers to really examine the issue of tube input noise very much
> > in these times of opamps and fets and so forth.
> >
>
> But it is something WE should examine as tube users.
Indeed. But only if we need to. I test tubes for noise to make sure my
customers are not upset by noise....
Many sellers and traders of tubes have no technical abilities and they
trade NOS tubes that are anything but NOS and are infact pulls that look
well in a box that looks new after having been carefully stored.
NOS tubes are often a complete scam.
>
> > But I found that afer several years of business I hade acquired a
> > collection of preamp tubes, mainly pulls from garbage gear, and I began
> > to wonder what was actually good amoung the pile. Some had to be good if
> > the garbage gear had had its tube replaced just before it was junked.
> > And I found some tubes were, low EIN, not very microphonic, nice bright
> > looking getter, hmm, maybe very close to NOS, so I went through the
> > piles and soon discovered what I could put in a "good" pile, ie, less
> > than 3uV for AX7, AU7, AT7 CG7, and what would be N, for noisy with the
> > average per section written on the tube, and M, for microphonic, ie,
> > well above a good sample. Some were quiet but M. Some are marked N,M,
> > meaning a bit noisy and microphonic but usable in a power amp second
> > stage for good while. Quite a few were just trash, more than 10uV and
> > micro, so out they went to the bin. Some had terrible hum.
> >
>
> Can you remember what proportion of the total had low noise and
> microphony?/
Maybe 20%.
One non technical guy had inherited an enormous number of 12AU7 and 6CG7
amoung cratefulls of stuff an old guy had left over from being a tech.
He offered them to me and I explained that old guys' collections of
tubes are usually crap and that I couldn't pay much.
OK, so we negotiated a deal and a price based on me paying a certain
agreed number of $$$ per tube if they tested OK. He would send them and
I would test them and any that failed my tests would be sent back with
noise and microphonic figures on the wrappers.
So a box arrived and I tested each and every one of about 150 tubes.
Took me quite some time, and at tech rates this added up to a
considerable sum. Most of the 6CG7 were in fact brand spanking new. But
only 20% of teh 12AU7 were OK, so a pile went back with my cheque for
the goodies.
The 6CG7 were made in Oz and I am content they sound as good as anything
by RCA or Seimans, but they don't fetch the big E-bay prices based
largely on the perceptual bullshit of the hi-fi glitterati.
In the earlier era of 1930s to 1960s, Oz had its own protected
electronics industry so we could be independant if need be.
So we made our own range of tubes and they were mainly very good because
the industry was honest and the working girls were well paid in nice
conditions so they did good work.
Now all that's gone, and won't be again.
Now the US is hard at it making thousands of robot thinges for soldiers
to use for its security such as something the size of a model aeroplane
and crammed full of secret electronics is the go.
No need for us to have a local electronics industry for national
security.
The terrorist in the open field waiting to make a hit now has a much
decreasing chance of succeeding. Some little gizmo will fly high above
and unseen, and guide in a little surprize package.....
But of course we have seen superior technical force defeated during
history......
Patrick Turner.
Ian Bell
Ah, that explains why you need the extra gain.
That's a very good point. The screen grid is clearly much smaller than
the anode. However, the papers I have read say shot noise is due to
changes in the rate of arrival of electrons which should be unchanged.
Maybe it does but maybe at audio frequencies other mechanisms begin to
dominate. I don't know. What I do know is they did not do their testing
at audio frequencies so we cannot be certain the formula applies.
> And wouldn't frequency vary things? as F rises, Rin drops, due to input
> C.
Yes, perhaps, which would make things worse a low (audio) frequencies.
Interesting. I would have expected a lower figure but 20% is quite
encouraging.
> One non technical guy had inherited an enormous number of 12AU7 and 6CG7
> amoung cratefulls of stuff an old guy had left over from being a tech.
>
> He offered them to me and I explained that old guys' collections of
> tubes are usually crap and that I couldn't pay much.
> OK, so we negotiated a deal and a price based on me paying a certain
> agreed number of $$$ per tube if they tested OK. He would send them and
> I would test them and any that failed my tests would be sent back with
> noise and microphonic figures on the wrappers.
> So a box arrived and I tested each and every one of about 150 tubes.
> Took me quite some time, and at tech rates this added up to a
> considerable sum. Most of the 6CG7 were in fact brand spanking new. But
> only 20% of teh 12AU7 were OK, so a pile went back with my cheque for
> the goodies.
>
> The 6CG7 were made in Oz and I am content they sound as good as anything
> by RCA or Seimans, but they don't fetch the big E-bay prices based
> largely on the perceptual bullshit of the hi-fi glitterati.
>
I recently bought some 6CG7 off eBay and all the RCA ones tested better
than the others.
> In the earlier era of 1930s to 1960s, Oz had its own protected
> electronics industry so we could be independant if need be.
>
> So we made our own range of tubes and they were mainly very good because
> the industry was honest and the working girls were well paid in nice
> conditions so they did good work.
>
> Now all that's gone, and won't be again.
>
> Now the US is hard at it making thousands of robot thinges for soldiers
> to use for its security such as something the size of a model aeroplane
> and crammed full of secret electronics is the go.
>
> No need for us to have a local electronics industry for national
> security.
>
LOL
> The terrorist in the open field waiting to make a hit now has a much
> decreasing chance of succeeding. Some little gizmo will fly high above
> and unseen, and guide in a little surprize package.....
>
> But of course we have seen superior technical force defeated during
> history......
>
Oh, how true.
heers
Ian
>
> Patrick Turner.
Patrick Turner
The only time I figured I'd need to have a lower range for voltage
measurement is for distortion measuring.
So when I measure THD on 1kHz, I filter the THD out with a bridged T LC
notch filter and then amplify the THD with a bandpass amp based around
an opamp with gain = 10 and band from 1.6Khz to 10kHz with sharp roll
off below 1.6kHz to get some extra 1kHz attenuation, and obstruction to
all noise below 1kHz.
Therefore if I have 0.1mV of THD, it appears as 1mV on my meter, and if
the signal tested is 1Vrms, this means I have 0.01% THD, which is a
common typical amount of THD for an SE triode preamp with a ľ-follower
topology.
If V tested = 10Vrms as from say a good measuring SS amp the THD is
0.001%, a fairly typical result.
But apart from THD measurements, having a range 0 to 1.0mV is a bit
useless, and very prone to noise.
Probably right. Temperature needs to be kept low though or else the
secondary emission may make noise higher. So the Ia using g2 as anode
cannot be too high, which means gm will be on the low side as it varies
a lot with Ia.
Unfortunately we don't have access to the references nominated in RDH4
about such issues around tube manufacture and R&D.
Many people have a copy of RDH4, but that's all, and they don't have the
library full of other material which delves deeper into tube behaviour.
Most has been chucked out of libraries long ago.
Page 783 of RDH4 states equivalent input noise 2.5/gm and that this is
an expression of shot effect noise based around the idea that such noise
is equivalent to an R between grid and cathode. It is of course an
imaginery R because grid input resistance is in megohms,
although grid input impedance is falling as F rises due to C.
A useful table appears on page 938 with input noise resistances given
for various tubes. The lowest EINR is for 1/2 a 6J6 which is 470 ohms.
And so if you used both triodes of a 6J6 you'd have EINR = 235 ohms and
about a third of the noise of a 1/2 of 12AX7.
And I will eat all my 12AX7s if you can find me a 6J6 with noise much
lower than a 12AX7. EC97 should also be good, but samples I have are
not.
I don't see where RDH4 says the noise depends on frequency, or that the
EINR was calculated based on measurements at RF.
The rules about noise are valid right down to dc once you have said that
the noise is due to a resistance.
So where you have bandwidth of dc to 20kHz, and say you have an R giving
2uV of noise, then if R is reduced by 10 then noise reduces by 1/3.16 or
about a third. And if the bandwidth is effectively reduced from 20khz to
50 as it is with an RIAA filter then noise is reduced the square root of
the amount of bandwidth reduction, ie, sq.rt ( 50 / 20,000 ) = 1 / 20,
or -26dB.
>
> > And wouldn't frequency vary things? as F rises, Rin drops, due to input
> > C.
>
> Yes, perhaps, which would make things worse a low (audio) frequencies.
Depends on the resistance. The lower the R the less effect C has.
Tubes tend to have higher noise as F is lower, and this spoils their
effective use for phono inputs because of how RIAA eq means bass signals
off records are low and amp gain must be high at LF.
j-fets on the other hand have less LF noise and very little rumble or
flicker noise.
Now if the record industry had settled for 18" records with 45 RPM,
maybe things could have been different....
They ended up finally settling for a 5" silver disc with 1 and 0 used in
a sort of morse code........
Now the disc is doomed, and its all done by file handling on memory
chips. Not even a hard drive disc involved.
But I still enjoy a good vinyl.
Patrick Turner.
Ian Bell
It doesn't AFAIK. However, the title of the Harris paper that derived
the 2.5/gm formula makes it clear he is talking about high frequencies.
Chech out the reference at the end of that chapter in RDH4.
> The rules about noise are valid right down to dc once you have said that
> the noise is due to a resistance.
>
> So where you have bandwidth of dc to 20kHz, and say you have an R giving
> 2uV of noise, then if R is reduced by 10 then noise reduces by 1/3.16 or
> about a third. And if the bandwidth is effectively reduced from 20khz to
> 50 as it is with an RIAA filter then noise is reduced the square root of
> the amount of bandwidth reduction, ie, sq.rt ( 50 / 20,000 ) = 1 / 20,
> or -26dB.
>
Where did that factor of 50 come from. According to Morgan Jones it is
nearer 1.
Cheers
Ian
Patrick Turner
Ian Bell wrote:
>
> Patrick Turner wrote:
> >
> > Ian Bell wrote:
> >> Patrick Turner wrote:
> >>> Ian Bell wrote:
> >>>> Patrick Turner wrote:
> >>>>> Ian Bell wrote:
> >>>>>> Patrick Turner wrote:
snip
> > I don't see where RDH4 says the noise depends on frequency, or that the
> > EINR was calculated based on measurements at RF.
> >
>
> It doesn't AFAIK. However, the title of the Harris paper that derived
> the 2.5/gm formula makes it clear he is talking about high frequencies.
> Chech out the reference at the end of that chapter in RDH4.
There's a reference to Harris, page 946, item B27, but no reference to
RF.
On page 937, there's a table attributable to Harris with formulas for
noise and the table has 4 sections, triode amps, pentode amps, triode
mixers, pentode mixer, multigrid mixer.
The only simple formula for noise resistance is the first one for triode
noise resistance, 2.5/gm., and there's no mention of RF specificness, so
I think that 2.5/gm is the official triode EINR formula regardless of F.
All the others look difficult to work out, but none have reactive
components in the formula.
If you use a 6BE6 as a mixer for two inputs to make one output then
it'll be noiser than if you have two triodes with a common anode load.
All that's important is that triodes are quieter than all other types of
tubes and that the higher the gm the better, and then all one needs to
consider is to have Ia high, and Ea low right at the input of a phono
stage, and hope it works out.
Usually it does for MM. And when Denon invented an MC with less than 20
ohms Rout, and with very low Vout, a transformer with 1:10 ratio was the
normal way to up the Vo and yet keep noise lower than with MM. The ZR of
100 meant that 20 ohms became only 2,000 ohms from grid to 0V of a tube
input stage. If EINR of 2,000 ohms gives 2uV of noise, then 20 ohms
makes about 0.2uV, and when transformed up this noise becomes 2uV, and
when added to the series 2uV of the tube you get 2.8uV. The cart output
of 0.3mV has become 3mV, and snr = -60dB, and thus just OK considering
vinyl noise will be higher.
Now if you have a 2sk369 j-fet with 0.14uV input noise ( typical ) and
0.4mV of rated MC cart Vo, then snr = -69dB, and there's less LF noise
so less grumble in the signal.
It was j-fets that allowed a whole new generation of bugs to be
developed for spying because it was far easier to conceal a little
battery and a fet than have to wire up a triode. AND the fet picked up
more clean sound without the noise of itself ruining the time spent by
CIA operatives in vans outside the russian embassy.
>
> > The rules about noise are valid right down to dc once you have said that
> > the noise is due to a resistance.
> >
> > So where you have bandwidth of dc to 20kHz, and say you have an R giving
> > 2uV of noise, then if R is reduced by 10 then noise reduces by 1/3.16 or
> > about a third. And if the bandwidth is effectively reduced from 20khz to
> > 50 as it is with an RIAA filter then noise is reduced the square root of
> > the amount of bandwidth reduction, ie, sq.rt ( 50 / 20,000 ) = 1 / 20,
> > or -26dB.
> >
>
> Where did that factor of 50 come from. According to Morgan Jones it is
> nearer 1.
50 is the new bandwidth after RIAA is added to the circuit because the
response is dc to -3db at 50Hz. BW was 20kHz before RIAA.
Therefore BW reduction factor = 50Hz / 20,000Hz = 1/400. The noise
reduction factor is the square root of bandwidth reduction = sq.rt of
1/400 = 1/20 = -26dB.
Now in vinyl recording they boost the hell out of treble above the low
bass reference, +40dB at 22 kHz compared to 10Hz.
So the RIAA filter flattens the music response and disposes of most of
the noise. The ears of most people don't notice LF noise as much as
noise above say 300Hz. Now because tubes make a fair amount of LF noise,
the theoretical -26dB noise reduction doesn't get achieved.
There will still be some noise. But on good vinyl its often tape hiss or
venue noise which is the highest noise of the recording. In FM radio HF
is boosted or "emphasised" and after detection the signal is
de-emphasized. Exit much noise.
The reference signal to make a final perception of all this is of course
at 1kHz at which most signal Vout figures for carts are quoted.
So if you have 3mV from an MM cart, there's only 0.3mV at 10hz and 30mV
at 22Khz An MC bass reference is 0.03mV.
I may not have all this 100% correct but Allen Wright's preamp cookbook
has a lot more.
Everything is relative though, and vinyl is very listenable when its
good and turn tables keep being designed and sold at very high prices in
hi-fi shops. Its nice to hear music straight from a black groove. But
TTs are also designed and sold in less salubrious emporioums where the
poor buy plastic contraptions for $100 to move their LP music to CD.
This is silly of course, but people like to feel they have saved the
past even though the vast majority of pop music has no relevance to life
being lived well now and never did 40 years ago.
So, at the end of the day if we want to use MC and get lowest amp noise
we may choose between a fet, a step up tranny, or carefully selected
triodes in about that order. Some say transformers veil the sound and so
forth. Vinyl is the land of claims.
Patrick Turner
Ian Bell
>
> Ian Bell wrote:
>> Patrick Turner wrote:
>>> Ian Bell wrote:
>>>> Patrick Turner wrote:
>>>>> Ian Bell wrote:
>>>>>> Patrick Turner wrote:
>>>>>>> Ian Bell wrote:
>>>>>>>> Patrick Turner wrote:
>
> snip
>
>>> I don't see where RDH4 says the noise depends on frequency, or that the
>>> EINR was calculated based on measurements at RF.
>>>
>> It doesn't AFAIK. However, the title of the Harris paper that derived
>> the 2.5/gm formula makes it clear he is talking about high frequencies.
>> Chech out the reference at the end of that chapter in RDH4.
>
> There's a reference to Harris, page 946, item B27, but no reference to
> RF.
>
You are quite right. In RDH4 is calls Harris' paper 'Fluctuations in
vacuum tube amplifiers and input systems'
However, the Burgess paper I have gives it its full title which is:
'Fluctuations in Space Charge Limited Currents at Moderately High
Frequencies, Part V - Fluctuations in Vacuum Tube Amplifiers and Input
Systems'
Clearly the RDH4 reference is to just one of a series of articles but
they all seem to be concerned with 'moderately high frequencies'. Indeed
Burgess talks exclusively about rf circuits.
> On page 937, there's a table attributable to Harris with formulas for
> noise and the table has 4 sections, triode amps, pentode amps, triode
> mixers, pentode mixer, multigrid mixer.
>
> The only simple formula for noise resistance is the first one for triode
> noise resistance, 2.5/gm., and there's no mention of RF specificness, so
> I think that 2.5/gm is the official triode EINR formula regardless of F.
>
I agree its official but I am not sure it is correct at af.
> All the others look difficult to work out, but none have reactive
> components in the formula.
>
> If you use a 6BE6 as a mixer for two inputs to make one output then
> it'll be noiser than if you have two triodes with a common anode load.
>
> All that's important is that triodes are quieter than all other types of
> tubes and that the higher the gm the better, and then all one needs to
> consider is to have Ia high, and Ea low right at the input of a phono
> stage, and hope it works out.
>
I agree it is important that triodes are quieter than all other types.
What is also important is how quiet they ahould be expected to be. If
you don't know this how can you possibly select tubes for low noise?
Sorry that makes no sense to me. How does the bandwidth suddenly become
reduced to 50Hz?
BW was 20kHz before RIAA.
> Therefore BW reduction factor = 50Hz / 20,000Hz = 1/400. The noise
> reduction factor is the square root of bandwidth reduction = sq.rt of
> 1/400 = 1/20 = -26dB.
>
> Now in vinyl recording they boost the hell out of treble above the low
> bass reference, +40dB at 22 kHz compared to 10Hz.
Depends what you are talking about. If you are talking about the
amplitude of the recorded groove then it is nearly constant with
frequency - there's just a small boost in the mid frequency range.
That's why ceramic cartridges need very little equalisation.
The reason a MM or MC cartridge gives a rising level with increasing
frequency is because its output is proportional to rate of change of flux.
> So the RIAA filter flattens the music response and disposes of most of
> the noise.
I am still not clear on this. I need to check what Jones says.
The ears of most people don't notice LF noise as much as
> noise above say 300Hz. Now because tubes make a fair amount of LF noise,
> the theoretical -26dB noise reduction doesn't get achieved.
> There will still be some noise. But on good vinyl its often tape hiss or
> venue noise which is the highest noise of the recording. In FM radio HF
> is boosted or "emphasised" and after detection the signal is
> de-emphasized. Exit much noise.
>
That's different and is done deliberately to reduce noise. RIAA is not
for that reason.
Cheers
ian
Paul Crawford
> noise resistance, 2.5/gm., and there's no mention of RF specificness, so
> I think that 2.5/gm is the official triode EINR formula regardless of F.
I have yet to find this document, but it is normally the case the noise
rises at 1/f below a certain point, typically 10-100kHz for MOSFET/GaAs
FET devices, or 100Hz-1kHz for BJT in oscillators.
> It was j-fets that allowed a whole new generation of bugs to be
> developed for spying because it was far easier to conceal a little
> battery and a fet than have to wire up a triode. AND the fet picked up
> more clean sound without the noise of itself ruining the time spent by
> CIA operatives in vans outside the russian embassy.
Nope, BJT work very well in those cases. Just wobble the bias of a BJT
oscillator and instant FM...
> Now in vinyl recording they boost the hell out of treble above the low
> bass reference, +40dB at 22 kHz compared to 10Hz. So the RIAA filter
> flattens the music response and disposes of most of the noise. The ears
The RIAA curve results in a boots of bass, as treble is cut in the
recording to tame the cutting head acceleration (and 'hiss' noise). One
way or another, the two effects conspire to produce more LF noise than
one would like.
> Everything is relative though, and vinyl is very listenable when its
> good and turn tables keep being designed and sold at very high prices in
> hi-fi shops. Its nice to hear music straight from a black groove. But
> TTs are also designed and sold in less salubrious emporioums where the
> poor buy plastic contraptions for $100 to move their LP music to CD.
> This is silly of course, but people like to feel they have saved the
> past even though the vast majority of pop music has no relevance to life
> being lived well now and never did 40 years ago.
I have the Linn LP12 and it works pretty well. Maybe others are better,
or at least different, but I enjoy it and hope to have more time to play
with valve hi-fi in due course.
> So, at the end of the day if we want to use MC and get lowest amp noise
> we may choose between a fet, a step up tranny, or carefully selected
> triodes in about that order. Some say transformers veil the sound and so
> forth. Vinyl is the land of claims.
My vote (so far) is for a JFET followed by a tube pre/power amp set. I
have quite a good MOSFET power amp of my own design, but these days the
treble area is of less importance due to age, and the midrange is an area
valves tend to excel in.
Paul
Patrick Turner
Well, Johnson noise in resistors is is a fact of life at all F.
If you just connect various values at the input of an amp you can listen
to it.
And when they say the input of a triode consists of an equivalent input
noise resistance between grid and cathode, then that's the model, and I
don't think there is any special consideration with RF. As bandwidth
extends up past the audio band, so does the noise and the noise voltage
at a rate of the square root of increase in BW. So thus having a bw =
2Mhz instead of only 20kHz, means noise should become 10 times greater.
Going to 20MHz would raise noise by 3.16 more times. Then if you have a
tuned circuit at say 1MHz, and BW = 18kHz as for a broadcast
AM radio station, then the reduction of BW is 18,000 / 1,000,000, or
about 1/50, so expect noise to be about 1/7 of when you have bw = 1Mhz.
>
> > All the others look difficult to work out, but none have reactive
> > components in the formula.
> >
> > If you use a 6BE6 as a mixer for two inputs to make one output then
> > it'll be noiser than if you have two triodes with a common anode load.
> >
> > All that's important is that triodes are quieter than all other types of
> > tubes and that the higher the gm the better, and then all one needs to
> > consider is to have Ia high, and Ea low right at the input of a phono
> > stage, and hope it works out.
> >
>
> I agree it is important that triodes are quieter than all other types.
> What is also important is how quiet they ahould be expected to be. If
> you don't know this how can you possibly select tubes for low noise?
You don't have to know all the theory; just general concepts a bit. Then
you find out what works by experimentation. Its the hands on work that
keeps the studio recordists happy with their gear.
A phono preamplifier is in fact a bandpass filter. Noise enters the amp
at the front end with wide bw maybe in excess of 50kHz.
But the RIAA filter has its first pole at 50Hz, -3dB point and
attenuation after that is first order for awhile, a bit of a flatter
curve around 1kHz, then back to first order at -6dB per octave again.
I suggest you set up a 6CG7 triode for test with RL about 100k and
bypassed Rk and about 3mA of Ia. You can measure the anode signal as we
have discussed to test noise. We can also place a 1M resistance from
grid to 0V to vastly increase the noise at the grid over having the grid
connected directly to 0V for tube noise testing. Now suppose the anode
load R plus Ra = 10k. If we place cap of say 0.318uF from anode to 0V,
will will see a dramatic reduction in noise.
Now with Rg = 1M at the the triode input, the noise bw of the 1M is
high, but then the R has some self C and then you have the miller C
between anode and grid of maybe 50pF so as F rises the noise measured at
the anode gets less because of the miller effect which is a negative
feedback effect. So don't expect much noise at the 6CG7 at above 10Mhz.
The capacitances beome low impedances which shunt the noise across
higher impedances involved.
But what we would find is that noise voltages above 50Hz would be
attenuated at 6dB/octave with 0.318uF from anode to 0V.
The RIAA filter acts very similarly to having just a singe 0.318uF from
anode to 0V on a 6CG7 set up as described.
The bandwidth of signal is measured between poles in the response. Now
the 6CG7 tested this way has BW from dc to 50Hz. But if we add a CR
coupling as we might use on a second amp stage then there may be a LF
pole at say 20Hz. So signal BW becomes 20Hz to 50Hz, or only 30Hz.
If the initial signal bw at the front end of the amp was dc to 20kHz,
then the bw reduction factor becomes 30/20,000 = 0.0015 and the noise
reduction factor = 0.038 = 28.2dB, a bit more than what we worked out
for bw = dc to 50Hz. The noise below 20Hz is missing.
Knowing about noise and its bandwidth and the value of capacitors to
reduce it means that if we have a noisy signal from some device under
test and we want to see the hum voltage in the blur of noise on the CRO,
than all we need do is place a shunt C of the right value to shunt all
the HF junk above say 150Hz, and we can see the hum clearly.
The RIAA filter cleans up the signal from the phono cartridge which
would otherwise sound like a rock being dragged down a canyon were it
not for the boosting of mid-treble 40dB above the 10Hz signal level
during recording and the filtering in the playback preamp. Along with
reducing noise in the recording, the RIAA also cuts the distortions of
the first am stage a little because higher harmonics of fundementals get
attenuated more than the fundementals.
Our ancestors were clever, doncha think?
>
> BW was 20kHz before RIAA.
> > Therefore BW reduction factor = 50Hz / 20,000Hz = 1/400. The noise
> > reduction factor is the square root of bandwidth reduction = sq.rt of
> > 1/400 = 1/20 = -26dB.
> >
> > Now in vinyl recording they boost the hell out of treble above the low
> > bass reference, +40dB at 22 kHz compared to 10Hz.
>
> Depends what you are talking about. If you are talking about the
> amplitude of the recorded groove then it is nearly constant with
> frequency - there's just a small boost in the mid frequency range.
> That's why ceramic cartridges need very little equalisation.
>
> The reason a MM or MC cartridge gives a rising level with increasing
> frequency is because its output is proportional to rate of change of flux.
Well OK, but they boost the signal sent to the cutting head for LPs.
Say you just recorded with a flat signal. You'd not be able to fit bass
grooves on the record very well. High F treble signals would be
seriously affected by competing levels of noise.
On playback, noise noise would be intolerable.
>
> > So the RIAA filter flattens the music response and disposes of most of
> > the noise.
>
> I am still not clear on this. I need to check what Jones says.
The disc recording process is a bit of a compromise.
>
> The ears of most people don't notice LF noise as much as
> > noise above say 300Hz. Now because tubes make a fair amount of LF noise,
> > the theoretical -26dB noise reduction doesn't get achieved.
> > There will still be some noise. But on good vinyl its often tape hiss or
> > venue noise which is the highest noise of the recording. In FM radio HF
> > is boosted or "emphasised" and after detection the signal is
> > de-emphasized. Exit much noise.
> >>
> That's different and is done deliberately to reduce noise. RIAA is not
> for that reason.
But RIAA reduces noise. It allows bass signals to "fit" onto the record
to give use 20 minutes a side from LP.
In fact if we use 1kHz as the reference, then the bass is cut, and
re-boosted i the playback amp and the noise in bass regions isn't much
reduced by RIAA; 100Hz signals are attenuated about -12dB compared to
1kHz. Noise at around 100Hz in the grooves and input amp isn't much
reduced but is amplified with the LF signal where most ppl agree the
extra bit of noise due to the vinyl and electronics does not matter due
to our ears being less sensitive to a given spl of noise as F goes low.
The noise in the bass music is low because its a small part of the whole
20kHz F range. But after 50Hz the signal recoverd from the record groove
has an increasing amplitude at about +6dB per octave. The snr of the
recovered signal gets better as F goes higher and then C of the RIAA
filter just flattens this signal and reduces noise enormously and
preserves a good snr. And we need all the noise attenuation we can get
because our ears become very sensitive to higher F.
I'm speaking of a concept. The exact numbers don't matter. Just use an
amp with low enough noise and keep other sources of noise low say by
cleaning records well and you'll enjoy a vinyl.
If your'e gonna build an MC amp its worth taking a little extra trouble
compared to the lack of trouble taken by many bottom end budget disc
replay systems.
Patrick Turner.
Patrick Turner
Paul Crawford wrote:
>
> > The only simple formula for noise resistance is the first one for triode
> > noise resistance, 2.5/gm., and there's no mention of RF specificness, so
> > I think that 2.5/gm is the official triode EINR formula regardless of F.
>
> I have yet to find this document, but it is normally the case the noise
> rises at 1/f below a certain point, typically 10-100kHz for MOSFET/GaAs
> FET devices, or 100Hz-1kHz for BJT in oscillators.
>
> > It was j-fets that allowed a whole new generation of bugs to be
> > developed for spying because it was far easier to conceal a little
> > battery and a fet than have to wire up a triode. AND the fet picked up
> > more clean sound without the noise of itself ruining the time spent by
> > CIA operatives in vans outside the russian embassy.
>
> Nope, BJT work very well in those cases. Just wobble the bias of a BJT
> oscillator and instant FM...
Fets are good mic amps. After getting a recovered clean audio signal,
then you use that to wobble the fm oscillator F to transmit it.
>
> > Now in vinyl recording they boost the hell out of treble above the low
> > bass reference, +40dB at 22 kHz compared to 10Hz. So the RIAA filter
> > flattens the music response and disposes of most of the noise. The ears
>
> The RIAA curve results in a boots of bass, as treble is cut in the
> recording to tame the cutting head acceleration (and 'hiss' noise). One
> way or another, the two effects conspire to produce more LF noise than
> one would like.
LF noise in the process is not too bad though. Most all people object to
high hiss levels and snap, crackles and pops with vinyl even when they
ain't eatin rice bubbles for brekki.
>
> > Everything is relative though, and vinyl is very listenable when its
> > good and turn tables keep being designed and sold at very high prices in
> > hi-fi shops. Its nice to hear music straight from a black groove. But
> > TTs are also designed and sold in less salubrious emporioums where the
> > poor buy plastic contraptions for $100 to move their LP music to CD.
> > This is silly of course, but people like to feel they have saved the
> > past even though the vast majority of pop music has no relevance to life
> > being lived well now and never did 40 years ago.
>
> I have the Linn LP12 and it works pretty well. Maybe others are better,
> or at least different, but I enjoy it and hope to have more time to play
> with valve hi-fi in due course.
>
> > So, at the end of the day if we want to use MC and get lowest amp noise
> > we may choose between a fet, a step up tranny, or carefully selected
> > triodes in about that order. Some say transformers veil the sound and so
> > forth. Vinyl is the land of claims.
>
> My vote (so far) is for a JFET followed by a tube pre/power amp set. I
> have quite a good MOSFET power amp of my own design, but these days the
> treble area is of less importance due to age, and the midrange is an area
> valves tend to excel in.
Indeed...
Patrick Turner.
>
> Paul
Ian Bell
Yes, but we are talking about shot noise which is a different beast
entirely. It is just convenient to model it as an equivalent inout
Johnson noise resistance.
> If you just connect various values at the input of an amp you can listen
> to it.
>
> And when they say the input of a triode consists of an equivalent input
> noise resistance between grid and cathode, then that's the model, and I
> don't think there is any special consideration with RF. As bandwidth
> extends up past the audio band, so does the noise and the noise voltage
> at a rate of the square root of increase in BW. So thus having a bw =
> 2Mhz instead of only 20kHz, means noise should become 10 times greater.
> Going to 20MHz would raise noise by 3.16 more times. Then if you have a
> tuned circuit at say 1MHz, and BW = 18kHz as for a broadcast
> AM radio station, then the reduction of BW is 18,000 / 1,000,000, or
> about 1/50, so expect noise to be about 1/7 of when you have bw = 1Mhz.
>
Agreed. I would not argue against Johnson noise, I am just saying I am
not convinced by the 2.5/gm formula.
>
>>> All the others look difficult to work out, but none have reactive
>>> components in the formula.
>>>
>>> If you use a 6BE6 as a mixer for two inputs to make one output then
>>> it'll be noiser than if you have two triodes with a common anode load.
>>>
>>> All that's important is that triodes are quieter than all other types of
>>> tubes and that the higher the gm the better, and then all one needs to
>>> consider is to have Ia high, and Ea low right at the input of a phono
>>> stage, and hope it works out.
>>>
>> I agree it is important that triodes are quieter than all other types.
>> What is also important is how quiet they ahould be expected to be. If
>> you don't know this how can you possibly select tubes for low noise?
>
> You don't have to know all the theory; just general concepts a bit. Then
> you find out what works by experimentation. Its the hands on work that
> keeps the studio recordists happy with their gear.
That's fine as long as the general concepts are accurate. So far I don't
fond 2.5/gm to be any where near reality.
Morgan Jones disagrees with you. He says the effective noise bandwidth
is 188Hz not 50Hz and also that you have to subtract the extra 19dB or
so gain you have to add to bring the response back at 1KHz which means
the overall noise improvement is only just over a couple of dB. He
cites the Texas Instruments Bi-FET manual IIRC as his source which I am
trying to find on the net now.
Not as a rule, only as they approach the inner grooves and the linear
velocity reduces.
> Say you just recorded with a flat signal. You'd not be able to fit bass
> grooves on the record very well. High F treble signals would be
> seriously affected by competing levels of noise.
>
Vinyl IS generally recorded with a near flat amplitude. Frequency
dependent boost on record is need because of the head characteristics
not to combat noise. Head cutters can cut bass into a record with no
problem. Its the playback carts that have the problem staying in the
groove because their compliance is frequency dependent.
> On playback, noise noise would be intolerable.
>>> So the RIAA filter flattens the music response and disposes of most of
>>> the noise.
>> I am still not clear on this. I need to check what Jones says.
>
> The disc recording process is a bit of a compromise.
>> The ears of most people don't notice LF noise as much as
>>> noise above say 300Hz. Now because tubes make a fair amount of LF noise,
>>> the theoretical -26dB noise reduction doesn't get achieved.
>>> There will still be some noise. But on good vinyl its often tape hiss or
>>> venue noise which is the highest noise of the recording. In FM radio HF
>>> is boosted or "emphasised" and after detection the signal is
>>> de-emphasized. Exit much noise.
>
>> That's different and is done deliberately to reduce noise. RIAA is not
>> for that reason.
>
> But RIAA reduces noise. It allows bass signals to "fit" onto the record
> to give use 20 minutes a side from LP.
No it does not. As I said, in amplitude terms, what is recorded on disc
is nearly constant amplitude versus frequency. A magnetic cart has a
rising output with frequency (just like a take head) because its output
is proportional to rate of change of flux. The top cut is necessary to
compensate for this.
Cheers
Ian
Patrick Turner
Somebody with far more insight than I have decided an imaginery
resistance between grid and cathode modelled tube noise OK.
I'm wondering that if I look further at this noise issue, will I make
more silent amps?
>
> > If you just connect various values at the input of an amp you can listen
> > to it.
> >
> > And when they say the input of a triode consists of an equivalent input
> > noise resistance between grid and cathode, then that's the model, and I
> > don't think there is any special consideration with RF. As bandwidth
> > extends up past the audio band, so does the noise and the noise voltage
> > at a rate of the square root of increase in BW. So thus having a bw =
> > 2Mhz instead of only 20kHz, means noise should become 10 times greater.
> > Going to 20MHz would raise noise by 3.16 more times. Then if you have a
> > tuned circuit at say 1MHz, and BW = 18kHz as for a broadcast
> > AM radio station, then the reduction of BW is 18,000 / 1,000,000, or
> > about 1/50, so expect noise to be about 1/7 of when you have bw = 1Mhz.
> >
>
> Agreed. I would not argue against Johnson noise, I am just saying I am
> not convinced by the 2.5/gm formula.
I'm not much convinced either. But when Rin is lower, there is less
noise. Even with bjts, paralleling a lot of them with very low base
current leads to more silent bjts. Less R.
>
> >
> >>> All the others look difficult to work out, but none have reactive
> >>> components in the formula.
> >>>
> >>> If you use a 6BE6 as a mixer for two inputs to make one output then
> >>> it'll be noiser than if you have two triodes with a common anode load.
> >>>
> >>> All that's important is that triodes are quieter than all other types of
> >>> tubes and that the higher the gm the better, and then all one needs to
> >>> consider is to have Ia high, and Ea low right at the input of a phono
> >>> stage, and hope it works out.
> >>>
> >> I agree it is important that triodes are quieter than all other types.
> >> What is also important is how quiet they ahould be expected to be. If
> >> you don't know this how can you possibly select tubes for low noise?
> >
> > You don't have to know all the theory; just general concepts a bit. Then
> > you find out what works by experimentation. Its the hands on work that
> > keeps the studio recordists happy with their gear.
>
> That's fine as long as the general concepts are accurate. So far I don't
> fond 2.5/gm to be any where near reality.
So what is a better solution?
OK, I thought signal BW was defined by the -3dB points.
But yes, if 1khz is the reference, and you have a low level bass signal
then snr at bass does not improve at bass.
But consider a microphone amp. Response is flat, and noise at input is
all amplified by a uniform gain.
So if there was 2mV of signal and 2uV of noise at the input, 20kHz bw,
snr = -60dB, and this will also be at the output, unless the amp
contributes considerable other noise in the path.
The same amp when used with an RIAA filter in the signal path has the
same noise below 50Hz, but the bass signals are very low, say 0.2mV at
low bass compared to mid F based around 1kHz at 2mV and so bass signal
snr is worse, but not a huge fraction of total. 2mV at 1kHz enters the
amp and noise and 1khz are both attenuated equally. Then the treble is
boosted so that 20mV appears at 22kHz so the noise above 1kHz begins to
be much lessened while the signal is flattened.
Just how quiet things sound depends on choice of low noise parts at the
amp input.
Allen Wright says a lot about how treble is boosted and bass is cut
acording to the RIAA recording curve which is the opposite to the
playback curve. He says its important to have just the right RIAA
playback filter to allow for the fact the treble isn't boosted
infinitely, but the only up to maybe 50kHz, or else they have
instability troubles with the cutting head amp and its loops of negative
feedback to ensure distortion is minimized.
>
> > Say you just recorded with a flat signal. You'd not be able to fit bass
> > grooves on the record very well. High F treble signals would be
> > seriously affected by competing levels of noise.
> >
>
> Vinyl IS generally recorded with a near flat amplitude. Frequency
> dependent boost on record is need because of the head characteristics
> not to combat noise. Head cutters can cut bass into a record with no
> problem. Its the playback carts that have the problem staying in the
> groove because their compliance is frequency dependent.
I've always thought there was considerable bass cut and treble boost in
recording.
But I have not recorded anything or cut a record. All I know is that
records sound fine with RIAA, and lousy without it.
>
> > On playback, noise noise would be intolerable.
> >>> So the RIAA filter flattens the music response and disposes of most of
> >>> the noise.
> >> I am still not clear on this. I need to check what Jones says.
> >
> > The disc recording process is a bit of a compromise.
> >> The ears of most people don't notice LF noise as much as
> >>> noise above say 300Hz. Now because tubes make a fair amount of LF noise,
> >>> the theoretical -26dB noise reduction doesn't get achieved.
> >>> There will still be some noise. But on good vinyl its often tape hiss or
> >>> venue noise which is the highest noise of the recording. In FM radio HF
> >>> is boosted or "emphasised" and after detection the signal is
> >>> de-emphasized. Exit much noise.
> >
> >> That's different and is done deliberately to reduce noise. RIAA is not
> >> for that reason.
> >
> > But RIAA reduces noise. It allows bass signals to "fit" onto the record
> > to give use 20 minutes a side from LP.
>
> No it does not. As I said, in amplitude terms, what is recorded on disc
> is nearly constant amplitude versus frequency. A magnetic cart has a
> rising output with frequency (just like a take head) because its output
> is proportional to rate of change of flux. The top cut is necessary to
> compensate for this.
Well, I'll have to think more about that but it won't alter how I make
amplifiers much.
Patrick Turner.
John Byrns
Ian Bell <ruffr...@yahoo.com> wrote:
> Patrick Turner wrote:
> >
> > But RIAA reduces noise. It allows bass signals to "fit" onto the record
> > to give use 20 minutes a side from LP.
>
> No it does not. As I said, in amplitude terms, what is recorded on disc
> is nearly constant amplitude versus frequency.
That depends on how you define "nearly constant amplitude versus
frequency"? During the recording process the RIAA curve attenuates the
frequencies above approximately 2 kHz by more than 12 dB, not my idea of
"nearly constant".
--
Regards,
John Byrns
Surf my web pages at, http://fmamradios.com/
Ian Bell
Depends. If you simply believe the 2.5/gm mantra without question then I
guess you probably won't.
>>> If you just connect various values at the input of an amp you can listen
>>> to it.
>>>
>>> And when they say the input of a triode consists of an equivalent input
>>> noise resistance between grid and cathode, then that's the model, and I
>>> don't think there is any special consideration with RF. As bandwidth
>>> extends up past the audio band, so does the noise and the noise voltage
>>> at a rate of the square root of increase in BW. So thus having a bw =
>>> 2Mhz instead of only 20kHz, means noise should become 10 times greater.
>>> Going to 20MHz would raise noise by 3.16 more times. Then if you have a
>>> tuned circuit at say 1MHz, and BW = 18kHz as for a broadcast
>>> AM radio station, then the reduction of BW is 18,000 / 1,000,000, or
>>> about 1/50, so expect noise to be about 1/7 of when you have bw = 1Mhz.
>>>
>> Agreed. I would not argue against Johnson noise, I am just saying I am
>> not convinced by the 2.5/gm formula.
>
> I'm not much convinced either. But when Rin is lower, there is less
> noise. Even with bjts, paralleling a lot of them with very low base
> current leads to more silent bjts. Less R.
I can understand paralleling because the signal is coherent so it
increases by 6dB when you parallel two devices but the noise, being
separate uncorrelated noise sources, add their powers so the noise
increses only by 3dB. Result S/N improves by 3dB. It is exactly the same
with track width in magnetic rtape recorders.
>>>>> All the others look difficult to work out, but none have reactive
>>>>> components in the formula.
>>>>>
>>>>> If you use a 6BE6 as a mixer for two inputs to make one output then
>>>>> it'll be noiser than if you have two triodes with a common anode load.
>>>>>
>>>>> All that's important is that triodes are quieter than all other types of
>>>>> tubes and that the higher the gm the better, and then all one needs to
>>>>> consider is to have Ia high, and Ea low right at the input of a phono
>>>>> stage, and hope it works out.
>>>>>
>>>> I agree it is important that triodes are quieter than all other types.
>>>> What is also important is how quiet they ahould be expected to be. If
>>>> you don't know this how can you possibly select tubes for low noise?
>>> You don't have to know all the theory; just general concepts a bit. Then
>>> you find out what works by experimentation. Its the hands on work that
>>> keeps the studio recordists happy with their gear.
>> That's fine as long as the general concepts are accurate. So far I don't
>> fond 2.5/gm to be any where near reality.
>
> So what is a better solution?
>
Now that is a really good question. Someone needs to do some work first
to verify that there is some relation between gm and EIN and secondly to
quantify it. Actually that's rubbish. All we need to know is what are
the achievable EIN values for popular audio tubes. So some experiments
are in order. I was planning to make the 40dB amp for measuring 6CG7
noise for my mic pres anyway so I guess I'll be the one to make a start.
Agreed.
I'll check him out.
>
>>> Say you just recorded with a flat signal. You'd not be able to fit bass
>>> grooves on the record very well. High F treble signals would be
>>> seriously affected by competing levels of noise.
>>>
>> Vinyl IS generally recorded with a near flat amplitude. Frequency
>> dependent boost on record is need because of the head characteristics
>> not to combat noise. Head cutters can cut bass into a record with no
>> problem. Its the playback carts that have the problem staying in the
>> groove because their compliance is frequency dependent.
>
> I've always thought there was considerable bass cut and treble boost in
> recording.
>
This is a common misconception.
> But I have not recorded anything or cut a record. All I know is that
> records sound fine with RIAA, and lousy without it.
That is true, but only for a magnetic cartridge of course.
>>> On playback, noise noise would be intolerable.
>>>>> So the RIAA filter flattens the music response and disposes of most of
>>>>> the noise.
>>>> I am still not clear on this. I need to check what Jones says.
>>> The disc recording process is a bit of a compromise.
>>>> The ears of most people don't notice LF noise as much as
>>>>> noise above say 300Hz. Now because tubes make a fair amount of LF noise,
>>>>> the theoretical -26dB noise reduction doesn't get achieved.
>>>>> There will still be some noise. But on good vinyl its often tape hiss or
>>>>> venue noise which is the highest noise of the recording. In FM radio HF
>>>>> is boosted or "emphasised" and after detection the signal is
>>>>> de-emphasized. Exit much noise.
>>>> That's different and is done deliberately to reduce noise. RIAA is not
>>>> for that reason.
>>> But RIAA reduces noise. It allows bass signals to "fit" onto the record
>>> to give use 20 minutes a side from LP.
>> No it does not. As I said, in amplitude terms, what is recorded on disc
>> is nearly constant amplitude versus frequency. A magnetic cart has a
>> rising output with frequency (just like a take head) because its output
>> is proportional to rate of change of flux. The top cut is necessary to
>> compensate for this.
>
> Well, I'll have to think more about that but it won't alter how I make
> amplifiers much.
>
No, please don't. From what I have seen your amps are pretty good and
RIAA is the right thing to use for magnetic cartridges.
Cheers
Ian
John Byrns
Patrick Turner <in...@turneraudio.com.au> wrote:
> Allen Wright says a lot about how treble is boosted and bass is cut
> acording to the RIAA recording curve which is the opposite to the
> playback curve. He says its important to have just the right RIAA
> playback filter to allow for the fact the treble isn't boosted
> infinitely, but the only up to maybe 50kHz, or else they have
> instability troubles with the cutting head amp and its loops of negative
> feedback to ensure distortion is minimized.
Sounds like he is talking about the ultrasonic time constant used in the
recording equalization network for velocity type cutter heads. We have
talked about this time constant here before, I think it is called the
4th time constant, or is it the 5th?
I don't know about instability issues, but it is obviously impractical
to keep the treble boost going up to indefinitely high frequencies, how
high would we go, 100 MHz? How much extra gain would that require? Out
of practical necessity we must include an ultra sonic pole in the
equalizer for a velocity cutter head.
This same problem exists with the preemphasis/deemphasis used in FM
broadcasting, only nobody worries about it because people are not as
picky about their FM reception as they are about vinyl playback.
Ian Bell
> In article <gp5gil$p22$1...@localhost.localdomain>,
> Ian Bell <ruffr...@yahoo.com> wrote:
>
>> Patrick Turner wrote:
>>> But RIAA reduces noise. It allows bass signals to "fit" onto the record
>>> to give use 20 minutes a side from LP.
>> No it does not. As I said, in amplitude terms, what is recorded on disc
>> is nearly constant amplitude versus frequency.
>
> That depends on how you define "nearly constant amplitude versus
> frequency"? During the recording process the RIAA curve attenuates the
> frequencies above approximately 2 kHz by more than 12 dB, not my idea of
> "nearly constant".
>
Mine neither. You are saying that above 2KHz the amplitude of signals on
the disc is attenuated by 12dB or more. That makes no sense to me. Do
you have a reference for this?
Cheers
Ian
Patrick Turner
Ian Bell wrote:
>
> Patrick Turner wrote:
> >
> > Ian Bell wrote:
> >> Patrick Turner wrote:
> >>> Ian Bell wrote:
> >>>> Patrick Turner wrote:
> >>>>> Ian Bell wrote:
> >>>>>> Patrick Turner wrote:
> >>>>>>> Ian Bell wrote:
> >>>>>>>> Patrick Turner wrote:
> >>>>>>>>> Ian Bell wrote:
> >>>>>>>>>> Patrick Turner wrote:
> >>>>>>>>>>> Ian Bell wrote:
> >>>>>>>>>>>> Patrick Turner wrote:
> >>>>> snip
snip,
> > Somebody with far more insight than I have decided an imaginery
> > resistance between grid and cathode modelled tube noise OK.
> >
> > I'm wondering that if I look further at this noise issue, will I make
> > more silent amps?
>
> Depends. If you simply believe the 2.5/gm mantra without question then I
> guess you probably won't.
I am a questioning man. I am a PITA for the extremist theoretician,
because I ask "And what do we get in practice?"
So, OK, to me I had a choice of persuing low noise with careful
selection of high gm triodes all paralleled up and then I thought I
could use one j-fet costing $1.0 instead. That iddy biddy j-fet, the
2SK369 will produce 1% THD per volt of output with most loads so that
with 5V out you'd get 5% THD. Its a quite terrible result compared to a
we set up 6CG7. BUT, the THD declines towards 0.0% at 0.0Vo, so that if
the Vo was 0.01V, THD should have declined to 0.01%, which is quite low.
And so I use a j-fet at my phono amp inputs. With MM, the j-fet has a
lot of local current FB from its larger Rs to keep its gain low and the
worst case is with the possible 50mV input at 22kHz. If the fet gain is
10, you have 0.5Vo, and the first harmonic is at 44khz and inaudible.
But in fact high levels of treble in music don't much occur.
With MC, input at 10Hz is perhaps 0.03mV, at 1khz, perhaps 0.3mV, and at
HF perhaps 3.0mV. So THD is low even with less local current FB and more
fet gain. When I test a phono preamp I apply a level sig gene signal
through a recording RIAA filter.
A fet can produce much higher output than required and still have low
THD.
If one wants less THD then one can use a balanced input with two j-fets
because most j-fets produce similar distortion dominated by 2H like a
triode.
snip,
> > So what is a better solution?
> >
>
> Now that is a really good question. Someone needs to do some work first
> to verify that there is some relation between gm and EIN and secondly to
> quantify it. Actually that's rubbish. All we need to know is what are
> the achievable EIN values for popular audio tubes. So some experiments
> are in order. I was planning to make the 40dB amp for measuring 6CG7
> noise for my mic pres anyway so I guess I'll be the one to make a start.
Getting less than 1uV noise at the input of any tube is problematical,
especially with 6CG7, 12AT7, 6DJ8, 12AX7, 12AY7, 12AU7 and many other
pentodes in triode mode.
But feel welcome to try things.
Maybe consider some higher gm tubes like 6C45pi or 417 or E80F/6BX6,6EH7
in triode and at various Ia and see what happens.
snip
> > Allen Wright says a lot about how treble is boosted and bass is cut
> > acording to the RIAA recording curve which is the opposite to the
> > playback curve. He says its important to have just the right RIAA
> > playback filter to allow for the fact the treble isn't boosted
> > infinitely, but the only up to maybe 50kHz, or else they have
> > instability troubles with the cutting head amp and its loops of negative
> > feedback to ensure distortion is minimized.
> >
>
> I'll check him out.
He was i think the first to use a high gm j-fet to drive a triode in
cascode.
Its very simple and it works.
snip,
> >
> > Well, I'll have to think more about that but it won't alter how I make
> > amplifiers much.
> >
>
> No, please don't. From what I have seen your amps are pretty good and
> RIAA is the right thing to use for magnetic cartridges.
I can't be too wrong then?
Patrick Turner.
Patrick Turner
John Byrns wrote:
>
> In article <49B66315...@turneraudio.com.au>,
> Patrick Turner <in...@turneraudio.com.au> wrote:
>
> > Allen Wright says a lot about how treble is boosted and bass is cut
> > acording to the RIAA recording curve which is the opposite to the
> > playback curve. He says its important to have just the right RIAA
> > playback filter to allow for the fact the treble isn't boosted
> > infinitely, but the only up to maybe 50kHz, or else they have
> > instability troubles with the cutting head amp and its loops of negative
> > feedback to ensure distortion is minimized.
>
> Sounds like he is talking about the ultrasonic time constant used in the
> recording equalization network for velocity type cutter heads. We have
> talked about this time constant here before, I think it is called the
> 4th time constant, or is it the 5th?
some records have low bass removed, 3,180uS, then you have standard
318us, then 75us, then ultrasonic at say 7.5uS, which isn't very
critical.
>
> I don't know about instability issues, but it is obviously impractical
> to keep the treble boost going up to indefinitely high frequencies, how
> high would we go, 100 MHz? How much extra gain would that require? Out
> of practical necessity we must include an ultra sonic pole in the
> equalizer for a velocity cutter head.
The HF cutting head amp boost is done by using NFB. Boost cannot be
infinite, and there is a level plateau in response above 50kHz.
So, in the playback amp we level our correction to suit.
Allen Wright talks about it in his Tube Amp Cookbook, and give anecdotal
evidence that such fine tuning of RIAA preamp networks makes the vinyl
experience optimal.
When vinyl was king for so many, there were agonising groans from the
hi-fi cognescenti if a manufacturer's RIAA R&C filter was even +/-
0.25dB inaccurate.
But some hi-fi sets I saw made in the 1960s just had one R&C filter with
a -3dB pole at 50Hz and that was that, and so you got slightly more bass
relative to treble, and it compensated for poor bass response elsewhere
in the amps and everyone thought the rock and roll sounded marvellous.
And jazz. And even Motzart. My mum used to use a Kriesler system like
this. I still have it, much modified though.
There was **so much** processing and equalisation done in studios that
what RIAA does is only ever "about right" according to what the studio
guy thought sounded right to get sales.
>
> This same problem exists with the preemphasis/deemphasis used in FM
> broadcasting, only nobody worries about it because people are not as
> picky about their FM reception as they are about vinyl playback.
Some ppl are very picky. But fact is without the treble
emphasis/de-emphasis, noise in FM would be worse.
And with a boosted HF, the waves are easier to decode when F gets higher
when trying to synthesize the L and R channels which are derived from a
sub carrier with a rather low 38kHz frequency. FM would have been better
if the sub carrier was at 100kHz and the pilot tone was at say 30kHz,
but the boffins on 1955 forsaw multiplexing and the possibilities of
other sub-carriers F at 67kHz and 96khz beckoned.
So the 1961 Zenith-GE agreed FM system was lowest common denominator.
But it sounds a lot better than AM radio. Live FM broadcasting when it
happens is still very very good to listen to. All the gear is in a
little suitacse type of box and the signal goes to satelites and its all
far more complex than live broadcasts in 1963, but it still sounds good.
My mum used to listen to live concerts on AM from Sydney Town Hall in
the 1960s, but the damn Kreisler gram-receiver was a bunch of crap with
AM audio BW of only 3kHz. We still enjoyed Motzart drifting around the
house.
Nobody I knew in 1960 ever had an AM receiver capable of a flat 10kHz if
it was being transmitted on AM. And it was transmitted then if it didn't
interfere with other stations, so AM could be almost as good as FM, if
you had the gear, but nobody did. Nobody had tweeters capable of a flat
response to 20kHz either.
When FM came to Oz in around 1972, the local electronics industry was in
its death throes as tariffs on cheap asian mades were being reduced.
FM sets were horribly complex like TV sets and so the only economic way
of providing receievrs for FM sets was to allow the cheap new chip
filled radio sets to flood the country. Integrated circuit chips
revolutionised the information and enetertainment industry. AM took a
real back seat and its bandwidth in most SS sets was a paltry 1.5kHz if
you were lucky and with poor selectivity. So SS gear really killed much
AM listening. But we still have AM stations and with a good set they
sound well, and there is any official emphasis except that some stations
boost bass and boost treble and then compress and de-ess and use MP3
music files and it is then OK on the poor cheap hi-fi sets and ghetto
blasters the lower orders use.
Only a few FM stations and AM station are worth listening to.
Patrick Turner.
Patrick Turner
The 75uS time constant in an RIAA network give a preamp response which
means the HF above 1kHz is down -3dB at 2,112Hz. above this F the rate
of attenuation becomes close to -6dB per octave so that if a 22kHz
signal enters the preamp it is about 20dB below the 1khz reference level
at the preamp output.
To get a flat response with vinyl and a magnetic cart, the signal from
the record contains a boosted HF so that a 22khz signal is 20dB above
the 1 khz reference level.
The +/-12dB points in the process are around the 10khz point, see the
official RIAA attenuation charts.
Patrick Turner.
>
> Cheers
>
> Ian
Ian Bell
That's what I say too, and my answer so far is 'way more than 2.5/gm'.
Cheers
Ian
Ian Bell
Nope. The reason the HF signal from a mag cartridge is boosted is not
because it is boosted on the record (it isn't) but because (as I have
said several times now) of the magnetic cartridge's rising output with
frequency (for a constant amplitude input) due to its output voltage
being proportional to rate of change of flux i.e. in this case frequency.
Cheers
Ian
John Byrns
Ian Bell <ruffr...@yahoo.com> wrote:
Hi Ian,
I don't have a reference handy, although this issue was discussed in
another group a few years back and someone actually found a reference on
the web.
Let me explain, as you said "in amplitude terms, what is recorded on
disc is nearly constant amplitude versus frequency." The operative bit
here is that your stated reference is "constant amplitude", not the more
commonly used constant velocity reference. To change the reference from
constant velocity to constant amplitude we must apply a correction
factor to the "RIAA" recording curve that rolls off across the entire
audio band at a rate of 6 dB per octave with increasing frequency. When
the conversion is made from a constant velocity to a constant amplitude
reference we are left with a recording curve with a pole at 500 Hz and a
zero at 2,122 Hz, ignoring the 50 Hz and ultra sonic time constants.
What this says is that on an amplitude basis the response starts rolling
off at 500 Hz at a rate of 6 dB per octave, and then flattens out again
at 2,122 Hz. Between 500 and 2,122 Hz the attenuation gradually
increases to approximately 12 dB above 2,122 Hz. In other words, the
response recorded on the LP is shelved down approx. 12 dB above 2,122 Hz
relative to the response below 500 Hz, on the basis of recorded groove
amplitude basis.
John Byrns
Patrick Turner <in...@turneraudio.com.au> wrote:
> Some ppl are very picky. But fact is without the treble
> emphasis/de-emphasis, noise in FM would be worse.
>
> And with a boosted HF, the waves are easier to decode when F gets higher
> when trying to synthesize the L and R channels which are derived from a
> sub carrier with a rather low 38kHz frequency. FM would have been better
> if the sub carrier was at 100kHz and the pilot tone was at say 30kHz,
> but the boffins on 1955 forsaw multiplexing and the possibilities of
> other sub-carriers F at 67kHz and 96khz beckoned.
Hi Patrick,
Another fact is that if a 100 kHz sub carrier had been used for FM
stereo as you suggest, the noise would have been much worse than it is
with the 38 kHz sub carrier. FM stereo is already something on the
order of approximately 22 dB noisier than monophonic FM. Using a 100
kHz stereo sub carrier frequency would make the FM stereo noise level 8
or 9 dB worse than it is with the 38 kHz sub carrier due to the effect
of FM's triangular noise spectrum. As a result you want the stereo sub
carrier to be as low in frequency as possible, 100 kHz is a bad choice
for this reason. FM's triangular noise spectrum is the tradeoff old
E.H. Armstrong made to achieve wideband FM's noise reduction in the
audio range below 15 kHz.
Patrick Turner
I thought the signal sent to the cutting head amp was passed through a
recording RIAA filter which cuts bass and boosts treble.
Thus the signal from the record has low bass and high treble and the
preamp applies a filter doing the opposite of what a recording filter
does and the music response is thus flat at the preamp output.
See RDH4, page 767 (v), Equalisation of the cutter.
They say equalisation is applied to the cutter to obtain the desired
recording characteristic.
There were 12 different characteristics before RIAA became the standard
for LP, but some were already close to RIAA.
If you do a Google on 'RIAA equalization recording' there are lots of
hits.
eg, see http://www.phonopreamps.com/faq.html
"""What is RIAA equalization?
Because of limitations in the LP recording process, an equalization
curve must be applied to the music or other sonic content prior to it
being cut onto vinyl, so as to reduce backround noise and sibilance.
Removing this equalization affect (called the RIAA curve) and restoring
the music's original frequency response curve during playback is an
important part of the phono preamp's job and differentiates it from
other preamps used for microphones and musical instruments, which
provide gain but no other modification of the original sound quality.
Proper RIAA re-equalization during playback is a must in faithfully
producing the original musical content without coloration or distortion.
(top)""""
Patrick Turner.
Patrick Turner
John Byrns wrote:
>
> In article <49B78F2B...@turneraudio.com.au>,
> Patrick Turner <in...@turneraudio.com.au> wrote:
>
> > Some ppl are very picky. But fact is without the treble
> > emphasis/de-emphasis, noise in FM would be worse.
> >
> > And with a boosted HF, the waves are easier to decode when F gets higher
> > when trying to synthesize the L and R channels which are derived from a
> > sub carrier with a rather low 38kHz frequency. FM would have been better
> > if the sub carrier was at 100kHz and the pilot tone was at say 30kHz,
> > but the boffins on 1955 forsaw multiplexing and the possibilities of
> > other sub-carriers F at 67kHz and 96khz beckoned.
>
> Hi Patrick,
>
> Another fact is that if a 100 kHz sub carrier had been used for FM
> stereo as you suggest, the noise would have been much worse than it is
> with the 38 kHz sub carrier.
Well the increase in bandwidth needed to have the subC at 100kHz would
increase noise.
> FM stereo is already something on the
> order of approximately 22 dB noisier than monophonic FM.
Hmm, I find that difficult to believe going by what i hear from better
FM tuners.
> Using a 100
> kHz stereo sub carrier frequency would make the FM stereo noise level 8
> or 9 dB worse than it is with the 38 kHz sub carrier due to the effect
> of FM's triangular noise spectrum. As a result you want the stereo sub
> carrier to be as low in frequency as possible, 100 kHz is a bad choice
> for this reason.
I was guessing, but 50kHz would give less noise.
> FM's triangular noise spectrum is the trade off old
> E.H. Armstrong made to achieve wideband FM's noise reduction in the
> audio range below 15 kHz.
Did EHA consider stereo?
Patrick Turner.
Ian Bell
Got you. It was clear to me there was a kink in the amplitude response
due to the separation of the two mid time constants between which the
response would fall at 6dB/octave. I had not realised they were two
octaves apart. So to summarise, the recorded amplitude response is flat
from 50Hz to 500Hz, then rolls off at 6dB per octave to 2.12KHz where it
is about 12dB down and thereafter is flat out to 20KHz.
Cheers
Ian
Ian Bell
It is, but because the recording head is also velocity operated, if you
had no equalisation, the amplitude recorded on the disc would fall
continuously at 6dB/octave. For constant amplitude on disk what you
really need is a rising EQ at 6dB/octave. Cutting the bass effectively
makes the frequencies below the bass turnover frequency constant
amplitude. Boosting the treble effectively makes the frequencies above
the treble turnover frequency constant amplitude.
> Thus the signal from the record has low bass and high treble and the
> preamp applies a filter doing the opposite of what a recording filter
> does and the music response is thus flat at the preamp output.
>
If you think in terms of recorded velocity then what you say is 100%
correct. I find that hard to do and prefer to think in terms of amplitude.
> See RDH4, page 767 (v), Equalisation of the cutter.
> They say equalisation is applied to the cutter to obtain the desired
> recording characteristic.
I am not disputing that EQ is needed.
> There were 12 different characteristics before RIAA became the standard
> for LP, but some were already close to RIAA.
>
> If you do a Google on 'RIAA equalization recording' there are lots of
> hits.
>
> eg, see http://www.phonopreamps.com/faq.html
>
> """What is RIAA equalization?
>
> Because of limitations in the LP recording process, an equalization
> curve must be applied to the music or other sonic content prior to it
> being cut onto vinyl, so as to reduce backround noise and sibilance.
Yes, I have seen many such references. Unfortunately most are just plain
wrong or at the very least blithely ignore the difference between
velocity and amplitude.
> Removing this equalization affect (called the RIAA curve) and restoring
> the music's original frequency response curve during playback is an
> important part of the phono preamp's job and differentiates it from
> other preamps used for microphones and musical instruments, which
> provide gain but no other modification of the original sound quality.
> Proper RIAA re-equalization during playback is a must in faithfully
> producing the original musical content without coloration or distortion.
> (top)""""
>
No argument there except it only applies when you use a velocity
sensitive transducer like a magnetic cartridge.
Cheers
ian
> Patrick Turner.
John Byrns
Ian Bell
It has always puzzled me why they did not attempt to get close to
constant amplitude from the start. It's the velocity operated head
cutters that gave them the excessive bass excursions that caused them so
much problem in the very early days. They were bright enough to know
what the head response was so why not correct for it? I guess there has
to be some practical factor that makes this harder than you would
expect. I was going to say it is no harder than driving a mgnetic tape
record head but I suspect the powers involved in a disc cutting head are
a lot higher.
Cheers
ian
Patrick Turner
Without being able to view relevant diagrams and response curves of real
world hardware, I am precisely confused why John is so precisely in
agreement.
Patrick Turner.
Ian Bell
OK, Here goes.
The (amplitude) response of the cutter head falls by 6dB/octave. The
inverse RIAA curve, which is used on record starts, with a rising curve
from 50Hz to 500Hz at 6dB/octave. The combination of these two means a
constant amplitude gets recorded on the disc between these two frequencies.
From 500Hz to 2.12Khz the RIAA curve is flat, so because of the head
response, the amplitude recorded on the disc falls by 6dB/octave for
about 2 octaves.
From 2KHz upwards the RIAA record curve again slopes upward at
6dB/octave which, when combined with the head response of -6dB/octave
means the recorded amplitude is constant versus frequency.
On playback the mag cartridge has a response which rise at 6dB/octave.
The RIAA playback curve falls at 6dB/octave from 50Hz to 500Hz. So the
constant amplitude on disc plus the rising characteristic of the mag
cart means the pre-amp input signal rises at 6dB/octave between these
frequencies and is compensated for by the falling characteristic of the
RIAA replay curve.
Between 500Hz and 2.12Khz the signal from the disc falls at 6dB/octave
which combined with the 6dB/octave rising characteristic of the mag cart
means the signal at the pre-amp input is constant between these two
frequencies and the flat part of the RIAA replay curve between these
frequencies leaves the input signal unchanged.
From 2.12KHz upwards the constant amplitude from the disc combined with
the 6dB/octave rising characteristic of the cart means the signal at the
pre-amp input rises at 6dB/octave which is compensated for by the
6dB/octave falling characteristic of the RIAA replay curve.
Hope that helps.
Cheers
Ian
John Byrns
Patrick Turner <in...@turneraudio.com.au> wrote:
Because I am talking about the RIAA "standard", not "real world
hardware". "Real world hardware" then attempts to meet the requirements
of the standard as closely as possible within the bounds of the relevant
economic and technical limitations.
John Byrns
Ian Bell <ruffr...@yahoo.com> wrote:
> It has always puzzled me why they did not attempt to get close to
> constant amplitude from the start.
It's not obvious to me why they would want to do that? Presumably there
are a lot of factors that go into choosing the best equalization curve,
including among others the spectral distribution of the energy in the
source material, and the capabilities of the recording medium vs.
frequency. Perhaps the high frequency amplitude shelving during
recording is because of limitations on the velocity that can be recorded
at high frequencies.
> It's the velocity operated head
> cutters that gave them the excessive bass excursions that caused them so
> much problem in the very early days. They were bright enough to know
> what the head response was so why not correct for it? I guess there has
> to be some practical factor that makes this harder than you would
> expect.
IIRC early electrical recordings were constant amplitude below the
turnover frequency, which I think was somewhere in the 250 Hz to 400 Hz
range, and constant velocity above the turnover frequency. This
equalization was mechanically implemented by the primary resonance of
the cutter head at the turnover frequency, so that the cutter was mass
controlled above the turnover frequency, and compliance controlled below
the turnover frequency. At least that is an approximation of the
explanation I have heard. I'm not sure about the earlier acoustic
recorders, but I think the equalization was similar. Eventually they
found that they could add some electrical boost to the constant velocity
high frequency region to allow reducing the high frequency noise during
playback with a corresponding high frequency rolloff.
> I was going to say it is no harder than driving a mgnetic tape
> record head but I suspect the powers involved in a disc cutting head are
> a lot higher.
Yes, the power involved is more like the power required to drive a
loudspeaker than it is like the power needed to drive a magnetic tape
recording head.
Patrick Turner
OK, so as F rises, there is less amplitude in the grooves, no?.
So the bass grooves are the biggest, so if you cut bass and boost treble
somewhere near 6dB/octave, then the groves of any F will all become
constant amplitude.
> The
> inverse RIAA curve, which is used on record starts, with a rising curve
> from 50Hz to 500Hz at 6dB/octave. The combination of these two means a
> constant amplitude gets recorded on the disc between these two frequencies.
> From 500Hz to 2.12Khz the RIAA curve is flat, so because of the head
> response, the amplitude recorded on the disc falls by 6dB/octave for
> about 2 octaves.
But the RIAA has about 5dB difference in eq between 500 and 2,112Hz.....
There is NO flat part on the RIAA recording or playback curves between
say 20Hz and 22kHz.
>
> From 2KHz upwards the RIAA record curve again slopes upward at
> 6dB/octave which, when combined with the head response of -6dB/octave
> means the recorded amplitude is constant versus frequency.
So the slight change inthe rate of RIAA attenuation/boost between about
500Hz and 2,112Hz means that there is a change in amplitude between F
ranges below 500 and above 2,112?
> On playback the mag cartridge has a response which rise at 6dB/octave.
>
> The RIAA playback curve falls at 6dB/octave from 50Hz to 500Hz. So the
> constant amplitude on disc plus the rising characteristic of the mag
> cart means the pre-amp input signal rises at 6dB/octave between these
> frequencies and is compensated for by the falling characteristic of the
> RIAA replay curve.
>
> Between 500Hz and 2.12Khz the signal from the disc falls at 6dB/octave
> which combined with the 6dB/octave rising characteristic of the mag cart
> means the signal at the pre-amp input is constant between these two
> frequencies and the flat part of the RIAA replay curve between these
> frequencies leaves the input signal unchanged.
But preamp input isn't constant between the two F.....
>
> From 2.12KHz upwards the constant amplitude from the disc combined with
> the 6dB/octave rising characteristic of the cart means the signal at the
> pre-amp input rises at 6dB/octave which is compensated for by the
> 6dB/octave falling characteristic of the RIAA replay curve.
>
> Hope that helps.
I hear what you are saying, and basically the faster the cutter has to
move side to side to make a groove then the less it travels.
And the faster the cartridge moves side to side to track gooves, the
higher the amount of voltage output and its all a fairly linear system.
And to avoid having high bass groove swings which would take up too much
room on the record they cut the bass signal to the cutting head and to
avoid treble grooves having too small an amplitude they boost the treble
signal to the cutter. The groove amplitude is thus more or less flat
across the whole range.
But the cart then puts out little bass and a huge amount of treble
because its response is linear with F and so the RIAA playback curve
reduces treble and boost bass in reference to 1kHz. Disc Noise is
reduced at HF above 1kHz at least, and amplifier noise somewhat more
methinks.
Patrick Turner.
Ian Bell
Yup.
> So the bass grooves are the biggest, so if you cut bass and boost treble
> somewhere near 6dB/octave, then the groves of any F will all become
> constant amplitude.
>
Yup
>> The
>> inverse RIAA curve, which is used on record starts, with a rising curve
>> from 50Hz to 500Hz at 6dB/octave. The combination of these two means a
>> constant amplitude gets recorded on the disc between these two frequencies.
>> From 500Hz to 2.12Khz the RIAA curve is flat, so because of the head
>> response, the amplitude recorded on the disc falls by 6dB/octave for
>> about 2 octaves.
>
> But the RIAA has about 5dB difference in eq between 500 and 2,112Hz.....
>
Yes, I was using the Bode straight line approximations - in practice the
real things are curves. In practice 500Hz and 2.12KHz are the 3dB point
so if you add 2*3dB to the 5dB you see, you get 11dB. However, there is
still 12dB amplitude difference on disc between frequencies removed from
the 3dB points.
> There is NO flat part on the RIAA recording or playback curves between
> say 20Hz and 22kHz.
In practice no but again I was using Bode approximations - see above.
>> From 2KHz upwards the RIAA record curve again slopes upward at
>> 6dB/octave which, when combined with the head response of -6dB/octave
>> means the recorded amplitude is constant versus frequency.
>
> So the slight change inthe rate of RIAA attenuation/boost between about
> 500Hz and 2,112Hz means that there is a change in amplitude between F
> ranges below 500 and above 2,112?
Yes
>
>
>> On playback the mag cartridge has a response which rise at 6dB/octave.
>>
>> The RIAA playback curve falls at 6dB/octave from 50Hz to 500Hz. So the
>> constant amplitude on disc plus the rising characteristic of the mag
>> cart means the pre-amp input signal rises at 6dB/octave between these
>> frequencies and is compensated for by the falling characteristic of the
>> RIAA replay curve.
>>
>> Between 500Hz and 2.12Khz the signal from the disc falls at 6dB/octave
>> which combined with the 6dB/octave rising characteristic of the mag cart
>> means the signal at the pre-amp input is constant between these two
>> frequencies and the flat part of the RIAA replay curve between these
>> frequencies leaves the input signal unchanged.
>
> But preamp input isn't constant between the two F.....
No, but see above.
>> From 2.12KHz upwards the constant amplitude from the disc combined with
>> the 6dB/octave rising characteristic of the cart means the signal at the
>> pre-amp input rises at 6dB/octave which is compensated for by the
>> 6dB/octave falling characteristic of the RIAA replay curve.
>>
>> Hope that helps.
>
> I hear what you are saying, and basically the faster the cutter has to
> move side to side to make a groove then the less it travels.
>
> And the faster the cartridge moves side to side to track gooves, the
> higher the amount of voltage output and its all a fairly linear system.
>
> And to avoid having high bass groove swings which would take up too much
> room on the record they cut the bass signal to the cutting head and to
> avoid treble grooves having too small an amplitude they boost the treble
> signal to the cutter. The groove amplitude is thus more or less flat
> across the whole range.
That's is basically it.
> But the cart then puts out little bass and a huge amount of treble
> because its response is linear with F and so the RIAA playback curve
> reduces treble and boost bass in reference to 1kHz.
That's right
Disc Noise is
> reduced at HF above 1kHz at least, and amplifier noise somewhat more
> methinks.
>
Amplifier noise certainly, disc noise I am not so sure about.
Cheers
Ian
Patrick Turner
Definately disc noise above 1 kHz is much reduced by the boosting of HF.
Just think if the amplitude of HF were to dwindle down with no RIAA
recording HF emphasis. The HF disc noise would still be high and signal
low and snr lousy. Someone may say that the higher HF amplitude means
more noise is generated but afaik, most experts say the disc noise is
much less when its done the way we know.
The luck with vinyl was that it was so much quieter than shellac
records.
And many people still take vinyl so seriously.
Patrick Turner.
Ian Bell
Agreed, but the boosting only compensates for the decline in HF due to
the cutter head to produce a flat amplitude response. In your FM radio
example, the response starts of flat and the 75uS emphasis boosts the HF
even more above the noise so at the receiver the de-emphasis improves
the S/N ration. It is well known that in music the amplitude of the
spectrum of the higher frequencies is lower than that of the lower
frequencies so you can boost the HF without danger of causing overloads.
I see no reason in principle why this could not have been applied to
vinyl with a consequent improvement in S/N.
Cheers
Ian