tube noise

[jhaible]

What is the typical noise of a good low-noise tube?

Ok, don't kill me for over-generalising. :wink:

I know it depends an a lot of different factors, and the circuit in which it's used.

What I'm looking for is the input-referenced noise (the nanovolts per square root Hz voltage source that would be connected to the grid on a noise-free triode, in common cathode mode with bypassed Rk, at the tube's optimal (in terms of noise) anode current and anode voltage.

Something like:

xxx nV/SQRT(Hz) for 12AX7 @ xxx mA @ xxx V

I'm looking for typical values, just to get an impression, of some common tubes, and also for recomendations of special low noise tubes.

(Background: Phono Preamp. Does it make sense to use tubes - without step-up transformer - for a cartridge with rather low output, like 1.5mV. )

Thanks in advance for any input. Feel free to just post a pointer, if it has been beaten to death before, somewhere else.

JH.

 

[gyraf]

In general, higher Gm > lower noise at relatively low source impedance.

For phono preamp, I'd use multiple parallel'ed e.g. ECC88/6DJ8.

Jakob E.

 

[jhaible]

[quote author="gyraf"]In general, higher Gm > lower noise at relatively low source impedance.

For phono preamp, I'd use multiple parallel'ed e.g. ECC88/6DJ8.

Jakob E.[/quote]

Hi Jakob,

thanks for the hints - I'll take a look at the ECC88.

But to clarify my question a little: I'm really looking for (rough) absolute noise level values. The "relative" stuff (improovement by paralleling, higher gm normally resulting in lower noise) is easier to find, i.e. I've learned about it in the forum.

JH.

 

[clintrubber]

[quote author="jhaible"]
thanks for the hints - I'll take a look at the ECC88.[/quote]

Hi,

Just a suggestion based on the various {a certain project}-info around here:
if you're going to make that phono stage with several ECC88's in // then it might be interesting (read: cheaper) to use the {a certain other European tube}.
Just a different ff-voltage. But most likely you'll already be aware of this alternative.
Seems like ECC88-prices have gone up because of hifi-craze and that they've forgotten the {a certain other tube} ?

// edited to avoid silly prices for the alternative / read a tube-databook or do a search and you'll know //


Regards,

Peter

 

[gyraf]

Indeed! :razz:

But please don't spread the word :green:

Jakob E.

 

[jhaible]

[quote author="clintrubber"][quote author="jhaible"]
thanks for the hints - I'll take a look at the ECC88.[/quote]

Hi,

Just a suggestion based on the various G-Pultec info around here:
if you're going to make that phono stage with several ECC88's in // then it might be interesting (read: cheaper) to use the :green: .
Just a different ff-voltage. But most likely you'll already be aware of this alternative.
Seems like ECC88-prices have gone up because of hifi-craze and that they've forgotten the :green: ?


Regards,

Peter[/quote]

Hi Peter,

thanks for reminding me. I think I've stocked up on :green: s last year, for just that reason (cheaper than ECC), but then forgot about it.

Do you have any idea what kind of input related noise level to expect?
I know good opamps or expensive transistor pairs have 1nV/SQRT(Hz), and I know tubes are a good deal worse - but how much worse?
For a ECC88 (or :green in particular? Or for a ECC83?

I'm asking because I'm _not_ looking for the best tube circuit for a 1.5mV cartridge phone preamp (not yet!) - I want to find out if it makes any sense to go tubes instead of silicon at all.

JH.

- ok guys, following the general paranoia, I've made some edits, too. :grin:

 

[clintrubber]

[quote author="jhaible"][quote author="clintrubber"][quote author="jhaible"]
thanks for the hints - I'll take a look at the ECC88.[/quote]
Hi Peter,

thanks for reminding me. I think I've stocked up on ...

JH.[/quote] [/quote]
Oops, I've edited my previous post but now the {a certain tube}-number is in your post Please remove ! :wink:
But no, I see, you stocked up and now would like to see prices rise :twisted:

Bye,

Peter

 

[clintrubber]

[quote author="jhaible"]
Do you have any idea what kind of input related noise level to expect?
I know good opamps or expensive transistor pairs have 1nV/SQRT(Hz), and I know tubes are a good deal worse - but how much worse?
For a {...} in particular? Or for a ECC83?

JH.[/quote]
I'm sorry, I haven't looked much at tubes in this regard. I thought I saw some equiv. input figures for the '83 somewhere though, probably somewhere here (maybe it was in one of PStamlers posts).

Regards,

Peter

 

[jhaible]

Edited.

We're a crazy crowd here, aren't we?

:twisted:

JH.

 

[jhaible]

I'm growing old. I've asked that question before and got good replies:
http://www.groupdiy.com/index.php?topic=8866&postdays=0&postorder=asc&start=0

I feel stupid. Well, at least I feel forgetful.
That's what you get when you plan too many projects at the same time.


JH.

 

[clintrubber]

[quote author="jhaible"]Edited.
[/quote]
Cool ! :grin:

We're a crazy crowd here, aren't we?

:twisted:

JH.

Sure ! :wink: :thumb:
I guess being a bit crazy is a prerequisite for all this stuff

 

[clintrubber]

[quote author="jhaible"]I'm growing old. I've asked that question before and got good replies:
http://www.groupdiy.com/index.php?topic=8866&postdays=0&postorder=asc&start=0

I feel stupid. Well, at least I feel forgetful.
That's what you get when you plan too many projects at the same time.


JH.[/quote]
Please don't feel obliged to change the {certain tube} to :green: there as well :wink:

 

[jhaible]

Please don't feel obliged to change the {certain tube} to :green: there as well :wink:[/quote]

:shock:

JH.

 

[pstamler]

First off, it's way easier to think of tubes and non-bipolar transistor circuits as having an equivalent noise resistance. With a FET opamp, you get that (roughly) by taking nV/rt-Hz, squaring it, and dividing by 1.656 x 10^-20. With a tube, you find the transconductance (gm) and divide 2.5/gm to get the noise resistance. It is relative, no way around it.

But what's nice about noise resistance is that you can compare it with source resistances. A good rule of thumb: if your circuit's total noise resistance is < 1/4 of the source resistance, you'll add less than 1dB of noise to the source's own thermal noise. That's s good number to shoot for.

Let's say your cartridge has a DC resistance of 1000 ohms. That means you want the preamp's input stage to have a noise resistance of <250 ohms. Sticking that into the upside-down tube noise equation, this is the equivalent of saying that the transconductance has to be > 2.5/250, or 0.01S, or 10mS (millimho for us old farts). Except...

...that the plate resistor also contributes a bit of noise. Let's say the stage has a gain of 31.6x (just to make the arithmetic easier) and a plate resistor of 100k. Divide the plate resistor by the square of the gain (100k / 1000 = 100); that's the contribution of the plate resistor to the total noise resistance.

Which means you need the tube's own noise resistance to be <150 ohms if you want the total to be <250 ohms, and that translates to a transconductance of 16.7mS or more. There aren't a lot of tubes that will do that; none of the 12a*7 tubes comes close. The only commonly-available tubes that will do that are in the 6DJ8 family, so if you're going for a preamp for lowish-output cartridges, I think that's the only place you can really look. Or you can accept a noise figure <1dB. [Edit: Or you can use a smaller plate resistor; that'll help a little, but not a lot.]

My gut feeling is that tubes make sense for higher-output cartridges (3.5mV at 5cm/sec and up) but they're marginal for cartridges with outputs lower than that.

Peace,
Paul

 

[jhaible]

Hi Paul,

thanks for posting this!

No problem to see noise described as resistance - I have printed out some
typical level values for 16kHz bandwidth, just to learn to get a feeling for this description.

But what's nice about noise resistance is that you can compare it with source resistances. A good rule of thumb: if your circuit's total noise resistance is < 1/4 of the source resistance, you'll add less than 1dB of noise to the source's own thermal noise. That's s good number to shoot for.

Fair enough. It shows the level of noise from which it doesn't make sense to go lower (for the amp), because the source resistance would dominate the overall noise. But if the noise from the mechanical tracing the vinyl is much higher than the thermal noise of the pickup coil (is it?), then I might end up paralleling a lot more tubes than necessary, or give up the idea of a tupe preamp early.

Let's say your cartridge has a DC resistance of 1000 ohms. That means you want the preamp's input stage to have a noise resistance of <250 ohms. Sticking that into the upside-down tube noise equation, this is the equivalent of saying that the transconductance has to be > 2.5/250, or 0.01S, or 10mS (millimho for us old farts). Except...

...that the plate resistor also contributes a bit of noise. Let's say the stage has a gain of 31.6x (just to make the arithmetic easier) and a plate resistor of 100k. Divide the plate resistor by the square of the gain (100k / 1000 = 100); that's the contribution of the plate resistor to the total noise resistance.

Which means you need the tube's own noise resistance to be <150 ohms if you want the total to be <250 ohms, and that translates to a transconductance of 16.7mS or more. There aren't a lot of tubes that will do that; none of the 12a*7 tubes comes close. The only commonly-available tubes that will do that are in the 6DJ8 family, so if you're going for a preamp for lowish-output cartridges, I think that's the only place you can really look. Or you can accept a noise figure <1dB.

I just looked it up: My new cartridge has 160 Ohms DC resistance.
(And 1.5mV nominal output voltage)

This means, I'm quite lost, even with 6DJ8 etc.

>My gut feeling is that tubes make sense for higher-output cartridges >(3.5mV at 5cm/sec and up) but they're marginal for cartridges with >outputs lower than that.

Mmm hmm - doesn't look good with my 1.5mV then.

I was wondering: Aren't there even MC amps out there, with tubes and
without input transformer? Talking 300uV for a typical MC cartridge ...
What kind of SNR would these achieve? (I shiver just thinking about it.)

But then again, the eqalisation of the phono preamp does a lot of damping on the high frequencies. So the noise I'll hear after the amp will not just be the V/SQRT(Hz) * SQRT(16kHz), but I'd have to integrate over frequency, with the equalisation being a factor under the integral. This should bring the SNR from abysmal into the range of barely tolerable, maybe. But, as the low frequencies are boosted, I'd asume low frequency noise may play more of a role than in linear amps. So the RIAA courve helps with thermal noise, but boosts low frequent noise components. That's why I'm not so sure looking at DC resistance or the gm equivalent in the tubes is the whole story here.

JH.

 

[pstamler]

It's not the whole story by any means...but it's close enough for government work. At least you can do back-of-the-envelope calculations.

I did a couple this afternoon. It comes down to what you're willing to put up with. The "1dB more than source resistance" criterion is, in fact, pretty stringent -- perhaps more appropriate for microphone preamps. So let's take a couple of less stringent criteria.

For a lot of people, the performance of a Shure V15VxMR into a 5534a opamp is quiet enough. That's a nice benchmark, so let's take it apart. If Rin = 100 ohms (that's the resistor from the - terminal to ground) and Rfb = 3.15k, and the source resistance is 815 ohms loaded by 47k, your total input resistance is about 900 ohms, and the input noise (assuming a 2122 Hz bandwidth) is about 0.24uV, for a signal-to-noise ratio of 82.5dB. (The Shure has a 3.2mV nominal output at 5cm/sec.) Not too shabby!

Okay, what is the equivalent noise resistance which would get you that noise level in a system with no current noise, just voltage and Johnson noise? Answer: about 1650 ohms. The cartridge is about 900 ohms of that, so the additional allowable noise resistance is about 750 ohms. Let's assume that a tube's plate resistor is 100k and the gain is 31.6 (just to make the math easy), so the plate resistor contributes 100k / 31.6^2 or 100 ohms to the noise. Now we can allow 650 ohms from the tube; turning the noise resistor equation inside out, that means its transconductance needs to be 3.85mmho -- excuse me, mS. Several tubes will do that, including the 6DJ8 and its derivatives. So a good tube design will achieve the same S/N as a 5534a on a Shure V15etc. cartridge.

What about your lower-output cartridge? For that, we have to divide the total acceptable noise resistance of the Shure, 1650 ohms, by the square of the output ratio, 1.5 / 3.2. That works out to 363 ohms. The cartridge is 160 ohms, leaving you only 203 ohms to play with. Even ignoring the plate resistor, this is the equivalent of using a tube with transconductance of 12.3mmho (I'm sticking with the old name, dammit). That pretty well dictates a 6DJ8, running pretty hot. And you do have to allow for the plate resistor...

Okay, let's take a looser criterion. I personally find the noise level of a Shure V15etc. operating into a Dynaco PAS preamp to be acceptable; it's certainly way, way below any surface noise. Playing around with the numbers to find total noise resistance, we get 800 ohms from the cartridge and its load resistor, 1k from the unbypassed cathode resistor, 1667 ohms from the tube itself, and a measly 83 ohms from the plate resistor. Add them all up, you get about 3550 ohms -- again, this is my acceptable figure for a Shure cartridge.

Once again, divide by ( 1.5 / 3.2 )^2, and the net result is 780 ohms. Subtract 160 ohms for the cartridge and you're left with 620 ohms. Even taking out 100 ohms for the reflected plate resustor, you have 520 ohms. That's the equivalent of about 4.8mmho, and again there are several tubes that will do this if you give them enough current.

So the answer is, given this slightly relaxed criterion (the Dynaco's S/N with the Shure is about 79dB, which isn't bad at all), yes, you can use tubes to amplify your cartridge and get quite acceptable results. I stand corrected, by my own numbers.

Again, these figures aren't perfectly exact, since they don't take into account the bass boost and they assume that the bandwidth is a sharp cutoff at 2122Hz instead of a 6dB/octave rolloff, but again, they're close enough for government work, or to at least give you the idea that yes, the experiment's worth trying.

Peace,
Paul

 

[bcarso]

[quote author="jhaible"]I was wondering: Aren't there even MC amps out there, with tubes and
without input transformer? Talking 300uV for a typical MC cartridge ...
What kind of SNR would these achieve? (I shiver just thinking about it.)

But then again, the eqalisation of the phono preamp does a lot of damping on the high frequencies. So the noise I'll hear after the amp will not just be the V/SQRT(Hz) * SQRT(16kHz), but I'd have to integrate over frequency, with the equalisation being a factor under the integral. This should bring the SNR from abysmal into the range of barely tolerable, maybe. But, as the low frequencies are boosted, I'd asume low frequency noise may play more of a role than in linear amps. So the RIAA courve helps with thermal noise, but boosts low frequent noise components. That's why I'm not so sure looking at DC resistance or the gm equivalent in the tubes is the whole story here.

JH.[/quote]

There is a good discussion of noise estimation for phono preamps with RIAA EQ in some National Semi documents, taking noise voltage and noise current into account (the latter generally negligible with tubes and FETs).

I don't think anyone has been crazy or excessive enough to attempt a transformerless MC tube preamp. We gassed about this a while back, with PRR pointing out that a massively paralleled arrangement would tolerate the odd tube going south without much effect, for most tube failure modes at least---unlike a tube computer, where losing a flip flop somewhere could bring computations to a grinding halt.

As also discussed somewhere, I think I mentioned that one of the supertriodes, the 6C45, had noise midband around 3nV/Hz, but also suffered from microphonic ringing with extraordinarily high Q (>3000 roughly) at about 9kHz in some samples I tested. And like all tubes the low frequency noise gets foul, just where you are boosting like mad for RIAA. Even with help from Fletcher, Munson, and their descendants you will get in trouble.

I guess one could envision a hybrid design, where the FET or whatever handled the lows and transitioned to tubes for mids and highs.

There is a good discussion of tube noise in Cherry and Hooper, Amplifying Devices and Low-Pass Amplifier Design, but the book is out of print and scarce. If you run across it snatch it if it is anywhere near reasonable---as mentioned the last web copy I saw offered was for 1.9 kilodollars US!

 

[pstamler]

[quote author="bcarso"]I don't think anyone has been crazy or excessive enough to attempt a transformerless MC tube preamp.[/quote]

Audio Research's SP10 tried. It had the reputation of sounding great on moving magnets; don't know if it was really quiet enough for MC's. It used 6DJ8s, I think. Their next generation, the SP11, was a hybrid design.

Peace,
Paul

 

[jhaible]

[quote author="pstamler"] (assuming a 2122 Hz bandwidth)[/quote]

Ah, that's what was missing for me. Something like an equivalent bandwidth for calculating noise level from noise density, taking the massive eqalising into account. The 2122Hz would give 9dB better SNR than on a flat amp with 16kHz bandwidth. Asuming the density is equal, which - as far as I've read here - it isn't.

is about 0.24uV, for a signal-to-noise ratio of 82.5dB. (The Shure has a 3.2mV nominal output at 5cm/sec.) Not too shabby!

Okay, what is the equivalent noise resistance which would get you that noise level in a system with no current noise, just voltage and Johnson noise? Answer: about 1650 ohms. The cartridge is about 900 ohms of that, so the additional allowable noise resistance is about 750 ohms. Let's assume that a tube's plate resistor is 100k and the gain is 31.6 (just to make the math easy), so the plate resistor contributes 100k / 31.6^2 or 100 ohms to the noise. Now we can allow 650 ohms from the tube; turning the noise resistor equation inside out, that means its transconductance needs to be 3.85mmho -- excuse me, mS. Several tubes will do that, including the 6DJ8 and its derivatives. So a good tube design will achieve the same S/N as a 5534a on a Shure V15etc. cartridge.

What about your lower-output cartridge? For that, we have to divide the total acceptable noise resistance of the Shure, 1650 ohms, by the square of the output ratio, 1.5 / 3.2. That works out to 363 ohms. The cartridge is 160 ohms, leaving you only 203 ohms to play with. Even ignoring the plate resistor, this is the equivalent of using a tube with transconductance of 12.3mmho (I'm sticking with the old name, dammit). That pretty well dictates a 6DJ8, running pretty hot. And you do have to allow for the plate resistor...

Okay, let's take a looser criterion. I personally find the noise level of a Shure V15etc. operating into a Dynaco PAS preamp to be acceptable; it's certainly way, way below any surface noise. Playing around with the numbers to find total noise resistance, we get 800 ohms from the cartridge and its load resistor, 1k from the unbypassed cathode resistor, 1667 ohms from the tube itself, and a measly 83 ohms from the plate resistor. Add them all up, you get about 3550 ohms -- again, this is my acceptable figure for a Shure cartridge.

Once again, divide by ( 1.5 / 3.2 )^2, and the net result is 780 ohms. Subtract 160 ohms for the cartridge and you're left with 620 ohms. Even taking out 100 ohms for the reflected plate resustor, you have 520 ohms. That's the equivalent of about 4.8mmho, and again there are several tubes that will do this if you give them enough current.

So the answer is, given this slightly relaxed criterion (the Dynaco's S/N with the Shure is about 79dB, which isn't bad at all), yes, you can use tubes to amplify your cartridge and get quite acceptable results. I stand corrected, by my own numbers.

Again, these figures aren't perfectly exact, since they don't take into account the bass boost and they assume that the bandwidth is a sharp cutoff at 2122Hz instead of a 6dB/octave rolloff, but again, they're close enough for government work, or to at least give you the idea that yes, the experiment's worth trying.

Peace,
Paul

I must admit that the method looking at equivalent resistance values is not that familiar to me yet, but something strikes me as very odd: If these calculations are right, then amplifying the low output system would be less demanding for the input tube (4.8mmho) than for the higher output system (12.3mmho) for similar results. That's hard to believe.

JH.

 

[jhaible]

[quote author="bcarso"]
There is a good discussion of noise estimation for phono preamps with RIAA EQ in some National Semi documents, taking noise voltage and noise current into account (the latter generally negligible with tubes and FETs).
[/quote]

This one?
http://www.national.com/an/AN/AN-104.pdf#page=5

JH.