The volume pot - The hidden villain of preamp

[MelB]

I'm sure a volume pot is not ever going to be an "ideal" resistor. Do people here have a favourite pot? (Less noisy or more reliable as far as not getting scratchy) Or does the stepped attenuator have less noise and rule the day?

Also pretty much every guitar amp on the planet places it's volume pot after the first valve. I would assume to reduce the noise introduced from said pot which Popilin showed so nicely in his first post.

[popilin]

Hi MelB

Nice to see you around here !

Originally Posted by MelB

I'm sure a volume pot is not ever going to be an "ideal" resistor. Do people here have a favourite pot? (Less noisy or more reliable as far as not getting scratchy) Or does the stepped attenuator have less noise and rule the day?

No resistive compound is exempt from thermal noise, given by the equation of post#1, only depends on the resistance value, not its composition.
I love my Alps Black Beauty, is made of a resistive polymer which minimizes other types of noise.
What makes a good pot so expensive, is its precision, which is harder if it is a dual device.
For the rest, there are no major differences with a stepped attenuator.
I prefer the pot, have that feeling that makes it so enjoyable, and without breaks.
Among the specs of my Alps, appears the momentum necessary for operation !

Originally Posted by MelB

Also pretty much every guitar amp on the planet places it's volume pot after the first valve. I would assume to reduce the noise introduced from said pot which Popilin showed so nicely in his first post.

Thanks for your feedback and kind words, encourages me to continue the thread.

[popilin]

Originally Posted by ruffrecords

Fair enough, but I think calling the volume pot a 'villain' is a little premature.

OK, if you prefer, you can think it's just marketing.

Originally Posted by ruffrecords

And the point is there is a whole range of noise sources in the pre-amp that you can control and pot noise is just one of them. You need to examine and quantify the others before you brand the volume pot as the 'villain'.

Cheers

Ian

As we have seen in the simulations and other estimates, the volume pot and ECC82, contributing almost 92% of the total noise, although it is a noisy valve, its contribution is of the same order as the pot.

Using high transconductance valves, and the final scheme of connection (3), valve noise is negligible compared to the pot noise, being valid the result for the last simulation of post#1.

Do not ask me to do the math on a weekend, my religion forbids me.

If it bothers you is the title of the thread, no problem, you can propound another and we ask the moderators to change it.

[counter culture]

Originally Posted by MelB

Also pretty much every guitar amp on the planet places it's volume pot after the first valve. I would assume to reduce the noise introduced from said pot which Popilin showed so nicely in his first post.

Most guitars have a volume pot built in, which itself is a source of noise, so that's not why. Think again...

[ruffrecords]

Originally Posted by popilin

As we have seen in the simulations and other estimates, the volume pot and ECC82, contributing almost 92% of the total noise, although it is a noisy valve, its contribution is of the same order as the pot.

Using high transconductance valves, and the final scheme of connection (3), valve noise is negligible compared to the pot noise, being valid the result for the last simulation of post#1.

In fact neither statement is entirely true. I have checked the thread carefully and can only find references to calculating the thermal noise of the resistor and the old 2.5/gm chestnut for the tube noise. It is not clear how you determined this is 92% of the total noise. However:

1. In the case with the pot between stages you have ignored the source impedance feeding the first tube. There would be no point using a 100k pot if this source was not of the order of 20k in which case the noise is no better than with the pot at the front. If the source is a lot less than 20k you don't need a 100k pot.

2. Tube noise is much worse than you think. 2.5/gm only applies to shot noise and only at rf. At audio frequencies flicker noise is at least as important and gets less as Ia is reduced, the exact opposite of shot noise. Minimising tube noise is not therefore simply a matter of choosing a high gm tube and winding up the idle current.

3. I have yet to measure a real cathode follower driven by a non zero source impedance that has less than 80uV output noise in a 20KHz bandwidth.

Cheers

Ian

[MelB]

Originally Posted by counter culture

Most guitars have a volume pot built in, which itself is a source of noise, so that's not why. Think again...

Humm...no idea then. Guitars have volume pots?

[counter culture]

So that you can overdrive the next stage, or not as the case may be...

Yes, they do.

[DF96]

Originally Posted by ruffrecords

In fact neither statement is entirely true. I have checked the thread carefully and can only find references to calculating the thermal noise of the resistor and the old 2.5/gm chestnut for the tube noise.

Back in post 18 I said:

Originally Posted by DF96

2.5/gm gives a useful estimate of shot noise for mid-frequency RF purposes. It fails at higher (e.g. VHF) and lower (e.g. audio) frequencies, because valve noise is then dominated by grid noise and flicker noise respectively. Flicker noise can't easily be calculated as it varies so much from sample to sample; it has to be measured.

It surprises me just how common is this myth that high gm guarantees low audio noise. I can't remember where I saw it, but someone did some measurements comparing 12AY7 (gm around 2mA/V?) and 6DJ8 (around 12mA/V?) and found very similar audio noise, with the 12AY7 being slightly better.

The funny thing is, 2.5/gm for triodes (which is only a rough approximation anyway) is most useful in the very region (IF/HF) where almost everyone uses pentodes - so partition noise dominates. Below that flicker noise dominates. Above that grid noise dominates.

[popilin]

Originally Posted by ruffrecords

In fact neither statement is entirely true. I have checked the thread carefully and can only find references to calculating the thermal noise of the resistor and the old 2.5/gm chestnut for the tube noise. It is not clear how you determined this is 92% of the total noise. However:

As I said before, the analyzes are only conceptual, and a reasonable approximation can be used in most cases.

Doing again the third simulation, of post#1 for a 20KHz bandwidth, we obtain

pot noise ≈ 35μV RMS, other sources noise ≈ 5μV RMS

Using

N = √[∑(Ni)^2]

And the wrong estimate of valve noise of post#16, we obtain

N ≈ 94μV RMS

So, noise is dominated by valve noise and pot noise, about 95%

Originally Posted by ruffrecords

1. In the case with the pot between stages you have ignored the source impedance feeding the first tube. There would be no point using a 100k pot if this source was not of the order of 20k in which case the noise is no better than with the pot at the front. If the source is a lot less than 20k you don't need a 100k pot.

I'll tell for a third time, for simplicity, we assumed a zero source impedance, in which case all simulations were made.

If it is necessary to place a source of about 20K, such as a phono stage, things can only get worse.

Leaving aside the fact that not everyone can afford to have a separate phono stage, and I say this for the second time, a 100K pot prevents us using an additional CF before pot, in approach (1).

The latter approach (3), is always the most favored, and I say it for the second time, that between the two stages where is located the pot there are high voltages, which require the use of coupling capacitors, again, a 100K pot avoid us using huge capacitors, regardless of the source impedance for the first stage.

If you put a 10K pot and 30μF capacitors, is fine, but it doesn't seems good engineering.

An implicit detail that is quite obvious, in this configuration is necessary to reduce the gain of the first stage, in order to reduce distortion, so a 20K source has less impact on noise.

Also tell for the fourth time, that the only decent pot I own, is a 100K Alps Black Beauty, which is the cause that motivated me to notice the problem and start this thread.

My Tarzan-English does not help much, but neither need a cryptographer.

Originally Posted by ruffrecords

2. Tube noise is much worse than you think. 2.5/gm only applies to shot noise and only at rf. At audio frequencies flicker noise is at least as important and gets less as Ia is reduced, the exact opposite of shot noise. Minimising tube noise is not therefore simply a matter of choosing a high gm tube and winding up the idle current.

This has been already clarified by DF96 in post#18, Morgan Jones also makes clear in his "Valve Amplifiers" third edition, at page 534 sic "Unfortunately, there is no way of predicting flicker noise for a valve as it is highly sample dependent, although its level is directly related to the level of thermal noise for a given valve".

Curiously, on page 533, he uses the same old 2.5/gm chestnut for triode noise.


Assuming that electrons are fermions, using the Fermi-Dirac distribution, along with the addition theorem of spherical harmonics...no, just kidding.


The only reference to noise, I could find is

The Sound of Silence: Lowest-Noise Riaa Phono-Amps: Designer's Guide - Burkhard Vogel - Google Libros

Which gives us for the ECC82, on a 20KHz bandwith, something like Vn ≈ 6.4μV RMS, well below the estimate of post#16.

Redoing calculations

N ≈ 37μV RMS

Now pot noise predominates, about 95%.

This result is consistent with some known line stages based on the ECC82.

Originally Posted by ruffrecords

3. I have yet to measure a real cathode follower driven by a non zero source impedance that has less than 80uV output noise in a 20KHz bandwidth.

Cheers

Ian

I just own a multimeter and an old 15MHz oscilloscope, so I can not measure noise.
However a CF asisted by a CCS could help, or even better, Allen Wright SLCF.

[popilin]

Originally Posted by DF96

It surprises me just how common is this myth that high gm guarantees low audio noise. I can't remember where I saw it, but someone did some measurements comparing 12AY7 (gm around 2mA/V?) and 6DJ8 (around 12mA/V?) and found very similar audio noise, with the 12AY7 being slightly better.

The funny thing is, 2.5/gm for triodes (which is only a rough approximation anyway) is most useful in the very region (IF/HF) where almost everyone uses pentodes - so partition noise dominates. Below that flicker noise dominates. Above that grid noise dominates.

This is enlightening, but also I have a good alibi.

I am gullible as a kid, maybe I should ask Morgan Jones to return my money.