• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

The volume pot - The hidden villain of preamp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Those who live far from major cultural centers, we have no access to live music and must settle for listening to a recording.
If the recording engineer is an OPAMP lover, or he was taken too seriously Niquist-Shannon theorem, we can do nothing, the die is cast.:rolleyes:
When designing a Preamp only remains for us, as design goals, the basic and standard

a) Low distortion, preferably predominantly second harmonic (the lesser evil)
b) Good dynamic response.
c) Good bandwidth.
d) Small phase shift.
e) Low output impedance.
f) Low noise.

Like many others, I always assumed that valves are noisy and little can be done about it.
Wrong ! noise may get worse ! much worse !

§- Thermal Noise

Thermal noise can be defined as the noise generated by thermal agitation of the charge carriers.
For an ideal resistor R at absolute temperature T over a bandwith Δf, the RMS noise voltage is given by

Vn = √(4kTRΔf)

Where k is Boltzmann’s constant.

The following analysis is only conceptual, and calculations are courtesy of the software.

As a happy owner of a 100K Alps Black Beauty, designing a line-preamp, found real difficulties and some surprises.

1.- A traditional approach

Input => Pot => CC + CF => Output

i) Phase shift

As a reasonable design goal, is expected +/- 5º phase shift from 20Hz to 20KHz.
The worst case regarding the pot is at half the resistance, ie 50K + 50K when the output impedance reaches a maximum, supposing an ideal source of Z=0 ohm.
The only valve that I know for this requisite and a reasonable low gain is the ECC82/12AU7.
In the first simulation can be seen that hardly achieve +/- 5º phase shift.

ii) Noise

Meanwhile, the thermal noise produced by the pot over a range from 10Hz to 100KHz is 6.4µV RMS (second simulation).
For a tipical gain of 22dB, at the output the noise is about 80µV RMS (third simulation), only due to the pot !

2.- Nice try

Input => CC + CF => Pot => Output

With this scheme, the noise of the pot is not amplified, but it is a disaster in terms of the output impedance.

3.- Another approach

Input => CC => Pot => CF => Output

i) Phase shift

In the fourth simulation we can see that the goal of +/- 5º is easily achieved over a range from 10Hz to 100KHz.
Not bad, right?:D

ii) Noise

In the fifth simulation, we see that now the output noise is about 7µV RMS !
Remember that the volume pot is not the only source of noise, neither the only resistor, and mathematical models of valves don't take noise into account.
However not a bad result...:cool:
 

Attachments

  • Phase Shift.JPG
    Phase Shift.JPG
    99.7 KB · Views: 1,974
  • Pot Noise.JPG
    Pot Noise.JPG
    98 KB · Views: 1,957
  • Preamp Noise.JPG
    Preamp Noise.JPG
    101.2 KB · Views: 1,946
  • Phase Shift 2.JPG
    Phase Shift 2.JPG
    104.8 KB · Views: 1,920
  • Preamp Noise 2.JPG
    Preamp Noise 2.JPG
    101.3 KB · Views: 1,916
There is a flaw in your pot first noise simulation - you forgot to include the driving source impedance. If this is low compared to the pot value then the worst case noise resistance due to the pot is one quarter of the pot value I.e 25k and the noise voltage is halved. In any case, 100k is too large for an input pot and only makes frequency response and phase shift distortion due to the Miller effect worse.

Cheers

Ian
 
Administrator
Joined 2004
Paid Member
I'm still using 100K stepped attenuators to maintain compatibility with older tube gear, but in general ruffrecords makes a good point.. A 25K or 50K pot or attenuator would be a much better choice wrt input tube miller capacitance provided that all sources contemplated perform adequately into the proposed load impedance that pot or attenuator represents. (What a mouthful.. :p )

Another obvious option is to choose tubes with low Cag and low mu, still a lower resistance pot would probably be a bigger win.
 
There is a flaw in your pot first noise simulation - you forgot to include the driving source impedance. If this is low compared to the pot value then the worst case noise resistance due to the pot is one quarter of the pot value I.e 25k and the noise voltage is halved. In any case, 100k is too large for an input pot and only makes frequency response and phase shift distortion due to the Miller effect worse.

Cheers

Ian

Hi Ian

Sorry, that you say is already contemplated in 1.- i) and in the simulation.
Pot noise is about 6.4µV RMS, then with a gain of 22dB we obtain about 80µV RMS.

I agree with you that 100K is a little high, but if the previous stage is a phono pre, can prevent another CF, also is the only one I have. :D
 

Attachments

  • ScreenShot001.JPG
    ScreenShot001.JPG
    87.1 KB · Views: 218
Hi Ian

Sorry, that you say is already contemplated in 1.- i) and in the simulation.
Pot noise is about 6.4µV RMS, then with a gain of 22dB we obtain about 80µV RMS.

I agree with you that 100K is a little high, but if the previous stage is a phono pre, can prevent another CF, also is the only one I have. :D

Understood but I am unclear why you use a bandwidth of 100KHz. In a 20KHz bandwidth the noise is 2.2 times less which gives less than 40uV at the output. This is about -90dBu and well below the level of the noise contributed by the tubes themselves.

The other thing that is unclear is the assumed source impedance when the pot is after the first stage. If that is low enough that after 22dB of gain to only give 7uV at the output then there was no need for a 100K pot in the first place.

Cheers

Ian
 
Understood but I am unclear why you use a bandwidth of 100KHz. In a 20KHz bandwidth the noise is 2.2 times less which gives less than 40uV at the output.

Sorry Ian, is a bad habit of mine, for me the standard audio range is from 10Hz to 100KHz.
This is due to the fact that there are MC cartridges arriving at 70KHz and very good tweeters too.
Hear it or not, the noise is there, the listening experience is more complex than we would like.
Your compadre Tim De Paravicini, says in a more elegant.
Also remind you that not always the noise density is constant as in the case of the pot.


This is about -90dBu and well below the level of the noise contributed by the tubes themselves.

We are talking about a preamp, and with high transconductance tubes, I would not be so sure. :confused:

The other thing that is unclear is the assumed source impedance when the pot is after the first stage. If that is low enough that after 22dB of gain to only give 7uV at the output then there was no need for a 100K pot in the first place.

Cheers

Ian

The source impedance was assumed low from the beginning for simplicity, as you request in post#2
When the volume pot is after the first stage, it is assumed that the voltage levels are high, you should put capacitors, with less than 100K, should resign the low end or use huge capacitors.
It is a matter of compromise, or taste. :)
 
Last edited:
Sorry Ian, is a bad habit of mine, for me the standard audio range is from 10Hz to 100KHz.
This is due to the fact that there are MC cartridges arriving at 70KHz and very good tweeters too.
Hear it or not, the noise is there, the listening experience is more complex than we would like.
Your compadre Tim De Paravicini, says in a more elegant.
Also remind you that not always the noise density is constant as in the case of the pot.

I can't see ultra violet light but it is still there. It is pointless including noise outside the audible range.



We are talking about a preamp, and with high transconductance tubes, I would not be so sure. :confused:

The only tube you mentioned is the ECC82 so there is no indication we are talking about hi gm tubes.

The source impedance was assumed low from the beginning for simplicity, as you request in post#2

In which case my point about 100K being the wrong value to choose still stands.

When the volume pot is after the first stage, it is assumed that the voltage levels are high, you should put capacitors, with less than 100K, should resign the low end or use huge capacitors.
It is a matter of compromise, or taste. :)

It is a matter of good or bad engineering.

Cheers

ian
 
I can't see ultra violet light but it is still there. It is pointless including noise outside the audible range.

I can not either see the background radiation of the universe, but there are people who it bothers.
Please, read Tim De Paravicini, about infra and ultra sounds, also are part of the musical experience.;)

The only tube you mentioned is the ECC82 so there is no indication we are talking about hi gm tubes.

Sorry Ian, I forgot to mention that this is a thought experiment that attempts to show the best place for the volume pot.
In the example above, I used the ECC82 because interelectrodic capacitances and low gain.
I also forgot to mention I'm trying to design a Hi-End Preamp, valves I use, for now I keep in secret.:D

In which case my point about 100K being the wrong value to choose still stands.

OK, if the preceding stage impedance is rather high, put a 10K pot if it makes you happy.
BTW. What part of "As a happy owner of a 100K Alps Black Beauty" and "also is the only one I have" you don't understand? :D

It is a matter of good or bad engineering.

Cheers

ian

I agree with you ! :p
 
10Hz to 100kHz might make sense when considering the bandwidth of a feedback amplifier but I think it's silly thinking this is relevant for hearing no matter what Tim has to say about it. Why not use a more 'conventional' range for now and then extend the findings later if the results are interesting enough to warrant it ?

....The quietest pre-amp is no-preamp.
 
Actually, it does make sense to keep the high end roll-off higher than one would normally expect, given the definition of 'audio band'. It has been shown that frequencies as high as an octave over 20kHz do have some (admitedly relatively small) impact, and in fact this is even built into high-res digital formats (one notable example is SACD), where there is a low order 50kHz filter that can be applied. I myself have been involved in a study about echo and phase shift perception where it was quite conclusively demonstrated that the human auditory system does indeed perceive phase shifts and time shifts that suggest audibility over 20kHz. But even leaving that aside, one would be well advised to look at the whole amplification chain and not only at one element - it forms a set of cascading filters. Simply put, you do not want the phase shift and roll-off to add up, so it pays to keep your passband quite significantly wider (within reason!) than 20kHz, especially when it's done easily enough at no real extra cost. In fact, one good rule of thumb for various components you expect to have connected in a chain, except perhaps the power amp is to keep amplitude response to within +-0.1dB 20Hz to 20kHz. This actually implies extending the lower and upper cut-off by nearly a decade (assuming a low order character to the slopes). Power amps (tubes especially) are a bit more problematic in that regard WRT output transformer, but even then, expecting an 'unknown in advance' or changeable amplification chain, you want the least possible elements to dominate the response, and normally these would be your speakers, certainly if there has to be another one, then let it be only one, and this would then fall to the power amp, where it's most difficult to extend the passband if tubes are involved.
 
Last edited:
10Hz to 100kHz might make sense when considering the bandwidth of a feedback amplifier but I think it's silly thinking this is relevant for hearing no matter what Tim has to say about it. Why not use a more 'conventional' range for now and then extend the findings later if the results are interesting enough to warrant it ?

OK, let's put aside the opinion of Tim.
Can you show me that the noise, seen as thermal agitation of electrons within the valve, say at 100KHz, has no influence on the behavior of the electron cloud in the audible spectrum ?
I dare not...
The next estimates will be in the range of 20Hz to 20KHz, happy ? :rolleyes:

....The quietest pre-amp is no-preamp.

And it is even quieter when you hear vinyl.
Vinyl without a preamp is the quietest thing in the world of audio. :D
 
Hi ilimzn

Thanks for your feedback, I can only agree with you.

With regard to valve amplifiers, when designing a transformer, I try to achieve at least 50KHz, which is quite difficult for the capacitances involved.
The last trick is to use NOMEX, due to its low dielectric constant.

For the same reason I'm working on a hybrid, valve voltage amplifier, current amplifier with BJTs.
 
popilin, I'll have to remember that trick :)

BTW interesting treatise, but it has been done before, mostly in the world of sand :) and in one case that is very dear to me, the 'pot' is in fact a custom made ALPS 'stepped' attenuator with 180 steps - they did not bother to make it actually have tactile steps, it acts as a pot - and it's resistance end to end is only 3k.
However, in the case of tubes, most standard tubes will overwhelm any pot generated noise by their own (via several mechanisms). As you know it's possible to combat this with a high gm tube, although it's not that simple. Tubes have several noise producing mechanisms and some are quite dependent on manufacturing quality, one important one being grid leakage current, which is mostly dependent on vacuum quality but not exclusively. This current is by definition 'noisy' and I am mentioning it because it can be significant in high gm tubes, especially as there is a trade-off between tube current and noise, but increased tube current can reduce noise but brings bias closer to zero, which increases grid leakage current, increasing noise. For this reason it's always advisable to use the lowest resistance pot that is feasible given the design goals and standard conformance.

One more area where choosing the right pot value is important is exactly the use of high-gm tubes, as these often have relatively higher input and reverse transfer capacitances. More than noise, the bandwidth of the circuit can vary a lot depending on pot position, which is one reason 'passive preamps' may sometimes have unexpectedly sub-par performance (usually this is more the case with solid state as capacitances are usually also nonlinear). From that standpoint your approach also has merit, and I would say more so than for noise considerations.
 
Hi ilimzn

My post is not intended to be a treatise, just another trick to the community. :)

In the world of sand, is simple, place a low value pot between two opamps, and the problem disappears.

Your post is almost a description of the problems encountered during the design, especially the variation of bandwidth with the position of the pot, however improved with approach 3.- and a suitable output stage, the Allen Wright SLCF there does wonders...;)

Regarding the noise, I can only deal with thermal noise, I'm just a TV repairman. :D

For triodes, RF Engineers use the approximation

r(eq) ≈ 2.5 / gm

In the worst case, the ECC82 has a transconductance

gm = 2.2 mA/V

Thereby

r(eq) ≈ 1136 ohm

Assuming that noise density is constant, and T = 313ºK, the equation of post#1 for a bandwidth of 20KHz, gives

Vn ≈ 62µV RMS

Of course, the use of high transconductance valves, greatly improves the noise reduction.
 
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.

Thanks DF !, you're always the reference. :)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.