Heaters Voltage

[Nikon1975]

Dear All,

I noticed that in more than one amp design the heaters of the drivers (12AX7 or 5842) are fed with a lower voltage compare to specs. Something like 5.5V instead of 6.3V. I understand that this is done intentionally. What's the theory behind this ? I could not find any reference about.

Thanks,

Davide

[kevinkr]

It's supposed to reduce thermal noise and other noise sources (mostly low frequency iirc) in tubes used to amplify very low signal levels like in phono and tape stages.

[EC8010]

Reducing heater voltage to reduce noise is another Hi-Fi myth with a kernel of truth. The original reference is "Signal, Noise and Resolution in Nuclear Counter Amplifiers" by A B Gillespie and refers to reducing grid current noise in electrometer amplifiers by reducing heater voltage (and therefore grid emission). The fact that reducing heater voltage crucifies gm and therefore raises shot noise within the valve was not a problem because noise in electrometer amplifiers was dominated by grid current. The only audio application that would benefit from this strategy is the head amplifier in a condenser microphone.

[kevinkr]

Originally Posted by EC8010

Reducing heater voltage to reduce noise is another Hi-Fi myth with a kernel of truth. The original reference is "Signal, Noise and Resolution in Nuclear Counter Amplifiers" by A B Gillespie and refers to reducing grid current noise in electrometer amplifiers by reducing heater voltage (and therefore grid emission). The fact that reducing heater voltage crucifies gm and therefore raises shot noise within the valve was not a problem because noise in electrometer amplifiers was dominated by grid current. The only audio application that would benefit from this strategy is the head amplifier in a condenser microphone.

That's very interesting, learn something new every day! Long ago I decided to heat using the nominal rated filament voltage because I was unable to establish that there was any measurable reduction in noise levels in a phono stage I was working on at the time. Subjectively I was convinced there had to be (audio dogma?), and I thought that the test equipment I had was just too noisy to reveal the difference. (HP 400GL, 330D, etc..) Now it appears that probably wasn't the case.

[tomchr]

Originally Posted by EC8010

Reducing heater voltage to reduce noise is another Hi-Fi myth with a kernel of truth. The original reference is "Signal, Noise and Resolution in Nuclear Counter Amplifiers" by A B Gillespie and refers to reducing grid current noise in electrometer amplifiers by reducing heater voltage (and therefore grid emission). The fact that reducing heater voltage crucifies gm and therefore raises shot noise within the valve was not a problem because noise in electrometer amplifiers was dominated by grid current.

Pardon my ignorance, but how does lowering the heater voltage lower the grid emission? Is the grid close enough to the heater to be heated to an extent where it emits electrons?

Also, I'm not clear on how lowering the gm would impact the shot noise. Shot noise is caused by the random movement of electrons, hence, is proportional to the anode current. Assuming the anode current is the same before and after lowering the heater voltage, the shot noise from the anode current should be the same. However, if lowering the heater voltage reduces the grid current, it would lower any shot noise caused by this current.

Am I understanding this correctly?

Thanks,

~Tom

[Merlinb]

Originally Posted by tomchr

Pardon my ignorance, but how does lowering the heater voltage lower the grid emission? Is the grid close enough to the heater to be heated to an extent where it emits electrons?

Yes exactly- lower voltage means lower temperature, and grid current noise is one of the (usually minor) sources on noise in valves.

However, EC8010 may be being too sceptical. Lowering the heater voltage to reduce noise does not just appear in computing texts, but also Cherry & Hooper's Amplifying Devices and Low-Pass Amplifier Design, 1968.

Shot and flicker noise is proportional to cathode temp', bandwidth and anode current^2 [yes!], and inversely proportional to gm, so it is a balancing act of minimising the cathode temp and and anode current, without reducing gm by the same factor (trial and error).
In fact, it might be fairer to say that the belief that running valves at high anode current automatically leads to low noise is the real audio myth! (well ok, not a myth, but an over-simplification).

[SY]

Well, remember that in consumer audio circuits, the source impedances tend to be low (e.g., phono cartridges) so that grid current noise tends to be negligible. Thus it really is worthwhile to run high current in input stages.

Is the grid close enough to the heater to be heated to an extent where it emits electrons?

Yes. That's why some low noise tubes use gold plating on the grid wires. Moreover, the grid can become contaminated by cathode emission, increasing its own propensity to emit.

[Shoog]

I think the main advantage of running lower heater temps is extended valve life. There is a chart somewhere that shows the relationship. Running at 10% over will considerably lower valve life, running 10% lower will have the opposite effect. Tungsten bulbs experience the same behaviour and within Britain light bulbs sourced from Europe (all of them now) have considerably shorter duty life than in main land Europe.

I have also heard it said that triodes run under voltage are more linear - but I cannot remember where I heard that and would not take it as anything other than hearsay.

Of course when small signal valves have lives of 5-10yrs this may not be an issue for you - but it would be ill advised to run them over-voltage.

Shoog

[EC8010]

Originally Posted by Merlinb

Shot and flicker noise is proportional to cathode temp', bandwidth and anode current^2 [yes!], and inversely proportional to gm, so it is a balancing act of minimising the cathode temp and and anode current, without reducing gm by the same factor (trial and error).

I'm afraid I can't agree with the anode current squared statement. If you think about it, the anode current is composed of the sum of all those quanta of charge (electrons) and because they have a Poisson distribution, the variance is equal to the total number of electrons. Their standard deviation is equal to the square root of the variance, so the relative standard deviation (which is effectively the "noise") is the reciprocal of the square root of the total number of electrons. Thus, shot noise is proportional to the inverse square root of anode current. Add some fudge factors in to make the real world fit our system of units, note that equivalent noise resistance is a more useful parameter, and include the effect of voltage gain within the valve, and you find that the inverse square root relationship magically transforms into the well-known r_eq.=2.5/g_m.

Just to add a date; Gillespie was 1953 and dealing with seriously high source impedances, so grid current noise was the problem, not shot noise.

[Merlinb]

Originally Posted by EC8010

I Thus, shot noise is proportional to the inverse square root of anode current.

I am only quoting Cherry, although it sounds like you are applying a simplified model of a triode.

Cherry provides the following for the shot/flicker mean-square equivalent input noise generator:

d(vin) = [3.5 x 10^-20 / gm + P(1 / f) ] df

The first part is shot noise and is related to the cathode temp (normally 1000K multiplied by the Boltzman constant and an empirical constant etc)

P is a flicker factor which depends on tube type, and is equal to KI^2 / gm^2, where K is a constant (something around 10^-8) and I is the anode current.

f is a given frequency.
df is the bandwidth.