Understanding "grid drive" to drive, 2, 4, 6 power tubes

[Helmholtz]

I agree that we should endeavour to ensure new designs don't exceed the limits of the tubes used, but whilst not condoning the practice, my understanding is that the amp in question already tends to exceed that g2 limit, when idling / low signal levels at least.
The 6CA7 (vintage manufacture) g2 voltage limit looks to be lower than that of the EL34.
Note that many 6L6GC equipped amps breach the 450V g2 limit whilst at idle.

Seems we agree that it's not wise to exceed manufacturers' limits. And these include the screen voltage limit of 500V (425V) irrespective of actual screen current/dissipation.
I don't think experience regarding out of spec operation with one tube type/brand can be simply transferred to a different one.
Also I notice that people tend to apply design max. ratings ignoring that these don't allow for supply voltage variations.

[Helmholtz]

voltage drop over resistor

Does this measurement include the AC component, i.e. true RMS (AC+DC)?

[nsubulysses]

Measurement is made with Fluke 27 FM which is true RMS

[Helmholtz]

Measurement is made with Fluke 27 FM which is true RMS

Not familiar with this meter. But if voltage and current are RMS (AC+DC), their product (=apparent power) will give too high dissipation, as their AC components are out-of-phase.

[nickb]

I don't know how to do it the proper way. Nickb told me a good way to go about it sometime in the past but I remember it being pretty complicated.

I have crudely done it before by measuring voltage drop over 1K screen grid resistors at clip and at full blast (basically square wave) and multiplying it by what the screen voltage is.

How would you do it? I figured by measuring voltage drop over common 1K 20W resistor in series with plate supply you could estimate screen current if you divide by 6.

Is either of these ways a good way to do it?

Nor can I remember it! If you measure the average current into the screen and average voltage on the screen you won't be too far off. If you have got a meter that reads only RMS even on DC ranges you would need to make a little RC low pass filter say 10Hz (100nF across meter and 15k in series with lead). Measure the drop across the screen resistor using the filter and calculate to get the current. Use the filter once again to get the screen voltage.

[Old Tele man]

This one approximation method:

Average value of combined zero- and maximum-signals is approximated as:

X(avg) = (Xq/2 + ∆X/4).

[Helmholtz]

I think an accurate method would be to scope screen current taken with a (calibrated) DC current probe on one channel and screen voltage on the other (DC coupled), let the software multiply both channel signals for instantaneous power and then average the result.

Might be easier to watch the screen's glow color at full tilt. Bright red is bad, yellow is worse and white is close to death.

[loudthud]

If you measure the average current into the screen and average voltage on the screen you won't be too far off.

If you double the current for half the time and have zero current for half the time, (thus the average is still the same) the power goes up by a factor of two because when you doubled the current, the power went up FOUR times for that portion. The RMS current goes up by square root of two. When you square it in the I squared R equation, the result is two times the power. Average Voltage or current does not equate to average power.

[J M Fahey]

Might be easier to watch the screen's glow color at full tilt. Bright red is bad, yellow is worse and white is close to death.

Not kidding, that is one state of the art approach.

Math and Simulators are an incredible Human advance, but need accurate data and models to provide useful rtesults, while actually watching the d*mn screens is the real thing.

I did that all the time in the old days, when developing new power amps based on TV tubes (only kind available)
after supply of EL34 and 6L6 dried up for good and I had to actually *design* ... not rehash some old Fender or Marshall circuit.

Got the dreaded "disco lights dancing in rhythm with the Music" effect all the time, followed by real Pyrotecnics, until I could tame those pesky 6DQ6.

[Mike Sulzer]

But surely we cannot allow people to just look at the screens. There is a market niche to be filled: A device that slips on the tubes, connects to your computer and displays plate and screen glow, analyzes it, and has audible and visible alarms for danger: finally it delivers a report on expected tube life every few minutes if there is no immediate danger.

[nickb]

If you double the current for half the time and have zero current for half the time, (thus the average is still the same) the power goes up by a factor of two because when you doubled the current, the power went up FOUR times for that portion. The RMS current goes up by square root of two. When you square it in the I squared R equation, the result is two times the power. Average Voltage or current does not equate to average power.

I agree with what you say up to a point but the method I describe does work surprisingly well under certain conditons. In retrospect I should have mentioned them. These are a sine input up to max clean power. At max power the power comes out about 10% too high.

It makes me wonder about putting a little micro in an octal bias probe type of arrangement that does the calulation correctly. Any takers?

[Helmholtz]

I agree with what you say up to a point but the method I describe does work surprisingly well under certain conditons. In retrospect I should have mentioned them. These are a sine input up to max clean power. At max power the power comes out about 10% too high.

Which method are you referring to (as you didn't seem to remember)?
How did you verify accuracy/error?

Edit: Seems you mean the method described in post #65.

[g1]

has audible and visible alarms for danger: finally it delivers a report on expected tube life every few minutes if there is no immediate danger.

If you can work "Will Robinson" into that danger alert, I'll buy one.

[nickb]

Which method are you referring to (as you didn't seem to remember)?
How did you verify accuracy/error?

Edit: Seems you mean the method described in post #65.

Correct. Yes I did verify.

[Helmholtz]

Correct. Yes I did verify.

Using which method?