Measuring Screen Grid Dissipation

[JazzGuitarGimp]

I've finally come back to my latest build;
https://tubeamparchive.com/viewtopic.php?t=28247

I scrounged up enough cash to buy the Belton Micalex octal sockets, and at the same time, I bought a matched quad of TAD 6V6GT-STR tubes. I had taken the amp to a gig back in November, and the thing died on me during the second song. During troubleshooting, I bumped one of the output tubes and realized I had a defective socket:
https://tubeamparchive.com/viewtopic.php?t=28627

Anyway, I got the sockets replaced yesterday and the amp is working once again. But I got to thinking about screen grid dissipation: I have had back luck with 6V6 builds: The 5E3 I built seems to chew through output tubes. So I decided to see if I could measure the screen dissipation on the new amp at various output power levels. Here is what I did:

I measured the DC voltage drop across the 470R screen grid resistor at idle, full un-distorted output, and full on square wave output power levels. I then calculated the DC current through the resistor, and multiplied by the screen voltage. I discovered that, as long as the output stage is not pushed into clipping, screen grid dissipation is within the 2.2W maximum. Once the output stages begins to clip, the current through the screen grid resistor increases rapidly. So I went on a quest to determine what resistor value would keep the screen grid close to the maximum rating.

The amp is producing 34 clean watts from a parallel, push-pull quad of 6V6GT's. Please note, I only changed (and measured) one of the four screen grid resistors for this test. The other three tubes kept their original 470R screen grid resistor value for the duration of the test.

The following results are based on a 285mV 1KHz sine wave applied to the effects return (the amp has a built-in Dubleator) with the master volume full up. This produces a nice square wave at the output. I didn't actually start recording my findings until I was at 2K-ohms, but I tried 1K first. I do recall that screen grid dissipation was about 5.5W for both 470R and 1K.

Resistor_____DC Drop____Screen DC V____Dissipation
2K_________28V________295V__________4.13W
3K_________34V________289V__________3.28W
4K_________39V________284V__________2.77W
5.1K_______42V________283V__________2.33W

I have ordered some 2.7K, 3.9K, 4.7K, 5.6K, and 6.8K @ 5W resistors to try. I'm curious to see how the amp sounds after the change. I assume I will lose some output power and most likely, the sound will be compressed. I want to believe there is a happy-medium, or compromise in resistor value between 470R and 5.6K where the amp doesn't sound awful, but is safe for gigging (and long tube life).

Questions:

What are your thoughts on this? - I suspect 5.6K is a pretty huge value for this spot in the circuit.

What's the highest value screen grid resistor you've used for 6V6's?

Many thanks,
Lou

[brewdude]

This is interesting to me. I don't have any expertise to share, but I have a few thoughts and questions.

I wonder if a shared resisted feeding all the screens of say 220R-1k and a small grid stopped at every screen grid of say 47R-100R would give better results.

What is the dropping string resistance feeding the screen node?
What voltage are the screens idling at?
Are the screens idling below the plates?[/list]

[JazzGuitarGimp]

This is interesting to me. I don't have any expertise to share, but I have a few thoughts and questions.

I wonder if a shared resisted feeding all the screens of say 220R-1k and a small grid stopped at every screen grid of say 47R-100R would give better results.

What is the dropping string resistance feeding the screen node?
What voltage are the screens idling at?
Are the screens idling below the plates?[/list]

Good questions, and I wish I had equally good answers. I'm using a 10H choke as opposed to a string resistor. It has a DCR of 280 ohms. And yes, the screen grid node is below the plate load, but as you might expect from a choke, it's only by a volt ir two. Plates idling at 375V.

But, expabding on your idea, perhaps a zener diode or two to get the screen grid node down a bit, or even a MOSFET reducer (or fixed VVR) circuit to bring just the screen supply down. I wonder how a screen grid node at 250V would do? Then maybe the usual 470R at each screen would keep the screens in safe territory....

Food for thought.

[Littlewyan]

I think Zeners can make the HT a bit noisy. Never used them but thats what I've heard. Just bump up the screen grid resistors a bit. I'm surprised there isn't any difference been 470R and 1K. I know its a custom build but have you looked at amps with similar output stages to see what they use for the screens and what HT? The TW Express for instance can take 6V6s but it uses a 1K resistor instead of a Choke and 1K Screen Grid Resistors. I measured the voltage when driving the amp hard and the voltage after the 1K Choke Resistor sags a LOT! Drops by 100V I think and the HT is at 420V. The screens will sag even further as there is another 1K resistor between them and the 1K Choke Resistor.

I've had the EH6V6s red plate on me in the Express but the JJ 6V6s take it just fine.

Also don't forget the tone and the way the amp responds will change as you mess with the screen grid voltage.

And what is the impedance of your OT?

[Malcolm Irving]

The screen dissipation limit is about how hot the screen is allowed to get (above which it starts to melt). The instantaneous ‘screen current’ multiplied by the instantaneous ‘screen to cathode voltage’ is the instantaneous value of heat being put into the screen. So we need an average value over a short period of time, if the screen current is varying. (The screen voltage is almost constant.)
(The thermal inertia of the screen must be very small, so the time over which the average needs to be taken is quite short, but probably a full cycle of a guitar signal is short enough.)
When the output stage is idling, the screen current is constant and so measuring the voltage drop across the screen resistor, using a digital meter, will give an accurate result for screen dissipation. Under full load sine wave, the screen current will be a spikey waveform, while under full load square wave the screen current will be close to a square wave with a duty cycle of 50%.
I’m not sure what kind of average reading a digital meter gives for these waveforms (perhaps it is an accurate time-average, perhaps not – maybe someone could chip in here?). It would be good to look at the waveforms on a scope, for comparison, but even then you would need to guestimate the time average of the screen current waveform. I suppose modern digital scopes might be able to calculate that average.

[Malcolm Irving]

Just to add something: In a true pentode (EL34 etc.) the screen is a coil of thin wire and will have a low thermal inertia. In a beam-tetrode (6L6 etc.) the 'screen' is actually beam-forming plates, which will have a much bigger thermal inertia.

EDIT: This is incorrect. See posts below!

[martin manning]

It is the suppressor grid that is replaced by the beam forming plates, not the screen grid. The screen grid is however carefully aligned with the control grid to minimize screen current.

[Malcolm Irving]

It is the suppressor grid that is replaced by the beam forming plates, not the screen grid. The screen grid is however carefully aligned with the control grid to minimize screen current.

Yes, you are absolutely right. I just remembered that and I was about to delete the post, but I'll leave it there as a reminder to 'think first - type later'

[JazzGuitarGimp]

I know its a custom build but have you looked at amps with similar output stages to see what they use for the screens and what HT?

Mainly Fender amps - and 470R seems to be all that was used for any of the 6V6 models.

I've had the EH6V6s red plate on me in the Express but the JJ 6V6s take it just fine.

I have an aversion to JJ products. Not to mention their 6V6 is generally thought of as a 6V6 / 6L6 hybrid. I'd like to be able to put any brand of 6V6 in the build and know it will work as designed.

Also don't forget the tone and the way the amp responds will change as you mess with the screen grid voltage.

Yes, and this is my concern. But I suppose I won't know how it sounds until I try it. I am probably going to build a VVR for just the screen supply - that way I can hear the difference quickly, rather than have to wait the time it takes to change screen resistors.

And what is the impedance of your OT?

4K

I’m not sure what kind of average reading a digital meter gives for these waveforms (perhaps it is an accurate time-average, perhaps not – maybe someone could chip in here?).

I know with signal present, there is an AC component riding on the DC screen voltage (on the screen node side of the resistor). At first thought, I assumed I would have to measure both the DC current as well as the AC current and add the two. But the more I thought about it, I hypothesized that, at least for symmetrical waveforms, the two halves of the waveform should average the energy about the DC reference point. I don't know if this is a valid hypothesis or not and I would appreciate anyone who has a better understanding of this stuff to give their perspective. But so far, I have measured, using the DCV mode on my DMM, the DC voltage both before and after the screen resistor, and taken the difference of these two voltages to arrive at the voltage across the resistor.

[Littlewyan]

Mainly Fender amps - and 470R seems to be all that was used for any of the 6V6 models.

Yeah but everyone knows Fenders beat the hell out of 6V6s .

4K

Seems a bit low but probably not the issue here as I believe its a higher impedance that makes screens work harder, not lower.

Have you tried different 6V6s? Just because I know when I've tested different EL34s in my Express I found one set that uses 2mA of Screen current per valve at idle, but another set (different brand) uses 8mA at idle.

[brewdude]

I'm wondering if using a VVR to find the right voltage to limit the screen current is going to give you misleading results with regard to current draw using a resister because of the current limiting nature of the VVR. (?)

I was alsowo wondering if this entire process is misguided because a guitar signal would be much different than a continuous sine or square wave. (?)

I am certainly no expert and mean no disrespect in my questioning your efforts. I'm actually following this thread with quite a lot of interest.

[jjman]

Using DC measurements under signal may or may not be numerically accurate depending on the meter. But the increasing reading is in the right direction since this is most likely class AB. A scope is needed to see for sure.

[tubeswell]

What's the highest value screen grid resistor you've used for 6V6's?

A shared unbypassed 4k7 for 2 x 6V6 screens. (Really noticeable compression)

[pdf64]

My Fluke has fancy rms ac+dc features.
When I've queried on MEF how to measure screen grid dissipation (ie if those features might give more accurate results), Merlin advised the method (ie straight Vdc) used in the OP as probably being the most reliable.

Unbypassed shared / individual dropper resistors can add a smoothing compression effect to the overdriven response.
Rather than increasing the individual screen grid resistors, in the first instance I suggest adding resistance in series with the choke, eg at least ~500 ohms.
If overdriven screen grid dissipation still seems excessive, then consider an unbypassed shared dropper resistor for the screen grids, as per the JTM45.

[JazzGuitarGimp]

It looks like my last post didn't go through - I have to be real careful when posting from my iPad and apparently, I wasn't paying attention this time....

Yeah but everyone knows Fenders beat the hell out of 6V6s .

Yes, understood. But apparently, 6V6's of yesteryear could take the abuse.

4K

Seems a bit low but probably not the issue here as I believe its a higher impedance that makes screens work harder, not lower.

Well, remember I am running a quad, not a pair of 6V6's. I think 8K is appropriate for a pair, so 4K for a quad should be the desired load (I think).

Have you tried different 6V6s? Just because I know when I've tested different EL34s in my Express I found one set that uses 2mA of Screen current per valve at idle, but another set (different brand) uses 8mA at idle.

The Weber kit came with a quad of unnamed Chinese 6V6's - two of which have now failed. Unfortunately, I wasn't looking at screen dissipation before those two tubes went south.

I'm wondering if using a VVR to find the right voltage to limit the screen current is going to give you misleading results with regard to current draw using a resister because of the current limiting nature of the VVR. (?)

It seems to me that a VVR doesn't limit current - it limits voltage. Please correct me if I am wrong.

I was also wondering if this entire process is misguided because a guitar signal would be much different than a continuous sine or square wave. (?)

I am not sure. It seems most guitar signals I've observed on the scope are symmetrical about their midpoint, so I think this causes a zero-net effect on the screen grid DC current.

I am certainly no expert and mean no disrespect in my questioning your efforts. I'm actually following this thread with quite a lot of interest.

No disrespect inferred. I am finding this subject most interesting as well. I'm looking at it as a chance to learn something new about tube circuits.

Using DC measurements under signal may or may not be numerically accurate depending on the meter. But the increasing reading is in the right direction since this is most likely class AB. A scope is needed to see for sure.

I am using a Fluke 87 mkIII, which is well regarded. It is class AB1. I do have a scope, but haven't yet spend much time looking at the waveform on the screen grids.

A shared unbypassed 4k7 for 2 x 6V6 screens. (Really noticeable compression)

Yes, this is the result I was thinking of. Thais may or may not be desirable and I just won't know until I've tried it. But I am thinking more and more about trying a VVR on the screen supply first. The stock 470R at each screen may be all that is needed if the screen node is taken down to 250ish volts. The trade off there, is the reduction in bias voltage. I'm currently at about -35V, and I like the way the PI is working with the PA. Taking the screens down to 250V will likely cause the bias to want to be at -20-ish (or less) volts, and this may upset the harmony between the PI and PA.

My Fluke has fancy rms ac+dc features.

Sounds like yours is a bit fancier than mine

When I've queried on MEF how to measure screen grid dissipation (ie if those features might give more accurate results), Merlin advised the method (ie straight Vdc) used in the OP as probably being the most reliable.

This is reassuring. It sounds like I am on the right path. Thanks for this.

Unbypassed shared / individual dropper resistors can add a smoothing compression effect to the overdriven response.
Rather than increasing the individual screen grid resistors, in the first instance I suggest adding resistance in series with the choke, eg at least ~500 ohms.
If overdriven screen grid dissipation still seems excessive, then consider an unbypassed shared dropper resistor for the screen grids, as per the JTM45.

These ideas sound interesting. I will likely come back to this advice - thanks again.