EL34 Excessive Screen Dissipation

[Littlewyan]

Lately I noticed on one of my EL34 builds that one of the screen grid resistors had discoloured from excessive heat. So I put a signal generator on the amp, sent a 1Khz 1V signal into the input and measured the screen grid current draw. Above 7 on the volume when the power amp starts to clip the screens start to draw 60mA and end up dissipating 21W! (348V was measured on the screen grid). The resistor itself only dissipates 3.6W at this point. The amp has 470V on the plates, a Marshall choke (120Ohms DC resistance), a 6.6K OT and 1K 5W screen grid resistors. I tried again halving the load and this time I measured a 40mA draw and the screens dissipated 14W (352V on the screens). It's still over spec but a good improvement and shows that my OT impedance is probably a tad high for this voltage.

I'm wondering if I should replace the screen grid resistors with 2.2K 7W. This should lower the voltage and in turn perhaps lower the current? But then I also thought, the Komet amps run a pretty high plate voltage (470-500V I think), with a choke and a higher than normal OT impedance. So why aren't those things suffering?! Be good to hear people's opinions on this.

[martin manning]

Is the idle screen voltage up near 470, and what is your Zpri nominally?

[Littlewyan]

Yes it sits at around 467V. By Zpri do you mean impedance? If so, 6,600 Ohms.

[martin manning]

Yes, correct primary impedance. 6k6 is good for 470V on the anodes as far as max plate dissipation goes. A quick check on that is Va^2/(Za-a * Pa max), which at 470V and 6k6 is 133%. The screen voltage is high, putting the static load line well below the knee of the Vg1=0 curve. That's the reason for the excessive screen dissipation. An easy fix is larger screen resistors, but you could also try adding a shared 1k dropping resistor like JTM45. Another possibility is to add a MOSFET screen voltage regulator, basically a VVR for the screen voltage, to reduce screen voltage to 300-350V. Then you can use small screen resistors.

[Littlewyan]

I think due to the limited space I have left I’ll try changing the screen grid resistors to 2.2K 7W first. Dissipation on the resistor itself should hopefully not be an issue, as when the screen voltage drops the current draw should too. It’s a bit of ‘suck it and see’. If it’s still an issue after this then I may have to look at using 5881s or 6L6s instead. Luckily this is a preamp distortion master volume amp that wasn’t designed with power amp distortion as a main feature. I just want to make sure this thing is bulletproof in case I do need to crank it in the heat of battle!

Thanks for your help Martin.

[Roe]

some el34s draw more screen current than others. RFTs/Siemens draw the most, Shuguang the least.

A 100w marshall may draw as much as 130ma, although the screens voltage tend to sag to around 325v at full power

[Littlewyan]

I tested some Groove Tubes EL34Ms that I had and they drew a bit less, but the dissipation was still high. I also tried a pair of 5881s and they drew much less screen current, but still too much for their spec I think. Maybe this is perfectly normal in guitar amps and designers just accept that they're running the valves hard. Although I measured screen current draw in my JTM50 and that was right on the money, 8W, which is the max spec. I can see the Trainwreck Express coping ok as it has a 1K choke resistor and a lower voltage overall, but I can't see how valves survive that well in Komet amps.

That can't be 130mA per valve on the screens though? That would be 42W screen grid dissipation!

[Roe]

I tested some Groove Tubes EL34Ms that I had and they drew a bit less, but the dissipation was still high. I also tried a pair of 5881s and they drew much less screen current, but still too much for their spec I think. Maybe this is perfectly normal in guitar amps and designers just accept that they're running the valves hard. Although I measured screen current draw in my JTM50 and that was right on the money, 8W, which is the max spec. I can see the Trainwreck Express coping ok as it has a 1K choke resistor and a lower voltage overall, but I can't see how valves survive that well in Komet amps.

That can't be 130mA per valve on the screens though? That would be 42W screen grid dissipation!

no, a quad, so 10+w per tube

[Littlewyan]

Ah that makes more sense. Personally I think over dissipation on the screens sounds bad. When testing my Airbrake recently I found that at high attenuation settings my amps didn't sound good and I'd lose headroom, even after I modified the Airbrake it to adjust what impedance the speaker sees. The issue was the impedance reflected back to the amplifier. At the highest attenuation setting the amp would see 9.5Ohms when it needs to see 4Ohms and this affected my JTM50 by making the screens dissipate 24W! At that level I found that I would hear this crackling static sound when hitting an E chord and the amp just sounded harsh. Once I modified the Airbrake a bit more and got the reflected impedance down to 6.9Ohms at the highest setting the amp sounded much better.

I did wonder if Ken only sold the Airbrake to Trainwreck owners originally because the Express or Liverpool could probably cope with the higher load. They have diodes on the valves that absorb high flyback voltages and the 1K choke resistor probably helps counteract the higher screen current.

[Roe]

it sounds bad on ac30s and on some jtm45s. but it may sound good on el34 marshalls unfortunately. Still, it affects the attack and response even more it seems

[wpaulvogel]

I’ve got some Tung-Sol CP EL34B’s that don’t draw as much screen current.
The shared 1k dropping resistor idea is a good one for what you have going on. It’s the same concept as the 2.2k 7 watt but it will put the heat in a different area than the socket which is good. The 2.2k per socket is going to put out twice the heat right under the tube. Remember that each tube only conducts for each half cycle so that added 1k dropper probably should be 7 watts rated. It’s going to be nearly 100% duty cycle. With 6.6k primary lower screen voltage is important. Trainwreck used a 1k “choke” resistor along with the 1k screen resistors and 6.6k primary’s probably for reliability.

[Littlewyan]

Unfortunately I don't think a shared dropper is an option as I don't have the space for a big enough resistor. 7W wouldn't be enough, the Trainwreck amps use a 25W resistor and for good reason! I'm sure when I measured my Express the 1K choke resistor was dissipating more than 10W. I'll try and get that amp out at some point this week and measure it again. I'm sure I had a pretty high reading when the amp was running flat out.

It'd be good if a Komet owner chimed in and measured their amp. They must be running the screens really hard.

[Roe]

if 5w is sufficient for the individual 1k screen resistors, then 25w is overkill for the shared screen resistor. 5w and 10 [20]w, respectively, should work, shouldn't it?

[martin manning]

Don't forget that with the additional dropper the screen voltage and peak current will both be reduced, lowering the resistor power dissipation.

[Littlewyan]

When I measured screen current draw at full tilt I measured 60mA on each valve and at one point 74mA (not sure why I had a different reading). If we assume for worst case that the current draw doesn't change after I add a 1K shared screen dropper resistor then that resistor will drop 148V (74mA for both valves over a 1K resistor). This means it will dissipate 21.9W. If the valves actually draw 60mA each then it's still going to dissipate 14.4W. If the current draw drops to 40mA per valve then the resistor will drop 80V and will dissipate 6.4W.

So there is a chance that a 7W would work, but I wouldn't know unless I tried. The hope would be that higher resistance on the screens would drop the voltage more, in turn dropping the current draw and at the same time not putting too much stress on the screen resistors. It's a bit of a suck it and see I think. I've ordered 2.2K 7W resistors so I'll try those first and see what results I get. If I still have issues then I'll look at a 470R/1K 7W shared dropper resistor. I do have some 470R 7W resistors so I could try one. I think from memory the turret that my screen supply comes from is very near one of the valve sockets, so I guess I could just bridge the resistor between the turret and valve socket.