Design considerations for high voltage, high power, tube output section/PSU

[nsubulysses]

They are dissipating about 14W at idle. about 25mA each at 575Va

So .025 x 575 = 14.375W, so 57% of 25W.

I would think if it's W thats killing them then the JJ logos would be darkened, but who knows. I guess that's just an assumption.

I will probe these for individual cathode current. I appreciate the suggestions thank you. Need a fresh perspective

Edit: actually I shouldn't say 25mA each because I have been lazily using the transformer shunt method to measure current draw per two tubes per each side of the OT. I have never hooked up the bias probe or installed cathode current sense resistors. Maybe pdf is on to something.....

[catalin gramada]

Hi. Just suggestion. Check the voltages over the screen resistor at idling and power condition. You can determine the current the screens draw in both situation. For screen P.max el34 data sheet say no more than 8w. I currently run my el34 screens at 12w when full throttle without problems but not expect to use this tube for life. So if your screens dissipate more at full power you can do two things- reduce the screen supply voltage somehow or change the screens resistors with bigger values to limit more current at full throttle. Check again then to be in range. Each method will provide different kind of power stage manifestation so is better to check both and choose whatever fit you demands.

PS - I was just curious how vt22. same as v4 you're talking about looks like and noted from my schematic... separate screen power supply 338V and low filtering without series screen resistors. That means safe and very dynamic conditions for the tubes-although some small values say 100ohm or so are welcomed to prevent oscillation here.
You will see also a 470 ohm/7w resistor into the screen supply then individual 470.that means a large estimation 2.35 k limiter for each screen which means other thing.
Just thinking...you decoupled those screen ''supply''node...No need it, otherwise those power resistor worth nothing from this point. That should be not a supply node but a common shared limiter resistor aso. I think is also too much for el34 with 560v plate. It will burn, but more slowly at least. The voltages on the screens are determined by this chain of resistors into dynamical conditions. But they used different tubes.so different screens can have different performances. What.s matter is the max power it support. For this reason you have to measure first at max power and fit the proper limiter in respect with max admissible value. say 12w for el34.no more.
Cheers. Catalin

[nsubulysses]

Please tell me if I have done anything wrong.

Voltage drop over 1.5K screen grid resistor is 3.5-4.5V per tube at idle and and 58-65V at max volume, essentially the amp pushing a signal so big it gets distorted to pretty much a square wave.

So worse case scenario 65/1500 = .043A
.043 x 65 = 2.816W

Or 65^2/1500 =2.816W

I get idles for cathode current that vary a little but not TOO bad. On one side I have a 27mA and a 17mA tube, the other side a 22mA and a 23mA tube

At clipping they all do about 125-135mA. At full throttle, essentially a square wave they do 180-200mA. This is measure over a 1 ohm cathode resistor I installed on all tube sockets.

[catalin gramada]

You say have over 1.5k resistor 65v when dimmed. that means 65/1500...43.3333 mA. but also you have to know the voltage in the g2 screen in those moment. 0.0043333 x screen voltage will give you exact the power disipate on those screen. So please repeat the test at full power and record the voltage dropped over 1.5k and the voltage on the g2 pin
You have to multiply the current with the g2 voltage. that means between g2 pin socket and ground
What you already measured was the power dissipated by resistor. Your interest is the power by screens so mesure screen voltage when dimmed directly to the tube socket

[nsubulysses]

At idle screen voltage is 569V

clipping 462V

full volume 410V .......x .043 = 17.6W

Catalin I don't understand what you mean by saying the screen supply should not have its own node. here is a crude schematic. B+ and screen supplies are totem poled but I didn't feel like drawing it.

[catalin gramada]

Well is over double the value recommended. You give no chance to that tube.
Now. best solution is to provide a separate low impedance supply decoupled with a low filter cap- 10uF or so. But that.s me. Limiting with big resistors will give you a very particular squashed sound. You have to try to see if you like it. But first thing I would do is to remove the filter cap between those big 470 ohm/7w and individual 1.5k resistors. Take the screen voltage directly from plate capacitor. Then big 470/7w then next node individual 1.5k screens. It will survive for a while
Wait a sec to find something to write in my cabin. I will draw a sketch for you.

late edit...
In respect with your sketch do that -

If you have not a choke replace with 100 ohm power resistor
PS. I.m very sorry you did not put a schematic and supposed you used v4 layout

[nsubulysses]

Hmmm, with 400 ohm 10W resistor installed in series with screen node as per your schematic gets me to 12W max dissipation at screen grid

50V over 1500 ohm screen grid resistor. At full volume screen voltage drops to 380.

So, 50/1500 = .033A .033 x 380 = 12.66W

Power output drops to 21V, 21^2/4 = 110W at clipping

My sine is a bit skewed now why is that?

[catalin gramada]

Hmm. ..is weird. Can you try a small cap in the node between 400 ohm and 1.5k resistors? Two 16u or 22u in series and see if looks the same ?
Also try to balance as much. The pair of tube from a phase should draw the same amount of current like the other.you have to match in this respect first.that means two by two then adjust the bias properly
I have no idea why you sine is tilted.I doubt the shared screen resistor have something in common with that.but can check fast if you put a cap in those node
Also check the signal balance of you inverter/driver stage.out of balance=power loses

[nsubulysses]

Catalin, you really helped me a lot. thank you so much.

For some reason I had a 3.9uF/630V film cap laying around so I used it to decouple the supply between the series dropping resistor in the screen supply and the 1.5K screen grid resistors. I tried a 300R and a 500R resistor and with the 500R I was able to get it below 12W max screen dissipation. Sine is symmetrical now as it should be.

For 6CA7 I got 9.8W screen dissipation at clip and 12.95W at full volume with 300R
8.75W at clip and 11.34W at full volume with 500R

It seems like my choke was kind of superfluous since it just made my screen voltage and screen current higher during idle and while driven, just killing the screens more. I wonder if I could just do a dropping resistor, say 500R, in place of choke and a smaller dropping resistor, maybe 200R in series with the screen supply if I were to build the amp again to cut down on cost, and still get similar results.

Since the new JJ 6CA7s I originally used were a little wore down and only doing 21-22V into 4 ohms I put in the old poorly matched GEs. They do 24V into 4 ohms. I will crank the amp full volume at practices and shows and what not and see if it maintains this full power output. I hope it will.

[catalin gramada]

No. The choke is not absolutely necessary but you don't want to drop so much voltage there into dynamical condition otherwise the operation points of your stages can vary to much. For this reason a ''parallel'' supply rail just for the screens is useful. You can keep under control the screens without to affect the rest to much. Well.if you can.t afford a choke or don.t want to use one a power resistor is the key. 100ohm or so...but beter to check yourself how much voltage it drops from idle to max and to decide yourself if is enough or not. Cheers. Catalin

[nsubulysses]

Another question then, if you or anyone else doesn't mind answering, since you recommend 100ohm or so dropping resistor. Perhaps I am misinterpreting something or have the math wrong.

I have a screen supply node of 110uF. So with 100R resistor this creates --
1/(2 x 3.14 x 100 x .00011) = 14.47Hz

I mentioned 500R dropping resistor in place of choke becasue
1/(2 x 3.14 x 500 x .00011) = 2.89Hz

I don't understand if this is significant to noise/hum rejection for the PS or perhaps not because these calculation are centered around a choke, not a resistor? My choke has a 65R DC resistance but as I understand it, a large (in the thousands) ohm impedance to AC, which is why chokes have better hum rejection but also better supply regulation as the DC voltage drops less. Is this correct? I assume it is not in some regard.

[catalin gramada]

In the moment you decoupled with those 3.9u those point becomes your supply screen node. The screens feding directly from this 3.9u cap
Those 100ohm is part of rc filter for all amp stages except the power anodes.
Of course 500ohm comes with significant noise reduction.but you have to put in balance what you win and what you lose. A 500ohm will come with a poor voltage regulation. For this reason we use chokes. Low dc resistance for better voltage regulation and high ac reactance for better ac ripple rejection
I said 100 ohm just to keep a minimum decoupling between stages I didn.t calculate nothing.You.re free to use some as large as 1k or so.but a choke is a better solution
If you are concerned by ripple reduction.here a link to a calculator. If scroll the pages are also for lc filters
https://www.ampbooks.com/mobile/ampl...ripple-filter/
The problem with those calculator it have just 50 and 60hz frequency options so values are not nominal for full wave rect
But I will not be worry to much.In your actual situation 100 ..or 500 ohm and 110uf is a suplimentary filter stage between main capacitor and individual decoupling stages so I think the ripple is not a real concern. I would be worry more about voltage regulation aspect instead
This is simplistic way of thinking cause the math is over my head a bit but if you catch the idea the calculation can confirm if works or not. I.m more 'measure and cut'

[nsubulysses]

This came out pretty good thanks for all the great suggestions. Preamp changed a lot in the process! It's still going but it's 95% there.

500R in series with screen supply was big thanks catalin

[nsubulysses]

Well is over double the value recommended. You give no chance to that tube.
Now. best solution is to provide a separate low impedance supply decoupled with a low filter cap- 10uF or so. But that.s me. Limiting with big resistors will give you a very particular squashed sound. You have to try to see if you like it. But first thing I would do is to remove the filter cap between those big 470 ohm/7w and individual 1.5k resistors. Take the screen voltage directly from plate capacitor. Then big 470/7w then next node individual 1.5k screens. It will survive for a while
Wait a sec to find something to write in my cabin. I will draw a sketch for you.

late edit...
In respect with your sketch do that -
[ATTACH=CONFIG]41430[/ATTACH]
If you have not a choke replace with 100 ohm power resistor
PS. I.m very sorry you did not put a schematic and supposed you used v4 layout

Hi Catalin, you have really helped me a lot with this build and now going through some of these things again I have two questions for you if you don't mind lending your insight. Thank you for your help so far.


when you say remove the filter capacitor for the screen supply and take the screen supply directly from the plate supply what is the advantage to this? Other than, It seems I could use a much smaller choke because screen current is no longer flowing through it and I would also have 2 less caps in the power supply. That seems good, but why is this screen supply node unnecessary? Because it gets decoupled by the 10uF cap anyway? If that is the case and my original screen node was 110uF wouldn't it be better to have a higher uF node for since my amp is for high gain sound? Or is it unnecessary and plate supply node is fine for both plate and screen?

Maybe I will experiment around but I overlooked this part of your post in the past and my current layout is as follows

power supply/output

[catalin gramada]

Hi. in the drawing above the second cap is your screen cap. What I sugested was a parallel rail suplied from first cap but with a smaller value cap eg.8-16uF for what I felt to work more dynamicaly operation if you want. You don.t realy need a 'stiff' suply point with over 580v into the el34 screens. But if you like you can try kt88 (6550)instead like Mesa did. Check 400 bass schematic. 540v into the screens with 2.7k screen limiters.
When they goes to next 400+ version 6L6 equiped they also changed to more friendly 480v supply





I remember I said less filtering will be my first choice and is a matter of taste for guitar aplication. If you looking for "high gain" sound ,as you said, stiff screen supply rules, you're right
From you sketch can reduce-reduce-reduce those 500 ohm resistor till you eliminate completly but have to raise the 1,5k resistors value accordingly. There are not guarantees the EL34 screens not fail at 580v arround even you bias very cold. KT88 is more robust from this point.
The screens draw very little current at iddle. If you share a node you have 4x factor, means a chance to drop some voltage over a common resistor to safe opperation (means max acceptable designed value)- then, individual resistors to limit curent under dynamical condition. These, in conjunction with common resistor ,will keep the screens in safe max power disipation limit. So, if you create a suplimentary decoupling point for screen ,like you did with those 10u, you can put a large value power resistors in between to drop more voltage over it in respect with current draw by screen, then raise the 10uf cap value to 50-100u, then dimension the individual resistors to not overdisipate into max condition
You are constrained to get a type of "sound" from reallity you have and I don.t see so many options here. Maybe some skilled gents on this Forum have a better view.
Cheers

try put large grid stoppers if you prefere like that. It works out of problem in one of my amp with 580v in screens in UL mode but with kt88


But , from you measurements, you really want to drop 200v arround for screens. 5,6k grid stoppers I think you'll need