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

[nsubulysses]

I drew soemthing too to verify it with you but now I see you edited your post to add a picture



It seems like my choke is unnecessary with the dropping resistor also. Bigger screen grid resistors seem to limit my max power output I think but I will mess with it in conjunction with lowering the screen dropping resistor value to see if I can get a similar result. Maybe I can just use choke and 5.6K grid resistors or so, and eliminate the dropping resistor?

[catalin gramada]

I feel obligated to warn you again. It is just a suggestion and never try it: I don.t believe a EL34 (6CA7) will not fail with more than 500v into the screens. It is better to ask someone who really knows. There are not big differences from a voltage point of view feeding the screens from second filter cap-what you did, or parallel rail. You will have a supply point close to the plate voltage which I consider dangerous for el34.(1k5screen resistors will not drop arround 500 v at idling.)
Why is happen that? cause at idling the screens draw very little current. I.m sorry if I cannot be clear....You screens no need to see more than 480-500v at idling for el34 and 380v full power(in respect with your plate voltage).
Take my considerations with a reserve, I.m not a tech. educated guy, it comes from a limited experience.

[nsubulysses]

I am really seeing what you mean now as I mess with the capacitance of the screen node in relation to screen grid resistors and finding the balance.

Increasing screen node capacitance from 10uF to 110uF with 1.5K screen grid resistors stiffens the supply a lot and puts it back in screen killing region. With 110uF screen voltage drops from about 580V at idle to 460V at clipping and does something like 10.5W dissipation on the screens. It will do about 14W screen dissipation full tilt, essentially doing a square wave output.

When I had the 10uF node with 1.5K screen grid it would do something like voltage drop from 580V at idle to 375V or so at clipping and do 8W screen dissipation at clipping and about 11.5W max output, square wave output.

With 110uF node the supply is too stiff to be nice to the screens. If I increase screen grid resistors to 4.7K to try to compensate it is still too much dissipation for the screens (10W at clipping, 14W full power output) and it also draws my power output for the amp down from 144W at clipping to about 115W.

So maybe the best balance is what I had, 10uF and 1.5K screen grids. Or maybe something like 22uF 2.2K screen grids. I tried 1.5K, 1.8K, 2K, 2.2K, 2.5K, 4.7K and it seems like if I go over 2.2K the amp puts out less power. I will have to do some arrangements and then do the math. So I can get 140-150W at clipping and still keep the power tubes in happy screen dissipation region.

I know it can work out. I was running this amp for the last few months full volume all the time and checking power output every week or two to make sure it was staying at about ~24V (144W) into 4 ohms and it was. That was with the looser 10uF screen supply and 1.5K screen grid. The power tubes didn't seem to be getting too taxed. The JJ logo remained bright red, not darkened at all

Now I just have a better power supply arrangement with less components since I have the parallel supply for the screen coming from the B+

[catalin gramada]

I just thinkig... you can use a divider with a price. Use a paralel rail for screens as I show and a voltage divider 2,7k series between plate cap and screen cap and 22k over 100UF screen cap will drop screen voltage arround 500v. you will need a 25w paralel power resistor. Take a look of bassman 135 schematic.

[catalin gramada]

Yes, is a thin balance. If you put a capacitor between whatever screen supply point you choose and screens load you will have not a true limiter network cause those cap will try to compensate. too big and get a stiff supply point, or to small and not decouple enough. the result is more evident in low freq range when the cap can or cannot sustain big energies cause it cannot topped up in respect of current necesities
you should have not care by plates. 600v for el34 plates is parfume. The screens are more sensitive. For unknown reason kt series seems to be more robust from this point. I recomanded to you kt88 and have to reconsider. Is wrong cause need more driver capabilities. Maybe kt66 or 77

[nsubulysses]

Hi Guys

The Custom Special (160W, 4-tube) and Bass Master (80W, 2-tube high-V; 40W low-V) both began as plexi circuits - like the high-voltage Marshalls of the day - nothing special. The only variation was the suppressor connection and the shared screen resistor, both hifi methods. The CS and the later Super CS had more elaborate preamps with passive and active EQ. TUT3 shows all the variations in schematic form, and these schemos are available on the web, too.

Regardless of your Traynor experience, the transformer set cited above will deliver 160W with no strain from the tubes. See the technical Article on my site of the expanded Hammond data, which includes a list of OT-PT pairs that work together..

have fun

Well the amp that might as well have been the example for my power supply is here. Traynor Custom Special. I've only seen one of these in person ever and fittingly it was when I played a show in Toronto.

This one does 160W into 8 ohms with 2 Tesla EL34 and 2 Sylvania 6CA7. I overlooked the shared screen grid resistor for the power tubes. What is the proper way to measure screen grid dissipation on this type of arrangement.

If there is a 1K 20W dropping resistor in the screen supply and the screen grid resistor is shared, it actually just looks like one power tube gets a 1047 ohm resistance in series and the other gets 1000 ohms.

Measuring voltage drop over 47 ohm screen grid resistor at clipping I get 1.84V / 47 ohms + .039mA. 423V at screen grid at clipping is .039 x 423 for a whopping 16.38W of screen grid dissipation for one tube. If it is the case that this is incorrect and must be divided by two to gauge actual screen grid current for each tube how is that so? 8.29W per tube ????? the 47 ohm is NOT in series with one power tube per push pull?



Measuring from screen supply side of 1K 20W resistor to screen grid pin I get about 98V voltage drop per side of push pull.

So, assuming that is the total drop for two tubes I get 98V / 1047 ohm = .093mA / 2 = .046ma per tube. and 424Va SO .046 x 424 = 19.84W of screen grid dissipation

This amp puts out 160W but it seems to be VERY stressful to the power tubes unless I am understanding and calculating something incorrectly. This PT in the custom special is also about 1.5X the size of my Hammond 278CX which makes sense becasue shortly after I bought it I realized 278CX only has 6A heater current rating. My 4 6CA7 is already 6A, plus a few preamp tubes and yeah......... still slays but probably not a good arrangement in the long run.

[g1]

Did you bias according to note 4 in the schematic?
Bet that will reduce your screen consumption.

[nsubulysses]

I biased it to 22mA at 580Va.........not at the shop right now

but also G, was I measuring screen grid dissipation correctly ??

[nsubulysses]

Obeying G1 and note 4 on the schematic and biasing the amp to 6.3V (closest I coudl get with resistance subsitution box) over 1K 20W shared screen grid resistor I get about 13mA at idle per tube. Now the amp puts out 33V into 8 ohms for 138W at clipping with a bit of a crossover notch, or maybe it is just a bias shift and that's normal.

Measuring screen dissipation is slightly different because I noticed my 1K 20W resistor was actually only 470 ohms. All the writing had worn off from heat I guess.

Replacing with 1K I get 76V measuring from screen supply side of 1K resistor to Pin 4 on the power tube.

So 76 / 1047 = .072mA

.072 / 2 = .036mA per tube

At clipping I measure screen voltage of 399V, so .036 x 399 = 14.48W screen dissipation per tube.

Ummm, so is the idea that this PS/output section arrangement runs the tubes with no strain internet BS. Am I missing something and measuring incorrectly? It does not stack up with the measurements because max screen dissipation on the datasheet is 8W. KOC has 1000X more experience than me so I want to believe his claim. I've been wrong before, as the saying goes. Running 4 power tubes at very high voltage for very high power with NO strain on the power tubes seems like a pleasant fantasy that I want to understand more, if it is possible this could actually be true. In reality it seems not so, unless I am ignorant

[nickb]

Obeying G1 and note 4 on the schematic and biasing the amp to 6.3V (closest I coudl get with resistance subsitution box) over 1K 20W shared screen grid resistor I get about 13mA at idle per tube. Now the amp puts out 33V into 8 ohms for 138W at clipping with a bit of a crossover notch, or maybe it is just a bias shift and that's normal.

Measuring screen dissipation is slightly different because I noticed my 1K 20W resistor was actually only 470 ohms. All the writing had worn off from heat I guess.

Replacing with 1K I get 76V measuring from screen supply side of 1K resistor to Pin 4 on the power tube.

So 76 / 1047 = .072mA

.072 / 2 = .036mA per tube

At clipping I measure screen voltage of 399V, so .036 x 399 = 14.48W screen dissipation per tube.

Ummm, so is the idea that this PS/output section arrangement runs the tubes with no strain internet BS. Am I missing something and measuring incorrectly? It does not stack up with the measurements because max screen dissipation on the datasheet is 8W. KOC has 1000X more experience than me so I want to believe his claim. I've been wrong before, as the saying goes. Running 4 power tubes at very high voltage for very high power with NO strain on the power tubes seems like a pleasant fantasy that I want to understand more, if it is possible this could actually be true. In reality it seems not so, unless I am ignorant

[ATTACH=CONFIG]44234[/ATTACH]

You can't calculate it that way. What you need is both the screen volts and current vs time and then integrate their product.

[nsubulysses]

Thank for helping me in the right direction. How would I get the screen volts other than probing the screen grid pin with DMM? I'm sorry but I don't know what you mean by current vs time and integrate their product. Measuring voltage drop over resistors in screen supply will get you the current draw (I thought), but how does time get factored into this measurement?

[nickb]

Thank for helping me in the right direction. How would I get the screen volts other than probing the screen grid pin with DMM? I'm sorry but I don't know what you mean by current vs time and integrate their product. Measuring voltage drop over resistors in screen supply will get you the current draw (I thought), but how does time get factored into this measurement?

The problem you are facing is that power = volts X current is only true when they are constant. Here, they are varying.

You could use your scope with a suitable probe to get the screen voltage over one cycle. Current is harder. If the screen supply is constant over the cycle you can simply calculate it from the drop across the screen grid resistor. If not you'd need have a current probe. This with give you a plot of voltage and another of current over time. You can approximate the integration of these by dividing into say 100 equal time intervals, find the product of voltage and current at each internal and then adding them all together and dividing by 100. As you can see, it's going to take some work. If your scope can dump the data into text file you can then use a spreadsheet to ease the process.

[nsubulysses]

Sorry for all the questoins but I really appreciate it. It really does further my understanding.

However, it makes a question come to mind. if I have master volume on 10, tone knobs left untouched and I bring the amp up to clipping with the signal gen, isn't this pretty constant? Maybe it's wavering very slightly but it seems to me that the voltage drop over screen grid resistor and voltage at screen grid pin vary by less than 1 volt when the amp is put up to clipping and I don't touch anything else.

With a very slight variation wouldn't my method would be 95% accurate, or some other high degree of accuracy?

Or perhaps waht you're getting at is that the cycle is important because it's push pull and the one power tube I'm measuring is cycling on and off so my measurements with DMM are not accurate? BUT with a sine wave isn't it not cycling on and off?

[nickb]

Your concern is max screen grid dissipation. This happens when driven hard and the plate voltage drops. As the plate drops so the screen grid current rises rapidly and in a non linear fashion. So in this case the screen current looks like this:



So, with that in mind and the screen voltage varying, I expected that you would need to do the calculation properly. But...if you take the product of the average screen current and average screen voltage the answer is surprisingly very close. Well, at least I was surprised

I think what made me think about the method and question it was the very high dissipation you came up with. So going back to your data of 36mA per tube, I suspect this is off. For this amp I'd expect average current should be more like 10mA. The peak will be more like the 36mA - is that what you measured and used instead of the average?

[nsubulysses]

When I measured screen dissipation I had the amp putting out full clean power, meaning amp cranked to the onset of clipping.

So if I understand what you mean, yes this could be called the peak since it's max clean power. It is not the true peak of screen dissipation becauses it's only cranked to the onset of distortion, not totally distorted.

I popped open my amp again and measured.

Putting the amp at the onset of clipping I get 33V output into 8 ohms

measuring voltage drop only over 1K screen grid resistor
21V /1000 ohms which is .021mA
screen voltage is 410V, so .021 x 410 + 8.6W of screen dissipation

measuring voltage drop over 1K screen grid resistor and shared 1K/20W resistor in screen supply I get
101V / 2000 ohms = 50.5mA / 2 = 25.75mA per tube
screen voltage is 410V at this setting so 25.75mA x 410V = 11.275W screen dissipation

at idle measuring voltage drop over 1K screen grids only
.75V / 1000 ohms = .00075 x 582Va = .44W

at idle measuring voltage drop over 1K screen grids and shared 1K 20W resistor in series with screen supply I get
5.5V / 2000 ohms = .000275 x 582V = 1.6W

so yeah I'm not averaging it. Just gauging screen dissipation at onset of clipping to compare the harshness to the tubes of these two amps