The Latest Push-Pull Build

[Cressy Snr]

I've sorted it now.
I decided in the end to just use the screen volts as they were, rather than mess around with the statistical regulator zener chain. It then became easier to sort out the op point. I have enough experience with pentodes by now to be able to make reliable judgement calls when I don't have the right characteristic curves.

The short answer was to leave the screens alone and simply juggle the cathode resistor value until I got the plate voltage in the right ballpark to centre up the operating point at approximately half the avalable HT, then check that the bias voltage made sense in that context, then check that the cathode followers were still operating properly.

It all fell out nicely. I ended up with each input pentode running at -2.8V bias at 2.8mA with 156Va and the 7044 cathode followers with 126V across them and biased at -6V and an Ik of 8mA. The plate voltage is now only 6V shy of centre with an HT of 300V; close enough I think.

Feedback is now tweaked up to where it should be with a parallel combination of 1M and 470K, 2W carbon film resistors.

Could now be ready to draw something up.

[Cressy Snr]

Here is the new schematic:



Wholesale changes have been made to the front end and these have been very worthwhile.

This circuit and the associated power supply, which will follow later, has been thoroughly tested and auditioned and gives excellent performance across a wide range of musical programme, doesn't it Mr Cholmondley Warner?

It certainly does Mr Grayson!

[Cressy Snr]

Oh dear,
That slow build to heart melting climax about two thirds of the way through Barber's Adagio for Strings, just frightened the life out of the dog. He climbed down off the back of the sofa and ran across to sit on Melanie's knee, with ears back and a very sad look on his face. Poor love; he normally enjoys classical. Had to throw him a Bonio to perk him up a bit.

Must remember to watch the volume.

[Cressy Snr]

Power Supply.

I'll leave these drawings for a while before putting them in "Stacks" so that any errors found can be corrected.
I think the PSU and audio circuit drawings are correct but best leave it a while for the panel to look over.
That's it now for this one. It's turned out well.
The radiogram project is now the one I'll be concentrating on.

[Mike H]

Sorry couldn't resist ..

[Thermionic Idler]

I guess if one were making mono blocks, you could substitute 3 x 0A2's for the VR tubes - 1 for the 6BS7 anodes/CF's and 2 for the 6BS7 screens - cos the current would be half I think?

[Cressy Snr]

I guess if one were making mono blocks, you could substitute 3 x 0A2's for the VR tubes - 1 for the 6BS7 anodes/CF's and 2 for the 6BS7 screens - cos the current would be half I think?

Yep.

[Cressy Snr]

A few minor mods to the circuit are underway in that MJ Statistical regs are currently being assembled
for precise regulation of the output stage screen grids.

All the wrinkles have been ironed out thanks to plenty of peer review and comment over the past two years.
We now have an excellent project.

The final circuit can be found in "Stacks" under "A Good Pentode Push Pull Amplifier." I created a new thread in there to avoid mixing this lovely circuit in with the
inferior versions

[Paul Barker]

Thank you?

[Cressy Snr]

Thank you?

Well that depends

[Cressy Snr]

Now that the front-end rebuild with the 6BS7, EF86 equivalents has settled in nicely, I spent a profitable couple of hours yesterday, sorting out all the currents and voltages throughout the amplifier.
In the end I didn't use the Statistical regs on the output stage. I have other plans for them as part of a possible line stage build.

Sorting out one or two resistor values in the power supply take off, I was able to get an exact 300V for the input stage HT, which gave me a spot on 150V at the 6BS7 anode; exactly half the HT value, which was nice.
Turning to the output stage, a couple of series resistors in the screen grid feeds, knocked down the voltage to 226V, which put the screens at exactly half my 452V of HT. Again that was nice.

There is what some call a "myth" that screens should be at half the HT to get best performance in terms of sound quality out of power pentodes in a hi-fi amp. I would say that this seems to be the case. However I would qualify that with the thought that the myth purveyors were probably talking about class AB, with high HT voltages and low current idling, NOT Class A operation, where having the screens at half HT would not make sense at all with your average KT88 or EL34.

I also stood the bottom of the output stage screen shunt reg zener string on a 10R resistor, making for easy current readings.
Measuring the current between the series resistor input and the top of the shunt, gave exactly 40mA, and at the bottom of the zener string there was 20mA, meaning that the output stage screens were drawing bang on the book value idle current of 20mA (5mA each). Again, a nice little outcome.

I then set the bias, for the first time knowing the exact contribution the screen currents were making to the total cathode current through each output pair. That was novel
The bias was set well towards the class B end of the chart with 35mA at each cathode giving 1.125W of idle screen dissipation and 13.5W of anode idle dissipation per valve. The 6CA7EH valves ought to have a long, easy life under those conditions. Oh yes, I'm back to them. They really are a steal at the price, having a clear, clean tone and a fine overall balance with just the right amount of warmth to let you know you are listening to valves.

The net result of all this "optimisation" work is a relaxed, musical, easy, low distortion sound, that makes for all-day listening with nary a hint of listener fatigue. It can kick booty when required and I reckon it has a good, honest 25WPC output.
Not much for an authentic pentode push-pull amplifier, you may say, but I've gone, with this design, for low distortion rather than sheer output power.
With the 6K a-a impedance of the OPTs, and the limitations of the mains transformer and choke, it wouldn't have been possible to go all out for power anyway, so why try? I have gone for the Quality outcome, working with the OPTs to gain reasonable power and low distortion, rather than trying to force the issue. Late in the day I have made a decision about what I want the Barbarik to actually be, rather than try to make it into something it cant be; not with the iron I have anyway.

The upshot is that the design will need to be paired with efficient, easy-to-drive speakers of around 93dB sensitivity, with a nominal 8 Ohm impedance that stays reasonably level across the frequency band. 4 Ohm systems? I can't say yet without trying some.

[Cressy Snr]

Full circle, right back to the start.



I haven't used EL34s in any recent versions of this amp; abandoning them early on for big power tetrodes.
Pleasantly surprised is the expression I think.
Running an HT of 465V and 232V screens at 35mA per valve, they really do the biz on vocals, brass and guitars, both acoustic amd electric.

The conclusion we can draw from this project is that with EF86 or equivalent input stage and 300-330K feedback resistors, it's a pretty versatile amplifier now the output stage is being driven properly by those 5687 type cathode followers. Simply choose the output valve you prefer, bias it up and balance the input and output and it'll play anything you want it to.

[Mike H]

There is what some call a "myth" that screens should be at half the HT to get best performance in terms of sound quality out of power pentodes in a hi-fi amp. I would say that this seems to be the case. However I would qualify that with the thought that the myth purveyors were probably talking about class AB, with high HT voltages and low current idling, NOT Class A operation, where having the screens at half HT would not make sense at all with your average KT88 or EL34.

Oo-er haven't heard / seen that. If I remember correctly tho there is a case (for 100W from single p-p pair) where EL34 can be run at quite a high anode Voltage, but the screen can't, that's the only thing I can think of where that idea might have come from. Otherwise I would be guided by the datasheet.

[Cressy Snr]

Aye, the high voltage thing is probably where I saw it. A lot of the EL34 data sheets put the screen at the same voltage as the anode, when running class AB, pentode mode, but then you need a lower impedance OPT (typically 3.5K a-a) than I have, in order to get the loadline steep enough to pass through the knee of the 0V grid line, which with 400V screens is way up in the air. Additionally you are over dissipating the valve every half cycle at anything above a whisper. Of course, being push-pull, this will average out so you don't blow up your valves, but personally, I don't think it is being very nice to them.

I have a 6K a-a OPT so I end up with a shallower loadline, which means a lower screen voltage to bring the 0V line down to get the knee at the right level. Putting the HT well over to the right hand side of the chart and at a low current gives me plenty of class B power, with low heat generation; far better IMO than thrashing the things senseless, and like Bruce Rozenblit has observed, there is negligible difference in sound quality between this mode and class A, providing care is taken to minimise high odd order distortion products. It just so happens that the best operating conditions fall with the screen volts at half the HT. So with higher impedance OPTs and shallow loadlines running almost class B, the half HT screen volts rule seems to work nicely. Running El34/6CA7s like this seems to sound good too.

[Thermionic Idler]

Presumably this design could be adapted to drive two paralleled pairs of EL34 in push-pull if one found themselves with less than efficient speakers (say, Harbeths) and needed a bit more welly (say 90W per channel). The CF buffer should cope with that shouldn't it?