SimpleSE goes Push Pull

I got put on a killer project at work recently so I have had no tube time. The boss asked me how much time it would take to do a particular project and I said 2 to 3 months. He said I had 3 weeks, I negotiated for 4. I have 3 weeks left, so I won't be here much. Most of my friends at work have been laid off, so I can't complain.

I was suffering some serious tube withdrawals, so I decided to spend my afternoon blowing up some tubes instead of mowing the lawn (we are having record breaking heat). I gathered up a couple of 22 volt sweep tubes, which I planned to wire into a SimpleSE board. I connected a SimpleSE board to a Fluke 407D, and put in my trusty pair of well worn 6L6GC's to test everything out. Then I had a brain flash. Why couldn't I make a push pull amp out of the SimpleSE board? It turns out to be pretty easy, but as with all simple experiments the snow ball started rolling, and my experiment that started at lunch time just finished (10PM) and I never touched the sweep tubes.

The first attempt at making a P-P amp involved using a driver transformer with a center tapped secondary, The UIS-1 from Triode Electronics. The center tap goes to ground and the hot leads go to each channel inputs. I connected a P-P OPT up to the OPT terminals of both channels. When driven with a signal generator the results were good. 20 watts in triode mode, 35 in UL. I connected this up to my CD player, but it sounded bad, real bad. My CD player didn't like being connected up to the transformer, so this experiment was over. I went into the kitchen and got lunch.

I thought about it while eating, and then tried this. I decided to feed the CD player into one channel like normal. Then I made a voltage divider out of the grid resistor on that channels output tube to drive the other channel, sort of a simple paraphase splitter. I replaced the 220 K grid resistor with a 200K resistor and a 20K trimmer pot in series, and wired the wiper to the second channels input. It worked! It worked GOOD!

I was using an 6K ohm OPT that I got from Handwound Transformers (the Lucas guy that ripped everybody off) that has never really sounded good, but I didn't care if I blew it up. I tried triode mode. With 450 volts of B+ I could get about 18 watts at clip. The measured distortion was low, 0.5% at 1 watt and 1.3% at 10 watts. I took out the Chinese 6L6GC's and installed EH KT88's. Power went up to 23 watts at clip. I switched to UL mode. I could get 40 watts at clip. Distortion was 0.2 % at 1 watt, and under 1$ at 20 watts. OK, I cranked the power supply all the way up, 552 volts. Clipping happens at 62 watts. The distortion numbers were even lower. I tried cathode feedback, which lowered the distortion even more, 0.165% at 1 watt, and 1% at 40 watts. The output impedance measured about 2 ohms without any GNFB.

I connected up the CD player and speakers. It didn't sound too good. I disconected the CFB. OK, much better. It sounded pretty good, much better than I expected. Good solid bass, but clipping sounded pretty gross, then again why do you drive a 60 watt amp into clipping?

I have been exploring several different P-P amp designs that will eventually become the SimpleP-P. I really want to come up with a simple P-P amp that sounds a lot like a SE amp with the bass, punch, and dynamics that a P-P amp can deliver. I know that It can be done, because I have an amp with that sound, but it is not simple or low cost (300B P-P). This amp has the simple part down, but the sound is not there....yet.

One more picture of the mod.

OK, now I am curious, and I haven't blown anything up yet. Could I manage to rewire the front end of a SimpleSE into a traditional P-P input stage? Didn't sound too hard. Just rip out a few parts, sky wire in a few parts, and now I have a typical input stage, feeding a split load phase splitter. The measured performance was similar to the previous mod, except the output impedance was about 3 ohms, I don't know why. I played with this for a while, but I couldn't make it sound as good as the simple mod. Again CFB sounded gross, and triode sounded better than UL. Time for something really wild!

Now that I had made a mess out of a perfectly good working SimpleSE board, I decided to try an idea that has been percolating in my brain for a while. I have recently re-read the paper written by O.H. Schade of RCA in 1938 called "Beam Power Tubes" after reading an article in Audio Xpress by Pete Millett. Both of these articles extoll the virtues of applying local feedback from the output tubes plate to its grid circuit. The claims are triode characteristics with pentode power output. For this to work you need a driver with a relatively high output impedance. This usually implies a pentode. My board has a dual triode. I remembered a post (#54 in this thread http://www.diyaudio.com/forums/showthread.php?s=&threadid=120888&highlight=
that showed a LTP / mosfet cascode. I decided to wire one of these up and apply some feedback to it.

It only took me a few minutes to blow the whole thing up, so I got to do it all over again. This is really an experiment in its initial stages, since it has been alive for about 2 hours. I am running the amp in pentode mode. It makes stupid levels of power, over 75 watts at clip. 4% distortion at 70 watts. 0.2% at 1 watt, output impedance is under 2 ohms without GNFB.

OH, It ROCKS! This amp plays everything that I throw at it very well, and verry LOUD! I played all of my usual test music and it sounded great on all of it. If this thing survives my usual torture tests, and it can be reliably and repeatably built, this could be the SimpleP-P. With 450 volts the power is about 40 watts, with 550 volts power is 75 watts. At 70 watts the distortion is 4.12%. When you pound it into clipping it just gracefully compresses. It never gets nasty. This picture was taken at an input level that should have produced about 90 watts.

I plan to develop this design further, as my schedule allows.

Any chance you have a link to one, or both, of these, or something postable?

Click to expand...

A link to the Schade paper is in the previous post. The interesting voltage feedback stuff is at the very end of the paper. This paper was essentially RCA's introduction of the 6L6 tube.

The Audio Xpress article can be found on their web site here:

http://www.audioxpress.com/magsdirx/ax/addenda/media/millet2903.pdf

I have been experimenting with a P-P amp that uses a similar technique using 6AU6's in LTP to drive EL84's. There is bunch more experimenting and testing to be done here before anything is ready to be called a TubelabP-P. I tend to post my experiments in case someone else wants to enjoy the smell of burnt parts!

I currently have 150K ohm resistors from the output tubes plate to the mosfet drains, which have 20K resistors to B+. Other configurations remain to be tested.

It was interesting to note that I tried the local feedback approach with the UL configuration. Adding the feedback actually raised the distortion, however the resulting signal sounded quite pleasing with music even though a sine wave looked ugly with up to 10% distortion.

Just make sure you listen to that LTP cascode with the gates ref'd to the cathode node. I've found it a distinct improvement...

Click to expand...

Thanks, I think that I have seen you write those words before. I will definitely try it. My last experiment Sunday night was to try injecting some feedback from the output tubes into the gates. The result was a pair of dead fets. It was far too late to be playing around with 550 volts, so I shut it off. I may get some time to experiment further in the coming weekend.

My last experiment Sunday night was to try injecting some feedback from the output tubes into the gates. The result was a pair of dead fets.

Click to expand...

I had about an hour to experiment today, so I replaced the Toasted Toshibas and retried the same experiment at a lower voltage. The reason for the blown fets is a 300 KHz square wave, so I decided to postpone these experiments until I had more time, and more mosfets!

Just make sure you listen to that LTP cascode with the gates ref'd to the cathode node. I've found it a distinct improvement...

Click to expand...

I can't do any listening tests today since Sherri has friends over, but I made some measurements. First off I found out that disconnecting the bypass cap results in the previously mentioned 300 KHz square wave, but no toasted fets this time.

Connecting the bypass to ground results in lower distortion levels below 10 watts (0.4% at 1 watt) but slightly higher numbers above 10 watts. The distortion increases monotonically as you increase power. This is typical for most amps.

Connecting the bypass to the cathodes raises the distortion below 10 watts (0.7% at 1 watt), but lowers it above 10 watts. The distortion is 2.0% from 10 watts remaining almost constant until clipping is seen. I get 2.1 % at 38 watts with a 450 volt supply, and 2.1% at 60 watts with a 550 volt supply.

I have always known that distortion numbers do not always correlate with sound quality. If they did we would all have sand powered stereos. This amplifier and the associated experiments have made this perfectly clear. I will try to do some listening tests on Monday.

The early (simple mod) amplifier with cathode feedback had extremely low distortion numbers (0.12% at 1 watt), but sounded horrible.

The experiment with UL mode created some funny looking distortion at high power but it was almost impossible to make it clip. The distortion went up as I cranked up the drive, no clipping, the sine wave kept getting fatter but stayed round on the top and bottom. It measured bad, but sounded great.

How are things going with the Simple P-P?

Click to expand...

There are really two of them, and neither has received much attention lately.

It looks like the next couple of weekends will be spent removing all traces of a name that sounds like "Simple SE" from my web site. The text was the easy part, Microsoft stuff actually worked. Now I get to manually edit all of the jpegs that have text in them.

The smallest Simple P-P is still in the breadboard stages, and I haven't really chosen the output tubes or the circuit topology. I like the sound that I get out of the 6AQ5's but I only get 8 to 10 watts. The idea of building the usual EL84 - 6BQ5 amp hasn't been ruled out yet. I can get 15 to 20 watts from them, and there is this other cheap tube that fits in the same socket....

I made some headway with the Simple SE to Simple P-P conversion (the big one) until my momentary lapse of reason last week, when I tossed it in the trash. The next morning Sherri talked me into recovering it before the trash truck showed up, but most of those skywired parts were broken off. I just restored the Simple SE board back to its original function, and used it to test the transformer that I shipped off to a forum member with a broken amp.

http://www.diyaudio.com/forums/showthread.php?s=&threadid=123790

do you have means to do an FFT on the distortion? I am curious as to what its spectral content is. Also I find the shape of the distortion v. power extremely interesting.

Click to expand...

Yes, I do. I use WinAudioMLS with a 24 bit 192 KHz sound module. I can measure some pretty minute details with that setup. For now it resides in the same computer that I use for the web site, photo editing and my documentation, so it makes it hard to do everything at once. I have gathered all of the components for a dedicated computer for audio analysis and music storage and retreival / playback. I just don't have the time to build it.

I found the "constant distortion" thing a bit odd myself and will investigate it further at a later date.

The success of the Simple SE has been because builders of all skill levels put it together and it works. There are no adjustments needed and the design tolerates substantial tube and component variation without loss of sound quality. For me to put the Simple P-P name on a design, it must do the same. In my limited tinkering time with the hybrid cascode LTP, I couldn't get there. At low supply voltages (300 to 350) tube variations can starve the mosfets of voltage headroom.

My latest itteration uses an additional input / gain stage in front of a conventional LTP. The extra tube takes up less space than 4 mosfets. I have started a test PC board layout to see if everything fits.

My work schedule barely allows time for breathing for the next two weeks. I hope to resume the P-P activities then.