CatSkin Class A2 SET with Secondary Flux Cancellation

[jazbo8]

Dave Slagle of Intact Audio was the inspiration for this, he advocated using a reduced gap (from what I remember) and a tertiary offset winding to provide a wideband transformer suitable for the higher Ra tx types such as 75TL.

Reducing the gap in a toroid transformer? Or was he talking about another type of OPT?

 

[316a]

Reducing the gap in a toroid transformer? Or was he talking about another type of OPT?

Toroidals have no gap . Reducing the gap of an EI transformer increases inductance which I believe was the intention...

316A

 

[jazbo8]

Ok, that makes more sense.

 

[BinaryMike]

Nevermind. I got it. Number of turns on the 25V tap is actually more than the 70V tap because i need to refer to the voltage to the primary side.. So to convert 6.32Vrms (for 5W@8R) to 70V needs 11:1 ratio while to the 25V only needs 4:1 ratio.

6.32Vrms * 70/25 = 17.6Vrms = 25Vpeak.. which makes your 40V headroom more reasonable though you can reduce it to 30V which would cut the device dissipation by 25%.

I wanted to allow for excursions well over 10W output so the CCS wouldn't have any involvement in clipping behavior. That puts minimum CCS compliance closer to 35V, hence the 40V supply spec. Power supply simulation suggests that I will end up around 38V with the Triad F8-28 transformer, so there's not much opportunity for improvement here short of going to a switch-mode PSU.

 

[BinaryMike]

again it would be interesting to find out the rationale for it, it may boil down to the cost of a real SET traffo vs his contraption...

There was certainly a desire to publish a high quality SET amp design that would be attractive to wealth-challenged DIYers.

I also wanted to explore the high-frequency performance of low-ratio OPTs, where cheap iron can actually gain the advantage. We know that leakage inductance is proportional to the square of turns ratio, and the $12 MCM transformer does exhibit very low leakage inductance. Its Lpss is only 3.9mH at 1KHz in the configuration used here (11:1). At 2W output, using 12dB feedback, the resulting frequency response is down only -1dB at 26KHz and -3dB at 49KHz. Squarewaves are excellent; see attachments.

 

[20to20]

I wanted to allow for excursions well over 10W output so the CCS wouldn't have any involvement in clipping behavior. .

The reason I asked why you chose the 6p36s is because I found even less info for that than the xGB5 types, but what I did find about it as triode strapped signal output power was typically in the 1.5W range. How are you getting pentode range signal power out of it as a triode?

 

[BinaryMike]

How are you getting pentode range signal power out of it as a triode?

Mainly by driving the bejeebers out of it. Most tube datasheets don't show what the tube can do when the grid is driven into positive territory. That's because grid current begins to flow and conventional driver circuits can't cope. The control grid effectively becomes the anode of a vacuum diode. Now that we're able to lean on solid-state devices for help in this area, there's no reason to insist on keeping the grid negative wrt cathode. The CatSkin driver has enough grunt to pull 6P36S anode voltage down to about 40V in triode mode, making anode current peak at nominally 160mA in the 1K OPT primary.

FYI: I swapped out the Sziklai pair driver for an FQPF2N70 MOSFET (with gate stopper and Zener) during circuit development. There was no detectable change in performance, and no change in parts count.

The attachment shows what happens at the onset of clipping. Speaker output is the yellow trace, the control grid is aqua, and purple is THD analyzer residual. The grid is driven about 16V positive including a sharper hump where anode current limits out. That's a result of feedback attempting to make the output signal match the input signal. The correlated dip in the distortion waveform is simply the result of a flattened peak in the output signal. Limiting looks a bit different on the negative output excursion because the tube is approaching cutoff in this case, and cutoff occurs more gradually.

 

[TonyTecson]

There was certainly a desire to publish a high quality SET amp design that would be attractive to wealth-challenged DIYers.

I also wanted to explore the high-frequency performance of low-ratio OPTs, where cheap iron can actually gain the advantage. We know that leakage inductance is proportional to the square of turns ratio, and the $12 MCM transformer does exhibit very low leakage inductance. Its Lpss is only 3.9mH at 1KHz in the configuration used here (11:1). At 2W output, using 12dB feedback, the resulting frequency response is down only -1dB at 26KHz and -3dB at 49KHz. Squarewaves are excellent; see attachments.

again good job, we would encourage you to carry on....

me, i am going to try out your drive circuit on a conventional SET...consider this as asking for your permission to do so...

 

[BinaryMike]

Please steal anything you find useful from my work. I'm transitioning into retirement and having the time of my life doing this stuff. No dreams of profit. Just hoping for a little notoriety!

 

[20to20]

Mainly by driving the bejeebers out of it. Most tube datasheets don't show what the tube can do when the grid is driven into positive territory. That's because grid current begins to flow and conventional driver circuits can't cope. The control grid effectively becomes the anode of a vacuum diode. Now that we're able to lean on solid-state devices for help in this area, there's no reason to insist on keeping the grid negative wrt cathode. The CatSkin driver has enough grunt to pull 6P36S anode voltage down to about 40V in triode mode, making anode current peak at nominally 160mA in the 1K OPT primary.

FYI: I swapped out the Sziklai pair driver for an FQPF2N70 MOSFET (with gate stopper and Zener) during circuit development. There was no detectable change in performance, and no change in parts count.

The attachment shows what happens at the onset of clipping. Speaker output is the yellow trace, the control grid is aqua, and purple is THD analyzer residual. The grid is driven about 16V positive including a sharper hump where anode current limits out. That's a result of feedback attempting to make the output signal match the input signal. The correlated dip in the distortion waveform is simply the result of a flattened peak in the output signal. Limiting looks a bit different on the negative output excursion because the tube is approaching cutoff in this case, and cutoff occurs more gradually.

So you're puting 80v p-p in, it looks like on the scope, 10v/div.. That gives 200v p-p out with 200v on the anode?

 

[dady]

Thanks for share gentlemen

 

[BinaryMike]

So you're puting 80v p-p in, it looks like on the scope, 10v/div.. That gives 200v p-p out with 200v on the anode?

Grid drive reaches about 80Vpp max, which yields something over 300Vpp at the anode. The 200Vpp (70.7Vrms) figure on the schematic is the nominal value for 5W output, ignoring transformer insertion loss and actual turns ratio.

Speaker line matching transformers are apparently designed to compensate for insertion loss by turns ratio reduction, assuming constant voltage on the transmission line. The 5W tap on a '70V' xfmr should have an impedance of precisely 1K, but the MCM xfmr offers a calculated 968 ohms instead. That means it will actually take 5.165W off the transmission line, or 3.306% over nominal power, if the load is precisely 8.000 ohms. I'm not sure any of us really care much about this, but there it is.