• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Tube modulated SMPS

Tektronix made a couple of scopes with tube based SMPS inside 503 with 6DQ6 is an example.

It’s occurred to me that the same analogue to switching topology can be used to remove the diode bridge or tube rectifier from the audio power path.

AC comes in, compared to the triangle wave to give a sw+ and sw- that can then be used to drive A set of bridge valves directly to a current (pfc) chop both sides of the waveform. Biasing (akin to veer+ veer- thresholds) can then be used to prevent 90% of ac noise causing switching.

The output is then a full waveform pfc power wave that is current limited and spreads the current load over the ac waveform. This means the output is naturally a bi polar power rail.
Now this also means that the noise on the is not common nose as each rail is powered from each separate half of the ac waveform (and the switching is independent). If could be possible to rectify the output to only provide the B+ and then use a ground ref to create bi polar rails again to create common noise psrr opportunities.
 

Intersting thoughts. Frankly, though, I don't see any reason why the outputs should be out of phase in your schematics.
Best regards!

I went through this with a pen and paper last night whilst watching TV with the wife.

Kay - you're correct that the connection is connected to the wrong phase.

There's a bigger problem with the original paper op-amp 'comparator' configuration.

There are four comparators:
* top left - gives a single ended waveform by difference
* top right simply outputs the difference between the triangle minus the wave form.
* bottom left is a schmitt trigger - so that the when the voltage rises above the threshold (in this case ground) it drives a square rail-to-rail transition. When the voltage drops below it switches to the other rail. The 'verr' biasing means that point of switching can be adjusted to prevent switching due to noise. It depends on your comparator capability (ie if it works bi-polar)..
* bottom right is simply comparing the square wave and the triangle.

The problem is that the upper never sees a square wave. So the wave form coming out isn't correct (actually, correct is probably too strong, it's not a clean square wave).

Essentially they're doing what have been working towards.
a) a comparator of triangle wave gives when the wave is under/over the analogue level. This provides a window of time "on" at 250kHz. However this will give a product of the two during that on time not a clean square wave.
b) the Schmitt trigger is used make that a full square wave switch.

Their paper comparators work in both +v and -v hence a tube would have to be operating in class A to present the same capability with 1/2 the maximum rail - note switching is far less concerned with linearity of the tube curve - but the comparisons that are not binary would be impacted by the tube linearity in class A. If we switch to class B then we have a switching zero crossover distortion problem. However I suspect the configuration could be biased to operate AB and use the schmitt trigger biasing to simply 'gate' the square wave before the tube enters the lower murk of linearity that exists at the bottom of every tube load graph.

So TL;DR is that it's possible, the paper has a mistake in that it doesn't square switch the Sn waveform, it's possible and even improve linearity in the tube but the tube implementation needs a little more TLC given a tube pair comparator is not an ideal comparator (as per the crown paper).

Now that is also possible with full differential :D

So now going back to my previous post on AC... that got me thinking of a Valve PFC creating a 400V DC link using the above design - the switching driving a pair of valves as an active rectifier, with a series pass regulator (tube) dropping from 400V to 200V then running a second implementation of the above with the audio signal as the AC to switch drive the output.

The mains AC implementation needs some careful consideration - adding a current limiter sense resistor to switch off, also adding a voltage sense to switch off should the voltage exceed the DC link.

The audio 'class TID" (Class Tube I/D) can run either:
* SE (common ground) output stage (a push-pull totem for example) through a LPF and output cap.
* Full bridge (four triodes) configuration for a non-common ground with two LC LPF

Now the fun is - if we want to run digital differential. So each differential audio results in digital differential (ie 2 phases becomes 4 switching phases).. not sure of the benefit on that but I suspect combining all four digital phases could be used to essentially null any noise from amplification.
 
Last edited:
Tektronix made a couple of scopes with tube based SMPS inside 503 with 6DQ6 is an example.

6DQ6-B... looks very interesting. Normal max plate voltage 5kV pulse, average 65mA at 250+V (340mA at 60V). 18W max plate dissipation.

https://frank.pocnet.net/sheets/093/6/6DQ6B.pdf

That means highish transition speed, able to handle mains voltage transients and and a pentode may offer the ability to reduce the number of triodes.

I'm using triodes (and the ecc99) simply because I can focus on the concept easily (literally a ecc99 is 200V top, ~4K on top, ~210R for a pair on bottom so I don't have to think too much and plenty of current). I think once I have a triode (with way too many tubes) model running then I can optimise - I'm not familiar with pentodes.
 
This was the output from the corrected phase:
Screenshot 2021-05-14 at 12.29.17.png

So you can see the missing square wave function - however if the attachment is a quadrature (ie quadrant) based switching then that may explain it. I've not tried an output stage with quadrant switching.

For simple PP - that isn't going to sound very nice..
 
Screenshot 2021-05-14 at 19.04.37.png

Brain shot.. too tired from work. Will have a think over the weekend.

This is basically the concept of having a comparator on each phase (minus rail-to-rail caused by the schmitt trigger). I have a pen and paper drawing of that.

Still lots of work needing. The waveform output isn't too pretty (need the right LPF) - this is where I suspect going R-R square wave will solve some of issues in the FFT (by shifting the noise).
 
Last edited:
Started looking at creating a DSD output stage - the final stage would then combine the switching of the SMPS power.

Screenshot 2021-05-19 at 14.14.15.png

I still have work on this but this would allow (a) analogue input, (b) DSD input via an source switch. The output would then (not at he moment) essentially switch the 390V DC link. Currently a direct connection out this would then perhaps change to a output transformer (switching) before the LPF.