Cathodyne investigation - stop the presses!

[exeric]

I've been doing some further LTspice simulations using cathodynes (concertina, split load phase inverters). If I'm not wrong here, it seems myself and others may have been giving out erroneous info previously. It's a little hard for me to believe that such fundamental info could have been overlooked. I plead guilty also because I believed the conventional wisdom just like most people. This isn't a final verdict though because this just comes from LTspice simulations. It needs to be corroborated by others. I'll try to show what I'm getting at in a series of posts.

The bottom line though is that one may not need a high voltage to run cathodynes and get adequate output. It appears after much simulation that cathodyne output voltages will scale with the drive voltages and power supply voltages required for most output tubes. In other words, one might not need separate voltages supplies for cathodynes in most amps. This will depend on these simulations being correct. I used the miyumi model for 6sn7
in spice. I'm presuming its fairly accurate. More posts forthcoming.

 

[exeric]

Just as background, the original cathodyne invention was direct coupled from the previous amplification stage in the Williamson. It's really hard though to get the output of that previous stage down to a voltage where it will bias the cathodyne so the upper load shares half the voltage and the lower voltage shares half the voltage. That bias voltage needs to be at approx. 1/4 of the total PS voltage going to the cathodyne. So its generally referred to as 1/4, 1/2, 1/4 scheme where the plate sees 3/4 of the PS voltage and the cathode, 1/4 of the voltage. The two signals occupy bounce around between the two 1/4 points and are both stopped in their excursion at the middle and end points.

Because it's so hard to direct couple it at that bias point people often bias it at 1/3, 1/3, and 1/3 point.

Here's the circuit and the FFT for it. I didn't direct couple it because it's more convenient to make changes without it.

 

[soulmerchant]

Um, why am I not surprised?

Put a decent sized grid stopper on a cathodyne phase splitter, then don't forget to use decent drivers (preferably not a follower).

Tough thing to beat, darned tough to do it wrong.

 

[exeric]

As you can see everything is good at 120v p-p with only the 2nd harmonic significant, a characteristic of cathodynes. I'm including now the fft for 160v p-p and for 180 v p-p. As you can see it's still good up to 160 but the output gets destroyed at 180 with a spray of upper harmonics.

 

[soulmerchant]

Because it's so hard to direct couple it at that bias point

Oh that's what you mean? I guess if you used the same triode for input as for phase splitting duty you would have that problem.

I never tried that since it just never made much sense. Maybe it is easy to do it wrong if you bias it like you suggest. I wonder how Morgan Jones did it... darn.. have to dig around for his book again!

 

[soulmerchant]

As you can see everything is good at 120v p-p with only the 2nd harmonic significant, a characteristic of cathodynes. I'm including now the fft for 160v p-p and for 180 v p-p. As you can see it's still good up to 160 but the output gets destroyed at 180 with a spray of upper harmonics.

Exeric.. are these posts simulations with the plate at 3/4 voltage? What does your simulation look like if you introduce a 10k Ohm (or more) grid stopper?

 

[exeric]

That seems to be a characteristic of cathodynes. When either signal moves outside of its available envelope it creates that spray of intermodulation distortion. So I tried the same thing at the preferred 25% of PS voltage bias point to see how far I could get. Here's the results.

 

[exeric]

Soul Merchant,
What I'm getting at is that it's darn hard to direct couple a cathodyne from the previous stage and bias it at the optimum point. I couldn't get close to the maximum theoretical excursion when attempting to direct couple it. Also, I learned that the upper load deteriorates even into a high resistance before the lower load does. But this only occurs at very high signal levels. I played around with the bias level and found for this particular instance that 21.5% of the PS level achieved equal excursions into high load resistances and high signal level.

So I decided to bite the bullet and try the concertina with a 300 volt PS and biasing it at 21.5%. Heres the final circuit and the FFTs.

 

[rongon]

Just a helpful hint to make it easier for us to see what you're doing...

You posted your schematics and screen shots in .pdf format, which means we'll need to download all your attachments before we can see them. If you had uploaded them as .png or even .gif files (image files) then they would display right in the browser window.

Thanks.

 

[exeric]

So, the bottom line is it's a colossal waste of time to try to direct couple the concertina to the previous stage. It just is. You can get 105v PP out of a concertina supplied by just 300 volts if you do not direct couple. There is absolutely no need for a separate high voltage supply if you do that. If you are absolutely intent on removing a coupling capacitor then do it where it will do the most good: going into the finals using a power drive scheme, preferably a mosfet. Here's the output using that schemo and driving it with 120v P-P. It has an output of 105 v PP. There are tons of commonly available output tubes that require 300 v power supplies that can easily get by on that drive. And if you need more then at 450volts the concertina drive will scale up for the drive requirement for that.

One more thing. If you do not learn anything else then think of this. For most topologies you should use the concertina (not direct coupled from the previous stage) to drive the finals. The reason is that there will then be no deviation from phase symmetry by further amplification stages before the signal get to the finals.

 

[exeric]

Rongon,
Thanks, next time.

 

[exeric]

Exeric.. are these posts simulations with the plate at 3/4 voltage? What does your simulation look like if you introduce a 10k Ohm (or more) grid stopper?

soulmerchant, the bias levels are what actually show on the grid. In other words the cathodes will be about 4 to 8 volts more positive. But generally the .33 percent equates to 66.6% voltage on the plates. the .25 equate to 75% voltage on the plates. All of those voltage are approximate. For instance the actually voltage on the 21.5% is 23.7% of the PS voltage on the cathode and 76.3% of the PS voltage on plate.

Again, these are simlations. Not real corroborated results. I do not have the test equipment to corroborate it. Hopefully someone out there will get curious and try to prove me wrong. Good Luck!

 

[soulmerchant]

Exeric

Ok, so I learn nothing new? Or does it look to me that you mean that the bias for a concertina is not so important as well?

One more question: Are these individual phases simulated, or the sum of both phases? sounds crazy of me to ask, but one never knows..

 

[smoking-amp]

You could use a P channel Mosfet for the Cathodyne (Source-O-dyne?) and dirct couple at 3/4 B+. Like Fairchild FQP1P50. The higher the gm, the better it works.

 

[soulmerchant]

P channel MOSFET "source-o-dyne". I like that. It actually sounds like worth trying out.

The big grid stopper would probably not be needed then.

edit: mouser says FQP1P50 is obsolete.

 

[smoking-amp]

You may have to do a parmetric search for the lowest current HV P channel part. Maybe FQPF2N60C? FQPF2P60C? I bought a bunch of the FQP1N50 and FQP1P50 when they were available.

 

[exeric]

Smoking amp, so you could have your cake and eat it too? I think you are more up to date on the different types of mosfets and how they bias up than I am. Maybe you can post a schemo... Seems like a sweet idea. Definitely better than trying to direct couple a tube.

 

[smoking-amp]

Did a parametric select at Mouser: FQB1P50 500V P channel Mosfet Crss 6 pF
$2.06 each $1.75 per at 10 quantity
http://www.mouser.com/ds/2/149/FQB1P50-107746.pdf

Same hookup as the cathodyne basically. Source through top resistor load to B+, Drain through bottom resistor load to 0V, direct drive input through a 500 Ohm gate stopper resistor to the gate. May not need the stopper with P channel, but safer. Have to try. (source acts like the "cathode", but for a P channel tube) You will have most of the available voltage (down to 50% 50%, less a volt) for output.

Probably a good idea to put a small gate protection 15V Zener across gate to source. (Zener diode cathode to source, Zener diode anode to gate) Normally not conducting. The gate stopper resistor limits the current into the Zener when out of bounds. (maybe during warm-up) (so gate stopper resistor connects to the gate just outside the Zener anode connection)

 

[exeric]

Exeric

Ok, so I learn nothing new? Or does it look to me that you mean that the bias for a concertina is not so important as well?

One more question: Are these individual phases simulated, or the sum of both phases? sounds crazy of me to ask, but one never knows..

I checked visually both phases, but didn't combine them in something like a interstage coupling transformer. It was obvious visually on each FFT when either load crossed the line. I just didn't include in my visuals to the forum. That's how I arrived at the 21.5% bias level. Both sides didn't degrade and then when they did then both pretty much did it together. Above or below that one went before the other.

As far as putting in a high value resistor instead of the 1k, it doesn't make any difference. I just checked. Do it yourself on spice. What does make a difference is varying the ratio of the 1 meg resistor to the one below it. You can optimize everything if you aren't juggling operation of two stages. A waste of time.

 

[exeric]

You can optimize everything if you aren't juggling operation of two stages. A waste of time.

I guess I better clarify that I wasn't referring to the mosfet direct coupling scheme. I just don't know enough about it to comment or have an opinion on it. I was just referring to the traditional Williamson way.