Cathodyne Oscillation

[famousmockingbird]

I changed around the wiring in order to use a 6BL8 instead of the hard to find 7199 in an Ampeg B-18. Problem is I am getting a 120kHz oscillation at the cathode output of the phase inverter. The anode output is fine. When I open the feedback loop it disappears. 😕

To make matters worse I don't know if it was there to begin with. Basically the amp was making strange noises so I tested the tubes and all were strong except for the 7199. I performed surgery and popped in a 6BL8, the operating points look close to where they should be so it's wired up correctly.

https://frank.pocnet.net/sheets/168/7/7199.pdf

https://frank.pocnet.net/sheets/049/6/6BL8.pdf

 

[rayma]

I am getting a 120kHz oscillation at the cathode output of the phase inverter.
The anode output is fine. When I open the feedback loop it disappears.

Try adding a series capacitor-resistor across the pentode plate resistor,
something like 10k in series with 100pF.

Is the plug that grounds the output transformer secondary (pins 2 and 3) installed,
with a load resistor?

 

[Gnobuddy]

I am getting a 120kHz oscillation at the cathode output of the phase inverter. The anode output is fine. When I open the feedback loop it disappears. 😕

I suggest increasing those 1k grid stoppers on the output valves - a lot - and seeing if that does the trick.

The datasheet says input capacitance is only a few picofarads (typical for pentodes and beam tetrodes), and up to 500k resistance between G1 and ground is okay, so you can easily raise those 1k grid stoppers to 10k, 22k, 47k, or even 100k or 220k.

With 10 pF input capacitance, a 100k grid stopper won't begin to roll off treble response until you go above 160 kHz, so even 220k shouldn't make any audible difference at all. But there is a very good chance it will kill the oscillation problem.

-Gnobuddy

 

[gingertube]

Also,
The 65V label on the concertina triode grid is WRONG. Should be more like +95V. Are tou sure that 10M goes back to the anode. I would have expected it to go to a rail so that the 10M + 3M3 divider sets the concertina bias point.
The concertina triode could also do with a gridstop resistor.
Cheers,
Ian

 

[Enzo]

In fairness, that grid voltage is no doubt higher, but 65v might well be what the measurement is considering the meter impedance will affect the reading. If you read voltage to ground your meter impedance is parallel the 3.3 meg. Check the note lower left, all readings with 20,000 ohms per volt meter.

 

[famousmockingbird]

Yes I figured meter impedance is pulling down the voltage at the grid.

I will pop a grid stop resistor on the Triode grid, since there is less than unity I think I can use a large value. I got about a million 390k and 150k CC resistors so I will use one of them.

 

[gingertube]

For your reference, a 6BL8 Concertina in a Git Amp
Note the 6K8 + 220pF stability network similar to rayma's suggestion
In this case the concertina triode is self biased.
http://www.ozvalveamps.org/playmaster/playmast_116.jpg

Cheers,
Ian

 

[famousmockingbird]

Try adding a series capacitor-resistor across the pentode plate resistor,
something like 10k in series with 100pF.

Get THIS man an effin Cigar!!!!!

Dude you rock thank you so much 🙂

For your reference, a 6BL8 Concertina in a Git Amp
Note the 6K8 + 220pF stability network similar to rayma's suggestion

Yes rayma hit ball out of the park on this one. I tried a 390k grid stopper on the cathodyne (concertina) and no change. Then I did as Rayma suggested and bam, no more oscillation. I only had 220pF so I used that with 10k.

Is this due to the 6BL8 having less than half the plate to grid capacitance? The 7199 says .06pF and the 6BL8 says .025pF. Interesting stuff.

 

[rayma]

Get THIS man an effin Cigar!!!!! Dude you rock thank you so much 🙂 Yes rayma hit ball out of the park
on this one. I tried a 390k grid stopper on the cathodyne (concertina) and no change. Then I did as Rayma
suggested and bam, no more oscillation. I only had 220pF so I used that with 10k. Is this due to the 6BL8
having less than half the plate to grid capacitance? The 7199 says .06pF and the 6BL8 says .025pF. Interesting stuff.

Thanks, glad to be of service. Change that over-sized grid stopper back to 10k or so.
The problem is the entire amplifier circuit's phase shift. The added RC network reduces the gain
at ultrasonic frequencies to below unity before the phase shift becomes excessive, which causes
instability when using overall negative feedback.

 

[gingertube]

Not really,
It has to do with phase margins due to phase shifts (mostly in the output tranny) and open loop gain.

The 6K8 + 220pF (or 10K + 220pF in your case) decreases the high frequency gain whilst leaving the phase response largely unaffected.

You need to do vector (amplitude + phase) PITA calcs for exact effect
BUT
losely
The HF gain will be halved at the frequency when the 10K + 220pF impedance = the anode loade impedance.
If you just add a cap across the anode load to do thios then you get a 45 degrees phase shift at that frequency.
By adding the series 10K resistor that 45 degree phase shift occurs when the impedance of the 220pF is equal to that 10K series resistor instead. This will be at a fequency of 8 to 10 times higher.
That is we have reduced HF gain but left the phase response largely intact.

By the frequency that the total phase shift hits 180 degrees turning the negative feedback into positive feedback we have reduced the gain to less than 1 - so no oscillation.

This is why ideally that series resistor (10K in your case) should be about 1/10th of the anode load resistor.

Cheers,
Ian

 

[gingertube]

Busy typing whilst Rayma responded. He said the same as me without the "techo babble".

The "techo babble" was coz' one of my buttons got pressed. I see these stabilty networks done badly more often than ideally. Usually by using too high a series R in the network.

What you have ended up with is actually pretty good (not absolutely ideal) and does the job.

Cheers,
Ian

 

[rayma]

Busy typing whilst Rayma responded. He said the same as me without the "techo babble".
The "techo babble" was coz' one of my buttons got pressed. I see these stabilty networks
done badly more often than ideally. Usually by using too high a series R in the network.

Or in techno-babble, the series RC sets the dominate pole of the open loop amplifier circuit,
which is also called slugging. The RC series resistor reduces the open loop gain to below unity,
and the RC series capacitor restores the open loop gain up to as high a frequency as possible,
while still allowing the amplifier to remain stable when closing the loop.

 

[famousmockingbird]

Maybe I didn't ask my question properly.

I have a somewhat okay understanding of feedback and compensation networks. My question is not what causes oscillation but what made the difference between the 7199 and the 6BL8 to cause this.

Maybe the 6BL8 gave me more open loop gain which in turn gave me more feedback bringing it closer to instability?

Or, maybe because miller capacitance was lower with the 6BL8 there was more open loop gain at ultrasonic frequencies requiring the compensation network at the 6BL8 pentode anode?

So to rephrase my question; What changed to cause the instability when switching from 7199 to 6BL8?

 

[rayma]

What changed to cause the instability when switching from 7199 to 6BL8?

The 6BL8 version of the open loop amplifier may have more gain, and so worse stability
when the loop is closed.

 

[famousmockingbird]

The 6BL8 version of the open loop amplifier may have more gain, and so worse stability
when the loop is closed.

Makes sense. I guess regardless of whether it's open loop gain increase at all frequencies or just high frequencies the end result is the same when the loop is closed. The only way to know would have been to measure open loop gain before changing over to the 6BL8 and then compare it.


Anyway it's been a fun learning experience and if anyone else goes to do the same thing they will know what to do. Those 7199's are getting rare and expensive.

Thanks again to everyone for helping me!!

 

[nauta]

There is something seriously wrong with the schematic of the splitter. The grid-leak ties to ground instead of to the previous gain stage's plate or to a voltage reference. So, the concertina is now idling at microamps and its transconductance is practically zero, so no wonder there is oscillation but there should also be no output.

 

[PRR]

7199 is a fairly tame pentode, for audio where gain is easy.

6BL8 is a hotter pentode for TV Tuner work, where gain is hard, so tends to higher gain.

We don't know what *else* changed. If the cathode current increased due to differences between the tubes, that tends to more gain.

Theory can be fun, but sometimes you just put on the R-C network which generally "should" be there, and get on with more important things.

 

[famousmockingbird]

There is something seriously wrong with the schematic of the splitter. The grid-leak ties to ground instead of to the previous gain stage's plate or to a voltage reference. So, the concertina is now idling at microamps and its transconductance is practically zero, so no wonder there is oscillation but there should also be no output.

There is a voltage reference, it's from the triode's anode, the 10M and 3.3M form a divider biasing the triode. Not sure why they didn't use the B+ rail for the reference. The concertina idles at 2mA.

7199 is a fairly tame pentode, for audio where gain is easy.

6BL8 is a hotter pentode for TV Tuner work, where gain is hard, so tends to higher gain.

We don't know what *else* changed. If the cathode current increased due to differences between the tubes, that tends to more gain.

Theory can be fun, but sometimes you just put on the R-C network which generally "should" be there, and get on with more important things.

I agree and I didn't lose much sleep over it. The pentode section of the 6BL8 matched up close to the 7199; they both had 2v at the cathode but clearly there was more gain with the 6BL8. The RC network is a good thing so it's a win win.

I am happy with the results. 😀