diyAudio (
-   Tubes / Valves (
-   -   Problem with my amp, and I have run out of ideas (

HFGuy 11th February 2011 02:53 PM

Problem with my amp, and I have run out of ideas
Hey folks,

I am hoping you fellas can help me track down a problem with my just finished DIY amp. First alittle background. I started with the Plitron design

I then changed it to auto-bias on the output stage and build the following

I built the amp with Plitron torodial OPTs and used ASC motor run caps (bypassed with 1uF Solen films). Well here is the problem. I am getting massive oscillations on the output stage. I am getting a oscillation of about 9W on the 4ohm load at 488KHz !!! When I power up the amp everything seems normal until the tubes warm up, then the oscillations build until it basically runs rail to rail :(. I have disconnected the NFB and even pulled the 6SN7 tubes out. None of these has helped. I have gone over and over the circuit and I cannot find any wiring mistakes. I am using a 5k grid resistor and a 1k resistor on the screens, both of which are suppose to prevent this parasitic oscillation.

Frankly folks I am stuck, I don't know where to continue with tracking down the problem.

kevinkr 11th February 2011 03:02 PM

You need to go learn a little about HF stability margins and lag compensation networks. You have too much HF gain and too little phase margin. Start with Jones "Valve Amplifiers 3rd Edition" or Radiotron Designer's Manual 4th Ed..

I doubt that the change from fixed bias to cathode bias would cause this issue, but building designs not as intended can lead to this result. It is possible that ESL in the cathode bypass capacitors has affected the phase margin at very high frequencies, but the lack of any sort of compensation is questionable as you must assure that loop gain is below unity at the point where closed loop phase shift reaches 180 degrees or you will have an oscillator.

Cathode biased output stages are not a choice I would make and given just how critical DC balance is with a toroid OPT I would probably not have made this choice. Look into Baxandall's garter bias if you must use cathode bias. Note that the change from fixed to cathode bias probably significantly reduced the maximum available output power as well, and requires a commensurate boost in plate voltage to maintain the same quiescent and small signal operating points. (Large signal is a lost cause as the bias will shift towards class B at higher powers anyway.) Your plate voltage should increase by the amount dropped across those cathode resistors and I also suspect that the very high idle current is on the verge of cooking those tubes.. (I'll swag it is more than 100mA per tube)

Also noted that you changed the 100K grid resistors to 220K which has the effect of increasing open loop gain by several dB which helps make your problem worse and is also not a good idea with crummy modern KT88 due to their excessive grid current, and the possibility that as they get hot they will run away and self destruct.

HFGuy 11th February 2011 03:09 PM

If I am not mistaken both phase and gain margin only refers to the use of NFB, as the feedback becomes positive with the phase shift caused by passive networks. As I mentioned somewhere in my post I have disabled the NFB in my circuit, even going as far as to pull the 6SN7 tubes.

DF96 11th February 2011 03:14 PM

RF oscillation in UL outputs is not uncommon. It means your OPT is less than perfect. The usual solution is snubber networks between each output anode and g2. Try starting with 1K in series with 1nF, and adjust as necessary. This is nothing to do with overall global feedback stability.

On a separate issue, what happen when the balance slider lifts off the balance pot? Does it make the cathode decouplers go bang?

kevinkr 11th February 2011 03:15 PM

Please post pictures of the layout as your comments indicate a possible construction related issue.

Also recheck to make sure that there are no obvious wiring errors.

Those 1K grid stoppers should be right at the socket and I would recommend stoppers on the screens as well - quite possible this is contributing to the issue. Something in the range of 220 - 330 ohms ought to work well. DF96's comment shouldn't be the case here, but as a last resort a snubber can be used across the primary as he suggests. Given the very high quality of your OPT this should absolutely not be necessary.

Should the balance pot wiper go open it is possible for very high voltages to appear across the cathode bypass caps so depending on the voltage rating of those caps it is possible that something untoward could happen. The solution would be to put a couple of 220 - 470 ohm power resistors from each end of the pot to the wiper. (Just something to maintain a current path - operating point will obviously shift, you need to choose resistors large enough to not adversely impact adjustment range, but small enough to maintain reasonable current through the output stage.)

HFGuy 11th February 2011 03:30 PM

I might be confusing the tube terminology here, but in the schematic I built (the lower one) there is a 5k resistor on the grid and a 1k resistor on the screen.

I am really trying to avoid having to go back to the upper schematic, as it's a serious bit of work. I would really hate to rebuild the output stage only to have it still oscillate. And given the fact that at AC the cathodes are both grounded, I fear changing the output stage back to grid bias will have no effect other then to test my patience.

SY 11th February 2011 03:46 PM

I'm rather surprised that the upper circuit would be stable with no dominant pole compensation or lead network. That said...

With no feedback, you're asking for trouble unless your grounding is perfect (no such thing), you have everything properly bypassed to the correct points, and you have avoided any other seemingly negligible and subtle HF positive feedback path.

If it were my amp, I would restore the feedback (the gain will be ridiculously high otherwise), snub the UL connections as DF96 suggests, and put a dominant pole in the first stage (RC network across the plate resistor). The whole advantage to a toroid transformer is that it pushes the HF poles far enough up that you can get away with reasonable levels of feedback without having to drop open loop bandwidth too far.

HFGuy 11th February 2011 03:50 PM

I'll try establishing the feedback network when I get back home and give it a go. Ill also track down the components for the snubber network.

You know what the ironic part of this all is ? I design GHz PLLs and VCOs for a living :)

HFGuy 11th February 2011 04:00 PM

So I assume this is the network DF96 has recommended .... aka lowering the Q

DF96 11th February 2011 04:01 PM

It is just about possible (but, I admit, unlikely) that going from 1K to 5K for the g1 stoppers is part of the problem. Whatever effect the g1 stopper has, will happen at about 5 times lower frequency now. Remember, to get oscillation you need both gain and phase shift.

It is difficult to build an audio output using RF type practices, because lead lengths are naturally higher, but it is always good to remember that the valves don't know what frequency range they are meant to be amplifying. Is there any scope for capacitive feedback from OPT leads to the body of the coupling capacitors, for example? If you ground the output grids, either before or after the grid stopper, does the oscillation cease or change? A general rule: if changing it affects the oscillation, then it is part of the oscillator.

If you can do GHz, then 488kHz is almost DC!!

All times are GMT. The time now is 09:18 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright 1999-2017 diyAudio

Content Relevant URLs by vBSEO 3.3.2