diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Tubes / Valves (http://www.diyaudio.com/forums/tubes-valves/)
-   -   Need Help with New Amp Build, Ugly signals (http://www.diyaudio.com/forums/tubes-valves/207341-need-help-new-amp-build-ugly-signals.html)

SandyG 22nd February 2012 04:55 PM

Need Help with New Amp Build, Ugly signals
 
4 Attachment(s)
Just finished up building one of the EZ260A amps which is based on the Grommes 260A. It sounds very good and only on a whim did I decide to put it on the scope to see what kind of power and how clean its signal looked.

It looked bad!

The 1Khz sine wave looked like it had some oscillations in it, and at 100hz more visible.

Not sure where to start working on this as this is my first tube project. I'm waiting for another set of tubes to see if potentially I have a bad one, but not sure if the waveforms will give a clue to help get me messing with it.

One other thing I did note when bringing it up on the Variac was that without a load the amp seems to motorboat. With an 8 ohm dummy load, stable. With a speaker, stable and no noise while idle.

Any ideas where to start looking for the problem. (scope pics and schematics attached)

Thanks

Sandy

mississippi 22nd February 2012 05:07 PM

Your neg. feedback is doing this.

kevinkr 22nd February 2012 05:13 PM

The box with the RC you have labeled "Optional" isn't - it's the amplifier's lag compensation network for HF stability. You may fiddle with the values to optimize square wave response. (Minimal overshoot and ringing on edges without killing the HF response totally. A 1kHz square wave is usually suitable for doing this, then check at 10kHz to see what sort of performance you get.)

Also C14 may be applicable in certain cases, but depending on the output transformer used it may be better left out. First add the lag compensation and then recheck with C14 in and out of circuit.

SandyG 22nd February 2012 05:23 PM

Thanks Kevin,

The optional box comment should have been removed. I started with a 250pf cap in C17 and upped it to 500pf and it help clean things up a bit, but still the ugly wave forms persisted. In some documents I found they had a 250pf, other had 470pf, so was unsure what the value was. Seems 470/500pf is what was used.

Thanks on the tip (both Kevinkr/Mississippi) on the feedback, I'll remove C14 and see what the waveform looks like.

Also, the output transformer is an 100watt Edcor with 4200 ohm plate. I am not using the ultra-linear taps, but using the fixed screen option leaving the UL taps un-connected.

Sandy

kevinkr 22nd February 2012 05:32 PM

Hi Sandy,
Not sure why you aren't using the UL connections as running this amp in pentode connection definitely affects the overall output stage gain and phase margin, the UL connection provides local feedback between the screens and plates. The transformer design is probably optimized for UL operation as well, there being numerous tradeoffs between leakage inductance and interwinding capacitance to be made. In pentode connection leakage inductance is less of an issue than interwinding capacitance since the source impedance is high, in UL operation the feedback reduces the effective rp considerably and makes the output stage somewhat more sensitive to leakage inductance and less so to capacitance - the extreme case is triode connection where leakage inductance becomes a really big deal.

The above is grossly simplified but hopefully helps. I would strongly recommend the UL connection be tried.

I would disconnect the feedback loop and compensation and figure out the phase and amplitude response of the output transformer and see how the pentode connection and UL connection differ.

With the pentode connected output stage you may need to reduce the value of R22 slightly in addition to increasing the value of that cap.

DF96 22nd February 2012 07:19 PM

Do you have screen stoppers? The oscillation could be taking place in the output stage rather than the feedback loop. Temporarily disconnecting or reducing the feedback would show this.

Somewhat strange circuit. R23 appears to have no function, unless it is a warning of marginal output stage stability. The upper 12AU7 has strange bias arrangements, relying on grid current - OK for small signal inputs but could get bias shift with larger signals. The 12BH7 has too low mu for a LTP phase splitter without a CCS tail, so it needs manual adjustment for balance.

Miles Prower 22nd February 2012 10:49 PM

Quote:

Originally Posted by SandyG (Post 2917994)
The 1Khz sine wave looked like it had some oscillations in it, and at 100hz more visible.

Not sure where to start working on this as this is my first tube project. I'm waiting for another set of tubes to see if potentially I have a bad one, but not sure if the waveforms will give a clue to help get me messing with it.

You've got snivets there. Your schemo doesn't include screen stoppers for the finals. It's a Barkhausen oscillation that's occurring at plate current cutoff. Never used KT88s myself, but I've seen the same thing with 807s. (Damped 60KHz oscillations when they hit plate current cutoff.) Adding 1K5 screen stoppers damped out the oscillation. (6BQ6s didn't show these snivets, but I included 680R screen stoppers just in case.)

Quote:

One other thing I did note when bringing it up on the Variac was that without a load the amp seems to motorboat. With an 8 ohm dummy load, stable. With a speaker, stable and no noise while idle.
You're going to get problems like that with gNFB sometimes. Dropping the voltage with a Variac will reduce the open loop gain, and that, in turn, can cause instability.

Nyquist Criterion: "The number of unstable closed loop poles is equal to the number of open loop unstable poles plus the number of encirclements of the point: -1 + j0".

The denominator of the transfer characteristic is: D(s)= G(s)H(s), where G(s) is open loop gain, and H(s) the x-fer characteristic of the feedback network. Given that, reducing G(s) can sometimes cause the whole characteristic to hit that -1.0 + j0 point. Less open loop gain can make for instability. :eek: It's counter-intuitive, but it does happen, and some amps will have a burst of noise when you power off, cathodes cool, filter capacitors discharge, and open loop gain falls off. It's something you'd like to avoid if at all possible since that means a marginally stable design that might be put into full-blown oscillation if different speeks with different impedances are connected. It could also mean that you have rising gain over a narrow band of frequencies that could prove detrimental to sonic performance.

SandyG 22nd February 2012 11:45 PM

Quote:

Originally Posted by DF96 (Post 2918200)
Do you have screen stoppers? The oscillation could be taking place in the output stage rather than the feedback loop. Temporarily disconnecting or reducing the feedback would show this.

Somewhat strange circuit. R23 appears to have no function, unless it is a warning of marginal output stage stability. The upper 12AU7 has strange bias arrangements, relying on grid current - OK for small signal inputs but could get bias shift with larger signals. The 12BH7 has too low mu for a LTP phase splitter without a CCS tail, so it needs manual adjustment for balance.

I will toss the scope on the driver stages to see what it looks like. The original circuit is pretty close to what I have except for the missing internal impedance adjustment which Grommes removed back in the day.

After looking at the driver stages it's seemed more unconventional then some of the other drivers for KT88's (I have Triode Electronic in my Dynaco MK3's).

R23 Was a reminant of the old grommes internal impedance circuit and was left in. I had it in and also removed it with no difference in open load stability. I left the ability to stuff the PCB with it as my Dynaco MK3's have a 680ohm across the 16ohm/feedback output tap which I'm guessing is a stability fix, which is fine if it works. I'll work on the motorboating issue later as I think it may be un-related to the fedback issue, which might be the first issue.

[edit]
One last thing is that I used 10ohm 2 watt resistors on the cathodes of the tubes to measure the bias.

SandyG 23rd February 2012 04:35 AM

1 Attachment(s)
Miles - Go easy on the theory, I'm just starting out ;)

Sorry if I wasn't clear. I did bring up the power full on the Variac and on its watt meter you could see the power draw jump around a bit, so as I suspected it was motor boating without a load (And confirmed when tossed it on the scope)

I can easily try the screen resistors since in the PCB I have a jumper that can be cut.

And to learn as I go do the screen stopper prevent oscillation or just prevent a problem during an overload situation? The Screens are driven by the 6L6GC although both tied together, not sure how that might impact things, But will try a 1K5 in each of the screens to see what it does. My guess is that it might only help with the motor boating issue without a load and a safety net if over driven. (all newbie speculation of course)

I don't think the amp is that far off, it was an odd choice to base one on, but I had seen a few people comment on how good they sounded so figure hey, lets try something odd.

And I can say that I spent an afternoon listening to every type of music I had and with it and the modded Dyanco MKIII used for the other channel sounded very good. So I think with a bit of clean up on some parts values it will come out sounding very well (I might be too easy to please...) .

Attached is the original Grommes 260A for reference as well.

DF96 23rd February 2012 11:58 AM

Don't run a valve amp without a load, unless you have lots of money to keep buying new OPTs. Screen stoppers prevent the output valves from oscillating at RF.

If you scope the driver stage you may still see the oscillation, as it will come round the feedback loop. Perhaps attenuated by the compensation network, though. Remove or reduce the feedback. If that stops the oscillation then you have a loop stability issue. If it doesn't stop it, then you probably have output stage parasitic oscillation. Different problems, different cures.


All times are GMT. The time now is 04:44 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2