Unstable Eico HF-87

[Jim W]

Greetings -
I'm mostly a lurker on this forum, as I don't have much to add over and above the wisdom that the frequent posters display, but it's been an education to hang out here.

I've been working on my old Eico HF-87 and I'm trying to restore it to the original topology. I've had this amp since I was a kid in the 60's and it's been through several iterations of mods, including the most recent FET driver board setup. Now, it's back to the original schematic with pretty much all the same resistor values and the following changes:
- CLC power supply with much bigger caps, and polypropylene bypasses
- polypropylene caps throughout the signal path, generally larger values, mostly a step or two up in standard values.
- fixed bias on the output tubes with a 10 ohm resistor from each cathode to ground
- removed the level controls on the input
I don't think any of these things should make the amp less stable, but it is. Open loop, the amp will output a pretty good square wave with no sign of instability. Add the 7.5K feedback resistor/cap and it starts ringing. I've attached pics of a scope display of both. The "no feedback" picture is taken just after it starts ringing, turning up the 1.2 khz input signal makes the ringing much worse.

The feedback rolloff cap is a 200pf polystyrene. After reading the forum a bit, I added a 100 ohm screen grid resistor, then replaced that with a 1K ohm screen grid resistor, hoping that would tame the oscillation. Not much difference. Both channels are the same, so I doubt it's a wiring error, but I did try to trace everything just to make sure.

I'm hoping the brain trust on the forum will have some suggestions. I don't think I'm missing anything simple.
Thanks,
Jim

[Jim W]

Forgot to post this - here's a schematic link from the HF-87 refurb thread.

HF-87 Schematic

Jim

[djoffe]

It doesn't seem real far from stable...Maybe we could do an experiment to see just how far...

If you inserted another 7.5K resistor between the 16 Ohm tap and the 7.5K||225pF, you would drop the loop gain by 6 dB, and increase the amp's gain by the same amount. The decrease in loop gain might be enough to get you out of trouble...you could then sweep the amp for frequency response to see where the big rise is, and that would probably give us a good hint as to what might be going on.

Just some random thoughts...have you checked C12 and C17 (careful...lots of volts)?

Is your R16/R30 replacement wirewound? Might it be resonating and doing something funny? Are any of the new resistors wirewound? extra inductance could casue a problem.

[SY]

The asymmetry is the clue- your problem originates in the output transformer. The ones used in the HF87 were generously proportioned but not particularly well-balanced. C12 could be suspect or you might need to increase its value a bit. You might also try a step network across R9.

[Jim W]

Increasing the feedback resistance and with it, the gain, seems to help. I'd really like to go the other way though as the amp had too much gain from the factory. My original goal was to increase the feedback and reduce both the input sensitivity and (at least theoretically) distortion.

All resistors are metal film. C12/C17 are brand new polyprops, and I have the same problem on both channels so I'm guessing it's not the cap. I could lift one end and test it. But the fact that these caps are connected to one side of the transformer primary - couldn't that be the source of the asymmetry? R16/R30 have each been replaced by two 10 ohm resistors which should be metal film from each cathode to ground. (adjustable bias via a separate supply)

As far as frequency response, I could test that. The oscillation is way up there, I think around 200Khz if I remember correctly. I'd be surprised if the amp was still producing useful output up there, but I don't know.

"step network"? Could you elaborate on that?

I'll reduce the feedback and look at the frequency response. BTW - increasing the input signal voltage makes it less stable. At some point the entire top half of the square wave is just high frequency ringing, with the feedback resistor.

Thanks,
Jim

[Jim W]

Couldn't wait until tomorrow, so I just did some poking around with the scope. With a sine wave, the amp looks pretty stable up to about 2.2v P-P out, after which it will start ringing on the tops of the waves. I did a sweep with my audio generator and indeed there is a pretty big peak around 91Khz, where the gain is up by 2.5x or so. The amp was flat to past 80Khz - didn't think that was possible on old tube stuff. So is it just a case of rolling the open loop gain off at some reasonable frequency - say 30Khz? And if so, where's the best place to do that? Is that going to increase the distortion at the high end? (less open loop gain -> less feedback)

Jim

[djoffe]

You could measure the loop gain, or maybe just open the loop entirely, and measure the gain and phase. Then you could figure out just what the feedback could look like to be stable.

That it only happens at higher levels is sensible...output tube gain increases with plate current, increasing loop gain.

[SY]

I've moved this thread for you.

By "step network," I mean a series resistor and capacitor, with the series combination put in parallel with the plate load resistor. It's called a step network because it causes a step in the frequency response at a point determined by the RC combination.

In this case, trying to reduce the overall feedback to stabilize things won't address the fundamental problem, the output transformer asymmetry. The step network (and for that matter, the cap across half of the output transformer primary) reduces the feedback on only the side of the push-pull that needs it. I'd still replace C12 first (watch the voltage rating!), and if that doesn't help, perhaps increase its value. if THAT doesn't work, go for a step network, which should be adjusted while watching the square wave response.

[stokessd]

I agree with Sy, you have a high frequency oscillation. If you sweep your sine wave above the audio band, I bet you see the gain go up at high frequencies right before it drops. The phase compensation circuit that SY described will squash that flatter than a pancake. It will be something along the lines of 15K resistor and say a 270pF cap in series with the voltage amp plate load resistor. The values will be a good starting point (in my experience), and then tuned by eye using a scope. You could also do a bode plot, but that seems painful vs. tweaking it by eye.

Sheldon

[vintagesound]

As suggested by djoffe, check C12 and C17. The real Eico manual has those at 750pf rather than the 470pf on your Sams schematic.