Hello all,
I've just finished building a Bevois Valley amp from Morgan Jones' book, and have run into an issue with distortion in the right channel.
The DC operating points all checked out OK, after which I put an 8.2ohm 10W dummy load resistor across each channel's output and applied a 2Vp-p 1kHz sine wave on both channels to test them out.
With a scope on the left channel's output, everything looks fine - the sine wave is undistorted and around 12Vp-p. But when I scope the right channel's output, I see a distorted, attenuated sine wave - more like a triangle wave with the peaks clipped.
When I scope the right channel's input at the RCA input jack, I see the same distortion (around 0.4Vp-p, down from 2Vp-p). The unloaded waveform from my signal generator looks fine on the scope.
Does anyone have any ideas how I should fix this? I'm going to try resoldering everything on that channel's input, and see how I go. Possibly I should try swapping the E88CCs in each channel, and see if the fault follows the valve?
Oh, and with inputs shorted to ground and negative feedback hooked up, I see no oscillation on the outputs, so I don't think the distortion is due to me having the phase of the NFB incorrect.
Cheers!
I've just finished building a Bevois Valley amp from Morgan Jones' book, and have run into an issue with distortion in the right channel.
The DC operating points all checked out OK, after which I put an 8.2ohm 10W dummy load resistor across each channel's output and applied a 2Vp-p 1kHz sine wave on both channels to test them out.
With a scope on the left channel's output, everything looks fine - the sine wave is undistorted and around 12Vp-p. But when I scope the right channel's output, I see a distorted, attenuated sine wave - more like a triangle wave with the peaks clipped.
When I scope the right channel's input at the RCA input jack, I see the same distortion (around 0.4Vp-p, down from 2Vp-p). The unloaded waveform from my signal generator looks fine on the scope.
Does anyone have any ideas how I should fix this? I'm going to try resoldering everything on that channel's input, and see how I go. Possibly I should try swapping the E88CCs in each channel, and see if the fault follows the valve?
Oh, and with inputs shorted to ground and negative feedback hooked up, I see no oscillation on the outputs, so I don't think the distortion is due to me having the phase of the NFB incorrect.
Cheers!
Try swapping the E88CC's, just to eliminate them as a possibility. Re-check wiring around the input stage, and re-check DC conditions there.
Also, what happens if you use a smaller input voltage with NFB disconnected?
Is the grid stopper (330R) right at the grid pin? If you touch the grid or the cathode pin do the DC conditions change? I ask because VHF parasitic oscillation in the input stage might give the symptoms you have. (Be careful not to touch the anode pin!)
As another check, compare DC readings obtained directly with a DMM probe on the relevant point, with DC with a 1M resistor attached to the end of the probe (so in series with the DMM). This should give you 91% (assuming a 10M DMM). Include the first stage grid in this (should be 0V). This check avoids disturbing any oscillation, which normally probes may inhibit.
Also, what happens if you use a smaller input voltage with NFB disconnected?
Is the grid stopper (330R) right at the grid pin? If you touch the grid or the cathode pin do the DC conditions change? I ask because VHF parasitic oscillation in the input stage might give the symptoms you have. (Be careful not to touch the anode pin!)
As another check, compare DC readings obtained directly with a DMM probe on the relevant point, with DC with a 1M resistor attached to the end of the probe (so in series with the DMM). This should give you 91% (assuming a 10M DMM). Include the first stage grid in this (should be 0V). This check avoids disturbing any oscillation, which normally probes may inhibit.
Thanks for the suggestions. I found the problem - it was a short in the shielded cable at that channel's input. Too much heat when soldering had melted the insulation on the centre conductor where it met the shield.
I've replaced this, and now I get a clean sine wave on both channels when testing. But now, there's a new problem. With speakers connected and music playing, there's an intermittent clicking noise at approx 2-10Hz. At higher volumes, this breaks out into oscillation.
All my grid stoppers are right against their pins, but I'll try comparing the DC levels on the input grid with and without a 1M resistor in series with the DMM probe as you've suggested.
Could this be a symptom of incorrect gNFB phase? This is my first amp with negative feedback, so I don't know if the oscillation I'm hearing is due to this. I'll try swapping the phase (or disconnecting it altogether), and see what happens.
Or possibly it's motorboating? I have a Maida regulator supplying the 285VDC for the input stage - I might see if this rail is stable when playing music. Maybe an RC section for each channel following the Maida (I have some 10uF 450V caps spare) could be an idea?
I've replaced this, and now I get a clean sine wave on both channels when testing. But now, there's a new problem. With speakers connected and music playing, there's an intermittent clicking noise at approx 2-10Hz. At higher volumes, this breaks out into oscillation.
All my grid stoppers are right against their pins, but I'll try comparing the DC levels on the input grid with and without a 1M resistor in series with the DMM probe as you've suggested.
Could this be a symptom of incorrect gNFB phase? This is my first amp with negative feedback, so I don't know if the oscillation I'm hearing is due to this. I'll try swapping the phase (or disconnecting it altogether), and see what happens.
Or possibly it's motorboating? I have a Maida regulator supplying the 285VDC for the input stage - I might see if this rail is stable when playing music. Maybe an RC section for each channel following the Maida (I have some 10uF 450V caps spare) could be an idea?
2-10Hz probably means motorboating, or possibly a faulty intermittent component. Wrong feedback phase is more likely to create an audio frequency oscillation, so probably not that. Try it anyway, just to eliminate it.
When the oscillation occurs, try to find what parts of the amplifier are taking part. It might be all of it, or just part.
When the oscillation occurs, try to find what parts of the amplifier are taking part. It might be all of it, or just part.
OK, now I'm noticing some weird hissing and squelchy noises as well as the motorboating. There's also some ringing on the edge transitions of a square wave, so I suspect parasitic HF oscillations.
I've ordered some 10K carbon comp resistors to replace the 330R grid stoppers I'm currently using on the E88CC grids. Will let you know how it goes!
I've ordered some 10K carbon comp resistors to replace the 330R grid stoppers I'm currently using on the E88CC grids. Will let you know how it goes!
No bypassing at the input stage - both channels are fed directly from the 285VDC output of the Maida regulator. I have some 0.1uF 630V caps that I could use for this, keeping them physically close to their input stage. But what's curious is that I looked at the 285VDC rail with the scope tonight, and it was rock-solid even when I had it pushed into instability again.
I did probe the E88CC grids though, and saw ringing on the square wave edges. I hear a lot of interference through an AM radio when I hold it near the E88CCs - I tried this as well on a SE 2A3 amp I built, and it was much quieter by comparison. So I guess I'll try the 10K grid stoppers. Maybe series RC snubbers between plate and screen grid on the EL84s, too - I'm using Edcor GXPP10-8-8K OPTs.
I did probe the E88CC grids though, and saw ringing on the square wave edges. I hear a lot of interference through an AM radio when I hold it near the E88CCs - I tried this as well on a SE 2A3 amp I built, and it was much quieter by comparison. So I guess I'll try the 10K grid stoppers. Maybe series RC snubbers between plate and screen grid on the EL84s, too - I'm using Edcor GXPP10-8-8K OPTs.
oscillatin'
I had a similar problem with my build- it was a while back so I don't remember everything I tried, but what solved it was adjusting the negative feedback. You have to adjust on test because we are using different output transformers than MJ did.
There is a description of how to adjust the feedback network using variable capacitors in this post:
http://www.diyaudio.com/forums/tubes-valves/99218-kofi-annan-oscillatin-bevois-valley.html
I had a similar problem with my build- it was a while back so I don't remember everything I tried, but what solved it was adjusting the negative feedback. You have to adjust on test because we are using different output transformers than MJ did.
There is a description of how to adjust the feedback network using variable capacitors in this post:
http://www.diyaudio.com/forums/tubes-valves/99218-kofi-annan-oscillatin-bevois-valley.html
Try to measure the amount of GNFB you have (the difference of the sensitivity with and without GNFB).
Usually this kind of problem can be solved by reducing the GNFB to max. 10 dB.
This is usefull to try if you have not necessary equipment to analyze the actual reason.
Grid stoppers are not the reason.
Usually this kind of problem can be solved by reducing the GNFB to max. 10 dB.
This is usefull to try if you have not necessary equipment to analyze the actual reason.
Grid stoppers are not the reason.
Good news - Today I disconnected the GNFB and the problem has gone away, so at least I know it's related to this.
To be honest, it sounds pretty good without the GNFB! I'd like to get it going with feedback sometime, but for now I might just enjoy it as-is. As has been suggested, reducing the GNFB (it doesn't have a whole lot of closed-loop gain with GNFB connected), or playing around with the capacitor in the GNFB network might be my next plan of attack.
Thanks again for everyone's help!
To be honest, it sounds pretty good without the GNFB! I'd like to get it going with feedback sometime, but for now I might just enjoy it as-is. As has been suggested, reducing the GNFB (it doesn't have a whole lot of closed-loop gain with GNFB connected), or playing around with the capacitor in the GNFB network might be my next plan of attack.
Thanks again for everyone's help!
There is no such thing as a 'safe amount' of NFB, such as '12dB'. It all depends on the circuit.
Low frequency instability may mean the OPT is smaller and cheaper than the original design. You could try changing the size of coupling caps, to ensure that their bass rolloff does not coincide with the OPT bass rolloff. Try up or down.
Is the regulator good down to DC? If not, that could be a source of motorboating too.
Low frequency instability may mean the OPT is smaller and cheaper than the original design. You could try changing the size of coupling caps, to ensure that their bass rolloff does not coincide with the OPT bass rolloff. Try up or down.
Is the regulator good down to DC? If not, that could be a source of motorboating too.
I can say that safe amount of GNFB exists if the used circuit has even a little sense (and I think the ciruit we disscuss about is such). Safe amount is typically some 10...12 dB.
I have designed and built tube HiFi-amplifiers with different topologies, output tubes and especially with many different type of output transformers and my statement is based on practical experience I have got.
With exceptionally good output transformer together with successful design GNFB can be a lot over 20 dB before instability begins to show up.
I have designed and built tube HiFi-amplifiers with different topologies, output tubes and especially with many different type of output transformers and my statement is based on practical experience I have got.
With exceptionally good output transformer together with successful design GNFB can be a lot over 20 dB before instability begins to show up.
artosalo said:
You must use 'always' with a different meaning from me. I assume it means always; you use it for 'all sensible circuits' (IYHO i.e. all circuits for which 12dB of feedback is safe).artosalo said:
I don't dispute that 12dB of feedback is often/usually a safe amount. 20-30dB is possible with some circuits.
It seems my amp, for whatever reason, just doesn't have the open-loop gain to be traded for feedback - I suppose the concertina phase splitter, with slightly less than unity gain, isn't helping.
For the same 2Vp-p 1kHz sine wave input, I'm measuring 24Vp-p (without feedback), and 12Vp-p (with feedback) at the loaded OPT secondary. The problem is that, even with my input source at full volume and a fairly modest 6dB of feedback, I'm not able to drive my speakers to acceptable listening volume.
Again, the amp luckily sounds fine without feedback. But in the future, I might try increasing that feedback resistor to 22k or 47k and see how it goes.
For the same 2Vp-p 1kHz sine wave input, I'm measuring 24Vp-p (without feedback), and 12Vp-p (with feedback) at the loaded OPT secondary. The problem is that, even with my input source at full volume and a fairly modest 6dB of feedback, I'm not able to drive my speakers to acceptable listening volume.
Again, the amp luckily sounds fine without feedback. But in the future, I might try increasing that feedback resistor to 22k or 47k and see how it goes.
Sorry about the thread bump, but I finally decided that I wanted to add gNFB to this amp after all, and thought I'd let you all know how I got it working.
I had to reduce the amount of gNFB to 10dB, as well as adjust the feedback compensation network and add series RC snubbers across the plate resistor of the input stage. What I didn't appreciate five years ago (!) is that all of this really had to be adjusted on test for my particular OPTs.
For this particular amp, a 3k9 feedback resistor gave me 10dB gNFB. A 750pF cap in series with a 2k resistor was connected in parallel with the feedback resistor. And in parallel with the input stage E88CC plate resistor, I used 10k in series with 200pF mica.
After this, the amp is running with gNFB, and without stability issues.
I had to reduce the amount of gNFB to 10dB, as well as adjust the feedback compensation network and add series RC snubbers across the plate resistor of the input stage. What I didn't appreciate five years ago (!) is that all of this really had to be adjusted on test for my particular OPTs.
For this particular amp, a 3k9 feedback resistor gave me 10dB gNFB. A 750pF cap in series with a 2k resistor was connected in parallel with the feedback resistor. And in parallel with the input stage E88CC plate resistor, I used 10k in series with 200pF mica.
After this, the amp is running with gNFB, and without stability issues.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.