That's good news, the oscillation isn't feedback through the supplies. The only option is to reduce the gain at 1MHz+, which is done with the circled capacitor (C6?). As moschfet, said, increasing its value will reduce the bandwidth. 33-47pF should reduce the gain enough for that.
The capacitor across R14 reduced the open-loop gain a little, but also made the drive asymmetric which is why I was suspicious of it. It looks like the manufacturer had a stability problem that they 'solved' with the added capacitor, but C6 is a better place to control the BW.
The capacitor across R14 reduced the open-loop gain a little, but also made the drive asymmetric which is why I was suspicious of it. It looks like the manufacturer had a stability problem that they 'solved' with the added capacitor, but C6 is a better place to control the BW.
valgamaa & moschfet
After changed C4 > 10p (orig was: 22p) and Miller cap (C6) > 33pF (orig was: 10p), the osc. signal amplitude decreased to 0.97Vpp and 1.61MHz.So it is better but still exist. If i apply a 1kHz signal to the input AND connct the 8 ohm load, the osc. amplitude increases, so it is sensitive to the load. (?)
Interestingly, a later release (VX1200) added two capacitors and increased the miller capacitance to 56pF (!)
Nelson Pass:
Thank You! Yes, I removed the C6 from the original drawing earlier. Unfortunately it has no effect.
Thank You! Yes, I removed the C6 from the original drawing earlier. Unfortunately it has no effect.
C9 is really the only variable there to control stability. Try a large value to make the amplifier stable, then look at a square wave to check the performance. If there is ringing then you may need to increase the cap further.
Thanks valgamaa!
I am beginning to seriously believe that this is how it came out of the factory. In fact, all these models would produce something similar, but no one has looked at it with an oscilloscope 🙂
Anyway, I've ordered MPSA.. transistors and will try to replace them, see if the semiconductors can age 🙂
I am beginning to seriously believe that this is how it came out of the factory. In fact, all these models would produce something similar, but no one has looked at it with an oscilloscope 🙂
Anyway, I've ordered MPSA.. transistors and will try to replace them, see if the semiconductors can age 🙂
Highly unlikely, these were for many years a staple of the more established touring sound companies in the UK, known as a bomb proof workhorse. Before swapping caps, I suspect there is a genuine fault that has not been found yet.
If you are convinced it came from the factory inadequately stabilized, you need to go through the entire design to analyze and compensate properly; just slapping in a new cap value won't fix it.
I think SubSoniks is on to something; before you got it, somebody tried to "improve" it and changed things or something was damaged (I've seen food or drink spills contaminate PCBs that caused big problems).
Looks like this is amp has a much larger than normal voltage gain; need to keep that in mind for your stability analysis, too.
Good luck ...
mlloyd1
I think SubSoniks is on to something; before you got it, somebody tried to "improve" it and changed things or something was damaged (I've seen food or drink spills contaminate PCBs that caused big problems).
Looks like this is amp has a much larger than normal voltage gain; need to keep that in mind for your stability analysis, too.
Good luck ...
mlloyd1
@mickeyratt
It may be worth a visit to https://forum.speakerplans.com/amp-forum_forum24.html
There a few very experienced repair people on that forum who will know the HH V800 inside out👍
It may be worth a visit to https://forum.speakerplans.com/amp-forum_forum24.html
There a few very experienced repair people on that forum who will know the HH V800 inside out👍
Without hands on experience, it's dangerous to criticize. But what I see is a lot of doing the wrong things to stabilize an amp that maybe has too much loop gain, and other problems. Adding a lot of "HF suppression" inside the feedback loop is a recipe of failure.
The things that I see are:
1. C7 without a series resistor is not a proper Zobel network and probably a bad idea.
2. C8 and C5 are probably a bad idea. I have no issue with C6.
3. 1K gate resistors are a pole around 150KHz. 1K seems excessive. Maybe there is a local FET stability issue? Normally I would expect about 100 Ohms.
4. C4 without a series resistor is a bad idea.
5. TR1 and TR2 should probably have some degeneration.
Physical layout and grounding can cause nasty problems that don't show in simulations and the schematic. If that's the case, it can only be solved be a thorough hands analysis and well-chosen redesign. Maybe C7 shares a ground path with the input?
Good luck. I suspect someone else has tried and fail to solve this problem.
The things that I see are:
1. C7 without a series resistor is not a proper Zobel network and probably a bad idea.
2. C8 and C5 are probably a bad idea. I have no issue with C6.
3. 1K gate resistors are a pole around 150KHz. 1K seems excessive. Maybe there is a local FET stability issue? Normally I would expect about 100 Ohms.
4. C4 without a series resistor is a bad idea.
5. TR1 and TR2 should probably have some degeneration.
Physical layout and grounding can cause nasty problems that don't show in simulations and the schematic. If that's the case, it can only be solved be a thorough hands analysis and well-chosen redesign. Maybe C7 shares a ground path with the input?
Good luck. I suspect someone else has tried and fail to solve this problem.
Have you tried the following:
Load the amplifier with resistive load, feed it a 1kHz signal and let the output signal approach 50% before clipping. One probe/channel connected to the output binding post, the other probe in your hand... and then poke each gate with the probe; first with 1:10 and look for oscillations on either channel. If you do not see any oscillations, change to 1:1 on the probe in your hand, and repeat.
The bad MOSFET will show you oscillations. The oscillations will be visible on the output channel, but also on the probe you have in your hand, connected to the offending MOSFET gate... this is how you'll know which MOSFET out of all of them is faulty .. or starting to break the gate capacitance.... without the need to remove one by one.
The above never fails. I used this approach when I worked for Jands electronics... and built, tested, and soaked virtually thousands of amps, BJT and MOSFETs alike... for PA and high power HiFi.
The above will either show a bad MOSFET, or you'll know it is not the MOSFETs to blame.
Good luck.
SubSoniks | mlloyd1 | steveu | Extreme_Boky
Thank you all for your valuable suggestions! (diyaudio's system didn't notify me that there were new posts, so I was late in replying, sorry!)So I replaced all the BJTs (Q1,2,3,4,5) with new ones, of course no change. However, the 3 changes in the attached picture that you suggested here in the thread reduced the oscillation to 100mVp-p. I will try to see with a square wave (I only have a sine wave generator for now) how it behaves in pulse mode. Until I changed the C4 from 22pF to 10pF, driving with a 1kHz sine wave, I had serious instability problems in the quarter period after the positive and negative peaks of the sine wave. It was very spectacular on an oscilloscope, but unfortunately I didn't take a photo of it. This disappeared with the 10pF value of C4.
To be continued 🙂
Thank You!
@valgamaa
"C5 doesn't do anything at all (current source into a (nominally) virtual earth)."
Not really. If C5 were not there, R15 would act with Q4's Miller capacitance to create a lower frequency time constant. Including C5 minimises the phase delay that would cause.
However, the design does not look well designed, I would agree.
The capacitor across R14 and C6 are obviously (?) intended to reduce the gain of the differential pair at high frequencies, presumably because the designer thinks that would mean Q5 behaving more like a singleton VAS, which should make (I assume) stability easier with a single Miller capacitor (the one across Q5 circled in red in the post above). And suggested to be 47pF, while C4 is reduced. A resistor in series with C4 would limit the capacitive loading effect at high frequencies too. Too high a capacitance can indeed be a problem. And no series resistor with C7 either!
"C5 doesn't do anything at all (current source into a (nominally) virtual earth)."
Not really. If C5 were not there, R15 would act with Q4's Miller capacitance to create a lower frequency time constant. Including C5 minimises the phase delay that would cause.
However, the design does not look well designed, I would agree.
The capacitor across R14 and C6 are obviously (?) intended to reduce the gain of the differential pair at high frequencies, presumably because the designer thinks that would mean Q5 behaving more like a singleton VAS, which should make (I assume) stability easier with a single Miller capacitor (the one across Q5 circled in red in the post above). And suggested to be 47pF, while C4 is reduced. A resistor in series with C4 would limit the capacitive loading effect at high frequencies too. Too high a capacitance can indeed be a problem. And no series resistor with C7 either!
"1. C7 without a series resistor is not a proper Zobel network and probably a bad idea." -Steveu is right. The output stage filter looks like a kludge between one Ed Cherry proposed (L//R followed by C) and the usual R-C followed by L//R. I would hesitiate to remove C7, but that is what Cherry's approach would suggest. Instead adding a 4.7 or 10 ohm resistor (a couple of watts) in series with C7 and removing C12 might be a start.
Though it might be that C12 is no longer functioning.
Though it might be that C12 is no longer functioning.
The lack of drivers between the VAS and FETs is a double issue. I estimate Q5 and Q3 will run about (70V*10mA=) 700mW, just over the rated maximum. But even so, an estimated 20mA peak output from the VAS will slew limit about 3V/uS, ie about 6KHz. Bob Cordel is very clear, and not very polite about this mistake. I think it can work for smaller amps, but not one that runs +/-70VDC.
I would not parallel MOSFETs without a current sharing source resistance.
I suppose the manufacturers would be able to match devices as they would have purchased 100's at a time, but diy-ers might not be able to.
When MOSFETs first arrived there were claims that "parallelling can be easily achieved" but there are variations in Vt.
Anyone know what FETs are used? The gate capacitance might be high and with 1k gate resistors that may be a frequency limitation. As steveu also points out it is not a good idea to drive FETS from the VAS without driver transistors.
I suppose the manufacturers would be able to match devices as they would have purchased 100's at a time, but diy-ers might not be able to.
When MOSFETs first arrived there were claims that "parallelling can be easily achieved" but there are variations in Vt.
Anyone know what FETs are used? The gate capacitance might be high and with 1k gate resistors that may be a frequency limitation. As steveu also points out it is not a good idea to drive FETS from the VAS without driver transistors.
Yes, especially considering that there is no thermal compensation. The bias spreader is just resistors, and the VAS current appears to be slightly thermal positive. I thought maybe they depend on lateral FETs, but the bias divider works out to a range of about 2V to 12V, which implies vto~2x4V , ie not Lat-FETs. Assuming this amp has not blown up from thermal runaway, I have to wonder why? What are the MOSFET parts?