As part of the resurection project of the rock pig amplifier I've been encountering some stability problems (to be expected I understand)
I have replaced the output stage with exicon 10p20 &10n20 devices & also replaced almost every capicitor on the boards.
Now I have done quite some fiddling & experimenting in my search for stability, studdied several similar circuit diagrames and spoken to a couple of tech friends.
Now I probably could have cured this by some of the methods "typically" used But I am convinced that there are some universaly practised methods on these fet amps that while achieving results are done for the wrong reasons.
Almost without exception I see caps taged across the gate to source ( suposedly capacitance equalisation caps) but the comments have been that the things arent stable without them!???
I also believe that the exicon fets while possibly superior to the old hitachies in some ways are more prone to occilation.
I also have been told thet the 10p/n20 units are more prone than the 1p/n16.
Now I can tell you right now that the pcb layout is less than ideal but I can't do much about that and to a certain extent thats not the point.
I have seen a number of measures on various circuits that seem to be "good practise" or "a good idea" I have tested some of these & they don't seem to get the solid result I am expecting.
the most important factor most significant result without other effects so far has been making sure the rails are properly bypassed to earth at a number of pionts.
changing the gate resistors dosn't seem to make a blind bit of difference.
Bypassing the bias pot with a .1uF seems to have its moments both ways depending on what else happens.
and so on
What we are talking about here is a pretty typical 80's style straight out of the app notes circuit common source 8 TO3 output devices.
My quest here is not to simply fix the thing but to get some sound understanding of the causes.
From comments of much better traveled techs than myself, I seem to gather a lot of similar amps like this have inherent problems.
One bloke I spoke to ( who has amps like this across his bench every day) said "arrh mate you just gota poke arround & try things untill it works"
So what do you recon fellas
I have replaced the output stage with exicon 10p20 &10n20 devices & also replaced almost every capicitor on the boards.
Now I have done quite some fiddling & experimenting in my search for stability, studdied several similar circuit diagrames and spoken to a couple of tech friends.
Now I probably could have cured this by some of the methods "typically" used But I am convinced that there are some universaly practised methods on these fet amps that while achieving results are done for the wrong reasons.
Almost without exception I see caps taged across the gate to source ( suposedly capacitance equalisation caps) but the comments have been that the things arent stable without them!???
I also believe that the exicon fets while possibly superior to the old hitachies in some ways are more prone to occilation.
I also have been told thet the 10p/n20 units are more prone than the 1p/n16.
Now I can tell you right now that the pcb layout is less than ideal but I can't do much about that and to a certain extent thats not the point.
I have seen a number of measures on various circuits that seem to be "good practise" or "a good idea" I have tested some of these & they don't seem to get the solid result I am expecting.
the most important factor most significant result without other effects so far has been making sure the rails are properly bypassed to earth at a number of pionts.
changing the gate resistors dosn't seem to make a blind bit of difference.
Bypassing the bias pot with a .1uF seems to have its moments both ways depending on what else happens.
and so on
What we are talking about here is a pretty typical 80's style straight out of the app notes circuit common source 8 TO3 output devices.
My quest here is not to simply fix the thing but to get some sound understanding of the causes.
From comments of much better traveled techs than myself, I seem to gather a lot of similar amps like this have inherent problems.
One bloke I spoke to ( who has amps like this across his bench every day) said "arrh mate you just gota poke arround & try things untill it works"
So what do you recon fellas
How long are the wires to the FETs? They are very sensitive to inductance.
As for the gate/source caps; if you look in the app notes, it says that you can either use a capacitor to equalize the speed of N and P channel devices, or use unequal gate resistors. I usually use just the gate resistors to avoid an extra component and it has always worked alright.
What about a Zobel network on the output? Sometimes it helps a bit.
What type and value of capacitors are you using for supply bypassing? I would recommend at least 100uF electrolytic + 0.1uF ceramic. The 0.1uF capacitor must be low inductance, hence why I tend to use ceramic.
Could you post the schematic? Maybe there is some particular issue with it.
MOSFET oscillation can certainly be a pain.
As for the gate/source caps; if you look in the app notes, it says that you can either use a capacitor to equalize the speed of N and P channel devices, or use unequal gate resistors. I usually use just the gate resistors to avoid an extra component and it has always worked alright.
What about a Zobel network on the output? Sometimes it helps a bit.
What type and value of capacitors are you using for supply bypassing? I would recommend at least 100uF electrolytic + 0.1uF ceramic. The 0.1uF capacitor must be low inductance, hence why I tend to use ceramic.
Could you post the schematic? Maybe there is some particular issue with it.
MOSFET oscillation can certainly be a pain.
Yeah, post the schematic. Common source output stages can be more prone to instability due to the gain/feedback, but there's a couple of tricks I know to tame them 🙂
I'll see if I can scan the circuit at some time, Its prety run of the mill.
As I said Iv'e read the exicon notes and aa bit of other stuff, I'm concerned not only with this amp but this type of design in general, and wisdom that is not contained in std notes.
As I sort of mentioned this seems to be a problem with a number of "respected " brand amps as well.
as for the source leads there is a drive track runs down the array with source resistors ( leads reasonably short ) tacked across to the source pins.
The zobel net doesn't make a blind bit of dif at this point, It will when it gets connected to a real load Im sure.
Yeah I'm prety fussy about caps & adding more bypas to the rails has shown improvements (i think).
As I said Iv'e read the exicon notes and aa bit of other stuff, I'm concerned not only with this amp but this type of design in general, and wisdom that is not contained in std notes.
As I sort of mentioned this seems to be a problem with a number of "respected " brand amps as well.
as for the source leads there is a drive track runs down the array with source resistors ( leads reasonably short ) tacked across to the source pins.
The zobel net doesn't make a blind bit of dif at this point, It will when it gets connected to a real load Im sure.
Yeah I'm prety fussy about caps & adding more bypas to the rails has shown improvements (i think).
At what frequencies, what amplitudes and in what circumstances does it oscillate?
[Some circumstances to look for are : Near Zero current crossing coming from the positive and negative sides, near positive clipping, near negative clipping, light load, heavy load, capacitive load, no load, etc..]
Oscillations abobe 5Mhz are usually due to local feedback while oscillations below 2.5Mhz are usually due to global feeback. Local oscillations in MOSFET circuits happen usually between 10 and 100Mhz
Also, the higher the amplitude of the oscillation the higher the amount of '180 degree feedback' that is causing it and the harder to supress it
[Some circumstances to look for are : Near Zero current crossing coming from the positive and negative sides, near positive clipping, near negative clipping, light load, heavy load, capacitive load, no load, etc..]
Oscillations abobe 5Mhz are usually due to local feedback while oscillations below 2.5Mhz are usually due to global feeback. Local oscillations in MOSFET circuits happen usually between 10 and 100Mhz
Also, the higher the amplitude of the oscillation the higher the amount of '180 degree feedback' that is causing it and the harder to supress it
Without any "measures of restraint" it freely oscilates at about 8.5 Mhz, at an amplitude of about 20V peak to peak.
With the gates straped to the rails with 220pF caps it is fine with no load.
Currently it will behave at first but as I wind up the signal to about 10V P-P out it starts to break as it is driven harder it come back into line. ... When it is stone cold with a 4 or 8 ohm load.
Its seems to fail to misbehave reliably when warm?!!!
I havn't touched the thing for about a week But I've been doing some thinking.
I recon the problem is between the bias/drive chain & the gates of the fets.
The tracks leading to the gate resistors are a bit long & I recon there could be some inductance stuff going on.
Its also dawned on me that because the problem is at 8.5Mhz I need to be thinking "RF" which means some things may not be as they appear.
I recon some ferite beads on the gate resistors will help things. I grab some next time I do a parts order.
meanwhile I'd better get some paying jobs out the door.
cheers
With the gates straped to the rails with 220pF caps it is fine with no load.
Currently it will behave at first but as I wind up the signal to about 10V P-P out it starts to break as it is driven harder it come back into line. ... When it is stone cold with a 4 or 8 ohm load.
Its seems to fail to misbehave reliably when warm?!!!
I havn't touched the thing for about a week But I've been doing some thinking.
I recon the problem is between the bias/drive chain & the gates of the fets.
The tracks leading to the gate resistors are a bit long & I recon there could be some inductance stuff going on.
Its also dawned on me that because the problem is at 8.5Mhz I need to be thinking "RF" which means some things may not be as they appear.
I recon some ferite beads on the gate resistors will help things. I grab some next time I do a parts order.
meanwhile I'd better get some paying jobs out the door.
cheers
Wire wound resistors? no not on the gates.
there are wire wounds as current shares on the sources and one way back in the drive circuit. but nothingg else will stand up in those locations.
feedback stability..... dont know?
there are wire wounds as current shares on the sources and one way back in the drive circuit. but nothingg else will stand up in those locations.
feedback stability..... dont know?
Try changing those source resistors. Like I said, they are very sensitive to inductance, so it might be just enough to push them into instability.
With inductive wirewound resistors in the source leads you now have a Colpitts oscillator. I have seen that sort of thing band-aided by putting 100nF-1uF between one source and the next. Was a problem with the old ETI Series 5000 mosfet amp; the prototype was done with one brand of resistor and home constructors often used another brand with a different parasitic inductance.
LOWLEVEL,
I meant your Source resistors, wirewound resistor use to be high power resistors.
Talking about wirewound resistors;
The Elektor Crescendo amp suffered the same problem because of wirewounded resistors at the FET Sources.
A couple of issues later Elektor did actually have a small note that these resistors should be replaced by carbon film, but many don't seem to know this because I hear people sometimes talking about Crescendo NOT being stable, at least in the present...
It's all about engineering intelligence,
Cheers! 😉
I meant your Source resistors, wirewound resistor use to be high power resistors.
Talking about wirewound resistors;
The Elektor Crescendo amp suffered the same problem because of wirewounded resistors at the FET Sources.
A couple of issues later Elektor did actually have a small note that these resistors should be replaced by carbon film, but many don't seem to know this because I hear people sometimes talking about Crescendo NOT being stable, at least in the present...
It's all about engineering intelligence,
Cheers! 😉
- Status
- Not open for further replies.