Straightforward power amp per Self, using CFP outputs. Low power, TIP3x outputs. I can't seem to cure HF oscillations on the negative side when it clips. Any remedy people have found effective? I did install the extra transistor to prevent collapse of the Vas stage, but the oscillation remains. Elliot uses a B-C cap on the lower driver, but that didn't have any effect for me.
I use 220pf on the lower driver. What did you use ?
Oscillation can be introduced by various points in the circuit.
VAS can need larger BC cap.
Sometimes 10R base stopper resistors on output transistors are needed.
Oscillation can be introduced by various points in the circuit.
VAS can need larger BC cap.
Sometimes 10R base stopper resistors on output transistors are needed.
I tried up to about 370 pF on the lower driver, a 2N4037. I use a 2-pole Vas compensation, 100 pF and 1000 pF on a 2-transistor Vas, but going back to single pole didn't help any. Haven't tried base stoppers, but will give that a shot. In all honesty, I don't really understand the mechanism involved during clipping- overall loop feedback is (I think) out of the picture at that point, so it's more likely something local?
CFPs can have local oscillations if the driver transistor is too *slow*. Adding c-b capacitance often makes it worse, not better.
Without knowing the circuit you are using, here is a cunning albeit crude method to reduce or remove RFO I have used many times with mosfet power amps, that due to the high speed of the OPT's are prone RFO, sometimes for hard to understand reasons. Being a bit cautious of high voltages (I've never had an issue) place the pad of your big finger gently on the back of the circuit board making sure you are not accidentally earthing yourself and move it around until you see the RF on the scope change, hopefully reducing it substantially. Then try a 10pf cap across the same two tracks and see what happens. This is a last ditch method and is not ideal but it often works. Make sure you check the output into an inductive load with a low frequency square wave to make ensure you haven't replaced one problem with another. If you are reluctant to use a finger then solder a small cap across a pair of spikes with plastic handles and do it that way.
Ha! One of my favorite methods. Alas, in this case the oscillation isn't affected by changes in capacitance to any node of the circuit. That's what makes me think it's something very fundamental. Today I'll socket the drivers and try some different speed and capacitance parts.
OK, I'm at least half an idiot. I think I was putting the B-C capacitor on the driver to the wrong traces. Now I'm getting the expected improvement, but still not happy with the overall clipping performance on the negative rail compared to the positive. Also, with enough C to kill off the oscillation (330 pF) the response is rolling off slightly. Maybe the CFP output configuration isn't as wonderful as described when used in a very simple circuit. Maybe I'm too fussy about clipping quality.
Post a schematic a generic for starts may no values needed
then you will get far more advice
( we like schematic porn anyway Conrad but i expect you already know that )
Regards
Sakis
then you will get far more advice
( we like schematic porn anyway Conrad but i expect you already know that )
Regards
Sakis
Hi Conrad,
The key difference of the conditions at clipping from the normal operating conditions of the amplifier - NFB is not working properly. In this case, most likely, your amp is showing local oscillation an the output stage. It is suppressed by NFB in normal operation, but comes to the scene as soon as NFB loop is "broken" by clipping. If you publish your topology, it will be much easier to recommend something particular.
Cheers,
Valery
The key difference of the conditions at clipping from the normal operating conditions of the amplifier - NFB is not working properly. In this case, most likely, your amp is showing local oscillation an the output stage. It is suppressed by NFB in normal operation, but comes to the scene as soon as NFB loop is "broken" by clipping. If you publish your topology, it will be much easier to recommend something particular.
Cheers,
Valery
Dear Conrad,
I'm currently faced with exact the same problem as you described. Did you find a final solution for it?
Best Regards,
Oskars
I'm currently faced with exact the same problem as you described. Did you find a final solution for it?
Best Regards,
Oskars
This is usually due to the use of complementary feedback transistors.
As discussed in another post, sometimes reducing the loop gain helps. This can be done by adding a resistor to the emitters of the drivers. I don't know the circuit - as posts above say we need to see this - but if you have 100 ohm base resistors on the output then 10 ohms might be enough for the degen resistors.
My conclusion from seeing this myself is that CFP are always more trouble than straight Darlingtons.
Other than adding a b-c cap across the PNP driver or alternatively a VAS collector to ground capacitor of about 220pF there is not much else to try.
As discussed in another post, sometimes reducing the loop gain helps. This can be done by adding a resistor to the emitters of the drivers. I don't know the circuit - as posts above say we need to see this - but if you have 100 ohm base resistors on the output then 10 ohms might be enough for the degen resistors.
My conclusion from seeing this myself is that CFP are always more trouble than straight Darlingtons.
Other than adding a b-c cap across the PNP driver or alternatively a VAS collector to ground capacitor of about 220pF there is not much else to try.
Hello John,
thank you for the reply. I already played around with VAS B-C capicitance, VAS C-GND capacitance and the PNP drivers C-B capacitance. No success. The drivers have allready a 100R base, 10R emiter and 100R collector resistor. As you suggest, I will try to play around with these values.
The schematic is the Load Invariant Design from D. Self.
As Conrad wrote, it must be something fundemantal. The oscillation seems to be independant from frequency (10 Hz to 20 kHz show the same behavior), load (4 or 8 Ohms, no load looks better) and supply voltage from +/-15 to +/- 32V behaves the same.
In all other situations the amp is stable. Square wave even at full load (until clipping starts) and 20kHz shows no sign of ringing.
In this discussion the saturation of the output devices is discussed.
Google Groups
This make sense for me, but how can it be avoided? In digital logic one can use a schotkky diode from base to collector(baker clamp) to speed up the shut down. No idea if it can be used in a power amp.
Best Regards,
Oskars
thank you for the reply. I already played around with VAS B-C capicitance, VAS C-GND capacitance and the PNP drivers C-B capacitance. No success. The drivers have allready a 100R base, 10R emiter and 100R collector resistor. As you suggest, I will try to play around with these values.
The schematic is the Load Invariant Design from D. Self.
As Conrad wrote, it must be something fundemantal. The oscillation seems to be independant from frequency (10 Hz to 20 kHz show the same behavior), load (4 or 8 Ohms, no load looks better) and supply voltage from +/-15 to +/- 32V behaves the same.
In all other situations the amp is stable. Square wave even at full load (until clipping starts) and 20kHz shows no sign of ringing.
In this discussion the saturation of the output devices is discussed.
Google Groups
This make sense for me, but how can it be avoided? In digital logic one can use a schotkky diode from base to collector(baker clamp) to speed up the shut down. No idea if it can be used in a power amp.
Best Regards,
Oskars
Ok, the google search for Baker Clamp lead me to allready existing DiyAudio topics on the same problem - A simple Baker clamp for the Blameless
Here the VAS and not the Output devices are supposed to get saturated. Lets try it out...
Here the VAS and not the Output devices are supposed to get saturated. Lets try it out...
Yes that should work too. The problem is that if the VAS or current source saturate the phase changes as it no longer inverts, it becomes in-phase and therefore a problem when it is supposed to be inverted.
You may need to play around with one or two diodes to check (the standard single transistor clamp uses an extra diode in the base of the transistor it is clamping and then a diode that becomes forward biased). With the Darlington you may want to use the clamp on both transistors and need the extra diode, perhaps.
You may need to play around with one or two diodes to check (the standard single transistor clamp uses an extra diode in the base of the transistor it is clamping and then a diode that becomes forward biased). With the Darlington you may want to use the clamp on both transistors and need the extra diode, perhaps.
A quick try with a 1N4148 diode across the Cdom did not help much. Thanks for all the tips. Maybe next week will bring new insights...
Have a nice weekend,
Oskars
Have a nice weekend,
Oskars
Does anything here help?
60-80W Power Amplifier
It is important to route the high current output stage grounds carefully
and not mix them with the input. The output stage zobel grounding is
also critical.
60-80W Power Amplifier
It is important to route the high current output stage grounds carefully
and not mix them with the input. The output stage zobel grounding is
also critical.
It really would help to see the circuit in outline.
Oscillations can be caused by simple wiring inductances.
Many circuits need the two grounds (signal and power) to be connected with a low value resistor (2-10 ohms) on the PCB to make sure that the grounds are electrically close at high frequencies.
Oscillations can be caused by simple wiring inductances.
Many circuits need the two grounds (signal and power) to be connected with a low value resistor (2-10 ohms) on the PCB to make sure that the grounds are electrically close at high frequencies.
Hi,
here is the schematic. It is more or less the schematic from D. Selfs Book APAD 6 Figure 10.12. The overload protection is ommited for clarity, as it is not used in the moment. In the simulation the rail sticking due to saturation is also visible, but not the oscillation. Tomorrow I will post some real screenshots from the oscilloscope.
View attachment Load_Invariant.asc
Currently there are two grounds on the PCB - Input (input & feedback network) and power (decoupling Caps). Both grounds are connected by a separate wire to the star point on the PSU PCB. I will try to add a 10R resistor to input ground.
Best Regards,
Oskars
here is the schematic. It is more or less the schematic from D. Selfs Book APAD 6 Figure 10.12. The overload protection is ommited for clarity, as it is not used in the moment. In the simulation the rail sticking due to saturation is also visible, but not the oscillation. Tomorrow I will post some real screenshots from the oscilloscope.
View attachment Load_Invariant.asc
Currently there are two grounds on the PCB - Input (input & feedback network) and power (decoupling Caps). Both grounds are connected by a separate wire to the star point on the PSU PCB. I will try to add a 10R resistor to input ground.
Best Regards,
Oskars
- Status
- Not open for further replies.