I haven't been here in a bit, but I'm still tinkering with my amplifier design as time permits. Firstly I wanted to say thank you to everyone who offered advice in my original thread (https://www.diyaudio.com/forums/sol...sign-help-slew-rate-transistor-selection.html). We ended up redesigning it to drive the transistors directly with the opamp, which worked great to get the desired frequency bandwidth. However now I'm having trouble getting the current limiter to work correctly.
Long story short--this is a high voltage, low current Class A amp. Bandwidth 100Hz-100kHz. 350v pk-pk output swing, max 35mA into a 4.7k load.
The issue seems to be around C5--this cap was added to get good linearity at frequencies >20 kHz. Without C5 the current limiters imposed by Q5 and the bias point of Q3 work perfectly to clip the waveform across the entire frequency band. But then output waveform looks like junk at 100kHz.
With C5 in place, there is a 200uS current overshoot on positive and negative going waveforms, which becomes too large a duration for frequencies >2kHz. Basically the waveform is happening entirely in the overshoot region and I'm getting no current limit.
I've tried a few different version of current limiters--problem being they do not seem to work to clamp the input effectively. I've tried adding another NPN at the junction of R16/Q4b driven by an external voltage, but the opamp adjusts down and swings negative. Any clamp I put at the input of U2 has the same issue, only clamping half the input waveform + distorting the input signal.
I could try cutting the power rail, that might be the best solution. My other idea was a relay to cut C5 out of the circuit in case of overload (allowing the inherent limiters to work normally) but again, external logic to drive that and slow response time.
I'm sure there are lots of ideas I've not thought of--I'm open to suggestions!
Thanks,
Michael
Long story short--this is a high voltage, low current Class A amp. Bandwidth 100Hz-100kHz. 350v pk-pk output swing, max 35mA into a 4.7k load.
The issue seems to be around C5--this cap was added to get good linearity at frequencies >20 kHz. Without C5 the current limiters imposed by Q5 and the bias point of Q3 work perfectly to clip the waveform across the entire frequency band. But then output waveform looks like junk at 100kHz.
With C5 in place, there is a 200uS current overshoot on positive and negative going waveforms, which becomes too large a duration for frequencies >2kHz. Basically the waveform is happening entirely in the overshoot region and I'm getting no current limit.
I've tried a few different version of current limiters--problem being they do not seem to work to clamp the input effectively. I've tried adding another NPN at the junction of R16/Q4b driven by an external voltage, but the opamp adjusts down and swings negative. Any clamp I put at the input of U2 has the same issue, only clamping half the input waveform + distorting the input signal.
I could try cutting the power rail, that might be the best solution. My other idea was a relay to cut C5 out of the circuit in case of overload (allowing the inherent limiters to work normally) but again, external logic to drive that and slow response time.
I'm sure there are lots of ideas I've not thought of--I'm open to suggestions!
Thanks,
Michael
I don't think that capacitor is a good idea for several reasons - for a start it couples differential rail noise directly into the signal path.
Linearizing the voltage gain stage is probably best done by cascoding it (and the current source), and using larger emitter resistances for more degeneration (with 360V supply you can afford a few more volts for this).
Also the low loop gain of the feedback loop can be improved - rather than have it attempt x130 in one step, add an input stage with a gain around 10, and then you can increase the loop gain of the main feedback loop by 20dB or so, allowing it to do a much better job.
Linearizing the voltage gain stage is probably best done by cascoding it (and the current source), and using larger emitter resistances for more degeneration (with 360V supply you can afford a few more volts for this).
Also the low loop gain of the feedback loop can be improved - rather than have it attempt x130 in one step, add an input stage with a gain around 10, and then you can increase the loop gain of the main feedback loop by 20dB or so, allowing it to do a much better job.
Hmm, interesting idea. What is not shown is the preamp stage which has plenty of gain to spare. But I know from experience that this power amp does not do well once drive past 1.3vpk input. I've just been trying some things in simulation, it appears that my issue seems to stem from the opamp drive rather than the output transistors. If I reduce the loop gain by changing R1 to 50k and C4 to 15pF, the gain decreased, but if I try to drive the input higher it goes into distortion.
C5 seems to be keeping the opamp working nicely... adding additional emitter resistance doesn't seem to affect the linearity so far (up to 100R each).
Do you have any schematics of a cascode to help get me started? I assume it's not going to be as simple as adding a TIP50 with grounded grid above Q4 (and likewise for Q3a/b)?
C5 seems to be keeping the opamp working nicely... adding additional emitter resistance doesn't seem to affect the linearity so far (up to 100R each).
Do you have any schematics of a cascode to help get me started? I assume it's not going to be as simple as adding a TIP50 with grounded grid above Q4 (and likewise for Q3a/b)?
So I seemed to have stumbled upon a solution by accident... I was playing with the circuit in LTspice to add the cascode. I found adding a cascode to the TIP50 didn't help or hurt much, other than increasing complexity. Adding a cascode to the MJE5730 did help the high frequency response. Then I realized when I was biasing the upper cascode, I had moved C5 to act as a bypass capacitor on the base of the common-base amp. It dawned on me, why isn't C5 going to ground on the original design?
So I went back to the previous design, removed C5's connection from the opamp out and made it a ground bypass. Now I get high frequency response and no bizarre overdrive oscillation!
Does this make sense? Or have I found something that worked purely by accident? C5 was added to give the MJE's more base drive from the opamp. I'm not sure why it's working with just a bypass cap (I'll take it though!)
So I went back to the previous design, removed C5's connection from the opamp out and made it a ground bypass. Now I get high frequency response and no bizarre overdrive oscillation!
Does this make sense? Or have I found something that worked purely by accident? C5 was added to give the MJE's more base drive from the opamp. I'm not sure why it's working with just a bypass cap (I'll take it though!)