Hello all.
There is something that most builders of Class-D amplifiers may have suffered: turn on/off ticks/pops.
I am currently using a IR2113 for gate driving, using the SD pin for turn on/off. There is always an annoying click at the speaker, specially at turn on. First I used a relay to turn on/off several modules at the same time, shorting that pin to GND or driver's VCC, and I thought that the transient was due to the relay bouncing. But now I have done it with a transistor that has no bounce and it still remains the same.
I have measured it and consist of a dirty and damped oscillation that reachs around 20Vpp for 1 cycle or two, and its shape seems to depend on the load nature.
I think that the main cause is the LC filter transient response, and perhaps bootstrap circuit estabilisation. I am sure that some designs don't suffer from that, but can anyone point to a way of avoiding that? (apart from the obvious solution of using relays)
Thanks!
There is something that most builders of Class-D amplifiers may have suffered: turn on/off ticks/pops.
I am currently using a IR2113 for gate driving, using the SD pin for turn on/off. There is always an annoying click at the speaker, specially at turn on. First I used a relay to turn on/off several modules at the same time, shorting that pin to GND or driver's VCC, and I thought that the transient was due to the relay bouncing. But now I have done it with a transistor that has no bounce and it still remains the same.
I have measured it and consist of a dirty and damped oscillation that reachs around 20Vpp for 1 cycle or two, and its shape seems to depend on the load nature.
I think that the main cause is the LC filter transient response, and perhaps bootstrap circuit estabilisation. I am sure that some designs don't suffer from that, but can anyone point to a way of avoiding that? (apart from the obvious solution of using relays)
Thanks!
Yes, of course!
Please give me some time in order to sort it and draw a figure.
Meanwhile, a little description:
- Clock (triangle) based design, oscillating at 250KHz.
- Output filter: 20uH and 440nF, with 18 ohm + 120nF zobel in parallel with the output.
- Error amplifier: inverting configuration: 2k input resistor, 180k local feedback resistor. 6k8+330pF in parallel with that resistor. Feedback is taken AFTER output filter, through a 100k || 47pF network.
- Error amplifier is a NE5532 fed at +/-6V
- Output stage: IR2113, 10uF tantalum bootstrap capacitor.
- Gate resistors: 6.8 ohm in antiparallel with schottky.
That should give an idea of the circuit.
Anyway, I will do some test in open-loop (increasing the 2k resistor to something like 47K to reduce gain), to see if the click still happens.
Best regards,
Pierre.
Please give me some time in order to sort it and draw a figure.
Meanwhile, a little description:
- Clock (triangle) based design, oscillating at 250KHz.
- Output filter: 20uH and 440nF, with 18 ohm + 120nF zobel in parallel with the output.
- Error amplifier: inverting configuration: 2k input resistor, 180k local feedback resistor. 6k8+330pF in parallel with that resistor. Feedback is taken AFTER output filter, through a 100k || 47pF network.
- Error amplifier is a NE5532 fed at +/-6V
- Output stage: IR2113, 10uF tantalum bootstrap capacitor.
- Gate resistors: 6.8 ohm in antiparallel with schottky.
That should give an idea of the circuit.
Anyway, I will do some test in open-loop (increasing the 2k resistor to something like 47K to reduce gain), to see if the click still happens.
Best regards,
Pierre.
The thing is that I haven't tried with a smaller cap.
My fear is that, when the amp clips (i.e. one of the mosfets is always on while the other is always off), the behaviour won't be as good as now as the bootstrap cap will eventually discharge.
But take for granted that I will test as there is margin for a reduction.
IVX: How can I get it pre-charged? That won't be easy to achieve practically, is it?
My fear is that, when the amp clips (i.e. one of the mosfets is always on while the other is always off), the behaviour won't be as good as now as the bootstrap cap will eventually discharge.
But take for granted that I will test as there is margin for a reduction.
IVX: How can I get it pre-charged? That won't be easy to achieve practically, is it?
Thanks. I am all ears...
The amp is performing quite well, sound is very nice and seems also quite reliable (playing for some months at several parties with volume at full power during 7-10 hours). It is now using IRF640N mosfets. They did make the difference in my opinion.
But this "click" is quite annoying... let's see if we can fix it together ;-)
I will post an sch as soon as I have cleaned and updated it!
Best regards,
Pierre
The amp is performing quite well, sound is very nice and seems also quite reliable (playing for some months at several parties with volume at full power during 7-10 hours). It is now using IRF640N mosfets. They did make the difference in my opinion.
But this "click" is quite annoying... let's see if we can fix it together ;-)
I will post an sch as soon as I have cleaned and updated it!
Best regards,
Pierre
I had 10uF tantalum in parallel with 100nF ceramic capacitors as bootstrap ones.
I have removed the 10uF one, leaving only 100nF, as suggested by Charles, and the turn on transient is definitely smaller (around 10Vpp max. instead of 20Vpp). However, as expected, the operation is not correct with a 20Hz clipping sine. I have then added 1uF ceramic (instead of 10uF) and the transient is almost the same (aroun 12Vpp) and it also works properly at clipping.
It is now less audible, but still present.
I have also tried the pre-charge trick (30K resistor from bootstrap positive terminal, with 13V zener in parallel with cap), and I have measured no difference.
Well, it seems that the thing has improved, but I think there must be something more...
BTW: The turn off transient hasn't changed with the cap modification ¿?
I have removed the 10uF one, leaving only 100nF, as suggested by Charles, and the turn on transient is definitely smaller (around 10Vpp max. instead of 20Vpp). However, as expected, the operation is not correct with a 20Hz clipping sine. I have then added 1uF ceramic (instead of 10uF) and the transient is almost the same (aroun 12Vpp) and it also works properly at clipping.
It is now less audible, but still present.
I have also tried the pre-charge trick (30K resistor from bootstrap positive terminal, with 13V zener in parallel with cap), and I have measured no difference.
Well, it seems that the thing has improved, but I think there must be something more...
BTW: The turn off transient hasn't changed with the cap modification ¿?
Thanks, Workhorse.
However, your suggestions is a delayed start-up in order to allow for the rest of the circuit (specially the modulator and its supplies) to estabilize before allowing switching.
My circuit already has provisions for that, there is a 555 circuit that controls the SD pin and implements a delay of about 2 seconds. Everything is working by that moment, and then switching starts and the "click" is produced.
So I'm afraid that's not the cause of my small problem.
However, your suggestions is a delayed start-up in order to allow for the rest of the circuit (specially the modulator and its supplies) to estabilize before allowing switching.
My circuit already has provisions for that, there is a 555 circuit that controls the SD pin and implements a delay of about 2 seconds. Everything is working by that moment, and then switching starts and the "click" is produced.
So I'm afraid that's not the cause of my small problem.
Charles,
I recall it was quite big when no load is present (so much that its output falls outside the triangle amplitude and hence it doesn't start). That offset was caused by a negative voltage in the mosfets output, not the integrator itself.
With a reasonable load (<500 ohm or so), it starts ok. I will check the exact voltage with my current test load (speaker) before enabling the output, but I guess it is quite small (I am sure that it is below 100mV).
Best regards,
Pierre
I recall it was quite big when no load is present (so much that its output falls outside the triangle amplitude and hence it doesn't start). That offset was caused by a negative voltage in the mosfets output, not the integrator itself.
With a reasonable load (<500 ohm or so), it starts ok. I will check the exact voltage with my current test load (speaker) before enabling the output, but I guess it is quite small (I am sure that it is below 100mV).
Best regards,
Pierre
Hi Pierre,
Have you checked the Transient response in Open-Loop configuraton, and also try to disngage the error amplifier's output to the comparator's input and then check again for transient, you may aslo try to replace the error amp[NE5532] with some other type....
You can also share your schematic with us , so we can review it and might help you with that....problem
regards,
K a n w a r
Have you checked the Transient response in Open-Loop configuraton, and also try to disngage the error amplifier's output to the comparator's input and then check again for transient, you may aslo try to replace the error amp[NE5532] with some other type....
You can also share your schematic with us , so we can review it and might help you with that....problem
regards,
K a n w a r
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.