Hi Jaka Racman,
I am glad your simulations also show very good promise. I have not been able to work on the bench lately. I would like to try a p2p construct, maybe the fastest simplest way to get some actual experimental results. That could give a very compact circuit easy to modify.
The heavier load speeds up the oscillation, removing ripple. I suspect that the rms value of the ripple voltage is subtracted from the output voltage.
The circuit I would like to try would use the LM361 to switch two 6N137's which are biased on between the small signal power supply rails. The level shifted outputs would directly drive the inputs of the IR2110/IR2113. The open collectors of the optos would have pull-up resistors with values adjusted to provide the dead time.
I am glad your simulations also show very good promise. I have not been able to work on the bench lately. I would like to try a p2p construct, maybe the fastest simplest way to get some actual experimental results. That could give a very compact circuit easy to modify.
The heavier load speeds up the oscillation, removing ripple. I suspect that the rms value of the ripple voltage is subtracted from the output voltage.
The circuit I would like to try would use the LM361 to switch two 6N137's which are biased on between the small signal power supply rails. The level shifted outputs would directly drive the inputs of the IR2110/IR2113. The open collectors of the optos would have pull-up resistors with values adjusted to provide the dead time.
I finally remembered that I wanted to simulate the self-oscillating amp with a 1mH inductor in series with the 4 ohm speaker load. I found that the circuit becomes unstable as I feared. I am working on the phase compensation to try to resolve the problem. I am figuring that the inductance of an 8 ohm woofer is about 2mH.
schematic of another dig amp
Hi
I'm very new to this forum and also to pwm amplifiers.
My dream is to build one day a class d amp with 400w. I've looked around in the internet and picked up a interesting schematic of a pwm amp including pcbs.
if I hadn't found to that forum, i would have built the amp - now i'm not sure anymore!
here is the link: pwm amp by R. Balog
It is cheap but smart design. It hasnt got any error compensation. the pwm trigger signal section consists of low-cost elements unlike the MAX038 of Sergio's amp.
I'm interested to hear your opinions
and by the way: very interesting thread
Peter
Hi
I'm very new to this forum and also to pwm amplifiers.
My dream is to build one day a class d amp with 400w. I've looked around in the internet and picked up a interesting schematic of a pwm amp including pcbs.
if I hadn't found to that forum, i would have built the amp - now i'm not sure anymore!
here is the link: pwm amp by R. Balog
It is cheap but smart design. It hasnt got any error compensation. the pwm trigger signal section consists of low-cost elements unlike the MAX038 of Sergio's amp.
I'm interested to hear your opinions
and by the way: very interesting thread
Peter
Hi Peter,
only by fast looking:
A) VCC must be 18v and no more to avoid some parts damage. <40w@4Ohm is expected.
B) No any feedback...
C) triangle linearity for lm566 - 0.5%, Sergio will achieve better by opamp use.
D) dead time seems is very asymmetrical 500/150ns & complicated
E) 130kHZ enough for subwoofer applications only
F) P mosfet are used, slowest & more expensive
IMHO, it`s a very preliminary circuit-even 2 side PCB have top layer only, like education purpose weekend experiment are well...
Regards.
only by fast looking:
A) VCC must be 18v and no more to avoid some parts damage. <40w@4Ohm is expected.
B) No any feedback...
C) triangle linearity for lm566 - 0.5%, Sergio will achieve better by opamp use.
D) dead time seems is very asymmetrical 500/150ns & complicated
E) 130kHZ enough for subwoofer applications only
F) P mosfet are used, slowest & more expensive
IMHO, it`s a very preliminary circuit-even 2 side PCB have top layer only, like education purpose weekend experiment are well...
Regards.
Wau! ou ups - i think i better don't build this amp
I hoped that the cheap lm566 would be an alternative for max038. anyway, why not taking the CD4528BC (which is used in the Crest amp and is very cheap too) as an pulse wave generator?
i'm sorry for asking you probably stupid questions!
thanks, Peter
I hoped that the cheap lm566 would be an alternative for max038. anyway, why not taking the CD4528BC (which is used in the Crest amp and is very cheap too) as an pulse wave generator?
i'm sorry for asking you probably stupid questions!
thanks, Peter
Hi IVX,
I hope all is well with Charles. I have done a lot with the self-oscillating class D amp on the simulator. I wish I could build and test it now to see how it really works. The phase compensation and output inductors are optimized for what I can do with my simulator. The distortion appears very low when the input is low pass filtered at 300hz.
I hope all is well with Charles. I have done a lot with the self-oscillating class D amp on the simulator. I wish I could build and test it now to see how it really works. The phase compensation and output inductors are optimized for what I can do with my simulator. The distortion appears very low when the input is low pass filtered at 300hz.
Well. I have reached a point where my amp is almost finished, but still have a detail that I want to solve (with your help, if possible).
Here is a review of the current system, only for reference:
-PWM modulator: It is accomplished with a LM6172 dual opamp acting as a (very linear) triangle wave generator, using a LM319 comparator to compare with the error signal, that is passed via a NE5532 audio opamp.
-Level shifting: With two transistors, similar to Crest LT, but with 560 Ohm resistors instead of 1.5 to reduce switching times. It shifts the signal to 15Vpp referenced to the negative rail.
-Dead time control: using 4 XOR gates, the first 2 generate the direct and inverted version of the PWM signal and equaliza delays, and the other two act as buffers. The dead-time is implemented by delaying only the rising edge of both signals. Adjustable from near 0 to 500ns aprox.
-Power stage: Using a IR2110 driver IC that excites two N-channel mosfets with a 10 Ohms gate resistor. The mosfets are NTP35N15 from On-Semi's and there are no external diodes.
-Filter: 4-pole Butterworth filter with 2 inductors from Wilco and 2 ceramic 250V capacitors.
And now, the problem: I have measured a meaningful amount of switching noise (syncronized with the falling edge of the totem-pole's output) that is present in the audio input. It increases as the current to the load does so. I have found it to be present in the +/-15V supply and analog ground as well. I don't know if it is radiated or conducted. How can I decouple it? Perhaps something to do with the ground topology?
Thanks. If I solve this problem I can have a definitive version of the amp and publish it.
Best regards.
Sergio
Here is a review of the current system, only for reference:
-PWM modulator: It is accomplished with a LM6172 dual opamp acting as a (very linear) triangle wave generator, using a LM319 comparator to compare with the error signal, that is passed via a NE5532 audio opamp.
-Level shifting: With two transistors, similar to Crest LT, but with 560 Ohm resistors instead of 1.5 to reduce switching times. It shifts the signal to 15Vpp referenced to the negative rail.
-Dead time control: using 4 XOR gates, the first 2 generate the direct and inverted version of the PWM signal and equaliza delays, and the other two act as buffers. The dead-time is implemented by delaying only the rising edge of both signals. Adjustable from near 0 to 500ns aprox.
-Power stage: Using a IR2110 driver IC that excites two N-channel mosfets with a 10 Ohms gate resistor. The mosfets are NTP35N15 from On-Semi's and there are no external diodes.
-Filter: 4-pole Butterworth filter with 2 inductors from Wilco and 2 ceramic 250V capacitors.
And now, the problem: I have measured a meaningful amount of switching noise (syncronized with the falling edge of the totem-pole's output) that is present in the audio input. It increases as the current to the load does so. I have found it to be present in the +/-15V supply and analog ground as well. I don't know if it is radiated or conducted. How can I decouple it? Perhaps something to do with the ground topology?Thanks. If I solve this problem I can have a definitive version of the amp and publish it.
Best regards.
Sergio
hi Subwo1,
I have resimulated my circuit with 500uH and 1uF filter and 1mH 4R load. My feedback components are R12=39k, C1=680pF, R5=0. I have found no instability. You might also consider another comparator instead of LM361. I would suggest something with lower bias current and offset voltage. I had good results with Linear Technology LT1394.
Sergio,
do you have a 100MHz or faster scope ? Can you post pictures of totem pole switching ? Otherwise it would be very difficult to analyze where is your noise coming from. And grounding is not panacea for all problems in switching circuits. Anyway, if the circuit works otherwise, you might also consider adding lowpass filter at the input.
Regards, Jaka Racman
I have resimulated my circuit with 500uH and 1uF filter and 1mH 4R load. My feedback components are R12=39k, C1=680pF, R5=0. I have found no instability. You might also consider another comparator instead of LM361. I would suggest something with lower bias current and offset voltage. I had good results with Linear Technology LT1394.
Sergio,
do you have a 100MHz or faster scope ? Can you post pictures of totem pole switching ? Otherwise it would be very difficult to analyze where is your noise coming from. And grounding is not panacea for all problems in switching circuits. Anyway, if the circuit works otherwise, you might also consider adding lowpass filter at the input.
Regards, Jaka Racman
Thanks, Jaka.
What I did is adding a LPF in the feedback line and nothing, then I disconnected it and no change (but higher gain and noise, of course). So the noise doesn't come from the feedback.
I can try to add a low pass-filter to the input, but it should be connected backwards, with the capacitor facing to the input, shouldn't it? I will do some simulations so it doesn't affect to the overall frequency response.
Best regards.
Sergio
What I did is adding a LPF in the feedback line and nothing, then I disconnected it and no change (but higher gain and noise, of course). So the noise doesn't come from the feedback.
I can try to add a low pass-filter to the input, but it should be connected backwards, with the capacitor facing to the input, shouldn't it? I will do some simulations so it doesn't affect to the overall frequency response.
Best regards.
Sergio
Hi All.
Sergio,
It is strange... But how many you have amount (Db A weighed) of noise now?Source of noise triangle-gen may be?(imho for U4A a comparator is better choice). For amploiD I even did not measure owing to imperceptible hearing (probably bad idea). For example UcD-125db , MUETA-130db.
Regards.
Sergio,
It is strange... But how many you have amount (Db A weighed) of noise now?Source of noise triangle-gen may be?(imho for U4A a comparator is better choice). For amploiD I even did not measure owing to imperceptible hearing (probably bad idea). For example UcD-125db , MUETA-130db.
Regards.
Hi all.
Jaka Racman
I am also considering using an LM6172 so that I can use the U1a as a buffer and low pass filter and U1b as the comparator and driver for the 6N137. I am considering using just 1 optocoupler and generating the inverted signal at the negative rail by use of something looking like a 2 transistor monostable multivibrator. The collector of the opto would be the first of the 2 transistors and would still drive the IR2113 directly. The other transistor would drive the second input.
The simulated stability of the self-oscillating circuit improved when I added a .22uF, 15R zobel network across the 1mH, 4R load. The choke, L7 is also 500uH and filter the capacitor is 2uF. My R 12 is really too large to work in real life and then C1 would also have to be increased. I am also planning to set the LPF to roll off at around 300hz which does a lot for stability. I clip the input to keep the output at least 15 volts under the 90v power supply rails to give extra damping margin for the resonating speaker cone.
Sergio,
Make sure that when you connect your probe to ground that you don't still see the noise present. You may have to do some mental common mode rejection of the noise.
Jaka Racman
I am also considering using an LM6172 so that I can use the U1a as a buffer and low pass filter and U1b as the comparator and driver for the 6N137. I am considering using just 1 optocoupler and generating the inverted signal at the negative rail by use of something looking like a 2 transistor monostable multivibrator. The collector of the opto would be the first of the 2 transistors and would still drive the IR2113 directly. The other transistor would drive the second input.
The simulated stability of the self-oscillating circuit improved when I added a .22uF, 15R zobel network across the 1mH, 4R load. The choke, L7 is also 500uH and filter the capacitor is 2uF. My R 12 is really too large to work in real life and then C1 would also have to be increased. I am also planning to set the LPF to roll off at around 300hz which does a lot for stability. I clip the input to keep the output at least 15 volts under the 90v power supply rails to give extra damping margin for the resonating speaker cone.
Sergio,
Make sure that when you connect your probe to ground that you don't still see the noise present. You may have to do some mental common mode rejection of the noise.
Thanks, IVX.
The thing is that there is no audible noise, at least due to this reason, as it is high-frequency. The problem I see is only that I don't want any switching noise to couple to my A/V equipment, and it can easily reach 3V or even more peaks (over a 1V rms audio signal).
Refering to the possibility of the triangle generator as the source of noise, I don't believe that because the noise is only measurable when the current in the load is very high.
Subwo1,
I can even see the noise (peaks about half of that measured at the audio input) with the ground of the probe connected and measuring with the tip at the same point or some other ground point, but not when the tip of the probe is in the air. What do you think?
Do you think that a simple solution like the low-pass filter at the input can remove those transients?
The thing is that there is no audible noise, at least due to this reason, as it is high-frequency. The problem I see is only that I don't want any switching noise to couple to my A/V equipment, and it can easily reach 3V or even more peaks (over a 1V rms audio signal).
Refering to the possibility of the triangle generator as the source of noise, I don't believe that because the noise is only measurable when the current in the load is very high.
Subwo1,
I can even see the noise (peaks about half of that measured at the audio input) with the ground of the probe connected and measuring with the tip at the same point or some other ground point, but not when the tip of the probe is in the air. What do you think?
Do you think that a simple solution like the low-pass filter at the input can remove those transients?
Hello, Charles, nice to hear from you again.
Subwo1, do you mean that the noise may be radiated and couples to the probe thru the air? And do you think that the noise in the audio line is also that way? In that case, the best thing may be adding a LPF just at the input terminal of the amplifier.
Charles, what do you think of this? How can I make sure that the noise is there and how to decouple it?
Thanks to all for your help.
Subwo1, do you mean that the noise may be radiated and couples to the probe thru the air? And do you think that the noise in the audio line is also that way? In that case, the best thing may be adding a LPF just at the input terminal of the amplifier.
Charles, what do you think of this? How can I make sure that the noise is there and how to decouple it?
Thanks to all for your help.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.