Daily Report:
Good morning to everybody!
Well, yesterday I did the modifications (bypassing the supply of the IR2110, reducing the zener resistor... and things improved a lot.
In some moment, I managed to measure a much cleaner square wave at the output (with perhaps 8V overshoot and some fast ringing just after the rising and falling edges). With no load, the MOSFETS got only warm, so I decided to connect a filter and modulate with a sine wave: 33uH + 1uF capacitor. With a 10 Ohms/50W resistor, the output looked GREAT, and I was able to put all the power on it. Then I moved to 5 Ohms/100W, and everything went perfect.
The thing is that, after some tests and playing with the dead-time and DC adjustments, the mosfets started to get hot, and now I am in a point where they run very hot even with no load and filter. The waveforms at the input of the IR2110 have a lot of ringing and the average current on the + rail is about 170mA. (that disappears when I shutdown the IR2110). This must be due to shoot-through, musn't it?
Is there any other cause for the mosfets to remain conducting from one rail to the other due to degrading or something?
Well, at least, I saw it working once, so I must work again! Now I think I am in the right direction.
Best regards
Good morning to everybody!
Well, yesterday I did the modifications (bypassing the supply of the IR2110, reducing the zener resistor... and things improved a lot.
In some moment, I managed to measure a much cleaner square wave at the output (with perhaps 8V overshoot and some fast ringing just after the rising and falling edges). With no load, the MOSFETS got only warm, so I decided to connect a filter and modulate with a sine wave: 33uH + 1uF capacitor. With a 10 Ohms/50W resistor, the output looked GREAT, and I was able to put all the power on it. Then I moved to 5 Ohms/100W, and everything went perfect.
The thing is that, after some tests and playing with the dead-time and DC adjustments, the mosfets started to get hot
, and now I am in a point where they run very hot even with no load and filter. The waveforms at the input of the IR2110 have a lot of ringing and the average current on the + rail is about 170mA. (that disappears when I shutdown the IR2110). This must be due to shoot-through, musn't it?Is there any other cause for the mosfets to remain conducting from one rail to the other due to degrading or something?
Well, at least, I saw it working once, so I must work again! Now I think I am in the right direction.
Best regards
Hi ssanmor
Hard to say from remote where this comes from. But this is definitely increrasing idle power. Fed with this signal, your driver is despearately trying to turn the fets on and off several times for each transition, which gives a rise of dissipation in both the driver and the mosfets. I once managed to kill mosfet drivers this way !
What do your rail voltages look like ? How much ripple is there ?
Regards
Charles
Hard to say from remote where this comes from. But this is definitely increrasing idle power. Fed with this signal, your driver is despearately trying to turn the fets on and off several times for each transition, which gives a rise of dissipation in both the driver and the mosfets. I once managed to kill mosfet drivers this way !
What do your rail voltages look like ? How much ripple is there ?
Regards
Charles
Thanks, Charles.
I need to do more measurements on this. The power supply has some 100Hz ripple (it only has 3000uF per rail plus 470uF on-board), but this shouldn't be a problem when idle, even with a 170mA consumption.
Yesterday, when I connected filter+load, the ripple was very noticeable (about 4Vpp) when putting some power (70-80W) onto the load, but that was expected. In the real application, I plan to add BIG supply capacitors (I have two 0.1F/63V capacitors waiting to be used), with the proper inrush current limiting circuitry at the primary of the transformer, of course.
The problem of the strange waveform arises when I turn on the driver with the SD pin. I have tried with another IR2110 but no improvement. I will change the Mosfets also, perhaps they have some problem.
What are the diodes in parallel with the mosfets going to improve? Is it better to fine-tune all without them first?
By the way, the coil I used for the filter seemed to work very well. It was one of the Wilco chokes (33uH/20A). Ref: DC3-330. They sent me as samples.
I need to do more measurements on this. The power supply has some 100Hz ripple (it only has 3000uF per rail plus 470uF on-board), but this shouldn't be a problem when idle, even with a 170mA consumption.
Yesterday, when I connected filter+load, the ripple was very noticeable (about 4Vpp) when putting some power (70-80W) onto the load, but that was expected. In the real application, I plan to add BIG supply capacitors (I have two 0.1F/63V capacitors waiting to be used), with the proper inrush current limiting circuitry at the primary of the transformer, of course.
The problem of the strange waveform arises when I turn on the driver with the SD pin. I have tried with another IR2110 but no improvement. I will change the Mosfets also, perhaps they have some problem.
What are the diodes in parallel with the mosfets going to improve? Is it better to fine-tune all without them first?
By the way, the coil I used for the filter seemed to work very well. It was one of the Wilco chokes (33uH/20A). Ref: DC3-330. They sent me as samples.
Hi Sergio
You can improve the noise performance of the input logic of the IR2110 by having VSS and com bypassed seperately bypassed to ground and having them interconnected for DC with a low resistance path (thats why they have seperate pins !). You have to watch out however that the voltage between them is never greater than +- 5 Volts.
As IVX mentioned the unwanted transients can come via feedback loop. That's why I used an RC lowpass in the feedback. Also the overshoots have to be kept to a minimum* (our output rectangular almost looked like the one from a function generator).
Are these transients also present at the collectors of the voltage translators. Are they also present at the output of the comparator ? What about it's input signals ?
Regards
Charles
* the increased immunity to their own switching transients is one of the advantages of delta-sigma amplifiers IMO.
You can improve the noise performance of the input logic of the IR2110 by having VSS and com bypassed seperately bypassed to ground and having them interconnected for DC with a low resistance path (thats why they have seperate pins !). You have to watch out however that the voltage between them is never greater than +- 5 Volts.
As IVX mentioned the unwanted transients can come via feedback loop. That's why I used an RC lowpass in the feedback. Also the overshoots have to be kept to a minimum* (our output rectangular almost looked like the one from a function generator).
Are these transients also present at the collectors of the voltage translators. Are they also present at the output of the comparator ? What about it's input signals ?
Regards
Charles
* the increased immunity to their own switching transients is one of the advantages of delta-sigma amplifiers IMO.
What you say about separating VSS and COM makes sense, I will add a 100nF SMD capacitor to ground near each pin. I will connect VSS directly to the negative rail and COM with VSS via a 3.3 Ohm resistor. What do you think?
With no load, there shouldn't be any large currents that may cause transients in the feedback or digital inputs, however, I will measure all the signals to see which of them are "contaminated" and I will tell you.
Do you think that the 470uF capacitors on-board can be eliminated or reduced to, for example, 22uF tantalums, or on the other hand, increasing them will make a difference? They occupy a lot of space!
You will think that I want you to design my whole circuit, but what I really want is to learn as much as I can from the experiment with the help of people with much more experience.
Thanks.
With no load, there shouldn't be any large currents that may cause transients in the feedback or digital inputs, however, I will measure all the signals to see which of them are "contaminated" and I will tell you.
Do you think that the 470uF capacitors on-board can be eliminated or reduced to, for example, 22uF tantalums, or on the other hand, increasing them will make a difference? They occupy a lot of space!
You will think that I want you to design my whole circuit, but what I really want is to learn as much as I can from the experiment with the help of people with much more experience.
Thanks.
I think this is a good starting point, just keep an eye on the voltage accross this res for all possible load conditions.
You could use lower value cpacitors. It depends heavily on the wiring between amp board and PSU. But you might increase the RF bypass capacitors (BTW: paralleling low value ones decrerases the parasitic inductance).
You can have overshoot etc even when there is no load connected, because you always unintentionally build parallel and series parasitic resonant circuits.
For the reduction of transient ringing snubber networks can be used.
Regards
Charles
Daily Report...First of all, I changed both the MOSFETs and the entire power supply (transformer, bridge and caps). All reamained the same.
The signal at the output of the comparator also corrupts a bit, as well as the triangle.
I found that adding extra dead-time the average consumption, and hence mosfet dissipation is reduced. I have adjusted it simply by first adjusting the output of the comparator to 50% duty cycle exactly and then observing each input to the IR2110 so that the rising edge is delayed. With a delay of about 200ns in each, I have reached an average consumption of 90mA in the positive rail, a bit more in the other due to the zener, etc. That seems a lot to me, so I will get two high-speed probes to be able to measure them simultaneously just for the case they overlap, causing the problem.
I separated COM and VSS in the IR2110 and added bypass capacitors everywhere, and the signals improved slightly.
Thanks
Hi ssanmor,
For now, you might consider current limiting the supply to the IR2110 to limit its power dissipation until you get the quirks worked out of circuit. Some of the noise may be reduced by using a snubber network connected from maybe the middle of the totem pole to ground. With +/- 50v rails, you may use a robust one without too much power dissipation in the resistor. I would try something like 100 ohms in series with a 2200pf capacitor. If the resistor gets too hot, you can try a smaller capacitor.
Hi Charles,
Have you noticed that sigma-delta amps have lower distortion as the audio input resistor is increased in value and the audio input is preamplified to compensate? In other words, lowering the gain of the class D circuit, increasing the audio NFB, makes the output cleaner.
Hi IVX,
Your amp seems efficient. Does it use a reference triangle wave or sigma-delta. Approximately what is the switching frequency
?
For now, you might consider current limiting the supply to the IR2110 to limit its power dissipation until you get the quirks worked out of circuit. Some of the noise may be reduced by using a snubber network connected from maybe the middle of the totem pole to ground. With +/- 50v rails, you may use a robust one without too much power dissipation in the resistor. I would try something like 100 ohms in series with a 2200pf capacitor. If the resistor gets too hot, you can try a smaller capacitor.
Hi Charles,
Have you noticed that sigma-delta amps have lower distortion as the audio input resistor is increased in value and the audio input is preamplified to compensate? In other words, lowering the gain of the class D circuit, increasing the audio NFB, makes the output cleaner.
Hi IVX,
Your amp seems efficient. Does it use a reference triangle wave or sigma-delta. Approximately what is the switching frequency
?
Hi subwo1,
actually I already wrote about it few days ago at this thread.... probably mine attached gifs still invisible...and i see that my english bad so much, my posts almost not understandable...i am sorry.
This is a discrete circuit with a half-bridge power stage operating in a hysteresis switching (self-oscillating) mode. Feedback is combined from before and after the output filter.fficiency of 93 % 200w@2ohm (@4 ohm may be) @100Hz (in my case was designed for subwoofer use) THD .002-005 % (NOT SURE it lowest than my measurement limit) Demp.factor 1900 idle carrying freq 100Khz.
Regards.
actually I already wrote about it few days ago at this thread.... probably mine attached gifs still invisible...and i see that my english bad so much, my posts almost not understandable...i am sorry.
This is a discrete circuit with a half-bridge power stage operating in a hysteresis switching (self-oscillating) mode. Feedback is combined from before and after the output filter.fficiency of 93 % 200w@2ohm (@4 ohm may be) @100Hz (in my case was designed for subwoofer use) THD .002-005 % (NOT SURE it lowest than my measurement limit) Demp.factor 1900 idle carrying freq 100Khz.
Regards.
Hi subwo
What topology are you talking of ? What you write does not look logical to me from a first glimpse (so please refine).
If you add more gain to a stage outside the NFB loop and reduce inside instead you might decrease feedback instead of increasing it !
When I talk about SD amps I have the principle in mind that Sharp uses.
One advantage often stated by the ones marketing SD amps is the increased NFB factor (compared to classic PWM), made possible by the usually very high sampling rates (and high order feedback loops).
In my eyes there lies quite a lot of sales talk in this statement since it is not only a possibility but a BARE NECESSITY of this VERY COARSE conversion process to work decently at all.
While it is always possible to build PWM amps without any NFB at all, it is absolutely impossible to build an SD amp without any noiseshaping feedback loop (whether it includes the output stage or not).
Regards
Charles
What topology are you talking of ? What you write does not look logical to me from a first glimpse (so please refine).
If you add more gain to a stage outside the NFB loop and reduce inside instead you might decrease feedback instead of increasing it !
When I talk about SD amps I have the principle in mind that Sharp uses.
One advantage often stated by the ones marketing SD amps is the increased NFB factor (compared to classic PWM), made possible by the usually very high sampling rates (and high order feedback loops).
In my eyes there lies quite a lot of sales talk in this statement since it is not only a possibility but a BARE NECESSITY of this VERY COARSE conversion process to work decently at all.
While it is always possible to build PWM amps without any NFB at all, it is absolutely impossible to build an SD amp without any noiseshaping feedback loop (whether it includes the output stage or not).
Regards
Charles
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.