Scope is fine, bought it calibrated from Tektronix about 3-months ago, but I guess I tackled the problem it already. It is the switching frequency, by dropping down to 100KHz (increasing the integrator C value) the spikes have all gone, so I'll lower the carrier frequency to a more sane value (400KHz-ish), as honestly >700KHz is a bit much for a bread-board 
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Here's the amplifier at a 450KHz carrier frequency, looks a lot better, and considering this is a bread-board, with non-SMD components I'll leave it at that, unless you have a suggestion on how to improve it further without altering the layout, can't do much about that tbh, I made it all fit as tightly as I possibly could.
Did I already mention it sounds really good? I just hooked up a nice speaker than my beat up and battered testing ones and I was pleasantly surprised how good it sounded. But take that with a grain of salt for now, as of course I'd like to see all teh work pay off. I'll try to figure out a way to get some decent THD measurements done with RMAA or some other software tool.
Unfortunately my simulating skills are rusty, if not worse, so I'd love to simulate, but I'd need some help, last time I touched a sim for anything this complex was in univ. and that's almost a decade ago.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Did I already mention it sounds really good? I just hooked up a nice speaker than my beat up and battered testing ones and I was pleasantly surprised how good it sounded. But take that with a grain of salt for now, as of course I'd like to see all teh work pay off. I'll try to figure out a way to get some decent THD measurements done with RMAA or some other software tool.
Unfortunately my simulating skills are rusty, if not worse, so I'd love to simulate, but I'd need some help, last time I touched a sim for anything this complex was in univ. and that's almost a decade ago.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
It is very curious, Chris, it seems as we have been working today with very similar issues on our respective amps but at thousands of kilometers one from other!
Today I was looking at my output waveform and resembled a lot to the one you posted (the 700KHz one), but mine is running at about 250KHz. I don't really know where the spikes come from. The strangest thing is that I have two versions boards my amp, and one of them has the spikes while the other has almost none. They are exactly the same, (except for the 220uF/100V supply bypass caps that are 330uF in the good one), so ¿?
Sorry if I am a bit confused. You speak about two implementations of your amp, one that works fine and other that doesn't. What is the difference between them? are they from the same sch but the layouts are different? I assume that the latest waveform you have shown are from the good one, right?
Best regards,
Pierre
Today I was looking at my output waveform and resembled a lot to the one you posted (the 700KHz one), but mine is running at about 250KHz. I don't really know where the spikes come from. The strangest thing is that I have two versions boards my amp, and one of them has the spikes while the other has almost none. They are exactly the same, (except for the 220uF/100V supply bypass caps that are 330uF in the good one), so ¿?
Sorry if I am a bit confused. You speak about two implementations of your amp, one that works fine and other that doesn't. What is the difference between them? are they from the same sch but the layouts are different? I assume that the latest waveform you have shown are from the good one, right?
Best regards,
Pierre
BTW, Chris. Now that you ask, I did some tests to my amp with RightMark audio analyzer.
That's how I did it:
- First, prepare a loopback cable and calibrate your soundcard to see to what extent it influences distortion, what its noise floor is, etc.
- Now attach the amp: the input is direct from the output of the soundcard. Load the output with a 4 ohm (or 8 or whatever) power load.
- I took the signal back to the soundcard with a variable attenuator (to be able to look at different powers while still having the optimal amplitude at the soundcard input), and then a 3 stage RC filter, to filter the output ripple to the lowest possible value.
- You will find that you need to adjust the input level to get optimal readings.
I also used another program that measures spectrum and THD in real-time. That's very useful for adjustment and after all the THD readings are quite accurate (compared with other two programs). This program is SoundTechnologies "Spectra RTA".
The best program I have found so far for audio analysis is , however, Sample Champion, but its much more complicated.
That's how I did it:
- First, prepare a loopback cable and calibrate your soundcard to see to what extent it influences distortion, what its noise floor is, etc.
- Now attach the amp: the input is direct from the output of the soundcard. Load the output with a 4 ohm (or 8 or whatever) power load.
- I took the signal back to the soundcard with a variable attenuator (to be able to look at different powers while still having the optimal amplitude at the soundcard input), and then a 3 stage RC filter, to filter the output ripple to the lowest possible value.
- You will find that you need to adjust the input level to get optimal readings.
I also used another program that measures spectrum and THD in real-time. That's very useful for adjustment and after all the THD readings are quite accurate (compared with other two programs). This program is SoundTechnologies "Spectra RTA".
The best program I have found so far for audio analysis is , however, Sample Champion, but its much more complicated.
Pierre,
Same amp, just reduced the carrier frequency by increasing the value of the integrator cap, simple fix really. The last scope image is from the amp running with a 450KHz carrier frequency, the first image is from the amp running at a 770KHz carrier frequency.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Same amp, just reduced the carrier frequency by increasing the value of the integrator cap, simple fix really. The last scope image is from the amp running with a 450KHz carrier frequency, the first image is from the amp running at a 770KHz carrier frequency.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Now that I have re-read the thread, I realize that the versions I was refering to were 007 and 008. 007 Seems to be the working one (where you have increased the integrating cap to reduce sw freq.), isn't it?
Sorry, my previous post about Rightmark Audio Analyzer was for you, Sander, not Chris (well, it is for anyone, really, of course, but it was you the one who asked ;-)
Best regards,
Pierre
Sorry, my previous post about Rightmark Audio Analyzer was for you, Sander, not Chris (well, it is for anyone, really, of course, but it was you the one who asked ;-)
Best regards,
Pierre
Sander,
Is your amp (version 8) running?
I did bread-board one that is very similar to your design but with a diffrent feedback network. The feedback on my prototype board are connected to an op-amp before feeding it to the comparator.
I had similar problem as yours though did not fry the MOSFET but it did heat-up fast once a load is connected to the output.
I might be re-looking at my design again and see if I can get it to work.
Is your amp (version 8) running?
I did bread-board one that is very similar to your design but with a diffrent feedback network. The feedback on my prototype board are connected to an op-amp before feeding it to the comparator.
I had similar problem as yours though did not fry the MOSFET but it did heat-up fast once a load is connected to the output.
I might be re-looking at my design again and see if I can get it to work.
vt4c,
No, version 007 is the one that worked, 008 didn't. I'll revisit 008 when I'm confident that the switching stage is working properly hence the problem is in the stage before that which will make it easier to diagnose and possibly fix.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
No, version 007 is the one that worked, 008 didn't. I'll revisit 008 when I'm confident that the switching stage is working properly hence the problem is in the stage before that which will make it easier to diagnose and possibly fix.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
A couple of questions/comments for Sander...
- I have been carefully measuring my output with a 4R7 resistor across the output terminals, as short as possible, and with a good probe/oscilloscope. Now the peaks are much smaller (say, 500mVp-p).
Where are you measuring? Do you have long leads to your load and measure at it, or just at the amp terminals, where? It seems that the measurement is quite dependant on the setup, leads, etc. (HF involved ;-)
- In your best version (seems to be 007), do you have any starting issues? I explain: I was playing with my amp yesterday. It is designed for clocked operation (triangle wave), but I modified it so it was hysteretic and worked fine, although it doesn't start oscillating until input reachs a certain level. Then it holds with no problem.
For example, IRAUDAMP from IRF has a transistor that pulls-down the comparator output for a moment to force low side mosfet to go on, but yours (I know it is not exactly hysteretic), doesn't have provisions for that. Have you noticed that effect?
- Your mosfets are rated at only 55V, that limits your operation to around 25+25V maximum (80W at 4 ohm). I think your design can be easily adapted to work at much higher voltages, with no degradation, using more common and also nice mosfets such as IRF640(N). With that, it could go to around 400W/4ohm ;-)
And this is for classD4sure:
Why do you ask if that waveform is from an UCD? Is it bad? (I assume you refer to Hypex modules, not generic UcD, right?)
- Another thing about your design (that is very elegant by the way) is that it doesn't have provisions for dead-time apart from the diodes in parallel with the gate resistors. Do you thing that's enough?
Best regards,
Pierre
- I have been carefully measuring my output with a 4R7 resistor across the output terminals, as short as possible, and with a good probe/oscilloscope. Now the peaks are much smaller (say, 500mVp-p).
Where are you measuring? Do you have long leads to your load and measure at it, or just at the amp terminals, where? It seems that the measurement is quite dependant on the setup, leads, etc. (HF involved ;-)
- In your best version (seems to be 007), do you have any starting issues? I explain: I was playing with my amp yesterday. It is designed for clocked operation (triangle wave), but I modified it so it was hysteretic and worked fine, although it doesn't start oscillating until input reachs a certain level. Then it holds with no problem.
For example, IRAUDAMP from IRF has a transistor that pulls-down the comparator output for a moment to force low side mosfet to go on, but yours (I know it is not exactly hysteretic), doesn't have provisions for that. Have you noticed that effect?
- Your mosfets are rated at only 55V, that limits your operation to around 25+25V maximum (80W at 4 ohm). I think your design can be easily adapted to work at much higher voltages, with no degradation, using more common and also nice mosfets such as IRF640(N). With that, it could go to around 400W/4ohm ;-)
And this is for classD4sure:
Why do you ask if that waveform is from an UCD? Is it bad? (I assume you refer to Hypex modules, not generic UcD, right?)
- Another thing about your design (that is very elegant by the way) is that it doesn't have provisions for dead-time apart from the diodes in parallel with the gate resistors. Do you thing that's enough?
Best regards,
Pierre
Pierre,
Input shorted (no signal), load connected (either a speaker or a dummy-load). Then just measure across the speaker/load terminals.
No, runs at startup, immediately upon switch on the carrier-frequency kicks in.
I know, I picked these because I had them laying around and they're very easy to drive and remember this is just the prototype, constructed to prove whether the design works as planned. I'll substitute with IRFB4242 when I step the design up to the final version.
Yes, look at the pic of the 450KHz carrier, that's very nice switching behaviour for a class-D on a bread-board with a less than ideal layout, I see no need to change anything there.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Input shorted (no signal), load connected (either a speaker or a dummy-load). Then just measure across the speaker/load terminals.
No, runs at startup, immediately upon switch on the carrier-frequency kicks in.
I know, I picked these because I had them laying around and they're very easy to drive and remember this is just the prototype, constructed to prove whether the design works as planned. I'll substitute with IRFB4242 when I step the design up to the final version.
Yes, look at the pic of the 450KHz carrier, that's very nice switching behaviour for a class-D on a bread-board with a less than ideal layout, I see no need to change anything there.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Chris et al,
Here's a look at bottom of the bread-board so you have an impression of how I routed things. This is the ver 008, but 007 didn't change anything about the switching-section.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Here's a look at bottom of the bread-board so you have an impression of how I routed things. This is the ver 008, but 007 didn't change anything about the switching-section.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Wait a minute! My digital multimeter frequency readout is playing tricks on me! Upon looking at the scope screen and counting the sec/div and calculating the frequency it is way off. The actual circuit I started off with runs at 400KHz, the modified one, with the larger integrator cap at half that, 200KHz. So it is a 400KHz carrier I started off with, now the switching behaviour doesn't look as good, I'll see whether playing with the dead-time resistor is of any help 
400KHz carrier frequency
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Edit: added picture
400KHz carrier frequency
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Edit: added picture
That looks very similar to mine (which is running at around 260 KHz). However, I have about 4 times less carrier and about half the ringing.
Once I was told that the ringing could be due to self-resonance of the coil triggered by the switching events... I don't really know.
I also played with the dead-time and also with the gate resistor but it didn't change anything. The only thing that seemed to reduce ringing was the addition of a snubber between the sw. node and GND. But it only halved its amplitude.
Once I was told that the ringing could be due to self-resonance of the coil triggered by the switching events... I don't really know.
I also played with the dead-time and also with the gate resistor but it didn't change anything. The only thing that seemed to reduce ringing was the addition of a snubber between the sw. node and GND. But it only halved its amplitude.
Alright, narrowed it down even further, as obviously I was using a loudspeaker which is a reactive load. By replacing it by a pure resitive load, a 8.2-ohm/10-watt/MOX, resistor I get the below image, which looks a whole lot better. Hence, use a resistive load for debugging the design, not an actual loudspeaker.
400KHz carrier frequency, 8.2-ohm resistive load
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
400KHz carrier frequency, 8.2-ohm resistive load
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Hi guys,
I think it's caused by the reversal of the inductor current. I'd imagine a different mosfet with different body diode characteristics would make a difference. As would a snubber across it, as would turning off the body diode, or perhaps simply bypassing it, but it seems that's so tricky to do right it's probably not even worth it, turning it off would be the better option, and is more lossy.
Regards,
Chris
I think it's caused by the reversal of the inductor current. I'd imagine a different mosfet with different body diode characteristics would make a difference. As would a snubber across it, as would turning off the body diode, or perhaps simply bypassing it, but it seems that's so tricky to do right it's probably not even worth it, turning it off would be the better option, and is more lossy.
Regards,
Chris
Chris,
Any suggestions as to how I'd go about that? I'm willing to try, got the soldering iron heated up already and parts a plenty, so I'm all for a little experimenting.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Any suggestions as to how I'd go about that? I'm willing to try, got the soldering iron heated up already and parts a plenty, so I'm all for a little experimenting.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Not really I require further research and experimentation in that area, I don't dwelve into it as my only analysis tool is spice and it doesnt' model that sort of thing very well.
Send me your scope with a big red bowtie on it and I'll get back to you in a month
Ummm.. snubbers are one way, I can't even get into that there's too many options and then plenty of room to be inventive.
Turning off the body diode would be a sure fix if that is what's going on.
From: http://www.richieburnett.co.uk/mosfail.html
"MOSFET body diodes generally have a long reverse recovery time compared to the performance of the MOSFET itself. If the body diode of one MOSFET is conducting when the opposing device is switched on, then a "short circuit" occurs similar to the shoot-through condition described above.
This problem is usually eased by the addition of two diodes surrounding each MOSFET. Firstly, a Schottky diode is connected in series with the MOSFET source. The schottky diode prevents the MOSFET body diode from ever being forward biased by the free-wheeling current. Secondly, a high speed (fast recovery) diode is connected in parallel to the MOSFET/Schottky pair so that the free-wheeling current bypasses the MOSFET and Schottky completely.
This ensures that the MOSFET body diode is never driven into conduction. The free-wheel current is handled by the fast recovery diodes which present less of a "shoot-through" problem."
Or if you have other mosfets kicking around with a better "quality factor" body diode.. whatever is easiest for you.
I imagine any of the above would make a difference.
Regards,
Chris
PS:
I love these kinds of threads Nothing like the smell of perf board in the morning .
Send me your scope with a big red bowtie on it and I'll get back to you in a month
Ummm.. snubbers are one way, I can't even get into that there's too many options and then plenty of room to be inventive.
Turning off the body diode would be a sure fix if that is what's going on.
From: http://www.richieburnett.co.uk/mosfail.html
"MOSFET body diodes generally have a long reverse recovery time compared to the performance of the MOSFET itself. If the body diode of one MOSFET is conducting when the opposing device is switched on, then a "short circuit" occurs similar to the shoot-through condition described above.
This problem is usually eased by the addition of two diodes surrounding each MOSFET. Firstly, a Schottky diode is connected in series with the MOSFET source. The schottky diode prevents the MOSFET body diode from ever being forward biased by the free-wheeling current. Secondly, a high speed (fast recovery) diode is connected in parallel to the MOSFET/Schottky pair so that the free-wheeling current bypasses the MOSFET and Schottky completely.
This ensures that the MOSFET body diode is never driven into conduction. The free-wheel current is handled by the fast recovery diodes which present less of a "shoot-through" problem."
Or if you have other mosfets kicking around with a better "quality factor" body diode.. whatever is easiest for you.
I imagine any of the above would make a difference.
Regards,
Chris
PS:
I love these kinds of threads
Nothing like the smell of perf board in the morning .
Thanks Chris,
Very useful info! I actually got a whole bunch of the MBRS140 and MURS120 diodes used in the design, so I could use those and give it a try. I've also ordered new MOSFET's, the IRFB23N15D, which should arrive on monday.
Alright, I guess I'll have a go then, I'll report back if I manage to make any progress with the diodes. Otherwise expect an update on monday when the MOSFET's come in.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Very useful info! I actually got a whole bunch of the MBRS140 and MURS120 diodes used in the design, so I could use those and give it a try. I've also ordered new MOSFET's, the IRFB23N15D, which should arrive on monday.
Alright, I guess I'll have a go then, I'll report back if I manage to make any progress with the diodes. Otherwise expect an update on monday when the MOSFET's come in.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.