Chris, Bruno, et al,
I've been working on evaluating the various topologies over the holidays (I subscribe to the hands-on approach) and what I found regarding the output filter is similar to what I posted before and what Pierre posted. Below is the carrier-frequency from a UcD180 with a pure resistive load (resistor) and a reactive load (loudspeaker).
UcD180 resistive load
UcD180 reactive load
For comparison here's one of my prototypes driving a loudspeaker (reactive load) which is built up on perforated board, hence lacks proper EMI screening and PCB routing. I think the image is self explanatory.
Class-D prototype, reactive load
What is clear is that a pure resistive load results in the least distortion of the carrier frequency. Unfortunately all loudspeakers are reactive by nature, add to that the loudspeaker cable and you get a rather complex load that the amplifier must drive. From my experiments I conclude the following:
- carrier wavefrom is affected by reactive load thus negatively affecting THD.
- longer cables and higher order passive loudspeaker filters make matters worse.
- all topologies suffer from this effect, but post-filter feedback helps.
- any 'distortion' of the carrier frequency will negatively affect the THD.
Obviously I'm approaching this from a hands-on situation but feel free to shoot holes into the above provided you can supply ample arguments or examples.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I've been working on evaluating the various topologies over the holidays (I subscribe to the hands-on approach) and what I found regarding the output filter is similar to what I posted before and what Pierre posted. Below is the carrier-frequency from a UcD180 with a pure resistive load (resistor) and a reactive load (loudspeaker).
UcD180 resistive load
UcD180 reactive load
For comparison here's one of my prototypes driving a loudspeaker (reactive load) which is built up on perforated board, hence lacks proper EMI screening and PCB routing. I think the image is self explanatory.
Class-D prototype, reactive load
What is clear is that a pure resistive load results in the least distortion of the carrier frequency. Unfortunately all loudspeakers are reactive by nature, add to that the loudspeaker cable and you get a rather complex load that the amplifier must drive. From my experiments I conclude the following:
- carrier wavefrom is affected by reactive load thus negatively affecting THD.
- longer cables and higher order passive loudspeaker filters make matters worse.
- all topologies suffer from this effect, but post-filter feedback helps.
- any 'distortion' of the carrier frequency will negatively affect the THD.
Obviously I'm approaching this from a hands-on situation but feel free to shoot holes into the above provided you can supply ample arguments or examples.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Sander,
Inductive load by an UcD180: did you use the same cable? Big chance there is something resonating in the cable+load.
Try to connect the inductive load very close to the UcD180 module and measure at the output terminal of the UcD amplifier whereby your measurement cables are twisted.
Jan-Peter
Inductive load by an UcD180: did you use the same cable? Big chance there is something resonating in the cable+load.
Try to connect the inductive load very close to the UcD180 module and measure at the output terminal of the UcD amplifier whereby your measurement cables are twisted.
Jan-Peter
Jan-Peter,
The reactive load is simply a typical 8-ohm two-way loudspeaker attched to the amplifier with about 3-meter (10-feet) of loudspeaker cable. The resistive load is a 8.2-ohm 50-watt resistor connected with a 50-cm (1.2-feet) cable. I particularly use the loudspeaker as that's how these modules will be used, hence this is a common occurance, not many people will use their class-D amplifier to heat up a few power resistors 🙂
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
The reactive load is simply a typical 8-ohm two-way loudspeaker attched to the amplifier with about 3-meter (10-feet) of loudspeaker cable. The resistive load is a 8.2-ohm 50-watt resistor connected with a 50-cm (1.2-feet) cable. I particularly use the loudspeaker as that's how these modules will be used, hence this is a common occurance, not many people will use their class-D amplifier to heat up a few power resistors 🙂
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Yup, point is, the difference is not caused by the load but by the cable. The longer cable is quite nicely tuned to the frequency of the "blip" (around 100MHz). When we do EMI measurements we use various lengths of cable to make sure that this resonance is indeed taken into account in the emission plot.
Bruno,
I respectfully have to disagree, even with 50-cm (1.2-feet) of cable and the loudspeaker attached I get a similar image. From what I gather it is the reactive nature of the load that causes these, not 'just' the cable. Matters do get worse with longer cable lengths yes, but the essence of the matter is that any reactive load causes a distortion of the carrier frequency waveform, the cable only contributes slightly. But I'll be happy to cut a few lengths of cable an explore this further, actually, let me do just that right now, to check whether I'm on the right track here. I'll post new images with various cable lengths in a couple of minutes.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I respectfully have to disagree, even with 50-cm (1.2-feet) of cable and the loudspeaker attached I get a similar image. From what I gather it is the reactive nature of the load that causes these, not 'just' the cable. Matters do get worse with longer cable lengths yes, but the essence of the matter is that any reactive load causes a distortion of the carrier frequency waveform, the cable only contributes slightly. But I'll be happy to cut a few lengths of cable an explore this further, actually, let me do just that right now, to check whether I'm on the right track here. I'll post new images with various cable lengths in a couple of minutes.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Bruno et al,
I've measured the same 8.2-ohm resistor with 1-meter (3-feet), 3-meter (10-feet) and 5-meter (17-feet) of cable, with virtually no effect on the carrier frequency waveform. The resulting image on the scope can be seen below:
8.2-ohm resistor, UcD180, carrier frequency waveform
I've also measured the same 8-ohm two-way loudspeaker with 1-meter (3-feet), 3-meter (10-feet) and 5-meter (17-feet) of cable and saw no big differences, variations are comparable to the differences observed with the 8.2-ohm resistor and different cable lengths. The resulting image on the scope can be seen below:
8-ohm, two-way loudspeaker, UcD180, carrier frequency waveform
Test setup, using Tektronic 100MHz 10x probe
Hence I think this illustrates that a reactive load will slightly increase the THD of class-D amplifier, more so with designs that have pre-filter feedback (such as clearly visible in two of my prototypes) less so with post-filter feedback (such as the UcD180 module used here or my UcD prototype).
But as always, feel free to comment on the above.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I've measured the same 8.2-ohm resistor with 1-meter (3-feet), 3-meter (10-feet) and 5-meter (17-feet) of cable, with virtually no effect on the carrier frequency waveform. The resulting image on the scope can be seen below:
8.2-ohm resistor, UcD180, carrier frequency waveform
I've also measured the same 8-ohm two-way loudspeaker with 1-meter (3-feet), 3-meter (10-feet) and 5-meter (17-feet) of cable and saw no big differences, variations are comparable to the differences observed with the 8.2-ohm resistor and different cable lengths. The resulting image on the scope can be seen below:
8-ohm, two-way loudspeaker, UcD180, carrier frequency waveform
Test setup, using Tektronic 100MHz 10x probe
Hence I think this illustrates that a reactive load will slightly increase the THD of class-D amplifier, more so with designs that have pre-filter feedback (such as clearly visible in two of my prototypes) less so with post-filter feedback (such as the UcD180 module used here or my UcD prototype).
But as always, feel free to comment on the above.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Oh, and the differences between the images posted in my last post and those posted before is the simple fact that I took the module out of the chassis it was residing in and tested it on the bench (as pictured) to rule out any (EMI) issues with the chassis and its wiring. As is evident the 'naked' module has somewhat higher distortion, I'm guessing this is due to the fact that the output coil is now radiating freely instead of being encased in steel, affecting the speaker cable and scope probe directly. But I'll leave it up to Bruno to comment on this.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Perhaps noise pick-up by the probe?
Where the measurements should be done, at the amp side or at the load side?
Where the measurements should be done, at the amp side or at the load side?
Pierre,
Look at the test setup image, I try to be as comprehensive as possible by also posting that and you completely seem to ignore it?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Look at the test setup image, I try to be as comprehensive as possible by also posting that and you completely seem to ignore it?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
You should also take care of the "magnetic filed" probe that you unintentionally built with this setup.
Regards
Charles
Regards
Charles
No, Sander, I haven't ignored it. We all acknowledge that you have posted the photo.
I was only asking (perhaps thinking loud) where is the most sensible place to perform the measurement for comparison, at the load or at the speaker.
I see that you have measured at the speaker and I also see that pretty loop you have unintentionally build with your probe, picking up whatever is being radiated. I bet that you will measure quite a bit of ringing also if you connect your GND crocodile to the probe tip, "shorting" the probe.
HF measurements must be done with a proper HF probe, or with a regular probe adapted to have minimal loop area.
Mmm. I have just seen that Charles was saying just the same, sorry.
I was only asking (perhaps thinking loud) where is the most sensible place to perform the measurement for comparison, at the load or at the speaker.
I see that you have measured at the speaker and I also see that pretty loop you have unintentionally build with your probe, picking up whatever is being radiated. I bet that you will measure quite a bit of ringing also if you connect your GND crocodile to the probe tip, "shorting" the probe.
HF measurements must be done with a proper HF probe, or with a regular probe adapted to have minimal loop area.
Mmm. I have just seen that Charles was saying just the same, sorry.
Pierre,
I realize that, but that's not the issue here. If I go to great lengths to ensure that the ground loop is kept short I might as well solder the probe directly to the output coil, but that's not how these modules will be used.
Or rather, we're dealing with a practical issue here; with the previous images I was measuring at the output terminals of the amplifier chassis, with the same probe, which showed similar 'disortion' of the carrier frequeny waveform when a reactive load was connected.
Don't you agree we need to evaluate a situation that accurately represents how these modules (or other class-D amplifiers) will be used rather than a academic one where the measurement itself is under scrutiny?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I realize that, but that's not the issue here. If I go to great lengths to ensure that the ground loop is kept short I might as well solder the probe directly to the output coil, but that's not how these modules will be used.
Or rather, we're dealing with a practical issue here; with the previous images I was measuring at the output terminals of the amplifier chassis, with the same probe, which showed similar 'disortion' of the carrier frequeny waveform when a reactive load was connected.
Don't you agree we need to evaluate a situation that accurately represents how these modules (or other class-D amplifiers) will be used rather than a academic one where the measurement itself is under scrutiny?
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
Dear Sander,
You can´t measure in this way......
Please try the following: make a loop with your ground clip and the probe, i.e. put the ground clip to the tip of your probe. And use this as a coil, move this coil around the amplifier and you see that you pick a lot of HF garbish.
You should do this kind of measurement on other Class-D products, and you will see that by some products your scope has big problems to trigger.........
Jan-Peter
You can´t measure in this way......
Please try the following: make a loop with your ground clip and the probe, i.e. put the ground clip to the tip of your probe. And use this as a coil, move this coil around the amplifier and you see that you pick a lot of HF garbish.
You should do this kind of measurement on other Class-D products, and you will see that by some products your scope has big problems to trigger.........
Jan-Peter
Sander, I understand you, but I was referring to the same issue Jan-Peter says: noise pick-up by the probe.
If we want to compare different Class-D amps in terms of RFI residue, we should establish a test method: probe setup, point where measure is done, etc., so we know exactly that what we are measuring is IN the output signal (not in the air). (conducted EMI, not radiated EMI).
And once that points are clear, measure how many mV of ringing each module has, how many switching residue is left, etc.
If we want to compare different Class-D amps in terms of RFI residue, we should establish a test method: probe setup, point where measure is done, etc., so we know exactly that what we are measuring is IN the output signal (not in the air). (conducted EMI, not radiated EMI).
And once that points are clear, measure how many mV of ringing each module has, how many switching residue is left, etc.
Strange ...... if I do this measurement at the input of my loudspeaker (connected thru 3 meters normal parallel wire) I find only the basic waveform.
Equipment is DFR9000, not connected directly to receiver to not disturb the wife. 😉
465B; Tektronix; 100Mhz
Equipment is DFR9000, not connected directly to receiver to not disturb the wife. 😉
465B; Tektronix; 100Mhz
It seems Sander will be visiting me in the near future so we can have a serious look at these measurements here.
Apart from the amp, don't forget to bring the cables, the resistors and the loudspeaker.
Apart from the amp, don't forget to bring the cables, the resistors and the loudspeaker.
- Status
- Not open for further replies.