Hopefully I have not missed an older thread about this. I believe we could make some kind of listing of the majour types of distortion in a class D construction, and the cause of their nature.
I know there is a lot of people here with a lot of knowledge so this thread could actually be almost complete!
Please do not be afraid of writing anything wrong. It will probably be commented and corrected.
I know there is a lot of people here with a lot of knowledge so this thread could actually be almost complete!
Please do not be afraid of writing anything wrong. It will probably be commented and corrected.
Like the idea a lot, to get one place to find "all" distortion causes and ways to minimize these in class-d amps.
Classd4sure; I can't seem to find a thread which is dealing with this in depth!
Maybe we should try to also draw up the different topologies and descripe distortion causes for each, as BWRX pointed out you can't gereralize, but some issues will be common though.
I have at the moment problems with 3. harmonics in a full bridge self osc. design ..... the source of the distortion .... unknown!
2. harmonic is quite low, where 3. is about 1% at 10 W out. .... a thread like the suggested would help
Classd4sure; I can't seem to find a thread which is dealing with this in depth!
Maybe we should try to also draw up the different topologies and descripe distortion causes for each, as BWRX pointed out you can't gereralize, but some issues will be common though.
I have at the moment problems with 3. harmonics in a full bridge self osc. design ..... the source of the distortion .... unknown!
2. harmonic is quite low, where 3. is about 1% at 10 W out. .... a thread like the suggested would help
Google is your friend. Go to Google, click on Advanced Search then (this is important) search only in the domain diyaudio.com for the following terms:
"class d" distortion (the parentheses must be used as shown)
There are a ton of hits - about half are pertinent.
Among the hits is a recent thread that was on dead time distortion. Perhaps this is the one Chris was thinking of?
Regards -- analogspiceman
It isnt' the one I was thinking of, but I think this supercedes it!
You can append this to your search too site:www.diyaudio.com saves you a few clicks
It doesn't seem to bring up the old ones though?
You can append this to your search too site:www.diyaudio.com saves you a few clicks
It doesn't seem to bring up the old ones though?
Baldin said:Like the idea a lot, to get one place to find "all" distortion causes and ways to minimize these in class-d amps.
Classd4sure; I can't seem to find a thread which is dealing with this in depth!
Maybe we should try to also draw up the different topologies and descripe distortion causes for each, as BWRX pointed out you can't gereralize, but some issues will be common though.
I have at the moment problems with 3. harmonics in a full bridge self osc. design ..... the source of the distortion .... unknown!
2. harmonic is quite low, where 3. is about 1% at 10 W out. .... a thread like the suggested would help
-What kind of oscillation do you use?
-What kind of feedback do you use?
-Could it simply be some kind of 2. harmonic dist that adds up to 3. harm dist when bridged?
The dead time distortion linked to in a previous posting is very interesting. Here is the original post:
The text also include this image:
http://www.diyaudio.com/forums/attachment.php?s=&postid=872569&stamp=1142712297
The text also include this image:
http://www.diyaudio.com/forums/attachment.php?s=&postid=872569&stamp=1142712297
Hi Snickers-is
I'm using a self oscillating modulator, based around an integrator and an comperator (part of the HIP4080).
The modulator feeds the driver and output stage, and a feedback signal is taken before the reconstruction filter, and fed back to an sunning amp placed before the modulator.
Se my homepage for further details
I can't se how a second order distortion would be turned into a 3. order by the bridge. There is no modulation in the power stage, hense no frequency shift!
I think the higher level of 3. harmonics is a result of even order harmonics being suppresed by the differential output (H-bridge). Therefore the odd harmonics will be dominant ....
... but again a list of possible sources and improvements wrt. distortion would be quite nice
I'm using a self oscillating modulator, based around an integrator and an comperator (part of the HIP4080).
The modulator feeds the driver and output stage, and a feedback signal is taken before the reconstruction filter, and fed back to an sunning amp placed before the modulator.
Se my homepage for further details
I can't se how a second order distortion would be turned into a 3. order by the bridge. There is no modulation in the power stage, hense no frequency shift!
I think the higher level of 3. harmonics is a result of even order harmonics being suppresed by the differential output (H-bridge). Therefore the odd harmonics will be dominant ....
... but again a list of possible sources and improvements wrt. distortion would be quite nice
Feedback for teh modulator is taken from the output of the comperator not the output of the driver/power Stage (BHS/AHS).
Just measured without load, and no distortion in this case.
Measuring at different frequencies, seems to give more or less the same (with a 4 ohm speaker as load)
The output has some risidual switching spikes in it .... could that maybe be the course ... ?
I'll experiment with a snubber 8R2+100nF placed over e.g. the whole bridge.
The coil is not the best, and has too high loss and is not suitable for this application. I'm awaiting some Micrometals T106-2 cores, which should do the job somewhat better. ..... guess it could cause some distortion, as it is not linear at all.
Just measured without load, and no distortion in this case.
Measuring at different frequencies, seems to give more or less the same (with a 4 ohm speaker as load)
The output has some risidual switching spikes in it .... could that maybe be the course ... ?
I'll experiment with a snubber 8R2+100nF placed over e.g. the whole bridge.
The coil is not the best, and has too high loss and is not suitable for this application. I'm awaiting some Micrometals T106-2 cores, which should do the job somewhat better. ..... guess it could cause some distortion, as it is not linear at all.
So then you correct the signal twice. First in the oscillation. Second in the global feedback.
The global feedback is pre filter and also not able to correct the load dependent filter. The filter is emmulated by the ground capacitor at the loop, and therefore almost not dependent on load at all. I think you will find that the problem is caused by the peaks of the signal being compressed a bit. By redesigning the feedback and puting it post filter I believe the figures will change a lot. You could also try to change the value of the ground capacitor on the global loop.
The global feedback is not fast enough for the switching peaks, but anyway, I think you should threat them another way.
The global feedback is pre filter and also not able to correct the load dependent filter. The filter is emmulated by the ground capacitor at the loop, and therefore almost not dependent on load at all. I think you will find that the problem is caused by the peaks of the signal being compressed a bit. By redesigning the feedback and puting it post filter I believe the figures will change a lot. You could also try to change the value of the ground capacitor on the global loop.
The global feedback is not fast enough for the switching peaks, but anyway, I think you should threat them another way.
Think I need a common mode filter for my scope, as conecting the probe tip and the ground and holding it to the ground og the amp showsthe same spikes as on the output .... must be induced. Will try to use one of the cores I get and coil the probe cord around some turns to se whether this cures it! (but I don't think it has much to do with the distortion)
Snubber over bridge (Vcc to GND) gave no result (well I had to change to 10 nF as using the 100 nF produced a lot of heat in the resistor )
Snubber over bridge (Vcc to GND) gave no result (well I had to change to 10 nF as using the 100 nF produced a lot of heat in the resistor
)
What do you mean by peaks being compressed a bit??
Yes I know that a post filter feedback would probably give lower feedback. But I thought it would be easier to get this to work, as my first class-d
Yes I'll try changing C1 and C2, and se what happens.
And you are right the spikes must be dealt with in a totally different way
Yes I know that a post filter feedback would probably give lower feedback. But I thought it would be easier to get this to work, as my first class-d
Yes I'll try changing C1 and C2, and se what happens.
And you are right the spikes must be dealt with in a totally different way
I do not think you are slow, because this is some complicated.
Think of the chain of power FETs and the feedback loop as a voltage control loop. What this loop controls is the time and duration of the output directly after the FETs.
Think of the chain of power FETs and the output filter and load as a current loop. This is controlled by the theoretically ideal voltage from the output of the FETs.
The mission of the output coil is to make an average current identical to the audio signal. This filter is being mimiced by a parallell capacitor in the loop. This capacitor does not sense the output current or the load. The load will have a voltage drop equal to the current and the impedance of the load. Since the load is not resistive the output voltage will not be equal to the input signal as it is only controled by a theoretical feedback.
So the theoretical feedback is probably incapable of doing good enough corrections both to the variations caused by the filter, and the variations caused by running reactive loads.
Think of the chain of power FETs and the feedback loop as a voltage control loop. What this loop controls is the time and duration of the output directly after the FETs.
Think of the chain of power FETs and the output filter and load as a current loop. This is controlled by the theoretically ideal voltage from the output of the FETs.
The mission of the output coil is to make an average current identical to the audio signal. This filter is being mimiced by a parallell capacitor in the loop. This capacitor does not sense the output current or the load. The load will have a voltage drop equal to the current and the impedance of the load. Since the load is not resistive the output voltage will not be equal to the input signal as it is only controled by a theoretical feedback.
So the theoretical feedback is probably incapable of doing good enough corrections both to the variations caused by the filter, and the variations caused by running reactive loads.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.