Class D Design Issues

Quick question about IR2184 drivers. I can see that they have two versions of the chip, one has a common ground and the other seems to have a floating supply. Can I use Vss as the low signal ground (for the comparator, oscillator and audio) and COM as the separate, high voltage negative supply (say +-40V DC). I know I need level shifting for that, so does 21844 include them? Screenshot_Adobe_Acrobat_20180910-150002.jpeg
 
AC wave in class D output.

I tried the below given schematic from Apex (I guess.)
Amp is working and no heating but one issue.

Output have AC voltage swing of almost 1 volt all time even when input is grounded.

Power supply is 27-0-27 volt.

I am using 50uH inductor T106 -2 core and 2.2uF capacitor at output.
Earlier the LC was as mentioned in the schematic 0,47uF /30uH but the AC swing was even greater.

What is the cause of such a sine wave even after reasonably good LC filter.
 

Attachments

  • wa.jpg
    wa.jpg
    310.2 KB · Views: 465
  • brd.jpg
    brd.jpg
    373.3 KB · Views: 455
  • APEX UCD250.jpg
    APEX UCD250.jpg
    258.7 KB · Views: 466
  • D200_2_zpswoa29atm.jpg
    D200_2_zpswoa29atm.jpg
    101.3 KB · Views: 471
115KHz is quite low.
IRS2092 usually uses at least 250KHz.

I use 10uH and 470nF and get a little residual carrier but it doesn't cause a problem.

Core is heating up with higher frequency and less than 35uH inductance. (T106-2 core)

After trial and error I reached to this frequency by changing 56pf to 100pf in feedback after this change nothing heats.
 
I use a T106-2 10uH with 470nF at 250KHz and it barely gets warm.

I know on mine I got a little residual carrier but I don't think its anything to worry about. The speaker wont reproduce 100KHz.

I will try that again and come back.

Last time I tried 15uh / 0.4uF in Tl074 based amp (my first one), and mosfet were dead,so bit more cautious and reserved with this new trial.
 
I had also noted that while doing audio sweep at around 10khz to 13khz there is almost 3-4db peak, thats like this amp is behaving like a peak filter ??
Oscilloscope output was across speaker terminal and I was sweeping from 20HZ to 20khz manually from PC soundcard.

Could it be due to some speaker (4ohms) resonance or LC filter or feedback?

(But before that to be explored I had to sort out the sine wave in output.)
 
I had also noted that while doing audio sweep at around 10khz to 13khz there is almost 3-4db peak, thats like this amp is behaving like a peak filter ??
Oscilloscope output was across speaker terminal and I was sweeping from 20HZ to 20khz manually from PC soundcard.

Could it be due to some speaker (4ohms) resonance or LC filter or feedback?

(But before that to be explored I had to sort out the sine wave in output.)

Feedback before output filter, do post filter feedback.
 
Last edited:
I use a T106-2 10uH with 470nF at 250KHz and it barely gets warm.

I know on mine I got a little residual carrier but I don't think its anything to worry about. The speaker wont reproduce 100KHz.


Reduced the Fsw by changing he feedback resistor to 2.2k and 100pf.
OP LC remain original 30uH/0.47uF

Now the FLoating voltage diode Mur1100e heats considerably ,finger burns.
The waveforms are as in images below.

Now the new problem is amp becomes unstable beyond certain amplitudes ,
When input volume is increased beyond certain level , music begins to distort and suddenly dreaded HUm is heard which goes away when amp is restarted.
Probably amp starts to oscillate .

attached the image of wave when sudden hum is heard.

Also the output zobel 10 ohm resistor heats up too much almost to smoke.


Feedback before output filter, do post filter feedback.

Thank you ,I will try the filter and see.
 

Attachments

  • 1.jpg
    1.jpg
    86 KB · Views: 649
  • 2.jpg
    2.jpg
    110.5 KB · Views: 620
  • HUm wave.jpg
    HUm wave.jpg
    88.2 KB · Views: 616
Last edited:
Reduced the Fsw by changing he feedback resistor to 2.2k and 100pf.
OP LC remain original 30uH/0.47uF

Now the FLoating voltage diode Mur1100e heats considerably ,finger burns.
The waveforms are as in images below.

Now the new problem is amp becomes unstable beyond certain amplitudes ,
When input volume is increased beyond certain level , music begins to distort and suddenly dreaded HUm is heard which goes away when amp is restarted.
Probably amp starts to oscillate .

attached the image of wave when sudden hum is heard.

Also the output zobel 10 ohm resistor heats up too much almost to smoke.




Thank you ,I will try the filter and see.

You need to design the output low pass filter very carefully on the right speaker impedance, post feedback is not needed for cure the peak, this is because of output filter, maybe wrong core or such. Postfeedback can be used but fase shifts make feedback more difficult see pdf.

https://www.hobbielektronika.hu/forum/getfile.php?id=253790

regards
 
Last edited:
Hi All

I have a problem that I do not now how to calculate the feedback network on a 5 level class D amp, it is a postfeedback system.

beause the signal is 4 x carrier I need to now what to have to use, the triangle carrier or the multiplied carrier on the output filter who is now 1.2 Mhz, beause of this I can use a high bandwidth, but is this feasable, I can stick on for example 80 Khz or 100 Khz, then still a lot room for filtering out carrier.

I hope this is the place to ask, I include some examples, I need to calculate the unity gain point of amp, see pic. and it is a 5 level pic two. what does happen when I use prefeedback? because of the fractured output of switches, does this give problems, I now it can be done, with a 4th feedback network. with postfeedback and a fase shift, the working of this multilevel switching get be damaged and things get worse no more multiplying and bad output.

PS pic two is a selfoscillating three level, higher levels with self oscillation I do think will not work.

I do understand pic one what concerns the calculation but do not now what to do with a multiplevel version, maybe I need tread it like a normal pwm modulator (twolevel).

regards
 

Attachments

  • ScreenHunter_2510 Aug. 21 14.47.jpg
    ScreenHunter_2510 Aug. 21 14.47.jpg
    44.3 KB · Views: 333
  • ScreenHunter_2497 Aug. 10 12.46.jpg
    ScreenHunter_2497 Aug. 10 12.46.jpg
    382.6 KB · Views: 335
  • ScreenHunter_2511 Aug. 21 15.03.jpg
    ScreenHunter_2511 Aug. 21 15.03.jpg
    388.7 KB · Views: 174
Last edited:
I am not in a strong position to give you a definitive answer but from what I have seen from my own investigations elsewhere you will not be able to achieve above the proposed Nyquist distance for a single section using pure post filter feedback. In your case that would be 240K/PI [240K/3.25], the paper adds some margin for error, assuming Fs of a single section is 240KHz.

Rather than worry over things assuming that you trust the method used in the paper you are referring to try it out in your model. Design for FT/2PI and run the analysis. This is likely to be stable. Design for FT/PI. This is likely to be borderline unstable. Design for >FT/PI. This is likely to be horribly unstable.

If it is unstable at FT/2PI with the output filter properly loaded then you might start to question the rational of the paper or your implementation of the proposed method. I think you may be doing that already. I'm uncertain as to how your topology operates but you might be able to simplify things further. Attached.

In respect of the paper my view would be that it is not really proposing much over and above what is suggested elsewhere... adding a phase bump at crossover to render a second order system, the output filter, first order at crossover. The methods used to introduce that phase bump are better in that they are placed within the error feedback circuit rather than in the ouput to error input. There also seems to be the suggestion that it might be possible to cope with an unloaded, semi-unloaded, case by using a more complex network to achieve a faster phase transistion. Something akin to a twin-T notch/boost?

It's all likely to become very temperamental very quickly.
 

Attachments

  • Screenshot from 2019-08-23 12-09-24.png
    Screenshot from 2019-08-23 12-09-24.png
    33.5 KB · Views: 173
  • Screenshot from 2019-08-23 12-12-49.png
    Screenshot from 2019-08-23 12-12-49.png
    31.2 KB · Views: 153
  • Screenshot from 2019-08-23 12-23-54.png
    Screenshot from 2019-08-23 12-23-54.png
    17.6 KB · Views: 151
Thanks for your advice, I have this paper, I has contact with the designer of it, but you mention that one section is 240 khz is a answer already, because of the multiplication effect of multilevels, this was some confusing, but I did read properly, I need the section Fs.

Here is the paper of that postfeedback, however being just a single order i do not now if this is enough for low distortions, also multilevel amps does give higher order distortion, special the above the 3 level versions, higher order distortions is more nasty for our ears.

The feedback schematic you mentions is PI(D) controller, I have this schematic also and is mentioned in the paper, just read it because it is interesting stuff.

Postfeedback is quite a problem because of the faseshift caused by the filter, maybe a bessel filter 24dB is better for it has a more liniair faseshift respons over frequency, with a multilevel amp who does 300Khz x 4 = 1.2Mhz there is enough suppression left.

https://www.meng-engineering.ch/_do...er-Feedback-Halfbridge-Single-Supply-170V.pdf

B.T.W I have seen that I need to simulate much longer, if I do not it looks like I have much higher distortion.

Yes I co try your advises later on.

thanks
 

Attachments

  • prepostfeedback.jpg
    prepostfeedback.jpg
    436.5 KB · Views: 164
  • ScreenHunter_2027 Feb. 08 22.17.jpg
    ScreenHunter_2027 Feb. 08 22.17.jpg
    307.8 KB · Views: 149
Last edited: