Class D discussions

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Did you read this thread? My very first Class D pwm (switching) amplifier. I have begun working on a subwoofer amp in which the P-ch and N-ch source follower output mosfets have their drain voltages variably regulated to track 10-15v above the output. I think this way may produce better results and maybe lower distortion. I thought I may have been happy with the 0.1 to .5% distortion of regular class D, but after reading many views, I am not sure if I will be happy with the performance of the N-champ I designed.

Borrowing some design ideas from Johan's group at audioexperiment , I have been able to design an all N-ch
output class D which shaves off about 50nS off of the propagation delay, but am not sure if it is worth the trouble.

Could it be possible that class D becomes more unstable as the damping factor from the phase compensation is reduced? This could be one reason why it is not so popular. Another possible hitch is that the output is fed back to the input with phase shift, and this may be a cause of distortion which cannot be eliminated, but only lessened with increased switching frequency. I have found that with some of my class AB designs, the time delay between signals on the bases of the differential input pair is only about 2 nanoseconds while could class D be as much as half the switching frequency? I should like to investigate the possibility of the phenomenon.
 
I have not built one without dead time using the double inductor method instead. But simulations indicate a big loss of efficiency. I think one big source of distortion in class D is also the delay of the filter being fed back to the input stage. The non-feedback types may not suffer from the type of distortion I mentioned earlier, but they tend to color high frequency signals and are very sensitive to imperfections in the power supply.
 
Joined the audioexperiment yahoo group. Thanks for the info. They talked about dead time causing distortion.

The Class D that Sony made in the 70s (I think) did not have feedback and they switched at 300 KHz. I don't remember all the particulars. I don't believe they sold any of these units to consumers.

The propagation/switching delays from the input of the PWM to the input of the output inductor(s) should be minimized. Large delays relative to the switching frequency period show up as phase shift, I think. Pushing the switching frequency higher and minimizing delays, should enable feedback to be effective in lowering output impedance if nothing else.

Have you tried using any kind of peak or average current mode feedback with an outer voltage feedback loop?

Are you referring to the output filter delay or phase shift?
 
Yes, I think that distortion is caused by the mixing of the delayed feedback signal needed for oscillation with the audio input. The result may be a blurring of the original signal before it passes through the amp. If no feedback is used the issue does not exist because simple phase shift cannot be considered distortion in and of itself. That is one thing to beware of when doing distortion measurements. When measuring distortion, simple phase shift from a simple input filter, for example, is erroneously picked up as distortion.
 
Ran some simulations with switching frequency at 500 KHz. Phase shift about 35 degrees at 20 KHz input, with the LC filter values I chose. Tried 1 MHz switching frequency. Phase shift less than 5 degrees at 20 KHz input, with the LC filter values I chose.

If I chose to use a Class D amp, I would use it for the low frequencies, and use a linear amp for mids and highs. Using it for low frequencies would not require a high switching frequency and the efficiency should be very good.

Class D makes a good 50/60 Hz inverter for power.
 
Ran a simulation with a switching frequency of 25 KHz and 500 Hz input. Phase shift was less than 10 degrees, with the LC filter values I chose.

Maybe a bi-amp method would work better using 2 Class D amps. That way slower switching large die transistors could be used for amplifying frequencies below 500 Hz where most of the power is needed. Then mids and highs could be amplified using faster switching medium die transistors where less power is needed. Could go further with tri-amping of Class Ds.
 
hi all,
IMHO, class D (UcD like) rather at high frequency give more advantages(transparency, detail resolution), bass&midbass just identical to class AB(to my own AB). However the difference isn't easy audible in sounds with noise like harmonics distribution(hi-hat, snare etc), but hear the difference in vocals, violins or tambourines can my wife and kids even.:) It seems that class D have better spectral resolution, and i don't know why.. :whazzat:
 

Attachments

  • amplif_clased_max4295.zip
    17.9 KB · Views: 124

TNT

Member
Joined 2003
Paid Member
Agreed - exactly !

Can someone answer IVX question - this would be interesting to know !

/j


IVX said:
hi all,
IMHO, class D (UcD like) rather at high frequency give more advantages(transparency, detail resolution), bass&midbass just identical to class AB(to my own AB). However the difference isn't easy audible in sounds with noise like harmonics distribution(hi-hat, snare etc), but hear the difference in vocals, violins or tambourines can my wife and kids even.:) It seems that class D have better spectral resolution, and i don't know why.. :whazzat:
 
Some opinion does exist.. that amp have "zero feedback sound" because THD figures is undepend from frequency, and loop gain is constant from DC to 20kHz, and THD of the open loop system is constant from DC to 20kHz. I can do so that THD will increase with frequency (by loop components adjusting)..probably i'm not advanced listener too much.. IMHO the transparency isn't worse.
 
Welcome back to class D discussion, Ivan. That question is a burning one, TNT. I suspect that the very phenomena involved in causing the higher overall distortion in class D, may through some strange twist of physics, actually cause the improved definition and spectral response at higher frequencies! Apologetically, with my bad ear, I am not in a position to verify better performance of the UcD vs. class AB at high frequencies. However, could this difference disappear when a high frequency response class AB is used?

Best Regards
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.