The simple tweak for NFB

[phase_accurate]

Hi Lars

You may not only play with the RS/RP ratio (and their respective cap values) but also with the ratio RP/RF and RS/RF in order to keep the feedback at least partially "conventional".
You might even keep the conventional fraction dominant as John suggested.

Regards

Charles

 

[sovadk]

Lars, what is the connection between the error suppression from simulations, the frequency/phase response and your experience of the sound quality? Could you post some simulation results and give them comments based on actual listning test or something like that?
Lets make this scientific.

Regards Kaspar

 

[Lars Clausen]

Kaspar: Sorry i never use simulators 😉

 

[Bruno Putzeys]

Kaspar: Sorry i never use simulators 😉

For loop orders above 3 (like in my AD converters: 6 or 7) I too skip the simulation stage, going straight from the maths to the hardware. OTOH, for a simple PID loop around the output filter, simulation and some -uh- hand-optimisation works about as fast as getting at the theoretical values through symbolic maths.

 

[classd4sure]

Hi,

That's interesting. Exactly how close do those simulated values match those in final use. ?

Mine are by a rather large margin.

Thanks,
Chris

 

[Bruno Putzeys]

That's interesting. Exactly how close do those simulated values match those in final use. ?

I rarely simulate whole amps. Certainly not in the hope of getting realistic distortion figures. Most notably, even the best MOSFET models are only so-so when you're really pushing them.

Pspice is very good at simulating non-saturated-switching bipolar and small-signal stuff though, so the comparator and gate drivers can be reliably simulated.

The modulator is simulated using an "ideal power stage" for time domain simulations and using a linearised gain block for AC simulations.

In answer to your question I'd say that barring actual power switching, the simulation is quite reliable.

 

[Lars Clausen]

Below you can see our post filter NFB setup for the ZAPpulse. The values shown are the ones that work best for ZAPpulse, and will give a constant switching frequency independent of the load. For other Class D amplifiers other values might work better. The sound is best with plastic capacitors, compared with ceramic NP0 caps.

 

[Lars Clausen]

After presenting post filter NFB operation (UcD mode) to our listening board, we have now discarded any potential future products of this kind.

Even if the sound in some areas become more 'in control', especially the midrange, the sound stage of the post filter NFB amplifier also is restricted to 'within the loudspeakers'. There is no 'A Live' sense in the music, and nothing is coming 'out' towards the listener, to give a true sense of the music. In other words much of the experience of the music is missing.

So Compared with ZAPpulse in post filter NFB mode or other brands' post filter NFB amplifiers, we all found the original ZAPpulse (pre filter NFB) to be more alive and natural in the sound. And with true sound stage.

So from our side, post filter NFB for Class D amplifiers is officially dead.

 

[Bruno Putzeys]

So from our side, post filter NFB for Class D amplifiers is officially dead.

That's good news for us.

 

[ghemink]

After presenting post filter NFB operation (UcD mode) to our listening board, we have now discarded any potential future products of this kind.

Even if the sound in some areas become more 'in control', especially the midrange, the sound stage of the post filter NFB amplifier also is restricted to 'within the loudspeakers'. There is no 'A Live' sense in the music, and nothing is coming 'out' towards the listener, to give a true sense of the music. In other words much of the experience of the music is missing.

So Compared with ZAPpulse in post filter NFB mode or other brands' post filter NFB amplifiers, we all found the original ZAPpulse (pre filter NFB) to be more alive and natural in the sound. And with true sound stage.

So from our side, post filter NFB for Class D amplifiers is officially dead.

I don't understand this. With my UcD180 based active system, my soundstage is not limited to within the speakers. If I try for example the CD "Amused to Death" from Roger Waters, some of the sounds are coming from right side and left side of the listener, so a few meters before the speakers. For example the barking dog at the beginning of track 1 comes from right middle/back (using of course only two speakers in front). On another track, there is also a piano coming from that direction. Quite distracting to hear some of the sounds coming from the front while others come from the sides. So nothing wrong with the imaging I would say. In fact, I often use that CD to check these extreme ways of imaging, a kind of trademark of Roger Waters who does similar things on other CDs as well. With very good recorded SACD/CDs of for example TELARC, DMP, CHESKY I get wonderful sound and imaging. I don't think imaging is a problem that can be attributed to post filter feedback.

One questions, could you try to measure the freq. response of the ZAPpulse connected to the speaker in both cases, so with pre-filter feedback and with post filter feedback. I expect to see some difference. Maybe that difference explains why pre-filter feedback sounds different?

Best regards

Gertjan

 

[IVX]

🙂 nowadays everything that use after filter NFB =UcD mode?

 

[Lars Clausen]

IVX: No, many manufacturers use post filter NFB, Icepower, SONY S-Master and Tripath just to mention a few. We also not only used UcD for the comparations, also some of the other aboves.

I heard that at this point in time more than 50% of all power amplifiers in the world are now Class D. (!)

So i guess in this light Class A/B is not the 'normal' amplifier anymore but a rarity nowadays. 😀

Obviously many brands not so profiled here on DIY audio.com use also post filter NFB, if not for other reasons, just that it makes it possible to use small and cheap filter inductors. (Note that Jan and Bruno are not taking special advantage of this loophole). But others do.
A pre filter NFB (like ZAPpulse) has to use massive low distortion coils to avoid contaminating the signal with distortion from the ferrite core. For example the ZAPpulse core is 8 times bigger (in ferrite volume) compared to UcD, and maybe 16 times compared to Icepower. This costs money!

Ghemik: Yes i also use Amused to death as a test for soundstage. I don't remember which track, but there is one track with a stagecoach driving from left to right, in the rainy weather. That one should actually drive BEHIND you !!! ... I don't know how they do it!

 

[Jocko Homo]

Someone asks.........

One questions, could you try to measure the freq. response of the ZAPpulse connected to the speaker in both cases, so with pre-filter feedback and with post filter feedback. I expect to see some difference. Maybe that difference explains why pre-filter feedback sounds different?

I made reference to that here:

http://www.diyaudio.com/forums/showthread.php?postid=655964#post655964

 

[ghemink]

Ghemik: Yes i also use Amused to death as a test for soundstage. I don't remember which track, but there is one track with a stagecoach driving from left to right, in the rainy weather. That one should actually drive BEHIND you !!! ... I don't know how they do it!

Hi Lars,

That's interesting, I don't think there are that many people having that CD. Actually I hardly listen to it, it's quite old as well (1992) but when I want to check imaging out of the speakers I always use it. They call "the trick" QSound. It is obviously created by some clever manipulation that makes you believe that the sounds comes from a direction where it actually does not come from 🙂. I found that speakers, speaker placement and amps need all to be quite good the get the effect right.

There is also a rainy whether thing with some thunder and so on on track 3 I believe. The thunder starts somewhere at the back and than rolls around the listener and goes to the front from right to left. This kind of stuff makes me believe that you do not need 5 or more speakers to have surround sound. Nowadays, most if not all DVD players also emulate the speakers that you don't have with SRS surround. This also can work quite well with two speakers only.

Anyway, we are getting a bit off-topic here.

Best regards

Gertjan

 

[ghemink]

Re: Someone asks.........

I made reference to that here:

http://www.diyaudio.com/forums/showthread.php?postid=655964#post655964

Hi Jocko,

Yes, I know about that, that' s why I asked. It maybe that the post-filter feedback has a slow roll-off at the high end while the pre-filter feedback is slowly going up because of increasing impedance of some speakers in combination with the LC filter. This will obviously sound different with some people suggesting that a slowly going up freq response makes dull sounding recordings sound better. Because of this, I asked the freq. response to be taken with the speakers connected. Obviously, these effects would be very speaker dependent and may work for some speakers and not for others. If this is the case, then some form of tone control should be able to give the same effect. However, things like tone control in high-end world are taboo while things like frequency dependent output impedances of amps are being used as a kind of tone control and it is accepted.

Best regards

Gertjan

 

[IVX]

Lars, what i mean:
here http://www.diyaudio.com/forums/attachment.php?s=&postid=662340&stamp=1118660272 is not UcD, and also far from Charles suggestion, which he gave you here http://www.diyaudio.com/forums/showthread.php?postid=655739#post655739. 1M+1pf in serial (!), if you'll hear difference with/without - i'm sure it's placebo effect.
BTW, Rf=1M as feedback res and lm6172 input bias current 4uA + 3uA input offset current, in the worst case, it's will give several volts to output lm6172, and much more noise vs Rf=10kohm. Your zappuls really use Rf=1M?

 

[Bruno Putzeys]

IVX: No, many manufacturers use post filter NFB, Icepower, SONY S-Master and Tripath just to mention a few. We also not only used UcD for the comparations, also some of the other aboves.

It may be helpful to note that S-master is a fully open-loop digital class D with no feedback whatsoever, let alone past the output filter. Tripath has a patent description on how to close the loop around the output filter (a patent for a PID controller? cuuute!), but none of the commercially available T implementations has post filter feedback.

A pre filter NFB (like ZAPpulse) has to use massive low distortion coils to avoid contaminating the signal with distortion from the ferrite core. For example the ZAPpulse core is 8 times bigger (in ferrite volume) compared to UcD, and maybe 16 times compared to Icepower. This costs money!

The Amidon coils such as those used by JohnW are famously linear as well without being overly large.

Ghemik: Yes i also use Amused to death as a test for soundstage. I don't remember which track, but there is one track with a stagecoach driving from left to right, in the rainy weather. That one should actually drive BEHIND you !!! ... I don't know how they do it!

Yesterday I listened to the "ultimate 2-channel surround trip". It's a CD called "Live '93" by a group called "Orb". Highly recommended. It was mixed using Roland RSS10 boxes. Many Madonna CD's are mixed using Qsound processors. Both use HRTF (head-related transfer function) pre- and postcorrection to place sound sources outside the stereo base. Qsound is more subtle and only works really well in the frontal half, but it does so without sounding phasey. The RSS10 boxes can pan all over the 4*pi sphere around the listener very convincingly. However, in those days, processing power was at a premium. Early Qsound used off-line (overn¡ght) processing whereas RSS10 was realtime, severely restricting the complexity of the filtering. This produced a phaseyness that makes it instantly recognisable.

 

[ghemink]

Ghemik: Yes i also use Amused to death as a test for soundstage. I don't remember which track, but there is one track with a stagecoach driving from left to right, in the rainy weather. That one should actually drive BEHIND you !!! ... I don't know how they do it!

It is on track 8, it is some kind of slee with bells driven by horses. It is coming from left and going to right, it does not drive behind me however, is that the one you mean or is there another one?

For this QSound stuff to work well, you need to sit in the sweet spot, experimentely I found that the best spot is when the listener and the speakers form an equally sided triangle, so distance between speakers same as between speakers and listener. I guess the calculations that are done on the signal assume such a speaker/listener setup. Probably the same for SRS surround.

That whole CD has amazing imaging effects that become stable located when you sit in the sweet spot. Outside the sweet spot, they can besome a bit phasy.

Best regards

Gertjan

 

[ghemink]

Ghemik: Yes i also use Amused to death as a test for soundstage. I don't remember which track, but there is one track with a stagecoach driving from left to right, in the rainy weather. That one should actually drive BEHIND you !!! ... I don't know how they do it!

It is on track 8, it is some kind of slee with bells driven by horses. It is coming from left and going to right, it does not drive behind me however, is that the one you mean or is there another one?

For this QSound stuff to work well, you need to sit in the sweet spot, experimentely I found that the best spot is when the listener and the speakers form an equally sided triangle, so distance between speakers same as between speakers and listener. I guess the calculations that are done on the signal assume such a speaker/listener setup. Probably the same for SRS surround.

That whole CD has amazing imaging effects that become stable located when you sit in the sweet spot. Outside the sweet spot, they can become a bit phasy.

Best regards

Gertjan

 

[Bruno Putzeys]

That whole CD has amazing imaging effects that become stable located when you sit in the sweet spot. Outside the sweet spot, they can become a bit phasy.

Now completely off topic 😀

You also need fairly dry room acoustics to make the effect work well. Eelco Grimm told me that Madonna's Ray Of Light (ie. Qsound) has a complete wrap-around sound stage as well, IF you listen to it in an anechoic room. It's not necessary to have a particularly good stereo system, but the room is critical.