Time to buy a class D amp?

Would you say there is inherent limitation in the phase shift that is a factor of the comparator speed, digital logic delays, and turn-off time of the FET? What do you think a practical limit is on that? That may put a practical limit of a few-10 degrees at 20Khz, but that would be inaudible.

The phase shift is indirectly related to the speed of parts like the comparitor (which typically has plenty of more speed than the output section). It is directly related to the filter at the output.

    As I understand it from you and Bruno Putzeys as a layman, there is no such thing as too much feedback with class D amplification. The generally accepted concept that feedback negatively effects the sound quality of traditional linear amplifiers may be true, but this aversion to feedback doesn't apply to class D amplifier design. Very high levels of feedback, in an intelligently designed class D amp, are actually utilized to optimize the sound quality throughout the entire audible frequency spectrum.  
    Am I understanding this correctly?

That's it in a nutshell.

The problem is phase shift in traditional designs prevents the application of the required amount of feedback in such amps; otherwise oscillation will occur. For this reason **every tube and solid state amplifier made has had insufficient feedback**. This has resulted in the applied feedback adding its own distortion- which is interpreted by the ear as brightness and harshness. This is why every amplifier with feedback up until now has sounded brighter and harsher than real life, and has fueled the tubes vs transistor debate all these decades. The industry has also been complicit in this problem, so when you see harmonic distortion measurements the fundamental frequency is usually fairly low (60Hz is common) so as to not get in trouble with the gain bandwidth product limitations of the amp under measurement! And this also explains why zero feedback amplifiers (like our OTLs) exist, as by designing a circuit that is linear enough to run without feedback the harshness and brightness of feedback is avoided.


Transient Intermodulation Distortion is a symptom of poor feedback loop design coupled with poor gain bandwidth product and insufficient feedback.

Can you define the time domain characteristics of feedback in a Class-D amp and linear amp?

Yes. In a class D amp its all about propagation delay. In a conventional amplifier its all about phase shift as capacitive strays roll off the response (introducing phase shift). Effectively both have the same effect- at some high frequency the feedback is no longer negative so oscillation can occur. But unlike a conventional design, in a class D you can take advantage of that oscillation by using it as the switching frequency.

atmasphere,

     Thanks for the very good info on amp feedback. 
     On a totally different subject, I just bought a pair of pre-owned Magnepan 3.7i speakers to replace my current almost 25 year old 2.7QR panels.  I'm thinking my Levinson 326S preamp (very quiet, neutral, detailed while portraying a very good soundstage) and pair of D-Sonic M3-600-M class D monos (1,200 watts into 4 ohms) will be a good match with the 3.7i but I won't know for sure until they're installed tomorrow and can begin listening.
    I'm just going to listen, evaluate and hopefully enjoy this combination for a few months.  I have high expectations but, as we all know, the proof's in the sound quality and our subjective opinion.
    If I don't thoroughly enjoy the results, I'll probably be in the market for a new pair of fully balanced monos, preferably class D.  I'd love to audition your new class D amps if I'm in the market, before I purchase anything else and if they'll be available for audition and possible purchase in the next 3-6 month timeframe. I understand if you're currently reluctant to share any definitive information but would appreciate any info you're comfortable sharing at this point.  I'm very curious and interested in listening to your efforts.

Thanks,
  Tim

We're still working with prototypes. They have had 100 watts and double power into 4 ohms. Since last spring, they have had GaNFET output sections. We compare them directly against our tube OTLs as a reference. Mostly we've been sorting out layout issues; designing a class D from scratch using individual parts is challenging if you want high switching speeds! But we've been sorting things out, sometimes one at a time as problems are identified and solved.

Hello atmasphere,

     Thank you for your very good update on your class D amp development. Very interesting, although not very surprising, that you decided to utilize GaN FETs in your design given their ultra fast switching on/off capacity which enables the elimination of dead-time and lowers distortion. I’m not sure if you discovered yet what Merrill, of Merrill Audio, stated that he has discovered: that the newer GaN transistors don’t perform as well as some of the older examples.
     I remain very interested in listening to your new class D amps in my system. I recently upgraded my system with a pre-owned Levinson 326S preamp and Magnepan 3.7i speakers. I’m currently using a pair of D-Sonic M3-600-M mono-block class D amps rated at 1,200 watts @ 4 ohms that are performing very well driving the 3.7is.
     I understand the 3.7is likely don’t require 1,200 watts to perform optimally but I believe they benefit from a ready and steady availability of high current. My issue is that I’m uncertain if 200 watts @ 4 ohms would be sufficient for them. I’m also uncertain whether there would be a net overall sound quality gain in my system by utilizing very well designed class D amps, with reduced dead-time and distortion, but perhaps less than optimum current availability. I’ll likely only find out for certain by auditioning them in my system.
     Have you made a decision yet on whether you’ll be offering free limited in-home trial periods for auditioning?

Thanks,
Tim

Have you made a decision yet on whether you’ll be offering free limited in-home trial periods for auditioning?

We're not nearly that far along!
With regards to GaN performance, they can be variable depending on who made them and their intended application. Many of them have a reverse diode conduction phase which can really throw the designer for a loop. While they are faster, MOSFETs have been steadily getting faster too so at practica switching speeds we're seeing now in another year or two MOSFETs will probably work just as well. One problem we're seeing is its very easy to build a class D with a very high damping factor, much higher than any speaker should really see.