Class D Technology

Also what is the math relationship between bandwidth and phase shift? Where does 10X bandwith number come from? is there something akin to Nyquist that is used to determine this? Did the engineers making high end Class D amps miss the boat on this somehow? it does not sound like they did in practice based on listening, at least the good ones seem to know what they are doing.

From my perspective there is understanding the theory which is useful and how things sound, for which there is no mathematical equation to properly represent that. Pundits focus on the strengths of a supporting theory and skeptics the weaknesses. No approach is perfect and holds all the cards. That’s clear by the variety of successful amplifier designs out there today.

Ralph sounds like you are actively prototyping your own class D designs? that tells me you think there is in fact something worthwhile there if done right.

Dear Audio Friends,
I write this as one who is rather un-technical but possessing excellent 'ears'.;-)
As one of the few who have all 3 main amp designs in house and active in-system, and who owns 6 pair of priastine and extremely detailed Apogees, including a purely1 ohm Scintilla pair completely rebuilt by Rich Murry of True Sound Works, I am making a rare post here to clarify a couple of things re my experience with amps of all persuasions:
In my room as I write this I have 2 of Henry Ho's H2O SE amps driving my Scintillas, and the staging, imaging, frequency response and range are world class. I gave up a pair of beloved Class A Nelson Pass XA100.5s for them, as I could tell very little difference whiule getting some serious power benefits for my room and speakers. Please note that Henry Ho first made his name as a successful and well-reviewed Class A amp builder, and his Class D amps have Class A grade power supplies. They weigh 60lbs each, and mine have some extra-fine caps, to boot.
I also have a pair of custom Bob Carver built-for-me-by-Bob-Himself KT120 tube amps that he designed specifically to drive 1 ohm Scintillas. They have a 1 ohm tap, and sound exquisite, with a slight wider stage and image than the H2Os, but not as tight a control over the transducers, of course. These are NOT like my Kronzillas which, though superb and with unrivalled detail and finesse on my other Apogees, can NOT drive a 1 ohm Scintilla - nor would I ever try!
I have other amps, hybrid Vincents, Wyred4S SX500s, a fine Class A Coda 3.2 Stage, and the H2O and Carvers out-do them all by a wide margin in every particular.The only exceptions at all are the T1610-tubed Kronzilla DX MkII monos.

I hope my personal and ongoing experience with these amps might lend a bit of boots-on-the-ground substance to an otherwise thorny and subjective discussion.

My final and main point, really is this: From my experience it seems - and I have discussed this in detail with both Bob and Henry - that each style and design philosophy has its strengths and weaknesses: HOWEVER, as the designs and designers progress to the pinnacle of what is possible to each, the differences become a matter of diminishing returns. They each become less distinguishable in a blind test, with almost equally transparent and less 'visible' between listener and music.

That is and has been my experience.
All the Best

shibui,

Sounds to me like you pretty much nailed it in all regards based on actual experience seeking the best sound possible with some of the most challenging speakers to drive well ever.

That a Class D design can even compete in that arena says all that needs to be said really. There is no reason to categorically reject the approach. From there it comes down to personal preferences and case by case details that vary widely most likely. Plus things can only continue to get better as/if needed as bandwidth continues to increase over time. Better performance always tends to come for additional cost. class D is no different there except it lowers the price barrier for what most would consider good performance especially when more power is needed to get the most out of less efficient speakers. Ability to get the most out of more challenging speakers is the primary value added use case for Class D these days I would say though I find the newer ones to be top notch as well with easier load speakers I own. Class D has kept me from pulling the trigger on a tube amp now (and associated speaker changes that would be needed) for several years.

Also what is the math relationship between bandwidth and phase shift? Where does 10X bandwidth number come from?

Hi Mapman,

If I'm not mistaken class D amplifiers typically use an output filter consisting of a series inductor and a shunt capacitor.  Together with a primarily resistive load that will form what is known as a second-order low pass filter.  "Second-order" meaning a filter that increases the amount of attenuation it provides by 12 db/octave (12 db per doubling of frequency) above the frequency at which it has rolled off by 3 db (that frequency usually being what is referred to as the bandwidth of the filter). 

The equation defining the phase shift introduced at various frequencies by a second order filter is complex, and is shown (approximately!) as equation 3 on page 2 of this reference.

To provide some perspective, however, it may be helpful to consider the much simpler case of a first order filter (6 db/octave rolloff), which is what would be formed by the combination of a series inductor and a resistive load, without the capacitor.  A first order low pass filter will shift the phase of a given frequency f by an amount equal to:

Phase shift = arctangent (f/bandwidth)

So a first order filter having a 3 db bandwidth of 200 kHz would shift a 20 kHz signal by arctan(20/200) = 5.7 degrees.

The 10x figure is a rule of thumb, as Ralph indicated, chosen to limit the phase shift introduced at frequencies of interest (e.g., at 20 kHz and lower) to amounts that are presumably inaudible.

It should also be understood that while for a pure sine wave at a single frequency any amount of phase shift will be inaudible, a musical note consists of a combination of many frequencies that are simultaneously present.  And the goal is to achieve proper alignment of the timing of all of those frequency components relative to each other.

Best regards,
-- Al
       

Assuming nothing is there to start with isn't a phase shift of 0 = 0?

Nope! The point is that if you don't have the bandwidth in the amp it will cause phase shift at lower frequencies- down to about 1/10th the cutoff frequency is then its considered negligible (of course, 'negligible' by whom is a different story; obviously some designers don't care about that so much). IOW the ear might hear 20-20KHz but to reproduce that in an amplifier without phase shift you need 2Hz to 200KHz if you really want to do it right (the same rule applies on the bottom end; otherwise the bass loses impact due to phase shift). For this reason Stuart Hegman, who designed the h/k Citation 1 and Citation 2 was a big fan of wide bandwidth. 

I did not know vinyl has frequency bandwidth so high.   Most home hifi always talked of 20-20K frequency response.     Is this in practice or theoretical?   I'd agree theoretically vinyl could do more but practically its news to me.   The best digital (higher res) sounds as good as most vinyl to me these days.   RTR is better but look where that got us practically.

Its not just theoretical.

Its true that most LP systems don't spec past 20KHz but you would be very much mistaken if you think it stops there. Most phono cartridges and phono sections made since the mid 1960s go much higher than that (although it does not show in the specs as at the time there was no thought that it made any difference). Our cutter head is an early Westerex 3D, made about 1959 and it has no worries cutting 30KHz which can then be played back by a 1970s Technics on a 1970s h/k receiver (the Westerex mastering system has a filter that cuts it off at about 42KHz). By contrast RTR does not have this sort of bandwidth; about 25KHz or so is the best you're going to get at 15 i.p.s. IOW LPs have wider bandwidth than tape and its been that way since the inception of reel to reel. 

Every time a new system has come into the scene the previous knowledge of the prior engineering often goes by the wayside for a while until the new technology gets its pants on. This happened with digital; its only been recently that its begun pressing bandwidth past 20KHz and we're seeing the same thing with class D right now.  In time this will all get sorted out as the technology improves to the point that such bandwidths are routine; until then essentially what you will see is the industry collectively placing its head in the sand as if these facts don't exist. But this is not rocket science and we've known that bandwidth is essential going on half a century now. But people have short memories when new technology comes in... Look how long people put up with unruly fuel injection while carburetion was pretty figured out and actually performed better. But fuel injection was 'new' so people put up with it. Now days its sorted and no-one would consider a carburetor. But literally it took 3 decades to get there!