Class D amps that are superior to all or most?

I said

To those who want to understand, why it’s better to turn off your amps, here it is simply as possible.
(as they like to be drained and recharged otherwise a false permanent "empty floor" happens and your electrolytic power supply caps won’t have the uF storage ability they once had when new

That is not what I was told by an engineer from Cornell Dublier. He told me that to get the longest life the cap should remain charged, and otherwise kept at a low temperature- heat kills electrolytics. If the cap is experiencing high ripple current, this will cause it to heat up a bit- so in such cases it will run longer if shut down and allowed to cool.

In the case of a class D amp, the ripple currents are quite low and the amp tends to operate at a low temperature, so leaving it on all the time is fine insofar as electrolytic capacitor life is concerned.

I borrowed a friends class D with about 300 hours on it and didn't like it. Harsh is the best word.

Try a different amp. Not all class D amps sound the same, just as not all solid state or tube amps sound the same.

@coltrane1 I've explained earlier on this thread why class D has opportunity to do better than class A- and that has to do with how feedback is applied. My comments on this are near the bottom of the first page of this thread. Keep in mind my company has been making class A amplifiers for the last 46 years.

As to the rest of the reply, you cannot have one without the other, i.e. a speaker that will do it when the amp cannot accomplish the same goal and vice versa leaves you with the system not being able to do it so with all respect, it does have to do with the amplifier’s capabilities.

If the amplifier is unable to follow the waveform presented to it (which is the source of all dynamic contrast) then it will distort.

Perhaps a bit counterintuitively, if a particular amp is **more** dynamic than most others (for example SETs) its also because of distortion. In the case of SETs up to about 20% of full power their distortion is pretty benign. But above that point, the higher ordered harmonics become more pronounced. Since the ear uses the higher ordered  harmonics to sense sound pressure and because transients is where most of the power is, when the higher orders show up on the transients, the ear interprets that as 'louder on the leading edges' IOW more 'dynamic'.

Its a simple fact that when audiophiles talk about dynamics, about 90% of the time they are really talking about distortion, and you can safely replaced the word 'dynamics' in that conversation with 'distortion' without changing the meaning of the conversation.


Put another way, amps that sound more 'dynamic' are very likely to do so out of distortion and nothing else. The dynamic character of the music should come from the source, not the amp, that is if the amp is accurately reproducing the signal!

Additionally, when you look at the dynamic range capabilities of stereo source material (which is predominantly recorded at 24 bits) and the emergence of streaming high resolution two channel audio as well as Atmos surround sound, using an amplifier that can keep up with the dynamic range is increasingly beneficial.

While this is certainly true, it has nothing to do with a properly functioning amplifier. Usually it has more to do with making sure that the speakers used with the amp are such that when the dynamic range of the music is expressed, you hear everything without distortion, even if its at lifelike levels.

As an example, in 1959 RCA released the Soria series recordings. Dorle Soria came from Angel to work with RCA and produced upscale recordings and packaging. One such recording is the Verdi Requiem, which is recorded without any compression. The second cut on side one has dynamic range that can bring many systems to their respective knees. But if you have enough power and you have easy to drive speakers its no problem!

Challenge your self, nothing I said was aimed at you, yet you took it to heart, wonder why.

I'm not angry George, just disappointed :)   I often heard problems with class D some years back- weak bass and the like as you've described. But it was obvious to me even 10-15 years ago that class D was a tech that was on a research and development sigmoid curve- and a rising star on which we should keep our eyes. I commented on that often on this site in the past. Its now evolved to the point that a good class D amp is a player against any other amplifier technology, and has already eclipsed them.


I certainly don't/can't take it personally since at the very least none of your comments apply to any of Bruno Putzeys' designs (UCD, Hypex or Purifi) nor ours (and probably a few others for that matter).

but many hear some thing not quite right with Class-D in the upper mids/highs with many speakers, and this is it’s Achilles heel.
Technics (SE-R1/SU-R1000) claim to get rid of it by making the switching frequency 3 x higher, 1.5mhz, so that then the output filter is also 3 x higher and therefore it has little or no influence down into the upper mids/highs.
Most Class-D’s that have around 600khz switching frequencies, have output filters that have effects down into the upper mids and highs.
And yes throwing a **** load of feedback may fix it, but the best amp "theoretically" would be ones that use no feedback if it can be done without distortions, or very little or local not global.

Complete rubbish.

I've explained this many times on multiple threads. Its apparent that faced with facts, George is at a loss; his comments are far too inclusive.


The output filter will have an effect on phase shift in the audio band **if the feedback is insufficient**. So if you see a class D amp that has lots of phase shift in the audio band now you know why.


To prevent phase shift you can do it two ways: have loads of bandwidth (that's how we do it in our OTLs, by going well past 100KHz at full power) **OR** by using enough feedback that phase shift is corrected (which is in excess of 35dB). 


Some class D amps that switch at 500-600KHz sound great and others do not. Its very much like the difference between a type 45 SET and and 750 watt behemoth push-pull tube amp- they are two different things and sound different as a result. Class D in general is no different in this regard.


The last paragraph is really terrible. You don't 'throw a load of feedback' in an amp to fix the amp, you throw a load of feedback in the amp to *fix the feedback itself*. There is a profound difference; to start with you can't just keep adding feedback as you need gain in the amp to do that and each stage of gain degrades the phase margin of the circuit. So it requires a bit of finesse.


To get the required amount in a class D, you have to resort to self-oscillating circuitry (which is also helpful for keeping noise down as opposed to a zero feedback class D design). As a challenge to George: why not work out the oscillation criteria and get back to us here? If you can't don't feel bad, just allow the lesson to sink in that there is far more to this than you think!

No, You didn't, here's what you said:

However, the driven device forms part of the output filter and thus effects the frequency response.

In this case its hard to construe the FR being different from the bandwidth because we are talking about the response of the amplifier in its upper range and how its affected by the filter. But no matter: neither are affected if the amplifier is running sufficient feedback.

There is little change for some products as I showed. For others there is a lot.

It appears to me to come down to this.


The reason is the filter isn't the only variable- class D amps vary quite a lot; some have zero feedback, others have feedback and don't include the filter in the loop, others have feedback and do include the filter, still others are self oscillating and those that do might have the most feedback. The amount of feedback is the key. Less than 35dB in any amp simply won't work.


We've measured no change in bandwidth of our prototypes and Beta units from 20Hz to 20KHz, with less than 1 degree of phase shift over that range, with the load from 2 ohms to 16 ohms. No change.


So depending on the amp we are both correct. I think what is bothering me here is you do not draw a distinction with how the amp is designed and that makes quite a bit of difference to the outcome!

Here is an example of what I'm talking about:https://www.audiosciencereview.com/forum/index.php?threads/review-and-measurements-of-purifi-1et400a...:) FWIW the guy doing the measurements here isn't what I would call a 'fan-boy' :) 

@ieales  If you look at my post that you quoted and the information you quoted next to it above, you'll see that we are saying exactly the same thing. TI's information cannot be construed to be seen as a change in frequency response.

In that post you see me saying this:

Its true that the output filter is affected by the load.

To which you responded:

Methinks not.

Class D Output filters are load dependent.

If you look at these two statements they are quite similar!After that you followed with:

The variations can be small, but they will vary with every load presented. Hence, the perceived sound of the amplifier can vary with each speaker.

Its this conclusion that isn't quite correct. It can be true if the class D amp is using no feedback or insufficient feedback. If it has over 35dB, what happens is the feedback is so profound that it allows the amp to correct for minor changes caused by the Q of the filter with respect to the load impedance. The result is that the change to the audio passband bandwidth and phase shift is not measurable; i.e. there is no change.

when are you releasing your class d amps, ralph...?

@jjss49 We've run into covid-related issues with part availability. We've completed the AC line-related measurements (the concern is meeting EU directives so it can bear the CE mark; if it does this it will meet radiation directives pretty much world-wide); right now we're working on the RFI measurements. The plan is to have the first production occurring this summer.

Are you using tubes on the input....or is it discrete solid state or integrated circuits? Have you tried all kinds (types like toroid or enclosed, wire wound or foil...better wire quality, etc.) of different iron core and air core coils on the output? Everything makes a difference.

The earlier prototypes used no input circuit at all since our preamps have no problems driving a 2K load with plenty of voltage. But obviously we needed more gain so we set it up with opamp inputs. Initially we had used an input transformer but the opamps did a better job with wider bandwidth and lower distortion- and were more musical.


We found off the shelf SMPS to really be hard to work with. They usually have current limiting built in which was problematic. So we went with conventional power transformers and that problem was solved- and for less money by quite a long shot than a proper custom-built SMPS would have cost.


The filter chokes do pose their own issues but almost entirely about how well RFI and EMI is suppressed. If that isn't sorted out properly the amp simply won't sound right and no small part because it can affect other equipment if its leaking noise! If I were you I'd be very careful about replacing the filter choke if you don't have access to a good spectrum analyzer with a calibrated antenna as well as equipment to measure the noise the amp might inject on the AC line.

D tech and it's hybrids are still in their infancy in terms of development with regards to 'high end' audio. Since the latter social group is basically small population-wise, the equipment for now will demand a stratospherically lofty price for those adopting it.

This was true 20 years ago. But right now the tech is pretty mature if you look at it using the technology development sigmoid curve. We're not expecting our initial class D entry to be over $5000.00 yet its making similar power to class D amps that are $18,000.00 (Technics, for those keeping track). 

However, the driven device forms part of the output filter and thus effects the frequency response. Tube amplifier frequency response also suffers driving widely variant loads. Depending on the interaction, the results can be chalk or cheese.

I think you might have a misconception here! Its true that the output filter is affected by the load. But it won't change the FR significantly because of two factors. First, the load affects the Q of the filter. What this means is with lower impedances the filter broadens a bit and is less effective- so you might see bit more of the residual (sine wave at the switching frequency). The second is that in a self oscillating amplifier there is so much feedback that phase shift and the audio passband FR are unaffected. Quite literally there's enough feedback to correct for issues that might arise if the filter is operating at a lower Q.

In our prototypes and in our Beta production amps we're not able to measure a difference in frequency response between a 16 ohm load and a 2 ohm load.

A/AB IS THE WAY TO BE.

If you are talking about solid state, the problem you are up against is linearity of the circuit. Class A is used to help improve linearity, but a very real problem faced by audio designers these days is the semiconductor industry would rather not make linear devices.


So as a result unless one is very careful, feedback has to be implemented in the design to cause linearity. The problem here is twofold: first you have to be careful to not introduce oscillation by exceeding the phase margin of the amplifier circuit. The second problem is a lot trickier- you need to have about 35dB or more of feedback in order to prevent the feedback itself from adding distortion of its own (which tends to be highly audible as its mostly higher ordered harmonics).

To that end you need a lot of something called Gain Bandwidth Product, which to feedback is a lot like gas to a car: When you run out of it there's no more feedback. Put another way, the feedback is gobbling up gain in the amp and this gain has to be sufficient to allow the feedback to do the same job at all frequencies. I'm really oversimplifying this issue but that is because its a very tricky engineering concept.


For decades GBP has been sufficient in the bass region, which is why solid state has a good reputation in that regard. But its not been so good in the mids and highs- most amps made in the last 60 years have less and less feedback as frequency goes up, causing brightness and harshness as that is how the distortion is perceived. Incidentally, this is why distortion is usually measured at 100Hz. If distortion were measured also at 1KHz and 10KHz we'd have a far more accurate picture.


Class D offers a rather elegant way around this by allowing you to impose so much feedback that the amp's phase margin is exceeded, and so it goes into oscillation. The oscillation is then used as the switching frequency, thus killing two birds with one stone. Now you can have in excess of 35dB of feedback so the feedback can clean up after itself. In effect a class D amp is more likely than not to have lower distortion than traditional solid state amps (class A or AB). Its also impervious to weird speaker loads causing it to oscillate because its already oscillating!


Its very natural for a class D amp to make significant lower ordered harmonics due to how dead time and the encoding system make distortion. The result can be that those lower orders can mask any higher orders, resulting in an amplifier that sounds very much like a tube amp but with dramatically lower distortion and therefore more neutral. In addition its output impedance can be far lower than traditional A or AB amplifiers, allowing it to be a better voltage source.

So put another way A or AB isn't always the way to be.

there is something to the notion that many class d stages are a little too clean

If that is the case then the designer didn't pay attention to the distortion signature. It can be quite low in a class D amp, but if the higher orders are not masked by the lower ordered harmonics it won't sound right.