On Class D Amplifiers Measurements

[March Audio]

What's the tweeter inductance?That goes a long way toward reducing the current. Here's the idle of some Class D amps I have on hand, showing about 750mV of ultrasonic. Amount of tweeter heating and detected in-band intermod: below my measurement limits.

View attachment 39125

I do not understand why usual measurements are almost always restricted to audio band measurements, 20Hz-20kHz, maybe 50kHz. SINAD, THD vs. amplitude, THD vs. frequency. Is that enough, really? Isn't it pointless? Above some level of parameters, like SINAD = 80dB, do we find differences by listening? Why do we concentrate our efforts at those standard measurements only? Because they are easy to make? Why do not we try to find broader consequences? In the age where we are surrounded by EMI pollution, digital sources having MHz and tens of MHz rubbish at the output? Isn't it the key to perceived differences, these out of audio band interferences transformed as intermodulations and aliases into audio band? Why do we resist to investigate something else than a conventional, decades old audio band measurements?

It's very simple.

They are restricted to the audio band because that is what you can hear.

We are only interested in out of band signals if they cause problems in band. If these problems exist they
are indicated by our current measurements.

We do look at out of band signals, we don't ignore them, it's just that you need to put their consequences into perspective.

 

[restorer-john]

Why do we resist to investigate something else than a conventional, decades old audio band measurements?

Because current Class D amplifier manufacturers (and their proponents) are attempting to control the narrative on what is important and what isn't. Wide band amplification touted by everyone for many decades is no longer compatible with current Class D topology. Remember for a least 20 years we saw pictures of perfect square waves in advertisements, citing the superiority of wideband electronics, HiFT silicon, lightning fast rise times and perfectly resolved transients? Nobody wants to look at a Class D square wave do they- not exactly a thing of beauty?

Let's go back to the 70s and 80s when these types of bandwidths were common.

Fashions do full cycle and we are heading back to to bandwidth limited amplification...

 

[March Audio]

To answer the question whether something is being introduced into the audio due to RF stuff being present in signal?

I understood OP’s point to be introduce out of band signals to measure any audible artifacts and if there is proof that such artifacts exist then measure out of band to see if and which upstream devices send such signals.

So we know dacs do output a degree of RF. We play a dac into an amp and look at the in band noise and distortion. We see nothing untoward. Therefore we know its not a problem.

 

[March Audio]

Because current Class D amplifier manufacturers (and their proponents) are attempting to control the narrative on what is important and what isn't. Wide band amplification touted by everyone for many decades is no longer compatible with current Class D topology. Remember for a least 20 years we saw pictures of perfect square waves in advertisements, citing the superiority of wideband electronics, HiFT silicon, lightning fast rise times and perfectly resolved transients? Nobody wants to look at a Class D square wave do they- not exactly a thing of beauty?

Let's go back to the 70s and 80s when these types of bandwidths were common.

View attachment 39188

View attachment 39189

View attachment 39190

Fashions do full cycle and we are heading back to to bandwidth limited amplification...

There is absolutely nothing wrong with class d square waves. If you are concerned about band limited audio then you will abhor redbook digital audio....... And yet....

 

[March Audio]

Can you, on the other hand demonstrate it is not? The spurious/out of band output content of digital source components connected to the various Class D amplifiers is an unknown.

There will be 'audiophiles' with NOS filterless D/A converters putting out goodness knows what into their Class D amplifiers, just as there will be guys with SACD/DSD players with unknown HF content.

In real terms, PMA has done (unintentionally) what the high end companies do (and you heard it from here first). He has created, or discovered, a problem that really doesn't cause a serious issue on the whole, but is rooted in science, and all he has to do now, is manufacture and market a grossly overpriced solution in the form of nice wooden or extruded aluminium cased special filter with gold jacks and a nice velvet presentation box.

All it needs is a catchy name. Audiophiles will lap them up by the thousand, others will deconstruct and make their own, but it will become an essential add-on. "Do you want a filter with that Class D amplifier, sir?" Even if Purifi and others put perfectly adequate LPFs into their gear, the standalone filter will be a "belts and braces" approach no audiophile will be able to resist. Without one in the system, their friends won't take them seriously. In fact, each digital source component (an FM tuners) needs one too. I sense a new cash cow!

Yes. Every day I make measurements of class d amps and there are no significant in band spurious signals seen.

Again this is about design. If a company sells an amp, be it class a, ab or d without an input filter with DC to daylight frequency response, then expect potential problems.

I am aware of the potential for problems so I have implemented a mitigation. Problem solved.


It will be in a standalone box for $10000 tomorrow.

 

[March Audio]

PURIFI 1ET400A has a PSRR of 90 dB, very good data. Forced 1Vrms f≤1kHz ripple, either rail.

Hypex NC500 has a PSRR: 75 dB (min) and 85 dB (typ).

The Hypex module needs an aditional RF / EMI filter (if mains is noisy, usually in urban concentrations). And two stages filter work without problems. They are economical.

Maty for goodness sake shut up about your bloody mains filters. You bring this into every thread. This topic isn't even about RF on the mains.

@Thomas savage. Can you give maty a forum holiday to get get the message home that this behaviour of spamming every thread is not acceptable?

 

[restorer-john]

There is absolutely nothing wrong with class d square waves.

I know that. You feed Class D a nice little bandwidth limited square wave and all will look just as it should. But feed it a real square wave, say one with a >100KHz bandwidth, and compare input to output. Not so pretty huh?

Now try that with one of the amplifiers listed above and tell me what you see. That's the message sold to several generations of audiophiles through thousands of oscillograms, advertisements and brochures. Class D is trying to undo all that to suit its own narrative, let's be frank about that.

Consider every bit of discussion here is not remotely an attack on you, or your excellent products. There's no need to go into bat for everything Class D, and get so defensive because we all secretly (or publicly) want a few of your amps.

 

[blueone]

There is absolutely nothing wrong with class d square waves. If you are concerned about band limited audio then you will abhor redbook digital audio....... And yet....

Isn't there a difference between what DACs do, which is limit bandwidth with a filter that limits the output spectrum, and a digital amplifier, which has artifacts outside the audible frequency spectrum?

What is the mitigation you mentioned? A low-pass filter on the outputs?

 

[audimus]

So we know dacs do output a degree of RF. We play a dac into an amp and look at the in band noise and distortion. We see nothing untoward. Therefore we know its not a problem.

I am not disbelieving you. But this should be verified with a carefully designed procedure by objective interests that don’t have a skin in the game. The premise of this site.

It would be nice to establish what kind of out of band frequencies are emitted by upstream sources like DACs in frequency and levels by measuring their output in a wide band and make the worst case result part of the test input signals to audio band measurements of Class D amps to ensure that they behave well under such circumstances. This is my takeaway from the OP.

If it turns out that after a few reviews and measurements, this seems to be a non-issue then it might be retired. But the investigation would be worth it.

 

[March Audio]

I know that. You feed Class D a nice little bandwidth limited square wave and all will look just as it should. But feed it a real square wave, say one with a >100KHz bandwidth, and compare input to output. Not so pretty huh?

Now try that with one of the amplifiers listed above and tell me what you see. That's the message sold to several generations of audiophiles through thousands of oscillograms, advertisements and brochures. Class D is trying to undo all that to suit its own narrative, let's be frank about that.

Consider every bit of discussion here is not remotely an attack on you or your excellent products. There's no need to go into bat for everything Class D, and get so defensive because we all secretly (or publicly) want a few of your amps.

No, it will be perfectly fine within the bandwidth limits of the amp, which for Hypex is 50Khz and Purifi 60khz (-3dB).

Bandwidth limiting is bandwidth limiting. What's your point? Why are you not concerned about cd limiting a square waves bandwidth to 22kHz?

You can't be concerned about one and not the other, unless you are making up your own narrative


Thanks for the kind words. Whilst it may appear defensive, it's really about trying to correct a lot of the misunderstandings about class d. Yes I have a vested interest, but this is about the technical issues. I am just a very "no nonsense" talker.

 

[March Audio]

I am not disbelieving you. But this should be verified with a carefully designed procedure by objective interests that don’t have a skin in the game. The premise of this site.

It would be nice to establish what kind of out of band frequencies are emitted by upstream sources like DACs in frequency and levels by measuring their output in a wide band and make the worst case result part of the test input signals to audio band measurements of Class D amps to ensure that they behave well under such circumstances. This is my takeaway from the OP.

If it turns out that after a few reviews and measurements, this seems to be a non-issue then it might be retired. But the investigation would be worth it.

Well we do this everyday because we use dacs to input test signals into amps.

However certainly we can ask @amirm to look at the RF output of some DACs as he can measure up to 1MHz. He can also use them instead of the AP to play test signals into amps and see if any problems pop up.

 

[March Audio]

Isn't there a difference between what DACs do, which is limit bandwidth with a filter that limits the output spectrum, and a digital amplifier, which has artifacts outside the audible frequency spectrum?

What is the mitigation you mentioned? A low-pass filter on the outputs?

DACs can produce a degree of RF noise. Class D (Btw its not digital) uses a switching frequency around 450kHz in the case of Hypex. The OP is talking about the potential for other signals to intemodulate with this and create issues in the audio band.

I have an RF filter on the audio input to my amps

 

[DonH56]

As said above and many times in many posts on many fora: the "D" in Class D is not digital. We had A, B, and C amplifiers and D was just the next letter in the series. Now we've added E, F, G, H, I, etc. (note not all are used for audio amplifiers). Hand-waving high-level intro here: https://www.audiosciencereview.com/forum/index.php?threads/class-d-amplifiers-101.7355/

It would be nice to see some typical spectral diagrams of in-room (in-home) wideband noise. When I have done it in the past there was a lot more of it than I anticipated, and it has gotten worse since then. In addition to the SMPS in the amp, there are all those other little SMPS' around in little wall warts and other boxes, plus the ones in LED and fluorescent lights (CFL and all those "normal" fast-start lights -- they switch around 60 kHz and are a big problem in most test labs). Then there are dimmers new and old, various motors, etc. That on top of the guy living near a transmitter (or a HAM with one in the house) and so forth. The test loop we drove for a car radio way back in the dark ages intentionally passed near some AM/FM transmission towers in a nearby neighborhood where we measured wicked-high (~0.1-1 V!) signals.

What makes everything work in the audio world is usually a combination of shielded cables that hopefully limit the amplitude at the inputs, and input RF filters that used to be normal good design practice, but all too often seem to have been left out to save money or whatever.

Class D, or any similar methodology, adds the sampling component that can mix such noise back to baseband. Hopefully to be killed by input RFI filters as @March Audio includes. I think it is a real problem, as is the potential for lesser components picking up and amplifying and/or mixing the SMPS switching noise, and appreciate @pma 's measuring it and quantifying it for us. I don't think it should be blown out of proportion, as it is less likely an issue in the real world due to all the other filtering in the system, but I don't think we should neglect it, either.

Maybe I'll go into business making audiophile screen rooms...

 

[SIY]

No there aren't. A tweeter simply does not respond at 460kHz. You can't generate other spurious signals from a position of no response.

I'd add that the current in Pavel's measurements (and mine) of the outputs of several Class D amps is not necessarily even flowing in the voice coils- parasitic capacitance and (in many speakers) Zobels are two alternative current paths. Even with that, the currents are negligible.

Just because you can measure something doesn't mean it's of any consequence.

 

[restorer-john]

No, it will be perfectly fine within the bandwidth limits of the amp

That's what I said. Read my post again. The content has to suit the amplifier or it all goes pear shaped. That's the difference, preamplifiers and power amplifiers were able to cope with anything up to low level RF and do what they should- amplify without adding or subtracting anything. There was no 'rubbish' going on up top to cause trouble in the baseband.

Take the 1978 Kenwood L07c. I used to use one on my bench for high level square waves in between my signal generator and attached device as it could swing another >30V more than my signal gen. Up to 100Khz the rise time barely suffered. It was amazing- like a instrumentation amplifier. It's matching power amps were insane. There's a Perreaux 2150B in my storeroom which runs out to 3MHz at unity gain- that's insanity.

So, why is it OK for Class D proponents to justify chasing ever decreasing levels of THD as their headline point of difference, well below the limits of audibility, and at the same time throw water on people justifying wide bandwidth amplification? Hint, it's not OK.

Bandwidth limiting is bandwidth limiting. What's your point? Why are you not concerned about cd limiting a square waves bandwidth to 22kHz?

Read my post again, it's about the message and the justification sold to audiophiles and the attempt to subvert/divert and re-cast Class D in a different light by channeling the old UK amplifier designers of old where bandwidth limiting was the new black- until it wasn't.

Remember when CD-4 and other multichannel LP sources were devised? Bandwidths were expanded, capacitor coupling was thrown out and we got high speed diffused emitter, RETs, and the ability to do part time LW transmission with our power amps (kidding). Along came DC-Daylight for all things amplification related.

CD strawman nothwithstanding, is it OK to just say for once that Class D amplification does really, really well in the audible bandwidth and not outside it? Again, Class D doesn't have to be your own personal crusade, we already have one of them here (he's back).

 

[RayDunzl]

Class D (Btw its not digital)

Can we call them Binary?

 

[audimus]

Just because you can measure something doesn't mean it's of any consequence.

Ouch, you just blew most of the SINAD table on this site away.

 

[DonH56]

Can we call them Binary?

No, though I suppose you consider the amplitude to have only two states, the time at each state varies with signal. See e.g. https://www.audiosciencereview.com/forum/index.php?threads/class-d-amplifiers-101.7355/ for some pictures (and there is a plethora of better jobs on the 'net).

 

[RayDunzl]

Are they all On/Off? (one rail)

Are some +On/-On ? (two rails)

 

[DonH56]

Not sure what you mean by "all". The simplest scheme switches between V+ and V- (positive and negative power supply rails). Some amps also adjust the voltage rails dynamically (in steps or tracking the input) to save even more power. There is always a little "dead time" in the switches so that they do not short the rails together (i.e. a brief time when both switches are off). Some schemes interleave output stages to get a higher effective switching rate without having to raise the fundamental rate per amp output stage, some include a small class A or class AB amplifier to provide power at very low levels, etc.