On Class D Amplifiers Measurements

[amirm]

Here's an ancient Pioneer SA-708 (late 70s design) tested by me way back in 2009 at 76W@8R (rated 65W@8R) FR swept from 10Hz-20Khz just under clipping.

She's rolling off a touch at each end of the spectrum at 61.6W@8R/10Hz, but pretty good in real terms.

Why did you stop at 20 kHz if higher bandwidth is important?

But no, it doesn't matter because music doesn't have equal energy to 20 kHz let alone beyond.

 

[amirm]

What I find absolutely hilarious, is the same people who espouse "High Res", high sampling rate, DSD, wide bandwidth, ruler flat sources, and line level processing, are prepared to dismiss the same ruler flat requirements in their power amplification.

In this forum, we are more enlightened than just looking at specs. Here is the spectrum of 352 kHz sampled music from 2L label:

There is nothing beyond 35 kHz or so. What is there at 35 kHz is down 100 dB (130 dB - 30 dB). Even at 20 kHz, we are down 65 dB.

This is why I care about 40 kHz or so bandwidth for "high res" music. And it doesn't need to be flat above 20 kHz because we don't hear that anyway!

 

[amirm]

EDIT: As was posted in the thread already, Archimago did just that with this cheap amplifier, and demonstrated that it actually affects the output on the speaker https://archimago.blogspot.com/2019/11/measurement-yeeco-ti-tpa3116-class-d.html. Cheap class d on this level is obviously not high-fidelity

I don't see such a graph there. He used the Sony speaker as a load but then measured the amp at the input to the speaker, not output.

 

[amirm]

Good points. But I find it much more convincing when using real loads, rather than theoretical theorizing which goes over my head... This measurement here by Archimago, for example, tells me quite clearly that the high-frequency noise of this particular (cheap) class d amp actually does have an impact in the audible range around 1 khz:

That is not a 1 kHz signal. 1 kHz is the repetition rate of that square wave. A square wave has infinite bandwidth. See: https://www.audiosciencereview.com/.../composition-of-a-square-wave-important.1921/

That noise which you correctly state is "high frequency," cannot therefore be in audible band. For that, it would need to be low frequency.

Your speaker is a giant low pass filter. As is your ears. So what you see on the instrument is not what your brain is interpreting.

 

[amirm]

There does seem to be a lot of it. Not sure why. I've had three different Class D amps cycle through here in the past month, and the thing that leaves me marveling is how much better they performed in the audio range than nearly any old school amp- which is what counts.

That is it really. Because of their efficiency, they can have marvelous bass response which definitely come across as sounding better. Ditto for the higher power they can produce.

 

[Matias]

In this forum, we are more enlightened than just looking at specs. Here is the spectrum of 352 kHz sampled music from 2L label:

View attachment 40122

There is nothing beyond 35 kHz or so. What is there at 35 kHz is down 100 dB (130 dB - 30 dB). Even at 20 kHz, we are down 65 dB.

This is why I care about 40 kHz or so bandwidth for "high res" music. And it doesn't need to be flat above 20 kHz because we don't hear that anyway!

How about Pink Floyd's classic Dark Side of the Moon, song "Us and Them", extracted from the Blu-ray Audio remaster?

Yep, -70dB from 20kHz on...

https://forum.polkaudio.com/discussion/comment/1706967/#Comment_1706967

Music energy peaks are in the bass area. Exactly where class D are most efficient.

 

[pma]

I have another one, DVD-A Reprise 9362 48435-9, Steely Dan - Everything Must Go, Track 03 Blues Beach

 

[March Audio]

Facts matter? Sounds like religion, not science.

Yes facts matter.

That's why I present them instead of spreading FUD. You know, like demonstrating they don't act like radio transmitters contrary to your assertion above.

 

[restorer-john]

Why did you stop at 20 kHz if higher bandwidth is important?

It was just an example of a full power sweep on a vintage amplifier.

AudioLab is a very old piece of hardware and my log sweep is limited to 20KHz. Spot frequencies I can go higher. Linear sweep can be programmed in 0.5Hz minimum steps from 0.5Hz and up with around 500 steps, so you can zoom in on a troublesome area. As I mentioned before, it merely sees use for confirming basic responses of amplifiers, including their tone, filters etc. It is not an AP, nor do I pretend it is.

Amplifiers that test ruler flat (or close to it) on full power sweeps almost always translate to better sounding and measuring amplifiers in my experience. What in your vast experience tells you otherwise, Amir?

Basically as I see it, there is a concerted effort (vested interests notwithstanding) around here to paint Class D designs in light which serves to highlight only their strengths and attempt to hide or dismiss their weaknesses.

Acknowledging weakness is a good thing. It leads to respect, transparency and improvement.

 

[March Audio]

?

Basically as I see it, there is a concerted effort (vested interests notwithstanding) around here to paint Class D designs in light which serves to highlight only their strengths and attempt to hide or dismiss their weaknesses.

Acknowledging weakness is a good thing. It leads to respect, transparency and improvement.

Problem is John I haven't seen any cogent response as to why the "weaknesses" you refer to are actually a real world problem. Technical point of interest, yes, problem no.

It's not being dismissive, it's being objective.

Also, what evidence do you have, beyond anecdote that your statement is correct;

"Amplifiers that test ruler flat (or close to it) on full power sweeps almost always translate to better sounding and measuring amplifiers in my experience"

However, I do agree that an amp should ideally be flat to 20kHz. The Purifi ones are. Perhaps we should do a test a test with a filtered response the same as a Hypex amp and see what differences people can hear?

Just as a further point on this, in my 3 way dsp active speakers I use hypex nc250mp stereo modules for the woofer and mid. The tweeter however is driven by a Hypex nc100 mono add on module. This runs off the main board and delivers, as the name implies, only 100 watts compared to the main amp which is 250 watts per channel.

This reduced power capability makes not one jot of difference to the sound quality because more power is never needed.

 

[RayDunzl]

But no, it doesn't matter because music doesn't have equal energy to 20 kHz let alone beyond.

Curious:

What's the situation when there's a little HF wave riding on top of a big LF wave that pushes toward the limits?

 

[restorer-john]

Problem is John I haven't seen any cogent response as to why the "weaknesses" you refer to are actually a real world problem.

It's not being dismissive, it's being objective.

Unfortunately, and it's painfully obvious, objectivity and vested interests cannot exist together.

The weaknesses in Class D are well documented. I don't need to fill a thread with them again. Their strengths are well documented and acknowledged. From my perspective, the negatives outweigh the positives by a wide margin. I wish they didn't to be quite honest.

You however, have a different perspective. You sell them. I don't.

 

[March Audio]

Unfortunately, and it's painfully obvious, objectivity and vested interests cannot exist together.

The weaknesses in Class D are well documented. I don't need to fill a thread with them again. Their strengths are well documented and acknowledged. From my perspective, the negatives outweigh the positives by a wide margin. I wish they didn't to be quite honest.

You however, have a different perspective. You sell them. I don't.

That's no good John, you actually have to present justifiable cogent reasons why the weaknesses or negatives are a problem and outweigh the positives. Otherwise you are just saying "I don't Like" which is another phrase for bias.

I'm biased, I sell the things. However my arguments are well rooted in science and anyone is free to debate them and present contradictory evidence based views.

 

[LTig]

Curious:

What's the situation when there's a little HF wave riding on top of a big LF wave that pushes toward the limits?

View attachment 40147

Can't happen here (active speakers ).

I think we can rely on math here. Amps are linear systems and handle each wave of a mix in a linear way (disregarding IMD). I would expect that as long as the power limit is not reached for both waves and their sum the amp should deliver. At least in theory...

 

[RayDunzl]

If it's bandwidth limited and the little HF is above that limit, those little peaks disappear.

That clip is 16/44.1 from CD, and the question refers to what @amirm said.

Sorry if it triggered some Class-D defense mechanism.

He said something like "There is little HF energy in the music" (which is generally true), but... (I ask) the HF is riding at the peaks of the amplifier power from time to time, so what gives?

 

[March Audio]

Can't happen here (active speakers ).

I think we can rely on math here. Amps are linear systems and handle each wave of a mix in a linear way (disregarding IMD). I would expect that as long as the power limit is not reached for both waves and their sum the amp should deliver. At least in theory...

The peak amplitude is the peak amplitude. If you hit the power supply rails and clip theres is nothing you can do. Having the same power at say 20kHz ain't going to save the day in this scenerio. The power requirements will still be low at 20kHz. We are talking the difference between voltage excursion and power.

 

[March Audio]

That clip is 16/44.1 from CD, and the question refers to what @amirm said.

Sorry if it triggered some Class-D defense mechanism.

He said something like "There is little HF energy in the music" (which is generally true), but... (I ask) the HF is riding at the peaks of the amplifier power from time to time, so what gives?

No I misunderstood your point hence deleted that response. See the revised one above.

 

[LTig]

The peak amplitude is the peak amplitude. If you hit the power supply rails and clip theres is nothing you can do. Having the same power at 20kHz ain't going to save the day in this scenerio.

I didn't mean this scenario. Let's assume the maximum voltage is 50V at 20Hz and 5V at 20 kHz. Then a mix of a 45V 20 Hz sinus and a 5V 20 kHz sinus should work: both waves are below their individual limits, as is their sum.

 

[March Audio]

I didn't mean this scenario. Let's assume the maximum voltage is 50V at 20Hz and 5V at 20 kHz. Then a mix of a 45V 20 Hz sinus and a 5V 20 kHz sinus should work: both waves are below their individual limits, as is their sum.

Yes, sorry I was really replying to Ray's post. You are correct. We are interested in the sum of the waveforms.

 

[restorer-john]

@March Audio You were going to run some full rated power 20KHz tests weren't you? Say just 2 or 3 minutes so nothing vaporizes?