On Class D Amplifiers Measurements

[antcollinet]

Maybe. No solid proof ever given.

That is true - but it is difficult to prove a negative. And then there is also no contrary solid proof. You'd think that with all the vested interest amongst audiophiles against Class D (not to mention boutique manufacturers), that someone would by now have set up competent controlled blind testing to prove it damages the sound, such as your suggestion of the AHB2 V Purifi. Or perhaps they have, but failed to prove it.

I'd love to see the experts here band together to do just that.

 

[restorer-john]

Random amplifier I grabbed. Rotel RA-970BX. A well reviewed integrated in the 90s. Not super wide bandwidth, but nicely made and decent performance. Tone controls defeated for this.

Same 5kHz square, 50uS/Div on my trusty old 20MHz CRO for comparison.

Top trace: 150mV square output from the gen.
Bottom trace: 200mV per division output (loaded by a small 2 way Dali spkr) from the Rotel speaker outputs. Voltage.

 

[pma]

Random amplifier I grabbed. Rotel RA-970BX. A well reviewed integrated in the 90s. Not super wide bandwidth, but nicely made and decent performance. Tone controls defeated for this.

Same 5kHz square, 50uS/Div on my trusty old 20MHz CRO for comparison.

Top trace: 150mV square output from the gen.
Bottom trace: 200mV per division output (loaded by a small 2 way Dali spkr) from the Rotel speaker outputs. Voltage.

Sure. There is no question. The only difference in my measurement is that I monitored the current into the speaker, because there were arguments about no flow of the HF current into the speaker, with class D.
In case of a linear amplifier, there is certainly no reason to check the speaker current, if voltage shape is clean.

You know John, it is interesting as the test methods became "tailored" to the new topologies, to show them in the best light. It strongly reminds me to electric car campaign, that completely ignores serious issues like no effective method how to recycle tens and hundreds millions of used electric car batteries, not speaking about impact on a power grid when all cars will be the electric ones.

 

[restorer-john]

Rise time on that 5kHz square from the gen:

 

[restorer-john]

You know John, it is interesting as the test methods became "tailored" to the new topologies, to show them in the best light.

Look if you and I were manufacturing and selling Class D amplifiers, and had invested a ton of money and time into developing them, we'd be no doubt painting them in the best light we could. We'd be eschewing all manner of 'traditional' and 'outdated' tests in order to 'educate' people into what really matters- buying our amplifiers and making us rich.

But that's not us. I have no dog in the fight and neither do you, so all the marketing BS and attempts to re-cast the narrative are seen through as self serving BS. Sure, there is some truth to some of the claims, but endless 'white papers, and 'blog posts' are not education and many are full of errors and omissions.

I don't listen to square waves (often ), but I love gear that can reproduce them with perfect symmetry, no overshoot and no tilt. That's my kind of gear.

I'll grab a few other random amplifiers and post the scope shots.

 

[KMO]

Rise time on that 5kHz square from the gen:

14.6ns, so using a rough approximation of BW = 0.35/RT, that means it's got components up to around 23MHz.

I personally think it's reasonable for amplifiers to not reproduce the 20MHz parts of a signal.

 

[restorer-john]

14.6ns, so using a rough approximation of BW = 0.35/RT, that means it's got components up to around 23MHz.

I personally think it's reasonable for amplifiers to not reproduce the 20MHz parts of a signal.

It's a so-called 60MHz function generator. Not for squares, but it gets to 20MHz with reasonable symmetry and rise/fall.

That's the rise time of the source signal. Presented for the purposes of clarifiying its lack of effect in terms of waveform limitations.

All amplifiers that are correctly designed have filters at the front end at least to deal with such signal rise times, so you can throw whatever you like at them without fear of oscillation/slew rate limiting etc.

I personally think amplifiers should have at least an order of magnitude greater bandwidth than our hearing, don't you? That makes a 200kHz small signal bandwidth a bare minimum- something Benchmark have achieved, along with numerous other manufacturers over the decades. Class D may get there, just not yet.

 

[KMO]

I personally think amplifiers should have at least an order of magnitude greater bandwidth than our hearing, don't you? That makes a 200kHz small signal bandwidth a bare minimum- something Benchmark have achieved, along with numerous other manufacturers over the decades.

What about the transmission chain? Are you similarly concerned about the 20-24kHz bandwidth there? Should we be using 500kHz digital sampling to ensure an order of magnitude greater bandwidth?

Just wondering if this concern is specific to amps, or you do logically extend it to the entire system.

 

[sq225917]

I've always loved the sound of ncores on bass duty, but driving my tweeters there's an etched sound to them that I don't like. I certainly don't 5hink class d amps should get special filter treatment when being measured as that's not how they are used.

 

[SIY]

Sure. There is no question. The only difference in my measurement is that I monitored the current into the speaker, because there were arguments about no flow of the HF current into the speaker, with class D.

Could you describe your test setup for that measurement?

 

[maty]

Hummm...

If class D has no problem with high frequencies (tweeter), why does KEF use Class AB for the tweeter even in the new €6,000 KEF LS60 wireless?

 

[KMO]

Hummm...

If class D has no problem with high frequencies (tweeter), why does KEF use Class AB for the tweeter even in the new €6,000 KEF LS60 wireless?

Their answer in the white paper suggests bandwidth could be a consideration, but I suspect the "cheaper and simpler, and we don't need the heat/power saving" part is the main reason.

Based on the bandwidth and signal statistics a hybrid amplifier class arrangement was selected. The LF section of the system demands high amplifier power to deliver clean bass output. The MF section also requires high continuous power output on certain demanding tracks. Along with the constraint of housing the electronics in a separate enclosure at the bottom of the cabinet, this requires thermal efficiency to be a priority and hence Class D amplifiers were selected for these sections. Class D amplifiers are highly efficient due to their high-speed transistor switching design and their sound quality can be at a level required for high-fidelity reproduction.

The HF section has very different power statistics. Only low continuous output is required but the amplifier must have the ability to deliver sudden large peaks. High bandwidth is also an important consideration for true High-Res support. For this reason a class AB amplifier was selected. Class AB amplifiers have a much lower efficiency than Class D, but the thermal power generated is reasonably low under these signal conditions. Class AB has the added advantage of wide HF bandwidth and of not requiring an output filter.

 

[pma]

The more I measure class D amplifiers, the more I get puzzled. Below is the dependence of NC252MP module shut-down signal frequency on power, into 4ohm load. The roll-off of the plots is due to measuring filter used, the responses are in fact flat. This changes nothing of the phenomenon. Operating power is read at 1kHz and listed below the graph, for the individual plots. The shut-down frequency at individual power can be easily seen as a sudden breakdown of the frequency response. It is for one channel driven. Interesting. Measurement BW is 45kHz, 96kHz sampling frequency. 188.5W was the highest power without shut-down. Yes I know that I measure area above 20kHz. Still, I am interested in such behaviour.

 

[SIY]

So you're getting no protection kicking in at full power within the audio band?

 

[Head_Unit]

20kHz square sent through
1) A250W4R class AB amplifier
2) NC252MP class D
into tweeter, 1V, acoustic spectrum measured by microphone. No comments.

View attachment 196695

View attachment 196696

Some differences can be seen, bu what is the repeatability? If you repeatedly measure each does it look the same every time? Or does it vary almost as much as the differences we see here? Mmmm, and looking at such low levels (OK they're just FFT bins but still) how much is background noise from the room? Sorry if I missed a previous post where all this is explained...

 

[eliash]

The more I measure class D amplifiers, the more I get puzzled. Below is the dependence of NC252MP module shut-down signal frequency on power, into 4ohm load. The roll-off of the plots is due to measuring filter used, the responses are in fact flat. This changes nothing of the phenomenon. Operating power is read at 1kHz and listed below the graph, for the individual plots. The shut-down frequency at individual power can be easily seen as a sudden breakdown of the frequency response. It is for one channel driven. Interesting. Measurement BW is 45kHz, 96kHz sampling frequency. 188.5W was the highest power without shut-down. Yes I know that I measure area above 20kHz. Still, I am interested in such behaviour.

View attachment 265779

Looks like a self-protection function. Maybe they fear (self-)oscillation under certain load situations...

 

[restorer-john]

Looks like a self-protection function. Maybe they fear (self-)oscillation under certain load situations...

More like protection against excess disipation in the output filter capacitors at high power and high frequencies.

 

[eliash]

Good point!

 

[valerianf]

" maty said:
Hummm...
If class D has no problem with high frequencies (tweeter), why does KEF use Class AB for the tweeter even in the new €6,000 KEF LS60 wireless? "

Mostly to avoid the hiss in the tweeter.

 

[Matias]

" maty said:
Hummm...
If class D has no problem with high frequencies (tweeter), why does KEF use Class AB for the tweeter even in the new €6,000 KEF LS60 wireless? "

Mostly to avoid the hiss in the tweeter.

Hiss (noise) is lowest in high end class D amplifiers. Maybe KEF did not want to spend that much for the tweeter and chose a low power class AB instead.