On Class D Amplifiers Measurements

[March Audio]

For those who might be interested how the alias of 900kHz input frequency, described in the 1st post of the thread sounds, here it is in the zip file.

We have already discussed how your test signals are at unrealistically high levels.

 

[restorer-john]

The Purifi FR plot at 5W shows nothing of concern. I'd like to see the same plot at rated power when someone gets the chance.

 

[pma]

We have already discussed how your test signals are at unrealistic levels.

You are saying that and I do understand why. Would you make the test with input signal of mV amplitude and frequency close to 2Fs, which would be 900kHz approx., so that alias product Fin - 2Fs falls into audio band? This would be much more effective, even for the defense of your business interests, than countless repeating how unrealistic are my tests. I am measuring mV amplitudes from AM transmitters in my place. You keeping repeating how unrealistic tests I do does not neglect the fact that NC400 has had an alias issue which Bruno confirmed in a personal email to me ("Fundamentally the cause is that there is a direct path from the input to the comparator...."). Maybe your amp does not suffer from this (which is improbable), so go ahead and bring the proof in measurement. I hate endless word arguing. Too much words, little or no objective proof.

 

[March Audio]

You are saying that and I do understand why. Would you make the test with input signal of mV amplitude and frequency close to 2Fs, which would be 900kHz approx., so that alias product Fin - 2Fs falls into audio band? This would be much more effective, even for the defense of your business interests, than countless repeating how unrealistic are my tests. I am measuring mV amplitudes from AM transmitters in my place. You keeping repeating how unrealistic tests I do does not neglect the fact that NC400 has had an alias issue which Bruno confirmed in a personal email to me ("Fundamentally the cause is that there is a direct path from the input to the comparator...."). Maybe your amp does not suffer from this (which is improbable), so go ahead and bring the proof in measurement. I hate endless word arguing. Too much words, little or no objective proof.

I have already demonstrated you will not be getting these levels on the input of your amp connected to a source such as a DAC or CD player. In fact you provided some useful data that confirms it.

You seem to be under the impression its normal to hang a long wire antenna on the input of an amp.

The real world uses screened cable terminated into the source.

Amir has already tried this experiment and saw nothing.

You arent providing objective proof. I am countering the spreading of FUD

 

[restorer-john]

("Fundamentally the cause is that there is a direct path from the input to the comparator...."). Maybe your amp does not suffer from this (which is improbable), so go ahead and bring the proof in measurement.

The front end of the OEM NC 400 has an RC with an F3 of ~1.5MHz up front of the 5532 comparator doesn't it?

March may have some serious filtering on his buffer board- don't discount that, Pavel.

 

[March Audio]

Its not that John, its just that PAvel is creating pathological conditions that dont exist in the real world.

The proof of this is spectacularly simple.

Connect the amp to a DAC. Turn the amp on and measure the noise floor.

If there is a problem created by external RF beating with the switching signal, or for that matter RF from the DAC, we will see this IM in the noise floor.

Do we? No we dont.

Has anyone reported seeing this problem outside of contrived testing? Not to my knowledge.

Do you agree with the logic of that test Yes or No?

 

[restorer-john]

Do you agree with the logic of that test Yes or No?

No, I don't actually agree, the test is too simplistic, but I do see both points of view.

I think perhaps Pavel is looking at extreme situations and unlikely injections of RF. I also think you are dismissing everything here, without offering actual measured evidence to the contrary. I figured your front end would filter potential issues- you are, after all, an expert in this area- I don't think it would have slipped by your attention.

So, I think we should all go back to the drawing board, chill out a bit, and try to see other people's points of view- me included. I'm off to have a wine and listen to some music.

 

[March Audio]

No, I don't actually agree, the test is too simplistic, but I do see both points of view.

I think perhaps Pavel is looking at extreme situations and unlikely injections of RF. I also think you are dismissing everything here, without offering actual measured evidence to the contrary. I figured your front end would filter potential issues- you are, after all, an expert in this area- I don't think it would have slipped by your attention.

So, I think we should all go back to the drawing board, chill out a bit, and try to see other people's points of view- me included. I'm off to have a wine and listen to some music.

Can you explain how can it possibly be too simplistic? It is EXACTLY how you would use the equipment.

You can create problems for any amplifier if you inject enough RF.

I havent dismissed everything, my first response to Pavel was this can happen.

OK here is a P122 amp with motu dac feeding it. We see nothing in the noise floor.

 

[pma]

I have already demonstrated you will not be getting these levels on the input of your amp connected to a source such as a DAC or CD player. In fact you provided some useful data that confirms it.

You seem to be under the impression its normal to hang a long wire antenna on the input of an amp.

The real world uses screened cable terminated into the source.

Amir has already tried this experiment and saw nothing.

You arent providing objective proof. I am countering the spreading of FUD

OK, there will be no measurements posted by yourself.

I measured the issue HF spreading on the SE signal analog GND wire, depending on cables and configuration used.

Re DAC output, it seems you overlooked this measurement

This is DacMagic Plus output spectrum to 10MHz and it is not the only one DAC with similar HF behavior.

 

[pma]

The front end of the OEM NC 400 has an RC with an F3 of ~1.5MHz up front of the 5532 comparator doesn't it?

March may have some serious filtering on his buffer board- don't discount that, Pavel.

I do not discount that, John, however why not to make a measurement with input signal frequencies close to Fs, 2Fs and 3Fs? My communication with BP was very serious, but from understandable reasons I will not post it in full content. He says they took care to bring the matter to the attention of OEM users, so I assume March Audio is aware of it. So why not to admit? Is this supposed to be a scientific forum, or a forum where manufacturers display their products and use it for marketing purposes?

Calling my info FUD indicates there is not much rational reasoning left.

 

[solderdude]

@ 920kHz -104dBV = 0.00000631V = 6.3uV = -110dB opposite 2V
Consider that goes through an input filter that lowers it even more.
at 3FS this is higher but the input filter (certainly the one MA is using) will have attenuated this.
What is your point ?

 

[restorer-john]

Can you explain how can it possibly be too simplistic? It is EXACTLY how you would use the equipment.

OK, I'm really trying to be the reasonable person here. But it's an upward battle it seems.

I do not discount that, John, however why not to make a measurement with input signal frequencies close to Fs, 2Fs and 3Fs? My communication with BP was very serious, but from understandable reasons I will not post it in full content. He says they took care to bring the matter to the attention of OEM users, so I assume March Audio is aware of it. So why not to admit? Is this supposed to be a scientific forum, or a forum where manufacturers display their products and use it for marketing purposes?

Calling my info FUD indicates there is not much rational reasoning left.

Seriously, I'm not calling your testing FUD, you know that! I applaud your efforts wholeheartedly.

March is a staunch advocate of his own gear. He finds it hard to take criticism of anything from the Putzeys Stable of Thoroughbreds, or merely give credit and consideration to valid questioning. I'm merely giving him, and all of us, a chance to save face, pull back a bit, act like big boys and hang out together in the spirit of advancing high fidelity. That is after all, what ASR is about.

Is that a bridge too far?

 

[solderdude]

Regarding the BW of power amps.

Aren't analog amps with a 20dB/dec input filter and another one (most likely at another freq.) from the amp itself needs to be wider compared to digital filtering to get the same 'flat' response at 20kHz and a similar rise/fall time.

 

[pma]

@ 920kHz -104dBV = 0.00000631V = 6.3uV = -110dB opposite 2V
Consider that goes through an input filter that lowers it even more.
at 3FS this is higher but the input filter (certainly the one MA is using) will have attenuated this.
What is your point ?

My point are the spectral lines between 1MHz and 5MHz reaching to almost -60dBV. Fast edges, spikes.

 

[maty]

Human ear can not listen t

For those who might be interested how the alias of 900kHz input frequency, described in the 1st post of the thread sounds, here it is in the zip file.

It is very audible.

 

[March Audio]

OK, there will be no measurements posted by yourself.

I measured the issue HF spreading on the SE signal analog GND wire, depending on cables and configuration used.

Re DAC output, it seems you overlooked this measurement

View attachment 40472

This is DacMagic Plus output spectrum to 10MHz and it is not the only one DAC with similar HF behavior.

How was it terminated? Plug it into an amp and try again and measure the amp output.

So nothing near 450 or 900 kHz.

 

[solderdude]

Why doesn’t someone hook a class D amp to to typical speaker crossover and measure the output from the crossover to see if the high switching frequency is filtered before it ever makes it to the speaker?
I’m sure the behavior of the crossover could be predicted without actually doing the measurement, but it’d be a nail in the coffin at least.

There is little point in this.
1) There isn't a typical crossover
2) how much HF we will see there depends on the used tweeter as well as RC networks that may or may not be in parallel to the tweeter
3) how much of the resulting HF switching power is dissipated in the tweeter depends on the inductance of the tweeter used. Arguably a planar one will have a different impedance than say your average dome tweeter.

As has been shown the switching frequency output levels are very low and should at least comply to CE or FCC rules which are low.

 

[March Audio]

I do not discount that, John, however why not to make a measurement with input signal frequencies close to Fs, 2Fs and 3Fs? My communication with BP was very serious, but from understandable reasons I will not post it in full content. He says they took care to bring the matter to the attention of OEM users, so I assume March Audio is aware of it. So why not to admit? Is this supposed to be a scientific forum, or a forum where manufacturers display their products and use it for marketing purposes?

Calling my info FUD indicates there is not much rational reasoning left.

They are FUD if you use unrepresentative test signals.

My first response to you was to say it can happen. so quit the crap about my ulterior motives.

 

[pma]

Regarding the BW of power amps.

Aren't analog amps with a 20dB/dec input filter and another one (most likely at another freq.) from the amp itself needs to be wider compared to digital filtering to get the same 'flat' response at 20kHz and a similar rise/fall time.

The input RC filter is to be tuned somewhere above 120kHz/-3dB, otherwise the phase reponse is affected in the audio band.

 

[solderdude]

My point are the spectral lines between 1MHz and 5MHz reaching to almost -60dBV. Fast edges, spikes.

Yes I got that... but my question is how much attenuation of that signal happens by the input filter of the class-D module that receives these frequencies.
How much of this will actually fold back as audible / measurable content.
(I am just asking not trolling or wanting to know better)

Say you connected mentioned DAC and measured the actual analog output 0-20kHz with the mentioned device merely switched on and off what level of change in the audible band do you see ?