NC252MP (class D) vs. A250W4R (classAB) burst measurements into 4ohm//2.2uF load

[amirm]

I've seen. But you can not relate that power cube data to frequency, can you?

You can but the phase angle will be different.

 

[Deleted member 48726]

You can but the phase angle will be different.

I meant like frequency spectrum dependency.

 

[amirm]

Would you consider an equivalent speaker test jig. Like it could be three, "typical 2 way" "typical 3 way" "Rough Ruth"

For what power level? A few watts? If so, yes, I already have one. For more than that, it becomes a very different challenge.

 

[amirm]

I meant like frequency spectrum dependency.

Again, the load doesn't care. It is just that the nominal reactance of +-30 and +- 60 degrees will not hold at other frequencies. Kind of like a speaker with varying impedance.

 

[Doodski]

Let's not pretend that Class D topology as today has really proven test of time compared to the half of century A/B has.

It's a valid argument because we actually, factually do not know how they will sustain. And with the in-rush of quickly built amps in small enclosures with no cooling I would bet many do not last. But I don't know. So the use of FUD triggers me, as it's used as a bad word to negate whatever it's a response to.

I think it's natural to be Uncertain and Doubtful of the long term reliability of many current Class D's.

Also... The sandwich clamping of the PCB with FETs between the PCB and heatsink is a horrible method and is prone to all sorts of issues usually resulting in fried parts. I've worked that sort of sandwich method gear in quantity as I was a Warranty Depot for Canada for SoundStream and the failure rate was realllly horrible. I repaired them and they send it back blown up again. (Apparently SoundStream got their stuff together and made major improvements to the clamp design on newer models.) The pictures below show the complexity and indicate the time consuming assembly required. Even these heavy duty car amp PCBs where flexing a lot with all the screws that are used to clamp and thermal runaway was the norm. That is what I think will happen to the class D models that we have seen that clamp FETs between the heatsink and the PCB.

 

[Deleted member 48726]

For what power level? A few watts? If so, yes, I already have one. For more than that, it becomes a very different challenge.

Why though? I just assume you have like the cellar filled with coils and capacitors. Am I assuming wrong?

 

[ZolaIII]

It's my birthday at the weekend.

I was about to pull the trigger on a NC252MP based amp to power my Triangle Borea BR03s, because I feel my topping pa3s is struggling particularly on bass heavy tracks played a bit louder, midrange can distort a bit, feel the speakers just just need more power available.

Any issues with this pairing, I'm new and don't understand this thread fully, but leaning towards a hypecore NC***MP.

Advice for the uninformed?

Those speakers are more sensitive than average. Put foam corks in port's to cut the port resonance, buy a subwoofer and high/low pass them at 80~100 Hz instead. If subwoofer is acceptable for you of course and better siled enclosure one.

https://www.audiosciencereview.com/forum/index.php?attachments/triangle-borea-br03-speaker-measurement-driver-response-png.120208/

 

[PeterOo]

You haven't read the thread in it's entirety. I can tell.
Look at a phase / impedance diagram from a speaker and look at your post again..

Thanks for responding. Please bear with me. I think I followed the whole thread, but not being an EE I might have missed supporting measurements for steps that would be obvious (and not need evidence) to an expert.

I see phase/impedance graphs from two ESL speakers for which a 1-point fit was made for an R//C network. That network was then used as an equivalent circuit for the whole speaker, even though it would not fit the phase impedance curve for the whole audio BW, let alone BW up to the switching frequency.

It remains unclear how representative the two ESLs are for even difficult ESR speakers.

When BP said the capacity was not a problem because the ESL was connected through a step up transformer, this was modelled as a single inductor. Was BPs argument not worth investigating? Even in a simulation? Will the amp see a capacity through a transformer?
I only saw a comparison of impedance, but not phase in this post. Where there more comparisons of actual speaker measurements to equivalent circuits that I missed?

For the standard speaker an equivalent circuit was made where woofer resonance (a partly mechanical phenomena) was modelled by non-linearity in an inductor (hysteresis caused by the core material?). I would expect a few measurements from actual speakers validating you can make a standardised equivalent circuit that could be included in amplifier testing as @pma is proposing.

The whole thread looks like a great learning experience, but as a scientist I miss a systematic approach. Again it could be I am looking for proof for reasoning steps that all the experts here on ASR find self-evident.

 

[Geert]

Any issues with this pairing, I'm new and don't understand this thread fully, but leaning towards a hypecore NC***MP.

Advice for the uninformed?

Conclusion

With resistive load or with complex load of most usual speakers with dynamic drivers, the Ncore behaves very well.

@pma, although your experiments are very interesting from a technical perspective, it's clear from different comments they also cause confusion. That's a shame for a forum that aims to give insight. I think it would help if you could add your post with the conclusion about resistive and complex loads to the opening post?

 

[LTig]

It remains unclear how representative the two ESLs are for even difficult ESR speakers.

When BP said the capacity was not a problem because the ESL was connected through a step up transformer, this was modelled as a single inductor. Was BPs argument not worth investigating? Even in a simulation? Will the amp see a capacity through a transformer?

Yes. With an ideal transformer the impedance on the secondary side is seen on the primary side but in value divided by the square of the turn ratio. But transformers are not ideal, they have resistance and other leakage effects which cannot be ignored when we talk about frequencies outside the audible range.

 

[Rob Fens]

Back to the topic, load 4R//2.2uF. Unfortunately, the advice above does not work. I have added a small air-core inductor in series with the 2.2uF capacitor to get this measured impedance:

View attachment 275674
Please note the inductive magnitude rise above 75kHz and phase rising above 45kHz. Also note there is a minimum resistance of about 0.3 ohm.

The NC252MP does not like this load. It is even less stable than it was with 4R//2.2uF load. The quasi-periodic oscillations are moving through audio band above 1kHz.

View attachment 275676
In-band noise (BW22kHz) is 19mV.

Though the class AB amplifiers I have here work normally with this load and even with 4R//2.2uF load, NC252MP cannot be used with it. Below the comparison of THD+N vs. frequency into 4R7 // (2.2uF + Ls) at 7V output voltage, compared to small class AB amplifier PM-AB2.

View attachment 275677

Again Ncore is excellent with pure resistors, good with "usual" speaker load but hardly usable with speakers that have high capacitive impedance, even if in series with small inductance and resistance.

The class AB might be more stable than the class D, but i guess you are talking about a non existing problem.
The 2.2 uF you repeatedly mention is certainly way toooo much. I read some time ago that one should think about nF max.
And if you consider your example as beiing a practical one, this is the first time i ever saw the name of your ESL.
Why not take a Quad, Soundlab or Final, they are representative for the ESL and Hypex will drive them easily.

 

[MaxwellsEq]

Thanks for responding. Please bear with me. I think I followed the whole thread, but not being an EE I might have missed supporting measurements for steps that would be obvious (and not need evidence) to an expert.

I see phase/impedance graphs from two ESL speakers for which a 1-point fit was made for an R//C network. That network was then used as an equivalent circuit for the whole speaker, even though it would not fit the phase impedance curve for the whole audio BW, let alone BW up to the switching frequency.

It remains unclear how representative the two ESLs are for even difficult ESR speakers.

When BP said the capacity was not a problem because the ESL was connected through a step up transformer, this was modelled as a single inductor. Was BPs argument not worth investigating? Even in a simulation? Will the amp see a capacity through a transformer?
I only saw a comparison of impedance, but not phase in this post. Where there more comparisons of actual speaker measurements to equivalent circuits that I missed?

For the standard speaker an equivalent circuit was made where woofer resonance (a partly mechanical phenomena) was modelled by non-linearity in an inductor (hysteresis caused by the core material?). I would expect a few measurements from actual speakers validating you can make a standardised equivalent circuit that could be included in amplifier testing as @pma is proposing.

The whole thread looks like a great learning experience, but as a scientist I miss a systematic approach. Again it could be I am looking for proof for reasoning steps that all the experts here on ASR find self-evident.

You raise valid points. I think one of the things that holds the science back at this point is the "cost" of experiment. Taking a Popper-istic view, what we should do is propose a testable hypothesis, then identify suitable tests aimed at proving that hypothesis wrong, run the tests and see if the hypothesis survives. If it does, the hypothesis is sound until we identify a test that breaks it. I don't think this thread has a singular, specific hypothesis - so testing "it" is difficult. But the domain itself is difficult - building a black box that can soak up hundreds of Volt-Amps is not a small or cheap task. I've been involved in load-testing power generators and the test load is on a truck! Then, we need at least two test loads for either end of the speaker spectrum - electrostatics and low-impedance cone speakers. The obvious answer is to use a real world speaker, but the volume would be appalling and the speaker probably would not survive.

So what do I think some of the hypotheses being discussed in this thread include:

• Hypothesis1: 8 and 4 Ohm resistive tests don't simulate any real world speaker. Test - debate at length what real speakers are like [It's hard to know what the Stopping Rule is for this] Alternative test - seek prior art on speaker models and use that
• Hypothesis2: We need a better test than 8 and 4 Ohm resistive tests. Test - try different loads (derived from 1 above?) and seek corner cases that prove there is value testing with non-resistive loads. This is the direction amirm is adopting
• Hypothesis3: there are capacitive loads that this particular amp and PSU will become unstable with. Test - build a load, test the amp, prove it has problems. This is the approach pma is adopting
• Hypothesis4: the load in 3 above represents a real speaker in 1 above. Test - see 1 above. Alternative test - seek evidence of all known capacitive speakers and build real "average" and "worst case" loads
• Hypothesis5: the experiment in 3 "proves" all Class D has a problem. Test - either test all Class D amps with the load in 3, or agree a load (from 1,2 and 4) and test all Class D amps.

I'm sure there are more. The challenge is - there is an awful lot of expensive testing needed. Economics and consistency is why amps are tested with resistive 4 and 8 Ohm loads

 

[boXem]

Let's not pretend that Class D topology as today has really proven test of time compared to the half of century A/B has.

It's a valid argument because we actually, factually do not know how they will sustain. And with the in-rush of quickly built amps in small enclosures with no cooling I would bet many do not last. But I don't know. So the use of FUD triggers me, as it's used as a bad word to negate whatever it's a response to.

I think it's natural to be Uncertain and Doubtful of the long term reliability of many current Class D's.

Then I will be the one using the automotive argument. Gasoline and diesel injectors are PWM driven by FETs in conditions much harsher than playing music in a living room. Since decades. For decades. So the technology is reliable. The problem is more the competence of the people implementing the technology. Mainly at the end of the implementation chain (module in box slapping).

 

[MaxwellsEq]

Then I will be the one using the automotive argument. Gasoline and diesel injectors are PWM driven by FETs in conditions much harsher than playing music in a living room. Since decades. For decades. So the technology is reliable. The problem is more the competence of the people implementing the technology. Mainly at the end of the implementation chain (module in box slapping).

Absolutely. Hardy PWM is possible. There are lots of industrial use cases for PWM, PPM, PCM which have stood the test of time.

All tech has an Achilles heel!

 

[MaxwellsEq]

Magnetic planar speakers are purely resistive although they tend to not produce bass.

That aside, your implication that we are testing the easy stuff is not accurate. I run power tests at 1 kHz whereas real music has bulk of its energy in bass. If you equalize the energy at bass to 1 kHz, you would go deaf!

The crest factor of 1 kHz is also lower than for real music.

Net, net, we are conservative on one side, and aggressive on another. In balance, we get useful data.

I have started to test with reactive loads now so your complaint is no longer valid anyway.

Sorry @amirm if I come across as critical, that's not my intent. I'd add smileys, but I'm a bit too serious for that. I think what you are doing is brilliant. Most impressive is that you don't rest on your laurels and are continually expanding your tests and genuinely trying to find out what's worth testing and what is not. The reactive load graphs in the last few reviews are really interesting. It will be fascinating to see if patterns emerge from which we can draw distinctions, perhaps even identify why people hear some differences.

Some people posit a gross simplification of what you and others do - namely that all we need is frequency response, noise, a straight gain-line and distortion numbers and graphs to completely describe a piece of audio kit. For line-level analogue and digital devices, this may be sufficient. But when extracting signals from cartridges or moving air in a room, there is more to it than those measurements. And, importantly, you continuously work on identifying what's critical and improving your measurements to capture these aspects. I'm sure you'd love to go back and run your current test sets over devices you looked at 5 years ago!

 

[Deleted member 48726]

The class AB might be more stable than the class D, but i guess you are talking about a non existing problem.
The 2.2 uF you repeatedly mention is certainly way toooo much. I read some time ago that one should think about nF max.
And if you consider your example as beiing a practical one, this is the first time i ever saw the name of your ESL.
Why not take a Quad, Soundlab or Final, they are representative for the ESL and Hypex will drive them easily.

Oh no. You need to backtrack the thread a bit. Equivalent capacitance as seen by an amplifier can be much much higher than 2.2 uF on even an ordinary dynamic driver bookshelf speaker.

 

[Deleted member 48726]

...
I'm sure there are more. The challenge is - there is an awful lot of expensive testing needed. Economics and consistency is why amps are tested with resistive 4 and 8 Ohm loads

Yup, that's getting obvious why. But really, it shouldn't be _that_ expensive. You need to essentially build a crossover with some extra parts.

 

[Deleted member 48726]

Then I will be the one using the automotive argument. Gasoline and diesel injectors are PWM driven by FETs in conditions much harsher than playing music in a living room. Since decades. For decades. So the technology is reliable. The problem is more the competence of the people implementing the technology. Mainly at the end of the implementation chain (module in box slapping).

Well. Technologically PWM is old, yes. Used in this scale for sound amplification is not.

The two things you're trying to compare is by any means of my logic not comparable at all but the actual pulse width modulation tech.

And _of course_ it's up to the competence of implementation. That kind of underlines my point. Do you disagree with any of the things below?

Let's not pretend that Class D topology as today has really proven test of time compared to the half of century A/B has.

It's a valid argument because we actually, factually do not know how they will sustain. And with the in-rush of quickly built amps in small enclosures with no cooling I would bet many do not last. But I don't know. So the use of FUD triggers me, as it's used as a bad word to negate whatever it's a response to.

I think it's natural to be Uncertain and Doubtful of the long term reliability of many current Class D's.

 

[Deleted member 48726]

Thanks for responding. Please bear with me. I think I followed the whole thread, but not being an EE I might have missed supporting measurements for steps that would be obvious (and not need evidence) to an expert.

I see phase/impedance graphs from two ESL speakers for which a 1-point fit was made for an R//C network. That network was then used as an equivalent circuit for the whole speaker, even though it would not fit the phase impedance curve for the whole audio BW, let alone BW up to the switching frequency.

It remains unclear how representative the two ESLs are for even difficult ESR speakers.

When BP said the capacity was not a problem because the ESL was connected through a step up transformer, this was modelled as a single inductor. Was BPs argument not worth investigating? Even in a simulation? Will the amp see a capacity through a transformer?
I only saw a comparison of impedance, but not phase in this post. Where there more comparisons of actual speaker measurements to equivalent circuits that I missed?

For the standard speaker an equivalent circuit was made where woofer resonance (a partly mechanical phenomena) was modelled by non-linearity in an inductor (hysteresis caused by the core material?). I would expect a few measurements from actual speakers validating you can make a standardised equivalent circuit that could be included in amplifier testing as @pma is proposing.

The whole thread looks like a great learning experience, but as a scientist I miss a systematic approach. Again it could be I am looking for proof for reasoning steps that all the experts here on ASR find self-evident.

I was just referring to that you questioned that resistive loads isn't representative to real world measurements. And IR to that, there is several speaker measurements (IRL) used in this thread with impedance / phase that clearly is anywhere else than at 0 ° phase (resistive).

You can look at most serious speaker tests / reviews and find the impedance / phase plot and see that speakers are not resistive loads.

 

[boXem]

Well. Technologically PWM is old, yes. Used in this scale for sound amplification is not.

The two things you're trying to compare is by any means of my logic not comparable at all but the actual pulse width modulation tech.

And _of course_ it's up to the competence of implementation. That kind of underlines my point. Do you disagree with any of the things below?

Both ECUs and amplifiers use PWMed FETs in the audio band, so yes, they are quite comparable. Moreover many Ncore implementations use automotive FETs.

I was replying to

I think it's natural to be Uncertain and Doubtful of the long term reliability of many current Class D's.

When you look outside the very small audio world, it's not so natural.