Noisy Purifi amplifiers

[restorer-john]

What is needed is to plug an ADC, look at the noise spectrum to identify potential sources (mains, SMPS,...) then start working on the identified sources.

This is the key takeaway.

Let's look at noise spectrums and discuss what is audible and what is not.

 

[vco1]

Really have to wonder at some who buy from suppliers who cram power supply and amps into the smallest box possible, with terrible routing of the wiring and then complain that things aren't perfect and badmouth all of class D.

To start with, I didn't badmouth "all of class D". I was just asking a question about my new amps.

Also don't think it's fair to blame me for buying from Audiophonics. They seem to be a reputable company. And to be honest, I thought (and still think) it's more safe to buy from them, than from a single person building these in his spare time.

Some class D amps from other manufacturers are even smaller. I don't think the NAD's are that big for example. So size is not everything.

Cable management could probably be better. True. But that seems like an easy thing to fix.

And last but not least, even on ASR there are hardly any tests of this kind of amps. And regular shops usually don't have them. So it's not easy in this day and age to buy an amp like this and know upfront how it performs.

Apart from the fact that the slight hiss is a bummer, I'm still happy with the amps. Although it may not be an Accuphase. ;-)

 

[mocenigo]

In other words, mt42 is calling "hiss" what is actually a "hash" and in any decently designed Class D amp it should be inaudible with most speakers even by babies. This is fine. I also have some "history" with Class D amplifiers – I was involved in the Qualcomm modules developed in Canada (I was responsible for the security analysis of the firmware running on these chips, LOL, not the audio aspects, but, hey, I am a trained mathematician, with a PhD Thesis in Number Theory written under the supervision of one of the giants of the mathematics of the 20th century! hey hey!).

 

[mocenigo]

Dear @sarumbear, the picture I reposted shows the hissing spectrum quite clearly. Now, have a good day!

(Said with the voice of Brandon Roger...)

 

[sarumbear]

In other words, mt42 is calling "hiss" what is actually a "hash" and in any decently designed Class D amp it should be inaudible with most speakers even by babies. This is fine. I also have some "history" with Class D amplifiers – I was involved in the Qualcomm modules developed in Canada (I was responsible for the security analysis of the firmware running on these chips, LOL, not the audio aspects, but, hey, I am a trained mathematician, with a PhD Thesis in Number Theory written under the supervision of one of the giants of the mathematics of the 20th century! hey hey!).

Is “hash” the same as “brown” noise?

 

[mocenigo]

Is “hash” the same as “brown” noise?

Of course ;-) it is also very tasty

 

[mocenigo]

Additional service message to all the cable routing and box size armchair engineers: that's unfortunately not how things work.

It may not be the most important parameter but I had examples of amps with noise where I could reduce the latter by rerouting the internal cabling or using shielded ones in place of simple wires.

Of course proper design to start with is better, but cable routing is not very different from proper PCB design…

 

[BrokenEnglishGuy]

That... should not be possible?

What is your chain upstream?

why not? the purifi its a very veeeery clean amp
burson cv2 xlr adapter->purifi eval1-> kef r7

 

[Sal1950]

I am a trained mathematician, with a PhD Thesis in Number Theory written under the supervision of one of the giants of the mathematics of the 20th century! hey hey!).

Forest Gump? JK

 

[mocenigo]

Forest Gump? JK

No, Gerhard Frey. This said, my tone should have made it clear that this was completely irrelevant to the matter we are discussing. Apologies if the sarcastic tone was not properly conveyed through the text.

 

[boXem]

It may not be the most important parameter but I had examples of amps with noise where I could reduce the latter by rerouting the internal cabling or using shielded ones in place of simple wires.

Of course proper design to start with is better, but cable routing is not very different from proper PCB design…

Well, anybody having seen the inside of my amps knows that I pay a lot of attention to cable routing, for various reasons, including avoiding selling something built with the ... to my customers . The Purifi modules are very robust to external disturbances, so cable routing per se is much less important than with other amps.

 

[mocenigo]

Well, anybody having seen the inside of my amps knows that I pay a lot of attention to cable routing, for various reasons, including avoiding selling something built with the ... to my customers . The Purifi modules are very robust to external disturbances, so cable routing per se is much less important than with other amps.

I had a suspicion about the highlighted part, having had NCore amps in the past and now Purifi. I suspect then that the culprit in the Audiophonics case may be a suboptimal buffer realisation.

 

[cicastol]

I had a suspicion about the highlighted part, having had NCore amps in the past and now Purifi. I suspect then that the culprit in the Audiophonics case may be a suboptimal buffer realisation.

If the problem was the buffer realisation also mine should be noisy, instead is dead silent.........

Also tried on mine different cables routing positions around PSU and modules and in no cases it picked up noise.
IMHO the problem is outside the amp.

 

[Deleted member 27948]

In other words, mt42 is calling "hiss" what is actually a "hash" and in any decently designed Class D amp it should be inaudible with most speakers even by babies. This is fine. I also have some "history" with Class D amplifiers – I was involved in the Qualcomm modules developed in Canada (I was responsible for the security analysis of the firmware running on these chips, LOL, not the audio aspects, but, hey, I am a trained mathematician, with a PhD Thesis in Number Theory written under the supervision of one of the giants of the mathematics of the 20th century! hey hey!).

if you are "a trained mathematician" (although I do not understand how a mathematician can be "trained" ... you shall be born with this curse): start with an open-loop PWM modulator: assume ideal saw, switching, and rails. apply white noise to the comparator's input. Q1: What will be the noise at the output? add (the simplest) integrator (1/s) - as per feedback loop. close the loop. ~, "1/s" -> "s" shall apply to the noise / distortions. Q2: resulting noise spectrum before accounting for the integrator's self-noise? Q3: what high order feedback you can use while continuing to satisfy the self-oscillating conditions - and what spectrum the input white noise will translate into?

;-)

 

[boXem]

if you are "a trained mathematician" (although I do not understand how a mathematician can be "trained" ... you shall be born with this curse): start with an open-loop PWM modulator: assume ideal saw, switching, and rails. apply white noise to the comparator's input. Q1: What will be the noise at the output? add (the simplest) integrator (1/s) - as per feedback loop. close the loop. ~, "1/s" -> "s" shall apply to the noise / distortions. Q2: resulting noise spectrum before accounting for the integrator's self-noise? Q3: what high order feedback you can use while continuing to satisfy the self-oscillating conditions - and what spectrum the input white noise will translate into?

;-)

What if there would be both poles and zeros in the loop?

 

[Deleted member 27948]

What if there would be both poles and zeros in the loop?

nice Q4.

BTW, here are good illustrations of the amp author's way of thinking: https://www.researchgate.net/public..._PWM_Switching_Amplifier_with_Global_Feedback

(PDF) Simple Self-Oscillating Class D Amplifier with Full Output Filter Control

PDF | A stable and load-invariant self-oscillation condition is developed for a class D amplifier employing only one single voltage feedback loop taking... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

 

[boXem]

nice Q4.

BTW, here are good illustrations of the amp author's way of thinking: https://www.researchgate.net/public..._PWM_Switching_Amplifier_with_Global_Feedback

(PDF) Simple Self-Oscillating Class D Amplifier with Full Output Filter Control

PDF | A stable and load-invariant self-oscillation condition is developed for a class D amplifier employing only one single voltage feedback loop taking... | Find, read and cite all the research you need on ResearchGate

www.researchgate.net

Well, I was expecting you to answer this one . Because it's the key to a flat noise spectrum in the audible band.
I know this paper, it's the one that made me fall in love with these amplifiers.

 

[mocenigo]

if you are "a trained mathematician" (although I do not understand how a mathematician can be "trained" ... you shall be born with this curse): start with an open-loop PWM modulator: assume ideal saw, switching, and rails. apply white noise to the comparator's input. Q1: What will be the noise at the output? add (the simplest) integrator (1/s) - as per feedback loop. close the loop. ~, "1/s" -> "s" shall apply to the noise / distortions. Q2: resulting noise spectrum before accounting for the integrator's self-noise? Q3: what high order feedback you can use while continuing to satisfy the self-oscillating conditions - and what spectrum the input white noise will translate into?

;-)

Well, I was born with the curse and I also trained (PhD). However, I am not doubting the types of noise you describe. I only debate their audibility in a good design. And this is not about my ears. It is about the measurements, properly weighted, in the audio range. I know that there are O(1/f) and O(f) contributions, and in fact Putzey's Purifi design seems to have also a O(f^2) contribution. The devil is in the details, in this case in the hidden constants. That's all.

 

[Deleted member 27948]

Well, I was born with the curse and I also trained (PhD). However, I am not doubting the types of noise you describe. I only debate their audibility in a good design. And this is not about my ears. It is about the measurements, properly weighted, in the audio range. I know that there are O(1/f) and O(f) contributions, and in fact Putzey's Purifi design seems to have also a O(f^2) contribution. The devil is in the details, in this case in the hidden constants. That's all.

Well... let's start from the beginning. A valid theory can't contradict any of the reliably established facts, right? I hope you don't believe that the topic starter and many others who have complained about the Hypex noise are part of a worldwide conspiracy, or they are a bunch of schizophrenics with (exactly the same) auditory hallucinations? Thus you need to throw all of the theories of "no noise", "brown noise", 'flat noise", "good design" into a wastebasket, and ask yourself: Why this amp with positive feedback is so flaky and sensitive to any deviations from idealized conditions? (which is a rhetoric question, of course). Why this defect, first reported ~10 years ago, is still not fixed? Why there are still no factory tests? Well... maybe, it's because to fix it you need to spend a lot of R&D on everything around it; the power supply will become even more complicated than the amp itself, full-blown self-diagnostic and monitoring, and there is no commercial sense whatsoever in doing so.

But here there are too many people who are walking illustrations of the Dunning-Kruger effect who stuck on the very first step: acknowledging the problem.

Yet, the Hypex amps are good - because all the rest of the audio is pure snake oil and has nothing in common with science.

 

[boXem]

Still waiting an answer to my so called "nice question"...