Jump to content
IGNORED

Questions For Interview With Bruno Putzeys & Lars Risbo About Purifi Audio


Recommended Posts

1 hour ago, maty said:

A year ago I opened a thread in the diyaudio forums asking about the phase shift in class D technology and its influence on the sound and I still do not know the answer.

 

In class A and AB technology, this shift is less than 6 degrees at 20 kHz, which almost nobody specifies. But in class D it is much higher!  There is a new module, IcePower 1200AS2 that reaches up to 70 degrees!!!


Phase shift in class D amplifiers. How it affects the sound?

 

Specially with acoustic instruments and "natural" voices (without autotune, vade retro satana).


I take the opportunity to raise the issue directly, thank you very much from Spain.

Hi,

This is important. It seems that class D amplifiers despite amplifying, are acting like non-linear filters too.

 

Sounds like a job for MQA Ltd........

 

Sadly, they would require a separate MQA file that is optimised for every class D design.

 

Regards,

Shadders.

Link to comment
9 hours ago, mocenigo said:

 

GaN? I think Bruno Putzeys already said the Eigentakt modules are not GaN.

Hi,

As per article below :

https://semiengineering.com/electric-cars-gain-traction-but-challenges-remain/

 

The use of GaN and Silicon Carbide MOSFET's increases efficiency due to their faster switching speeds. Higher switching speeds can reduce the dead time, so maybe this is a future enhancement ?

 

Regards,

Shadders.

Link to comment
Just now, mocenigo said:

Very interesting article, but I do not see the connection between GaN and electric cars 🙂

 

 

Hi,

Electric cars are investing in GaN and SiC mosfets which offer higher switching speeds - more efficient. The class D modules can do this too - use higher switching speed devices.

 

Regards,

Shadders.

Link to comment
6 minutes ago, mocenigo said:

Yes and no. Class D amps have a phase rotation that is a linear function of the frequency. This is just a delay. So in fact this is often much better than irregular phase rotations as in A and AB amps. Luckily, the rotation of the latter are much more contained.

 

NCore also has much smaller rotations than ICEpower and, it seems, ICEEdge - I expect the same for Eigentakt

 

Hi,

Please see the screenshot of a simple class AB amplifier group delay.

image.thumb.png.067aca58c07cc2f722b34599a954409a.png

Despite the minimal group delay below 50Hz, the graph shows that the group delay is constant. The group delay is the dotted line at the bottom of the graph.

 

Can you provide your examples of class A or class A/B amplifiers with irregular phase rotations ? Thanks.

 

Regards,

Shadders.

Link to comment
30 minutes ago, psjug said:

The latest Qualcomm DDFA can run at 800kHz.  I think that is the frequency that the Sonos Amp uses.  I have one and there is a lot of ripple at 800KHz, so I think the frequency needs to be much higher for switching noise to be tamed by a simple LC output filter.

Hi,

I examined the product brief, and the THD is about 10x higher than the Purifi preliminary module. The problem is we do not know what THD measurement filter was used for the Qualcomm unit, so the Purifi and Qualcomm may be equivalent.

 

Amplifiers were generally measured with a THD filter of 80kHz, for class A/B designs, but for class D designs many reduce this to 22kHz, due to the excessive noise above 20kHz compared to class A/B. It makes the numbers look better.

 

Then, the proponents of class D respond that no one can hear any information above 20kHz, so the 22kHz measurement filter used is valid.

 

Classic case of moving the goal posts.

 

Regards,

Shadders.

Link to comment
8 hours ago, Matias said:

Everyone is so worried about out of band noise and forget that class D has the lowest amount of noise and distortions in band compared to other classes...

Hi,

Can you provide references to support this claim ?, and which classes of amplifier topology you are referring to ?.

 

Thanks and regards,

Shadders.

Link to comment
20 minutes ago, Matias said:

 

I refer to class A and AB with linear power supply, which together should account for something like 90% of the high end market.

 

As for references, check out the THD x power graphs, or 19+20kHz x frequency graphs, that Stereophile, SoundStage and Audio Science Review forums have, and compare nCore against similar prices or even ultra expensive amps.

 

Not that I am a hardcore objectivist but still.

Hi,

The new Purifi module preliminary data states that 1kHz THD for 10watts into 4ohms is 0.00026%. My basic class A/B design has a THD @ 1kHz into 4ohms of 0.000184% (simulated). My design is simple using a single pre-driver for the emitter follower output stage. I can easily enhance the performance by adding another pre-driver, creating a triple EF output stage.

 

Class A/B will always outperform class D, when comparing the best designs of both.

 

Regards,

Shadders.

Link to comment
1 minute ago, psjug said:

I don't see how you can know that either.  And anyway at what point do you say THD+N is good enough and start looking at performance in other areas?

H,

Since class D is an approximation and discontinuous system, whereas class A/B is continuous, and feedback can always be used to reduce further if so desired, while still allowing for the amplifier to be stable. Linearising techniques in class A/B are well known, and the performance in this area allows for minimal feedback in most basic amplifiers, which still out perform class D.

 

Regards,

Shadders.

Link to comment
3 minutes ago, opus101 said:

 

No, the plot on their website shows that number is THD+N - at 10W its clearly noise limited.

 

https://www.purifi-audio.com/eigentakt/

Hi,

Yes, but again, my design is basic, and for the Purifi plot, look at the step up in THD at about 15watts - you do NOT get this with a class A/B design. Why the sudden jump in THD at this point ?

 

Class A/B will always be superior.

 

Regards,

Shadders.

Link to comment
2 minutes ago, opus101 said:

 

Suggest you get acquainted with measuring real-world amps (not sims) and you'll find the answer for yourself. There's a very simple explanation for that which has nothing to do with the amp itself and everything to do with the measurement device (Audio Precision).

Hi,

I stand corrected - i just examined a 1990's DIY design with low THD and the slight rise for that amplifier at similar power output.

Regards,

Shadders.

Link to comment
11 minutes ago, mocenigo said:

I do not see why. With a higher switching speed also the noise between 22 and 80 Khz will be brought down, and also Class D measurements should look better. Maybe it won't, but I do not see how this should be a priori discounted.

 

At this level of 0.0000...% THD further improvements do nothing audible. At some point perhaps D and AB will get indistinguishable and what will count are weight, size, and power consumption. At which point Class D will win.

 

Note I do not care much about the winning topology. As long as I can get detailed, neutral, silent, dynamic amplifiers at reasonable prices, I do not care what's under the hood.

 

 Roberto

 

 

Hi,

Since class A/B is continuous - class D which is discontinuous will always be an approximation.

 

The measurements of class D amplifiers restrict the bandwidth to 22kHz, yet many class A/B amplifiers have a measurement bandwidth of 80kHz. Maybe this is why people state class D is better than the majority of class A/B........

 

Regards,

Shadders.

Link to comment
49 minutes ago, mocenigo said:

Sorry but this does not make sense. Class A/B may be "continuous" but it is STILL an approximation. A different one, that does not require quantisation, but it is still an approximation, because it adds noise nonetheless. Yes, a different profile of noise but it is still noise.

 

As I told above, a higher switching frequency would easily push the noise further up. Also when you write "the measurements of class D amplifiers restrict the bandwidth to 22kHz" this is not 100% clear to me as well. Re: the MEGAschino, for instance, I only read "bandwidth 150Khz" and then a THD+N measurement, not specified as in audio band. Similarly, I often see measurements of Class A or AB amps restricted to the audible band only...

 

 Roberto

 

 

Hi,

Class D is an approximation since it is a pulse width modulation signal. It requires filtering to average the signal between the PWM samples.

 

Class A/B is a continuous function - there are no discontinuities. The noise is either component noise, or derived from the continuous function (polynomial) that adds the harmonics.

 

If you examine the output of class D, then for a 1kHz signal, there is the signal, the harmonics of the signal, and a lot of non-harmonically related noise at a very high level. (neglecting component noise)

 

If you examine the output of class A/B, then for a 1kHz signal, there is the signal, the harmonics of the signal. (neglecting component noise).

 

Class A/B performance will always exceed class D.

 

Regards,

Shadders.

Link to comment
34 minutes ago, Matias said:

 

If you think about it, a pair of transistors per channel, one recreating the positive signal, the other the negative signal, and blending both with crossover distortions, is also not very continuous.

Hi,

If that were the case, then the THD would be very poor. It is not for a class A/B. Through output stage bias, multiple devices, and feedback, the crossover distortion can be reduced to insignificant levels.

 

The signal output of the class A/B amplifier is a continuous signal, whereas for class D, it is an average of the discontinuous signal.

 

Regards,

Shadders.

Link to comment
1 minute ago, barrows said:

Ahhh, simulations do not count!

Hi,

Of course they do. Examine books by Robert Cordell and Douglas Self - all their work is confirmed by simulation versus experimentation.

 

What does seem to be happening on this thread is people using real world measurements of class A/B at 80kHz measurement bandwidth, against the bench testing of a module with a measurement bandwidth of 22kHz - not like for like.

 

Regards,

Shadders.

Link to comment
1 minute ago, barrows said:

You cannot fairly directly compare a sim with a working amplifier on the bench.  Simulations are good tools for development, but are not entirely accurate, as I am sure you realize.  This apples to oranges.

Have you ever done a fair comparison, a listening test, of a well implemented Ncore amplifier vs. a comparable class A/B design (approximately the same power output in the same system)?  If not, I would suggest that you do so, as you appear to have a theoretical bias against class D, you might just change your mind if you listened.

Hi,

I am not disputing that class D sounds good, or people like it more than class A/B.

 

The point of referring to the texts is that the simulation reflected closely the real world design. This is how the authors have been able to determine the distortion mechanisms.

 

In addition, class D measurements are restricted to 22kHz measurement bandwidth, and many amplifiers tested are measured using 80kHz bandwidth.

 

I have never claimed anything about the subjective impressions of class D.

 

Regards,

Shadders.

Link to comment
3 minutes ago, psjug said:

To get nitpicky with you, the unfiltered Class D signal is continuous as well, just switched quickly from rail to rail.  A digital music file is truly discontinuous so this must really bother you.

Hi,

The switching is not perfect, so this is why there is so much noise in the 20kHz+ band for class D. It is a sharp edge at each switch, so approximates a discontinuity, and is designed to be a discontinuity. The "continuous" aspect is a flaw in the implementation, which can only be reduced, never eliminated.

 

Regards,

Shadders.

Link to comment
2 minutes ago, Matias said:

 

Not that we hear or tweeters respond to much more than the 22kHz anyway...

Hi,

I have not referred to any subjective aspect of class D. I simply stated that a class A/B design will always outperform class D.

 

Regards,

Shadders.

Link to comment
4 minutes ago, psjug said:

Just as crossover distortion is a flaw in Class AB.  I don't understand why one is cause to dismiss the technology and the other is not.

Hi,

Crossover distortion is a non-linearity of a continuous function, which can be reduced through feedback, output stage bias, or the use of multiple output devices. The reduction in noise is such that it can be reduced to below the noise floor of a THD measuring device. If you examine the output of a class A/B amplifier, all signals are harmonically related to the input signal.

 

Class D switching is a discontinuity of binary level signal. The noise on the output of a class D amplifier is ALWAYS there - see slide 11 and 12 :

 

www.slideshare.net/shenpei/class-d-and-linear-power-amplifier-testing

 

Here is the simulated output of a basic class A/B with a triple EF stage into 4ohms load, 1kHz signal. Peak power into the 4ohm load is 500watts.

 

image.thumb.png.d63745a801b72f22a1321818469a0729.png

 

 

The peak signal is at 30dB (1kHz), the second harmonic is -135dB down, and third harmonic is -124dB down. The noise profile from 1kHz to 1MHz is non-existent. Compare that to the class D amplifier in slide 11 whose noise is at the -50dB level at least, across the band 100kHz to 700kHz.

 

Whether you can hear it or not, class A/B amplifiers exceed class D in performance. I have made no claim that people can hear above 20kHz.

 

Regards,

Shadders.

Link to comment
27 minutes ago, psjug said:

Just as crossover distortion is a flaw in Class AB.  I don't understand why one is cause to dismiss the technology and the other is not.

Hi,

Not sure why you are indicating that i am dismissing class D technology. I never said that.

 

All i have said is that class A/B will outperform class D, and always will do.

 

I stated "flaw" in that class D implementation can never achieve the perfect implementation, and as such your statement that the switching is continuous, is what i am stating is a flaw of class D implementation.

 

I am following your lead here. I would not state that class D switching is continuous, it is a discontinuous signal, although it does have a finite rise and fall time.

 

Regards,

Shadders.

Link to comment
1 hour ago, crenca said:

Now that THD measurements have been driven so low, what else is Bruno hoping to achieve in this and future designs that relate to the unmeasured quality of the sound of Hi Fi amplification gear?  Not looking for a high level debate about measurements, but I am assuming that not everything related to SQ has been measured and/or is not seen in the standard measurements.  Instead, I am looking for a specific inteligible question for Bruno around this:  How do these new amps sound?  What is your goal around sound?  and the like...

Hi,

Yes, to some extent for class D. If you examine the preliminary data sheet - figure 10, shows that THD of 1% for various supply voltages. It shows that for 65volts power supply voltage, that for 4ohm load, the power output is 475watts. The THD is quite high at 1%, but very low at lower powers.

 

For a class A/B design, using a +/-65volt power supply, triple EF output, three power transistors per positive and negative swing, and two pre drivers in parallel, then for a 4ohm load with peak output power of 500watts, the THD is 0.000083%.

 

So, class D is 1% THD for a 4ohm load, and class A/B is 0.000083% THD for a 4ohm load.

 

This specific class D module does not seem to perform very well for the given power supply voltages. So my question is, why for such a high power supply voltage, does this unit perform so badly at high power levels.

 

To be noted, 475watts peak into 4ohms is 43.59volts peak across the 4ohm load, so if 65volts is the power supply voltage, is there a stepping down of the voltage in the design, or does the design achieve low output impedance at the expense of THD and output voltage ???

 

Regards,

Shadders.

Link to comment
2 minutes ago, Matias said:

Shadders, all amplifiers eventually start to clip. Industry standard defines that a given power spec is when THD reaches 1%. You want more power, see another module, like NC1200 for example. All of them eventually THD graph rises to 1% and above, even your simulated amp would too in real life.

Now back to Bruno's questions...?

Hi,

Class D does not clip. The peak output voltage is ALWAYS being applied to the loudspeaker, even when there is no input signal. The peak output voltage for 475watts is 43.59volts, and the module is using 65volts power supply.

 

That is why i asked the question - why the degraded performance ?

 

Despite the stated low output impedance of 0.05miili-ohms (1kHz, 1A current), there is an issue for higher currents - which indicates that for 475watts peak output power, that the output impedance rises significantly.

 

This is why i asked the questions.

 

Regards,

Shadders.

Link to comment
1 minute ago, psjug said:

I really feel that you are being disingenuous.  Since you looked at the product info I'm sure you saw Figure 2 that shows low THD+N to over 300W into 4 ohms.

Hi,

This is not being disingenuous. There is no clipping in class D as per the clipping mechanism of class A/B (which is the voltage rails, assuming max current can be supplied from power supply). What Figure 10 shows is that for a 4ohm load, that the power output of 475watts generates 1% THD.

 

The table on page 3 for the output impedance, shows a very low output impedance value.

 

My questions are fair - you would not expect a class A/B to begin to clip at 43.59volts for an amplifier that has +/-65volts power supply.

 

What seems to be happening is that there is an output voltage versus current capability restriction. 475watts into 4ohms generates 1% THD, yet 500watts into 2ohms generates approx 0.00045% THD. The 500watts into 2ohms generates more current.

 

Regards,

Shadders.

Link to comment
9 minutes ago, psjug said:

Class D amplifiers clip.  If you'd like I will capture a clipped waveform from a little 20W T amp and send it to you tomorrow.

 

I don't know the specifics but in general amplifiers clip at a voltage below the nominal rail voltage, so there needs to be some allowance there.

Hi,

Yes - if you can post here - would help, thanks. What i am referring to, is that they do not clip as per class A/B. It seems to be output stage voltage/current capability - based on output device, and not that the output voltage is nearing the voltage rail.

 

Regards,

Shadders.

Link to comment
3 minutes ago, psjug said:

Here are specs and measurements of a Neurochrome Modulus 686 Class AB amplifier, running on +/- 36V rails, and bridged.  Why does it start clipping at just over 300W into 4 ohms?

https://www.neurochrome.com/modulus-686/

Hi,

300watts peak into 4ohms is equivalent to 34.64volts peak across 4ohms. The amplifier performs very well, at within 1.36volts of the supply rail.

 

Regards,

Shadders.

Link to comment
5 minutes ago, psjug said:

It's bridged.

Hi,

OK - it is a chip amplifier - but the LM3886 specification states that the clipping voltage is about 4.5volts below 36volts. The specification from the integrators page states onset of clipping at 225watts into 8ohms, but the THD is 0.01% for 240watts into 8ohms. I think the integrators definition of clipping is different to the Texas Instruments definition.

 

Regards,

Shadders.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...