Class D amplifier - the real characteristics vs specification

[PENG]

Stating - "you don't seem to have a clue" - is not really 'trying to understand' my view about class D amplifiers. Nevertheless:
The Graph below shows this anomaly I first spoke about, in a different way, which may be more helpful.

This is taken from an Infineon chip at random, but all chips seem to have a similar anomaly. Being a chip, does not automatically make it worse or better than having discrete components (or a combination for higher power) - so the 'chip' term is not important - as you may well know, there are two typical ways to design a class D - one is PWM using a triangle wave and a comparator, and the other is free running without the triangle wave. There may be others, or some with added signal processing. I'll take a wild guess that some really high speed DSPs or MCUs (ARMs etc.) can do the same job simply by inputting an analogue signal into an 24 or 32 bit ADC, then outputting it as a PWM (PWM DAC), using it's internal timer, and some MOSFET Drivers.

So, many (not necessarily all) non-processor based class D configurations are going to have a similar characteristic.
The graph clearly shows that at 6 kHz, the distortion is remarkably higher. You could claim for <0.02% at 1 kHz, 1 to 20 Watts in the amplifier specification, whereas at 6 kHz, the distortion is close to 0.1% - this is a big difference. Seeing as the harmonics of a 6 kHz sine is at and beyond 12 kHz, for many of us, this is not that much of a problem. This is why I wasn't that keen on Class D, even if the specs claimed 0.001% THD at 1 kHz.

But now I have seen the Hypex range, with the help of this thread, my view is a little different - these do still have this increase in distortion at around 6 kHz/7 kHz [with an 8 Ohm load], but it's low enough not to worry about. So I'm going to get one of their products to try out. On the other hand, is this sudden increase in distortion indirectly noticeable or influential? - that's another question - I'll soon find this out for myself.
View attachment 43685

Even if the distortions increase at around 6,7 kHz were much higher it shouldn't matter because music don't have much contents pass 20 kHz anyway. That is assuming those peaks were originated in the ultrasonic spectrum due to the way their filters work.

Below is a good example, though in this case the increase in THD+N were not due to the class D amp's filters, but the DAC reconstruction filters, the principles are the same as far as I could understand (obviously I could be wrong..)

Review and Measurements of Marantz AV8805 AV Processor | Audio Science Review (ASR) Forum

 

[lovinthehd]

Stating - "you don't seem to have a clue" - is not really 'trying to understand' my view about class D amplifiers. Nevertheless:
The Graph below shows this anomaly I first spoke about, in a different way, which may be more helpful.

This is taken from an Infineon chip at random, but all chips seem to have a similar anomaly. Being a chip, does not automatically make it worse or better than having discrete components (or a combination for higher power) - so the 'chip' term is not important - as you may well know, there are two typical ways to design a class D - one is PWM using a triangle wave and a comparator, and the other is free running without the triangle wave. There may be others, or some with added signal processing. I'll take a wild guess that some really high speed DSPs or MCUs (ARMs etc.) can do the same job simply by inputting an analogue signal into an 24 or 32 bit ADC, then outputting it as a PWM (PWM DAC), using it's internal timer, and some MOSFET Drivers.

So, many (not necessarily all) non-processor based class D configurations are going to have a similar characteristic.
The graph clearly shows that at 6 kHz, the distortion is remarkably higher. You could claim for <0.02% at 1 kHz, 1 to 20 Watts in the amplifier specification, whereas at 6 kHz, the distortion is close to 0.1% - this is a big difference. Seeing as the harmonics of a 6 kHz sine is at and beyond 12 kHz, for many of us, this is not that much of a problem. This is why I wasn't that keen on Class D, even if the specs claimed 0.001% THD at 1 kHz.

But now I have seen the Hypex range, with the help of this thread, my view is a little different - these do still have this increase in distortion at around 6 kHz/7 kHz [with an 8 Ohm load], but it's low enough not to worry about. So I'm going to get one of their products to try out. On the other hand, is this sudden increase in distortion indirectly noticeable or influential? - that's another question - I'll soon find this out for myself.
View attachment 43685

You were the one making general statements about class D amps and chips instead of specific statements about particular class D amps that you've looked into with particular chips, thus my wondering what you were on about. So what specific class D amps have you reviewed for this blip of thd? This does not seem a significant amount of distortion to worry about, your speaker would of course have far more. You might look into Purifi instead of Hypex, tho, they've come out a bit better in testing.

 

[Paul DS]

I have had 3 class D receivers throughout the years (all Panasonic). I listened as closely as possible and I could never tell one bit of difference between audio quality in the D amps and standard A/B receivers. I actually prefer class D receivers as they are far lighter than their counterpart and take a hell of a lot less electricity.

 

[Ren Kitchener]

...in this case the increase in THD+N were not due to the class D amp's filters, but the DAC reconstruction filters, the principles are the same as far as I could understand (obviously I could be wrong..)

Thanks PENG - this must be the reason for the increase in distortion in the 5 kHz region. I have a feeling that there are compromises made between the parasitic issues i.e. the MOSFET gate capacitance and the filtering etc., which happens to end up with the least distortion increases at or around 6 kHz. I'll follow you in that I could be wrong.

 

[Ren Kitchener]

You might look into Purifi instead of Hypex, tho, they've come out a bit better in testing.

Yes, the Purifi does have fantastic distortion figures - I'm now tempted - but the price is high - maybe I'll try those later for my Mk2 amplifier.

 

[mazersteven]

In five years we'll all be using class D amps

 

[PENG]

Thanks PENG - this must be the reason for the increase in distortion in the 5 kHz region. I have a feeling that there are compromises made between the parasitic issues i.e. the MOSFET gate capacitance and the filtering etc., which happens to end up with the least distortion increases at or around 6 kHz. I'll follow you in that I could be wrong.

I am only guessing though, if you/we want a definite answer we can try contacting Bruno Putzeys. He used to chime in once in a while but probably got too busy with his new Purifi amps company now to visit this site.

If my guess is right, or at least partially right, then it would be reasonable to say some class D amps may exhibit such characteristics, but many don't. Again, if I guessed right, its not a practical issue unless we listen to ultrasonic contents.

 

[Will Brink]

I know and love Class A and Class AB amplifiers, where I've always considered Class D to be inferior until now, but I still have a slight reservation that I may be wrong about.

You see the Class D chip manufactures claiming HiFi quality, with distortion figures close to 0.005%. This is true for 20 Hz to 3 kHz (making it past the 1 kHz reference), but for some reason, every chip or hybrid chip/discrete solution seems to have a sharp increase in distortion to around 0.02% to 0.05% at around 6 kHz, which sharply drops back at around 7 kHz. It's still very good, but it's not really 0.005%, 20 Hz to 20 kHz.
Some also see a distortion increase at 10 kHz (but the even and odd harmonics are inaudible at that frequency).

I have little idea why this should be (not being familiar enough with the physics behind PWM switching techniques).

So I wonder if every Class D has this strange anomaly, and it's not really ever published in the amplifier manufacturer's data? Or do they filter that band out? - Which would affect the response curve I guess.

Or, technology may have advanced so much, that the likes of AD, ST and Ti have not yet caught up?

I'm in the same boat you are. Everything I'm reading suggests D has finally come into its own and there's some great sounding D based amps out there now. Some say there's a difference still in the sound between A/B and D, but not inferior per se, just different. I have not gotten to listen to any of the D amps getting good reviews yet myself, but keeping an eye on it. Having just purchased an integrated (Anthem STR), I'm not in the market for new amps, but always interetsed to see what's developing.

 

[lovinthehd]

I'd love to see someone show in a proper blind test they can pick out amps by class/type. I sure as heck haven't been able to tell a difference with mine.

 

[PENG]

Yes, the Purifi does have fantastic distortion figures - I'm now tempted - but the price is high - maybe I'll try those later for my Mk2 amplifier.

How high is high?

How about Buckeye amps?

Buckeye Amps: New US based Hypex multichannel amplifier builder, line-up announcement! | Audio Science Review (ASR) Forum

It didn't measure as good as the top tier Hypex or Purifi based amps, but for $649 you get 2X500 W into 4 ohms.

 

[PENG]

I'd love to see someone show in a proper blind test they can pick out amps by class/type. I sure as heck haven't been able to tell a difference with mine.

I would be willing to pay up to $200 (depending on the what's on the DUT list and who's the organizer) participation fees if one such proper blind test is held within driving distance.

 

[Ren Kitchener]

How high is high?

How about Buckeye amps?

Mhh - those are a good price for the power. I'm glad I started this thread because it's completely changed my mind about class D - I'm making an integrated amplifier with some nice front (not back) lit meters (e.g. Glar-Ban) - 3 of them. The enclosure was to house a large toroidal and a bridged class AB (plus heatsinks), restricted to 65 W. However, with the Class D modules, I now have space enough to actually wire it neatly.

 

[Ren Kitchener]

I would be willing to pay up to $200 (depending on the what's on the DUT list and who's the organizer) participation fees if one such proper blind test is held within driving distance.

I'm going to 'stick my neck out' a bit here (and probably get 'hammered' by lovinthehd as a consequence), but this is why class D may sound better in a blind test for the older folk, if you can actually hear the distortion:

At 6 kHz, the pleasant 2nd (octave) harmonic distortion will just about be audible (for the older folk) at 12 kHz. The not-so-nice 3rd harmonics, will be banished to the 18 kHz band. The point is, there may be a preference by older folk, for class D, that is if you can hear the distortion of course. Having said that, is an octave harmonic tone at 12 kHz nice, or just a bit more 'airy'? I'm not 100% sure - but I will find out at some point.

 

[PENG]

Mhh - those are a good price for the power. I'm glad I started this thread because it's completely changed my mind about class D - I'm making an integrated amplifier with some nice front (not back) lit meters (e.g. Glar-Ban) - 3 of them. The enclosure was to house a large toroidal and a bridged class AB (plus heatsinks), restricted to 65 W. However, with the Class D modules, I now have space enough to actually wire it neatly.

I plan on diy one of those Purifi or Hypex based power amp too, but I still have a F5 diy kit idling, waiting for me to start. By the time I am done, I will have a class A amp to compare with a class D amp. I did build Passlab's ACA class A amp but that's only 7 WPC, not very practical but it was fun building it.

 

[mazersteven]

Which class-D kit is the preferred build?

 

[PENG]

I'm going to 'stick my neck out' a bit here (and probably get 'hammered' by lovinthehd as a consequence), but this is why class D may sound better in a blind test for the older folk, if you can actually hear the distortion:

At 6 kHz, the pleasant 2nd (octave) harmonic distortion will just about be audible (for the older folk) at 12 kHz. The not-so-nice 3rd harmonics, will be banished to the 18 kHz band. The point is, there may be a preference by older folk, for class D, that is if you can hear the distortion of course. Having said that, is an octave harmonic tone at 12 kHz nice, or just a bit more 'airy'? I'm not 100% sure - but I will find out at some point.

My point was, that spike on the THD vs F graph is not really there because the X-Axis is the test frequency, not the actual frequency you would be playing when listening to music.

For clarity, when you apply a 6.5 kHz signal, you get that spike, but the magnitude of the spike, looks to be about just over 0.002% is the "Total harmonic distortions" so it included all the harmonics in the ultrasonic band that is likely well pass the audio band of 20,000 Hz. ASR uses 45 kHz or higher, but in Hypex's, that did not specific the test bandwidth, and that's part of the reason I can only take an educated guess.

Another point to note is, that spike was there when tested at 200 W output. There are no such THD+N spikes at 2 and 20 W output level.

It would have been nice if Hypex published the FFT too, like ASR typically do, then you can see the frequency of the actual distortions. As it is now, you are only seeing the total, you cannot tell what that 0.002% is made up of, and as I said, if I guessed right, it would be made up of harmonics at frequencies much higher than 20,000 Hz so you wouldn't have to worry about the 2nd, and 3rd (12, 18 kHz).

Hypex:
DIYclassd.com

Let me use the AV8805 as example again, take a look of the FFT:
You can see the spike at about 35 kHz, that's why you have that rise in THD+N shown in the THD+N vs freq graph I linked in post#21. Now that's an extreme case, the much higher rise was due to the DAC reconstruction filter Marantz picked. For the Hypex amp, as you said a couple times, the "rise" is so small (just a touch more than 0.002%) that even if it is due to the 2nd and 3rd harmonics, I highly doubt golden ears can hear the difference let alone "normal" ears.

 

[PENG]

Which class-D kit is the preferred build?

Ask @ryanosaur , he did his research before pulling the trigger.

 

[Ren Kitchener]

Which class-D kit is the preferred build?

I'm not sure - I hope that someone here can help if it's a kit of parts - but I'm looking at the Purifi and the Hypex ready made modules. I think that the Purifi is the better, but Hypex is the less expensive, but still really good.

 

[mazersteven]

Ask @ryanosaur , he did his research before pulling the trigger.

If I remember correctly he did the Hypex

 

[mazersteven]

I'm not sure - I hope that someone here can help if it's a kit of parts - but I'm looking at the Purifi and the Hypex ready made modules. I think that the Purifi is the better, but Hypex is the less expensive, but still really good.

This is another kit from Parts Express that guy's build

Home

ICEpower 200ASC Class D Audio Amplifier with Power Supply Module 1 x 200WThe 200ASC combines an efficient forward type power supply with a 1x 200W high quality class D amplifier. It includes a DC hanger bus for powering up to two 200AC hanger amplifiers. On its own it is therefore perfect for...

www.parts-express.com