Class D audio amplifier distortion

It is well known that the class D audio amplifiers do have the severe 
nonlinear distortion problems. This happens due to the fundamental 
properties of the PWM process, as well as due to the non-idealities of 
the real amplifier. The perception of those distortions is quite 
different from that of the distortions of an analog amp. Due to the 
sampled nature of the PWM system, the efficiency of the negative 
feedback is very limited. The HiFi quality PWM amp implies operating at 
the switch clock rate about 1 MHz. If the clock is running that high, 
the efficiency of the switch is going to be low, and even lower than 
that of the classic linear amp. Also, the EMC is going to be a hell of a 
problem.

Do you know of someone who did a good analysis of the problem?
Can you suggest a good reading or discussion on this topic?

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Vladimir Vassilevsky wrote:

> It is well known that the class D audio amplifiers do have the severe
> nonlinear distortion problems. This happens due to the fundamental
> properties of the PWM process, as well as due to the non-idealities of
> the real amplifier. The perception of those distortions is quite
> different from that of the distortions of an analog amp. Due to the
> sampled nature of the PWM system, the efficiency of the negative
> feedback is very limited. The HiFi quality PWM amp implies operating at 
> the switch clock rate about 1 MHz. If the clock is running that high,
> the efficiency of the switch is going to be low, and even lower than
> that of the classic linear amp.

This may have been true 10 years ago. In the mean time most audio
amplifiers made for professional usage use PWM or related technologies
to deliver immensely high power in extremly small and light weight
housings (e.g. 2 x 1500W in 2HE housing). The THD figures may be
a bit higher than good hifi amplifiers with much less power but are 
usually not higher than 0.1% which seems to be good enough for
most purposes.

> Also, the EMC is going to be a hell of a problem.

I don't think so since otherwise they would not be allowed to sell their
stuff, at least in Europe (due to the required CE label).

> Do you know of someone who did a good analysis of the problem?

May be you can find something in the white paper sections of
the pro amplifier producers (Crown, QSC, EV, KME, Peavey, 
Camco, Dynacord).

> Can you suggest a good reading or discussion on this topic?

Try the NG rec.audio.pro

bye
Andreas
-- 
Andreas H�nnebeck | email: acmh@gmx.de
----- privat ---- | www  : http://www.huennebeck-online.de
Fax/Anrufbeantworter: 0721/151-284301
GPG-Key: http://www.huennebeck-online.de/public_keys/andreas.asc
PGP-Key: http://www.huennebeck-online.de/public_keys/pgp_andreas.asc

Andreas Huennebeck wrote:
> Vladimir Vassilevsky wrote:
> 
>>It is well known that the class D audio amplifiers do have the severe
>>nonlinear distortion problems. This happens due to the fundamental
>>properties of the PWM process, as well as due to the non-idealities of
>>the real amplifier. The perception of those distortions is quite
>>different from that of the distortions of an analog amp. Due to the
>>sampled nature of the PWM system, the efficiency of the negative
>>feedback is very limited. The HiFi quality PWM amp implies operating at 
>>the switch clock rate about 1 MHz. If the clock is running that high,
>>the efficiency of the switch is going to be low, and even lower than
>>that of the classic linear amp.
> 
> 
> This may have been true 10 years ago. 

The fundamental problems do not change with the time. Today it was 
decided that the boombox quality is overweigthed by the obvious 
size/cost advantages, that's the only difference.

> In the mean time most audio
> amplifiers made for professional usage use PWM or related technologies
> to deliver immensely high power in extremly small and light weight
> housings (e.g. 2 x 1500W in 2HE housing). The THD figures may be
> a bit higher than good hifi amplifiers with much less power but are 
> usually not higher than 0.1% which seems to be good enough for
> most purposes.

The 0.1% THD at 1kHz is a very poor figure indeed. And, besides, the 
distortions of the PWM amp can't be characterized by THD since many 
components of the garbage are not harmonically related to the fundamental.


> 
> 
>>Also, the EMC is going to be a hell of a problem.
> I don't think so since otherwise they would not be allowed to sell their
> stuff, at least in Europe (due to the required CE label).

:))))
Unfortunately, I have to deal with this kind of problem professionally. 
You will not be able to listen to the car radio if the PWM amp is on.

> 
>>Do you know of someone who did a good analysis of the problem?
> 
> May be you can find something in the white paper sections of
> the pro amplifier producers (Crown, QSC, EV, KME, Peavey, 
> Camco, Dynacord).

Been there, seen that. "We are the best of the best of the best" and such.

>>Can you suggest a good reading or discussion on this topic? 
> Try the NG rec.audio.pro
> 

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Vladimir Vassilevsky wrote:
> 
> It is well known that the class D audio amplifiers do have the severe 
> nonlinear distortion problems. This happens due to the fundamental 
> properties of the PWM process, as well as due to the non-idealities of 
> the real amplifier. The perception of those distortions is quite 
> different from that of the distortions of an analog amp. Due to the 
> sampled nature of the PWM system, the efficiency of the negative 
> feedback is very limited. The HiFi quality PWM amp implies operating at 
> the switch clock rate about 1 MHz. If the clock is running that high, 
> the efficiency of the switch is going to be low, and even lower than 
> that of the classic linear amp. Also, the EMC is going to be a hell of a 
> problem.

Though I do not have concrete data/evidence/knowledge on this area, I
can throw my 2 cents in here.

This last sentence really doesn't go with my experience --- a while
ago we installed a car audio system for a friend;  he had bought a
PWM amplifier, and we were so profoundly surprised by how cold it
was, that we had to open it;  it had an array of little 2N29XX
transistors  (I forget the code of that casing --- was it TO95,
perhaps?  Anyway, the little ones, black plastic).

I can tell you, we were throwing Watts at the speakers to the point
where it would bother us of how loud it was, and you could touch the
amplifier and *barely* notice that it was warm.

Another thing --- what are those "fundamental properties" of the
PWM process that you refer to?

As far as I know, the unpleasant distorsion is mostly due to the
Hysteresis (sp?) present in the low-pass filtering coil at the
output;  but that occurs only if it has a magnetic core --- if it
is empty, then I don't see why it would have an excessive amount
of inter-modulation distorsion  (then again, I'm not sure how those
are actually made).


Carlos
--

I concur with Vladimir. 

The argument for class D is power efficiency. 

There are several obvious problems with a switching amplifier. Assume 
that the transistors can not turn on and off instantaniously. The 
distortion created will not be simple harmonic distortion. Harmonic 
distortion is the least audible distortion, but the easiest to measure.
You can assume that complex intermodulation products will be created and 
these are a bigger problem.

You also have essentially no power supply rejection with a switching 
amplifier since the power supply is connected directly to the load. No 
amplifier has a power supply that is a perfect voltage source so 
imperfections in the supply are coupled to the loudspeaker.

Another problem is that you have the switching frequency to get rid of. 
Since a loudspeaker is a low impedance load, the filtering is not 
trivial. The Cs & Ls used for this filtering are also less than ideal.

For high performance systems, one might consider using a switching 
amplifier for sub woofers and woofers and use a linear amplifier for the 
higher frequency components. This gives you many of the power efficiency 
advantages with less audible consequences since the high frequency 
drivers need significantly less power.

-- 
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com
  






Vladimir Vassilevsky <antispam_bogus@hotmail.com> wrote in
news:NLiCh.59307$QU1.20798@newssvr22.news.prodigy.net: 

> 
> 
> Andreas Huennebeck wrote:
>> Vladimir Vassilevsky wrote:
>> 
>>>It is well known that the class D audio amplifiers do have the severe
>>>nonlinear distortion problems. This happens due to the fundamental
>>>properties of the PWM process, as well as due to the non-idealities
>>>of the real amplifier. The perception of those distortions is quite
>>>different from that of the distortions of an analog amp. Due to the
>>>sampled nature of the PWM system, the efficiency of the negative
>>>feedback is very limited. The HiFi quality PWM amp implies operating
>>>at the switch clock rate about 1 MHz. If the clock is running that
>>>high, the efficiency of the switch is going to be low, and even lower
>>>than that of the classic linear amp.
>> 
>> 
>> This may have been true 10 years ago. 
> 
> The fundamental problems do not change with the time. Today it was 
> decided that the boombox quality is overweigthed by the obvious 
> size/cost advantages, that's the only difference.
> 
>> In the mean time most audio
>> amplifiers made for professional usage use PWM or related
>> technologies to deliver immensely high power in extremly small and
>> light weight housings (e.g. 2 x 1500W in 2HE housing). The THD
>> figures may be a bit higher than good hifi amplifiers with much less
>> power but are usually not higher than 0.1% which seems to be good
>> enough for most purposes.
> 
> The 0.1% THD at 1kHz is a very poor figure indeed. And, besides, the 
> distortions of the PWM amp can't be characterized by THD since many 
> components of the garbage are not harmonically related to the
> fundamental. 
> 
> 
>> 
>> 
>>>Also, the EMC is going to be a hell of a problem.
>> I don't think so since otherwise they would not be allowed to sell
>> their stuff, at least in Europe (due to the required CE label).
> 
>:))))
> Unfortunately, I have to deal with this kind of problem
> professionally. You will not be able to listen to the car radio if the
> PWM amp is on. 
> 
>> 
>>>Do you know of someone who did a good analysis of the problem?
>> 
>> May be you can find something in the white paper sections of
>> the pro amplifier producers (Crown, QSC, EV, KME, Peavey, 
>> Camco, Dynacord).
> 
> Been there, seen that. "We are the best of the best of the best" and
> such. 
> 
>>>Can you suggest a good reading or discussion on this topic? 
>> Try the NG rec.audio.pro
>> 
> 
> Vladimir Vassilevsky
> 
> DSP and Mixed Signal Design Consultant
> 
> http://www.abvolt.com
> 

Carlos Moreno wrote:

>> It is well known that the class D audio amplifiers do have the severe 
>> nonlinear distortion problems.

[...]

> I can tell you, we were throwing Watts at the speakers to the point
> where it would bother us of how loud it was, and you could touch the
> amplifier and *barely* notice that it was warm.

I have the results of the real measurements on the linear and class D 
amps, and it is not all as good as claimed by the D class supporters.

> Another thing --- what are those "fundamental properties" of the
> PWM process that you refer to?

Basically, if the audio signal is mixed with the triangular wave on the 
comparator, it will produce all kinds of intermod products. Some of 
those products are falling into the audible band. This effect can be 
cured by the sophisticated DSP processing, however it is fairly 
demanding computational task, thus it is cost prohibitive.

> As far as I know, the unpleasant distorsion is mostly due to the
> Hysteresis (sp?) present in the low-pass filtering coil at the
> output;

The nonlinearity of the L and C at the output is only one of the many 
problems of the class D. The output devices are not identical, the 
impedances are not identical, the switch timing is not identical and 
varying with the voltage/temperature.


>  but that occurs only if it has a magnetic core --- if it
> is empty, then I don't see why it would have an excessive amount
> of inter-modulation distorsion  (then again, I'm not sure how those
> are actually made).

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Al Clark wrote:

[...]

> For high performance systems, one might consider using a switching 
> amplifier for sub woofers and woofers and use a linear amplifier for the 
> higher frequency components. This gives you many of the power efficiency 
> advantages with less audible consequences since the high frequency 
> drivers need significantly less power.

The elegant concept is sharing the common load between the switching amp 
and a small linear amp. The purpose of the linear amp is only to 
compensate the distortions of the switcher, hence the power of the 
linear amp can be quite small. So the total efficiency is determined by 
the switching amp. However this solution is cumbersome and the cost is 
too high for the consumer market.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

Vladimir Vassilevsky wrote:
> 
> It is well known that the class D audio amplifiers do have the severe 
> nonlinear distortion problems. This happens due to the fundamental 
> properties of the PWM process, as well as due to the non-idealities of 
> the real amplifier. The perception of those distortions is quite 
> different from that of the distortions of an analog amp. Due to the 
> sampled nature of the PWM system, the efficiency of the negative 
> feedback is very limited. The HiFi quality PWM amp implies operating at 
> the switch clock rate about 1 MHz. If the clock is running that high, 
> the efficiency of the switch is going to be low, and even lower than 
> that of the classic linear amp. Also, the EMC is going to be a hell of a 
> problem.
> 
> Do you know of someone who did a good analysis of the problem?
> Can you suggest a good reading or discussion on this topic?

The first class D amp I saw was the audio output of the world's first 
all transistor (apart from the tube) colour TV - the Thorn 2000. That 
was about 1967. It used only a few transistors, and was obviously 
designed for low cost in a TV, rather than hi-fi. However, a pair of 
those simple amps, with quite a modest switching frequency, sounded 
better than many quite expensive stereo amps sold in the Hi-Fi market. 
Its seems when done with care, pretty basic class-D can sound very good.

Steve

Steve Underwood wrote:


>> It is well known that the class D audio amplifiers do have the severe 
>> nonlinear distortion problems.

> The first class D amp I saw was the audio output of the world's first 
> all transistor (apart from the tube) colour TV - the Thorn 2000. That 
> was about 1967.

The first listeners of the Edison's phonograph were quite exited about 
its sound also :)

VLV

On Mon, 19 Feb 2007 15:12:45 +0000, Vladimir Vassilevsky wrote:

> This happens due to the fundamental 
> properties of the PWM process, as well as due to the non-idealities of 
> the real amplifier. The perception of those distortions is quite 
> different from that of the distortions of an analog amp. Due to the 
> sampled nature of the PWM system, the efficiency of the negative 
> feedback is very limited. The HiFi quality PWM amp implies operating at 
> the switch clock rate about 1 MHz.

I don't belive that most of the currently available class D amplifiers are
strictly PWM at all.  My understanding was that at least some of them were
essentially sigma-delta modulators: could easily be wrong, though -- it's
not my field.

I can say that I've heard demos of experimental versions of both B&O's
ICEpower amps and something by Sony the name of which escapes me, and they
seemed quite fine.  Certainly HiFi quality.

I hadn't heard that any of the currently available class D amps were
switching at much more than 500kHz, but maybe they are, now.

-- 
Andrew