Achieving low distortion in audio amps

[Chris Carlen]

Hi:

My recent experiences purchasing an audio amp afford some points for
reflection on the electronics of such.

I purchased a NAD C325BEE amplifier, basically the cheapest 2-channel
integrated amp on the market ($399). One might surmise then that I
don't accept the idea that amplifiers "sound different" presuming they
are of less than some threshold of distortion considered to be "hi-fi"
(let's say about 0.1%) and not operating at clipping.

Now I discover that the amp runs very hot when doing nothing. I
recorded temperatures on the top cover over the centrally placed
heatsinks of about 47C when the front panel power button was off, and up
to 52C when operating (but not producing any meaningful output power, so
basically idling).

This seriously bugs me. I have a Kyocera A-710 which is twice the
power, and it is barely perceptably warm almost always, but particularly
when idle. Granted, the Kyocera has two massive sinks on either side of
the chassis, with total area probably 2-4 times EACH the area of the
single sink in the NAD used for all four output transistors (two for
each channel, 4 total all on one sink).

I am worried that the amp was not correctly adjusted for bias when it
was made. Another possibility is that it was designed this way.

I posted in rec.audio.high-end:

http://groups.google.com/group/rec....f7d35?lnk=st&q=&rnum=4&hl=en#835d2256f3ef7d35

"I have personally designed a very high power (480W cont., 650W peak)
class-D amp and some single ended motion-control servo amplifiers in the
several hundred Watt level. When designed thoughfully, thermal cycling
is not a detrimental reliability factor. I'm afraid perhaps the
designers erred on the side of greater bias current rather than higher
negative feedback and open-loop BW to servo out crossover distortion.
While this might make some audiophiles (not myself, I think low
distortion can be acheived with low quiescent) happy, it is not ideal
from an idle efficiency standpoint. "

To which "BEAR" responded:

BEGIN QUOTE:
Secondly, you can not achieve the same results with (as you say)
"negative feedback and open-loop BW to servo out crossover distortion"
as you can with a higher bias point (there may be exceptions, of
course). In general, higher levels of negative fb yield undesireable
artifacts in terms of higher order harmonics.

It is certainly a good idea to have a wider than not wide open loop BW,
whenever possible. Of course.

Overall better sound comes when each stage in the amplifier is made as
linear as possible before applying feedback (if you do at all) and when
the details of each stage's operational pros and cons are considered
fully. It's not a simple thing, even though there are "standard ways" of
designing each stage of an amp. The "standard ways" do not always
yield the intended or optimal result.

(Let's ignore for the benefit of this discussion those who would posit
that all amplfiers not clipped, and below some threshold of distortion
are not audibly different)
END QUOTE.

As I implied my limited design experience is more toward motor amps
rather than audio. Yet I have read considerable material on the subject
over the years. All I really have to go on is the fact that my Kyocera
amp is twice the power yet runs extremely cool, and sounds no different
than the NAD amp when not at its limits.

What is the view on the fellows assertion that NFB is bad? Even if the
THD harmonics are not distributed evenly, what difference does it make
if the amp still acheives 0.02% or something like that?

Comments appreciated.

P.S. All this is making me curious to try some experiments sometime,
starting with a butt-simple class-B puch pull emitter follower driving a
speaker, then followed by an op-amp power booster topology with the
op-amp cleaning up the crossover distortion. I wonder if just going
that far would be enough to where the distortion is no longer audible.


--
Good day!

________________________________________
Christopher R. Carlen
Principal Laser&Electronics Technologist
Sandia National Laboratories CA USA
[email protected]
NOTE, delete texts: "RemoveThis" and
"BOGUS" from email address to reply.

 

[BobG]

This argument still is being waged on rec.audio.pro. The golden ears
will tell you their expensive amp/preamp/mic/speaker sounds much better
than the el cheapo musical instrument version that sells at the local
music store. They record all their masterworks at 192Kbits and 24 bits
and listen to them on the expensive playback system once before
dithering it all down to 128Kbit MP3s to listen to with the windows
down with 10 dB music to wind ratio. My experience is that even
harmonics sound 'warm' and 'bluesey', odd harmonics sounds 'harsh' and
'raspy'. Walt Jung spent a decade telling us that the garbage coming
from first generation opamps was 'Transient Intermodulation Distortion'
due to slew rate limiting in the opamp feedback loop. I think you cant
hear a couple of percent THD, but you can hear .01% TIM, depending on
the test setup, etc. There are test recordings for audibility of
certain distortions on rec.audio.pro as I recall....

 

[GregS]

Hi:

My recent experiences purchasing an audio amp afford some points for
reflection on the electronics of such.

I purchased a NAD C325BEE amplifier, basically the cheapest 2-channel
integrated amp on the market ($399). One might surmise then that I
don't accept the idea that amplifiers "sound different" presuming they
are of less than some threshold of distortion considered to be "hi-fi"
(let's say about 0.1%) and not operating at clipping.

Now I discover that the amp runs very hot when doing nothing. I
recorded temperatures on the top cover over the centrally placed
heatsinks of about 47C when the front panel power button was off, and up
to 52C when operating (but not producing any meaningful output power, so
basically idling).

The ideal situation is go to the store where it was bought and feel how hot
the floor model opprates, or ask the store manager. I'll asume it was not
bought under the ideal situation, so you have to test it yourself or send it
back. I would send it back.

greg

 

[Frank Raffaeli]

This argument still is being waged on rec.audio.pro. The golden ears
will tell you their expensive amp/preamp/mic/speaker sounds much better
than the el cheapo musical instrument version that sells at the local
music store. They record all their masterworks at 192Kbits and 24 bits
and listen to them on the expensive playback system once before
dithering it all down to 128Kbit MP3s to listen to with the windows
down with 10 dB music to wind ratio. My experience is that even
harmonics sound 'warm' and 'bluesey', odd harmonics sounds 'harsh' and
'raspy'. Walt Jung spent a decade telling us that the garbage coming
from first generation opamps was 'Transient Intermodulation Distortion'
due to slew rate limiting in the opamp feedback loop. I think you cant
hear a couple of percent THD, but you can hear .01% TIM, depending on
the test setup, etc. There are test recordings for audibility of
certain distortions on rec.audio.pro as I recall....

I've always wondered - and maybe this is already done ... Do audio
amplifiers ever use feed-forward circuits for harmonic cancellation as
some RF amplifiers do? This would reduce or eliminate the need for the
common negative feedback configurations.

Frank

 

[[email protected]]

This argument still is being waged on rec.audio.pro. The golden ears
will tell you their expensive amp/preamp/mic/speaker sounds much better
than the el cheapo musical instrument version that sells at the local
music store. They record all their masterworks at 192Kbits and 24 bits
and listen to them on the expensive playback system once before
dithering it all down to 128Kbit MP3s to listen to with the windows
down with 10 dB music to wind ratio. My experience is that even
harmonics sound 'warm' and 'bluesey', odd harmonics sounds 'harsh' and
'raspy'. Walt Jung spent a decade telling us that the garbage coming
from first generation opamps was 'Transient Intermodulation Distortion'
due to slew rate limiting in the opamp feedback loop. I think you cant
hear a couple of percent THD, but you can hear .01% TIM, depending on
the test setup, etc. There are test recordings for audibility of
certain distortions on rec.audio.pro as I recall....

TIM was Marshal Leach's "cause". I don't recall Walt Jung getting in on
it, but that was a long time ago.

http://users.ece.gatech.edu/~mleach/

I have no problem with negative feedback, but the amplifier needs to
maintain feedback. Amplifiers that get "stuck" when clipping are
running open loop. This is one reason for having power to spare, or
perhaps soft clipping. If the amplifier is slew limited, I consider the
feedback to be broken, though this is arguable. However, having front
end passive filtering and a sufficient slew rate should take slewing
induced distortion out of the picture.

The high heat is due to the level of class A biasing in the class AB
output stage. If you read Randy Sloane's book, "High Power Audio
Amplifier Construction Manual" he is not in favor of heavy class A in
the output stage. I did an Amazon search and noticed Douglas Self and
Marshall Leach have similar books. I've run into Self's papers doing
research. He has an interesting paper discussing distortion in passive
components. Sloane's book is very practical. I haven't read the other
books.

 

[[email protected]]

Chris Carlen wrote:
[experiencing a new stereo amplifier...]

Now I discover that the amp runs very hot when doing nothing.
... I am worried that the amp was not correctly adjusted for bias when it
was made. Another possibility is that it was designed this way.

[and an audiophile newsgroup said feedback bad, more ideal
amplifier stages are good]

What is the view on the fellows assertion that NFB is bad? Even if the
THD harmonics are not distributed evenly, what difference does it make
if the amp still acheives 0.02% or something like that?

An audio amp can change its output impedance (and its speaker drive
characteristic) according to its output current. This leads to
intermodulation,
where a high current at 40 Hz into the woofer causes, through output
impedance
changes, a couple of sidebands on the 2 kHz piccolo solo. Usual
testing
of amplifiers with single-frequency test signals won't always catch
this
(your hearing might, by narrow-band sensing, hear the annoying artifact
more loudly than the bass fundamental, and those 'percent' distortion
figures
don't reflect this kind of effect).

Narrow-band amplifiers and low frequency amplifiers don't have these
problems,
thus the audio situation is arguably 'special'.

So, techniques like feedforward won't generally work; you want a
signal-sensitive
(distorted) feedback to be the 'correction'. And negative feedback,
while a valid technique,
has problems due to the inherent phase lags in power transistors (base
spreading
resistance). Power MOSFETs and vacuum tubes are better than bipolar
junction transistors in this ONE respect.

Lots of extra current in the bias of the output transistors can be
used, to keep
the impedance low (and the heat loss high at zero signal conditions).
Since the
output impedance of the power transistors is low at high signal
currents, keeping the
impedance low at low signal conditions acts to make the impedance more
constant.

Practically, however, the usual faults of power amplifiers aren't in
their modulation
distortions, but in their peak-handling and
how-do-I-fit-this-into-my-living-room
character. If you really care about intermodulation and such, consider
biamplification (separate power amplifiers for each speaker element)
instead
of exotic amplifier redesigns. Me, I want a better remote control.

 

[[email protected]]

I've always wondered - and maybe this is already done ... Do audio
amplifiers ever use feed-forward circuits for harmonic cancellation as
some RF amplifiers do? This would reduce or eliminate the need for the
common negative feedback configurations.

IIRR the Quad 405 used pretty much this approach - rough Class B
supplied the power, a resistor bridge pulled off the error signal, and
a fast class A ampliifer added in a correction.
Seeemed to work pretty well.

 

[[email protected]]

Hi:

My recent experiences purchasing an audio amp afford some points for
reflection on the electronics of such.

I purchased a NAD C325BEE amplifier, basically the cheapest 2-channel
integrated amp on the market ($399). One might surmise then that I
don't accept the idea that amplifiers "sound different" presuming they
are of less than some threshold of distortion considered to be "hi-fi"
(let's say about 0.1%) and not operating at clipping.

Now I discover that the amp runs very hot when doing nothing. I
recorded temperatures on the top cover over the centrally placed
heatsinks of about 47C when the front panel power button was off, and up
to 52C when operating (but not producing any meaningful output power, so
basically idling).

<snip>

It sounds as if a class A/B output stage is set up for a fairly high
current. The ampliifier may need a high class A bias current, or the
pot-twiddler at the factory may have had a carelss moment.

Checking the temperature against an example in a showroom sounds like a
good idea.

 

[Eeyore]

I've run into Self's papers doing
research. He has an interesting paper discussing distortion in passive
components.

Does he indeed.

Do you have a link ?

Graham

 

[Eeyore]

Now I discover that the amp runs very hot when doing nothing. I
recorded temperatures on the top cover over the centrally placed
heatsinks of about 47C when the front panel power button was off, and up
to 52C when operating (but not producing any meaningful output power, so
basically idling).

IEC 60065 allows a max temp rise of 40C on user accessible parts IIRC in 'normal use' which means producing useful output..

I suspect you have a misaligned example.

Graham

 

[Eeyore]

If the amplifier is slew limited, I consider the
feedback to be broken, though this is arguable.

Hardly arguable. It's clear as day !

However, having front
end passive filtering and a sufficient slew rate should take slewing
induced distortion out of the picture.

Most modern amps have slew rates that make the issue almost academic.

I haven't designed one in recent times that didn't have a full power bandwidth
of at least 45kHz.

Graham

 

[Eeyore]

IIRR the Quad 405 used pretty much this approach - rough Class B
supplied the power, a resistor bridge pulled off the error signal, and
a fast class A ampliifer added in a correction.
Seeemed to work pretty well.

Is that really feed-forward ?

Graham

 

[Eeyore]

I've always wondered - and maybe this is already done ... Do audio
amplifiers ever use feed-forward circuits for harmonic cancellation as
some RF amplifiers do? This would reduce or eliminate the need for the
common negative feedback configurations.

Simply not practical to do so.

The amount of correction will vary with load impedance and phase angle and this
is such a collosal variable that the idea's doomed to failure.

Graham

 

[[email protected]]

<snip>

It sounds as if a class A/B output stage is set up for a fairly high
current. The ampliifier may need a high class A bias current, or the
pot-twiddler at the factory may have had a carelss moment.

Checking the temperature against an example in a showroom sounds like a
good idea.

When I googled on C325BEE I found a review in Dutch that talked about
the amplifier as using feed-forward correction, which would presumably
take the same form as the old Quad 405 which I've referred to elsewhere
in this thread. The Quad 405 design patents would have expired by now -
the unit was introduced before 1978

http://quad405.com/lips.pdf

so NAD would be free to copy it.

This approach does call for a fairly healthy Class A amplifier to fill
in the bits that the Class B amplifer can't source, and the quiescent
dissipation would be correspondingly high.

 

[Christopher Carlen]

When I googled on C325BEE I found a review in Dutch that talked about
the amplifier as using feed-forward correction, which would presumably
take the same form as the old Quad 405 which I've referred to elsewhere
in this thread. The Quad 405 design patents would have expired by now -
the unit was introduced before 1978

http://quad405.com/lips.pdf

so NAD would be free to copy it.

This approach does call for a fairly healthy Class A amplifier to fill
in the bits that the Class B amplifer can't source, and the quiescent
dissipation would be correspondingly high.

Too bad they don't specify idle power consumption in the specs.

I'm going to try sending a question to NAD support, posting in another
audio group about the temperature, and seeing how much it would cost to
buy the service manual for this thingy.

Here's an interesting question:

Assuming one had access to the bias adjustment pot of a typical modern
class AB amp, what would happen to the amps performance if it were
simply turned down a notch or two?

I know this question is probably impossible to answer without having
much more intimate knowledge of the circuitry than we have at present.

Oh well.

Thanks for input!

 

[Eeyore]

Too bad they don't specify idle power consumption in the specs.

I'm going to try sending a question to NAD support, posting in another
audio group about the temperature, and seeing how much it would cost to
buy the service manual for this thingy.

Here's an interesting question:

Assuming one had access to the bias adjustment pot of a typical modern
class AB amp, what would happen to the amps performance if it were
simply turned down a notch or two?

I know this question is probably impossible to answer without having
much more intimate knowledge of the circuitry than we have at present.

Typically, crossover distortion will increase.

Graham

 

[Phil Allison]

When I googled on C325BEE I found a review in Dutch that talked about
the amplifier as using feed-forward correction, which would presumably
take the same form as the old Quad 405 which I've referred to elsewhere
in this thread.

** That is FAR too big a presumption to make.

The Slowman is a champion of the foolish presumption.

The Quad 405 design patents would have expired by now -
the unit was introduced before 1978

http://quad405.com/lips.pdf

so NAD would be free to copy it.

This approach does call for a fairly healthy Class A amplifier to fill
in the bits that the Class B amplifer can't source, and the quiescent
dissipation would be correspondingly high.

** That is ALSO far too big a presumption to make.

Here is the original Quad 405 schem:

http://www.geocities.com/quad_esl63/images/schematic/power405.jpg

The class A stage operates at a mere 45mA and hence dissipates only 2.3
watts in the transistor ( TR7) at idle.

(45mA into 8 ohms = 360mV = more than enough to " fill in " the missing
crossover region of the output pair.)

All Quad "Current Dumping" power amps all run with LESS idle
dissipation than the vast majority of comparable class AB designs.



........ Phil

 

[[email protected]]

** That is FAR too big a presumption to make.

So list a couple of alternatives.

The Slowman is a champion of the foolish presumption.

Yep. On account of most of them turn out to be not so foolish after
all.

** That is ALSO far too big a presumption to make.

Here is the original Quad 405 schem:

http://www.geocities.com/quad_esl63/images/schematic/power405.jpg

That is a pretty horrible circuit diagram. The Peter Walker leter in
Wireles Word includes a much clearer schematic.

http://quad405.com/currentdumping.pdf

and enough discussion to make it a lot easier to follow what is going
on.

The class A stage operates at a mere 45mA and hence dissipates only 2.3
watts in the transistor ( TR7) at idle.

That is what the Quad engineers settled on.

(45mA into 8 ohms = 360mV = more than enough to " fill in " the missing
crossover region of the output pair.)

The missing crossover region could be as high as two Vbe drops - closer
to 1.3V - if you want to make other foolish presuppositions ....

All Quad "Current Dumping" power amps all run with LESS idle
dissipation than the vast majority of comparable class AB designs.

But I'm hypothesising about a NAD current dumping amplifier - Quad's
engineers certainly sounded pretty competent when I interviewed for a
job there, and they were competent enough that I didn't get hired.
NAD's engineers have looked pretty competent over the years, but they
have been designing different sort of amplifiers for production in
rather different volumes for rather different markets, with the NAD
guys doing the job some thirty years later.

 

[Phil Allison]

<[email protected]

So list a couple of alternatives.

** Make no presumptions.

Make only good ones.

Yep.

** Thankyou.

The missing crossover region could be as high as two Vbe drops - closer
to 1.3V - if you want to make other foolish presuppositions ....

** Another completely wild presumption.

The Quad 405 et alia actually work as advertised.

Autistic, fucking idiots never learn - do they ?

But I'm hypothesising about a NAD current dumping amplifier

** Another completely wild presumption.

Autistic, fucking idiots never learn - do they ?




........ Phil

 

[Eeyore]

The Quad 405 et alia actually work as advertised.

Funny how everyone said they sounded better when they started fitting faster
Japanese output devices then isn't it ?

Graham