Mr Zarella

You are correct, there is always a trade-off between a "clean" amplifier
and
a "dirty" amplifier. But the "dirty" amplifier will always loose in terms
of reliability, distortion, sound quality etc. All at the expense of good
Engineering practices and a handful of parts.

I have to disagree with you on this point. But I suppose much of it depends
on what you're referring to exactly when you use the term "dirty". I assume
you're talking about distortion and noise, and maybe frequency response.
Much of the distortion characteristics of an audio amplifier are due to the
input stage circuitry, even moreso than the components. You can further
reduce harmonic distortion trhough additional circuitry, however statistical
reliability is sacrificed by doing so. PSRR methods are usually sufficient
in even the simplest IPS designs, and noise is rarely an issue. The biggest
issue to tackle is linearity when compensating for freq response
imperfections. But I don't know of any modern-day class AB amplifiers that
don't perform adequately in this respect.


On a side note, what would you rather be driving, a Yugo or a Mercedes?
Ok,
not the same, but how about a Yugo or a Ford Focus? But if your life
depended on its proper operation all the time, which would you choose?

You're talking about reliability concerns I think, which I haven't disagreed
with anyone on in this thread. The analogy when applied to sound quality
does not apply, unless of course you're talking about driving 5 MPH down the
road, in which case you wouldn't notice a difference. The reason I modify
your analogy is because it completely neglects the measuring device: that
is, the human brain.

Here is another case in point. Old school PPI amplifiers. Some of these
amplifiers are over 15 years old and they are still operating and making
great sound. In fact, these units are still highly sought after. Can you
say the same for an equally aged Pioneer or Sony? I think you are perhaps
using a shifting baseline as to what is "clean" today.

Not sure what reliability has to do with "clean"/"dirty".


Could you also be confusing distortion with noise? They are not the same.
You should do some investigation on these subjects. What you say is a
measurement of distortion is actually called "THD +N" a.k.a. Total
Harmonic
Distortion + Noise.

There are several sources of distortion (but the resulting types of
distortion can most easily be narrowed down to harmonic distortion,
intermodulation distortion, and phase distortion), and I don't believe I've
narrowed it down to any of them. I've generally been referring to
distortion, noise, frequency/phase response as a whole in the discussion of
sound quality.

Then to make matters worse, some audio companies use
filters (read A-weighted) to make there units appear to have low THD+N

Are you referring to the amplifier designs or the testing methodologies?

To
add even more confusion into the mix, this "THD+N" is almost always
computationally measured using FFT methods. There are more than a dozen
FFT windows ( read measuring envelope ) that can be used to make the
measurement. Each will give a return a different value for "THD+N".

The estimation procedure is not vital in this case. A confidence interval
can be derived from any estimation procedure, and if the sample size is
large enough (and given the limited bandwidth in audio) it's quite easy to
limit the bias and variance to negligible levels, especially with proper
windowing when applicable.

There
is one more thing about measuring "THD+N". One must also specify the
power
of the signal being measured. One can measure at 1W, 5W, Full rated
power,
etc. "THD+N" is not constant at all power levels. It is usually higher
at
very low power levels and very high power levels. At very low power
levels, the major component comes from "noise". At very high power
levels,
the major component comes from "THD". Really "clean" amplifiers have "TH
+N" values nearly equal at all power levels under clipping.

Right. Distortion levels vary significantly with different power levels.
Good distortion measurements (like the kind I use ) take this into
account by performing the analysis at different power levels and with
different signals. Harmonic distortion is easy. It's IMD where you start
having to make seemingly arbitrary (though not necessarily incomplete)
decisions.

I can take even the "dirtiest" amplifier and measure extremely low values
of
distortion. On the order of 1/1000 of a %. That is, if I can choose my
own method and power level of distortion measurement.

That's why it's important to use representative techniques and keep them
uniform.


There is also harmonic distortion caused by the sum and difference of two
signals. Let's say a 40 kHz square wave from a power supply and a 1 kHz
sine wave. There will be strong harmonic distortion at 39 kHz and 41 kHz.
There are also artifacts called sub-harmonics of these two signals that
will be present. These are the artifacts that can appear down in the
audio
spectrum. I would also like to point out that music isn't just 1 kHz, but
a plethora of frequencies.

I have told this group many times not to be concerned with "THD+N" as a
figure of merit when buying an amplifier. The human ear doesn't mind even
1-3% THD.

Well, that can depend on a number of other factors, but your point is well
taken.

It is a "Marketing Tool" designed to sway the masses. Go listen
to a high quality pure tube amplifier and a high quality transistorized
amplifier of the same output power specifications. Do this test blindly.
I am absolutely certain that you will always pick the tube amplifier over
the transistorized one as having the better sound. But, that tube
amplifier will have distortion figures SEVERAL HUNDRED TIMES WORSE than
the
transistorized amplifier. Instead, try to buy an amplifier that has Low
"TIM" and low S/N. TIM stands for Transient Intermodulation Distortion.
Good audio distortion analyzers can find this kind of distortion using the
"IMD" settings.

I agree completely. But note that the discussion at hand is picking apart
the difference, rather than the more "pleasing" sound. Several reports have
demonstrated that many listeners actually prefer even order harmonics, which
tend to be especially high in tube designs. However, other reports say
quite the opposite, and asymmetrical clipping which often results in high
even order harmonic content is generally viewed as unfavorable. The most
telling paper I've read on the subject is Farrimond, Perceptual and Motor
Skills,1990,70. He found that in the subjective evaluation of audio
components, it was quite easy to implement paradigms where the users could
tell the difference between two sounds but could not reliably label which
was more or less distorted.


Now, onto the subject of noise. Noise is ever present in the audio world.
It comes from the amplifiers SMPS. It comes from the Automobiles charging
system. It comes from solar wind. It comes from the power transmission
lines. In short, it comes from just about everywhere. In my last post, I
stated several things that make for a "clean" amplifier. Those things are
the items or methodologies used to eliminate or diminish the ingress of
noise into an amplifier. I forgot one them. Decoupling capacitors. Once
again, do not do any or all of those items, and more effective noise will
make it into the audio path. Do all of them right and you will diminish
the ingress of noise into the audio path and have a "clean" amplifier.

I don't know of any modern amplifiers where the inherent noise is a problem.
In fact, the noise resulting from less-than-stellar installation tends to be
much larger than that which cannot be overcome in the amplifier. I said
earlier in the thread however that some amplifiers are more prone to noise
than others, but the significant portion can be eliminated with good
installation techniques. It wasn't too long ago that -70dB to -80dB was
considered good. And IMO, in the car that still is perfectly sufficient.
Now people quibble over the difference between -100dB and -92dB. Again,
it's inconsequential.


I have measured and tested amplifiers from many of the "top" car audio
companies. Some had signal to noise ratios of -50dB, some had signal to
noise ratios greater than -100dB. In other words, the best amplifiers S/N
was 64 times better than the worst. The reason, one amplifier was more
"clean" than anothers. Once again, you can fudge these numbers too. S/N
is different depending upon the position of the gain pot. At max gain,
S/N
to noise will be different than min gain. The difference, a handful of
parts and good Engineering practices. QED

Our disagreement isn't in the electrical aspect of the amplifier. We agree
on just about all of it. Rather, it's the psychophysical end of things.
Pushing distortion down past the noise floor is unnecessary. Trying to
obtain -100dB noise levels is unnecessary. Trying to minimize freq
response dips to below 1dB is unnecessary. This can all be demonstrated in
psychophysical experiments. And these so-called "dirty amplifiers" (even
when, in many cases, the cheaper amplifier is actually "cleaner" than the
more expensive one), do not exhibit deficiencies in any of these categories
that can be detected by us listeners, especially in an electric and acoustic
noise box like a car.