Re: What's a transient?

On Fri, 28 Apr 2006 05:03:19 GMT, "Ian Iveson"
<IanIveson.home@xxxxxxxxxxxxxxxx> wrote:

Stewart Pinkerton wrote

This distortion is not easily characterised, and its importance is not
easily
quantified, although it is easy to measure as an error at any point in time.
The
problem is what to sensibly compare it with to arrive at a proportion.
Relatively speaking, it is probably safe to say that quicker transient
response
equates to less of this distortion, but that says nothing of its character,
of
what it sounds like.

In general it doesn't have a sound, although the associated poor
frequency response may.

You are wrong, as above. As long as the input is changing, and for some time
after any change, there will be error arising from the system's transient
response.

Not true. So long as the musical transient is slower than the
transient response of the amplifier, there will be no effect whatever
due to the transient response of the amplifier.

No, Stewart. You may say it is insignificant (I don't know about that yet), but
not that it ceases to exist.

OK, I'm an engineer, I work in adequate approximations.
Mathematically, you are correct in that there may be a microsecond or
so of group delay through the amplifier, but it is certain that a
band-limited input signal can pass through an amplifier with say five
times that bandwidth, with less than 0.0001% distortion due to
bandwidth limitations. Stability is of course another matter, I favour
an 'overdamped' response at the expense of bandwidth, others will
allow a little overshoot into difficult loads.

Let's imagine you are chasing a fly with a stick. Your objective is to track the
movement of the fly with the end of the stick. You find by experiment that you
can make your stick move from one stationary position to another in less time
than the fly takes for the same trip. Does that mean you can track the fly
without error? Actually, to track without error requires a transient response of
zero.

Look up the classic servomech equations in any control engineering
text, they'll tell you the theoretical limits of tracking capability.
I've designed many track/hold circuits, and the tracking part was
never the problem.

All changes in the music invoke the transient response of the amp. Since this is
a property of the amp and unrelated to the music, it must be distortion.

Hence, it must be measurable. Hint - you'll have lots of trouble
measuring it with an unconditionally stable amp.

No, you are wrong if you include the signal processing in the path from sound
to
perception. It's the same with sight...if you hold your eyes still you can't
see
stationary things. All of our senses are relative, and need to be if you think
about it.

Allegedly true of dinosaurs, but you are saying that if we stand
perfectly still, the Mona Lisa will disappear. Untrue.

No, I said hold your eyes still. You can do it with practice and discipline (or
opium or other drugs), or you can clamp your eyeballs. Yes, she will dissappear,
along with everything else. You just "see" blackness. Oddly, the blackness
encroaches from the periphery to the centre, like an old tv when swithced off,
and it is hard to get the little white dot to go, or decide when it's gone.

Perhaps you need to seek professional help for your unusual
condition....

Otherwise your efforts are routinely defeated by REM. Note REM in sleep is
associated with visual dreaming. It's hard to have visions without REM.
Actually, once you know about its key role, it's hard to have visions at all.
Knowledge really fucks up your trip.

So does lack of knowledge. I first trained in Psy at Aberdeen Uni
before changing to physics - and I'm not aware that psychologists have
any trouble with dreaming. Of course, if you are seeing *visions*,
perhaps you're more in need of professional help than I thought!

I would say an amp should be no faster than it is beautiful. To the extent
that
it must have some transient response, it should be elegant...simple and
direct,
but like a flute rather than the saxophone or piano seemingly preferred by
some.

Not just quick, but lithe.

Zero meaning to this paragraph, I'm afraid.

None that you have been able to divine, apparently.

None that any rational person could divine...

I was trying to be poetic. Must have been at least half way there...

You certainly have an artistic bent, but your medium of choice appears
to be liquid biological waste.

Actually there is more to what I'm on about than you have seen, I hope. I am
circling around the issue of the quality, rather than the quantity, of
distortion arising from the transient response, and how this is effected by the
order, or complexity of the transfer function.

It isn't. A more complex transfer function simply requires more
attention to detail.

Feedback, for example, tends to shorten transient response. To what extent does
that reduce the distortion arising from the response, and how is the nature of
that distortion changed?

Feedback widens bandwidth, but it can have bad effects on the
transient response if carelessly done. Transient response is generally
measured in settling time, i.e. the time it takes the signal to return
to within a small percentage of the steady-state condition after a
transient input.

But if you don't believe it exists as long as the transient response is as short
as the signal is quick, then I guess I've lost you.

Only because I have vanished ahead of you.....

Stewart, is there a graph of this in your head? I mean, let's put distortion due
to transient response up one axis and "speed" of amp along the other. We draw a
line showing how distortion reduces as speed increases. What do we do as we get
towards the point where you say the amp is as fast as the signal? If at that
point distortion is literally zero, then there must be a discontinuity in our
line, otherwise we would get negative distortion were we to speed up the amp
some more. Surely the best for your case would be an asymtote? In which case it
may be close to zero, but never quite.

Indeed so, it's an asymptote, but you must also consider stability.
Some very fast amps have poor transient resonse into certain loads,
indeed they may burst into uncontrolled oscillation after a transient
hits. Some wags refer to settling time approaching infinity in this
case.

An amplifier should have a transient (and slew rate) response capable
of dealing with a full power signal at the highest frequency of which
the source is capable. Any more is nice but unnecessary.

Why nice?

Because it allows some headroom - like 24-bit recording.

"Capable of dealing with" merely begs the question. It goes nowhere towards
answering it.

Ideally, I'd design for a bandwidth five times that of the highest
frequency of interest

To track any changing input precisely, strictly, with no error, requires a
linear system with no transient response at all. Considering that is impossible,
the real issues are about the relative merits of various compromises.

No, it requires a linear system with a perfect transient response. We
can certainly get very close to that ideal.

As for what this has to do with valves, it's harder to get a quick circuit
when
it must include parallel capacitance and series inductance . As in
mechanical
engineering, the direction of progress is always towards stiffer, lighter
structures. But there again, engineers don't have to make music.


ANd what have valves to do with parallel capacitances and series
inductances? You seem a little confused here.

Most valve circuits have both.

So do SS circuits. So what?

So what indeed? Someone said valve amps tend to have a long transient response.
Is it true? Why?

It's generally (but not necessarily) true, and the output transformer
is the primary cause. Valve amps tend to ring due to limited
stability.

Perhaps I should have said they have greater *effective* parallel capacitance
and series inductance. That is, they have more as a proportion of gain. They are
less able to defeat their own open-loop transient response. I was just trying to
offer some explanation for delay, for a long transient response.

An extended settling time is due to a combination of low bandwidth
and/or poor stability, nothing more. Can be just as band in some SS
amps, the early Naim power amps spring to mind, as they suffered both
problems!
--

Stewart Pinkerton | Music is Art - Audio is Engineering
.

Quantcast