"Wally" wrote in message
...
Peter Scott wrote:


Issa testosterone-fueled disco on wheels, innit?

Definitely. Only got to look at the mags. All pink puppies and
rear ends.



In a similar way, a heavily clipped output stage can be said to be
producing
loads of HF harmonics, but it isn't really - there's no signal generator
producing the notes any more than there's a frequency doubler in a class-A
valve stage. It just so happens that a square(ish) wave can be modelled by
taking a fundamental and adding loads of harmonics, presumably again of
specific relative amplitude and phase. The fact that some of these
harmonics
might be well beyond the frequency response of the amplifier is neither
here
nor there - they aren't passing through the amp to begin with, they're an
artefact of overdriving the output devices. However, they *are* present at
the outputs, and are getting sent to the speakers - if you put a suitable
high pass filter in, you'd have loads of HF signal at high (amplifier's
maximum?) amplitude.

I'm sure someone who knows what they're talking about will chime in
presently...


What you've said is interesting and makes sense. Clipping after the
amplifying stage does not require the amplifier to produce the
frequencies. Yes I'm sure that's right. I was in a mindset of
Fourier synthesis.



I guess it's easier to drive a wide cone with a wide, relatively
short-throw
coil than a narrow cone with a narrow, long-throw coil. Easier to make
rigid
cones, IOW, than to make a narrow long-throw motor assembly that can
deliver
similar undistorted output (stuff to do with the coil tending to move
outwith the magnetic field, and the limitations of the cone surround - if
the latter is too flexible to cope with the increased movement, then I
guess
there's a risk of 'sideways slop').

I suspect that there is a real physical effect here, not just throw
problems.
A larger area allows the cone to couple with a greater mass and volume
of air, making it more efficient in energy transfer.

Some of the long-throw sub units are real beasts. They weigh around 50kg
and health and safety means two blokes have to deliver them! The outputs
are awesome. I was in a local ICE shop and chatting, as you do. They said
that someone had just traded in a four unit 15" sub for a six-unit one. Did
I want to listen?

He'd had it in a Fiat Brava. Took out the rear seats. Anyway they
started this thing up. The shop is large as its a converted pub. Everyone
in the shop just froze in their tracks. The air sort of condensed and
moving around became difficult. All the panels and bits for sale started
to rattle. When it went off again people took quite a few seconds to
start thinking and moving again. Bizarre, but an extraordinary
sensation! Treacly air.


Still waiting for the design book to arrive so I can crack out the
calculator.

Don't know what calculator you have, but I've tried a program called
BassBox, and thought it was pretty good. Takes plenty of specs for the
drivers (all that Thiele-Small stuff), lets you define the enclosure as
one
of several types (sealed, reflex, various resonant pipes), does
multi-driver
setups, including isobaric, and also has a bunch of stuff for defining the
room's characteristics.

Thanks for that. I'll chase up BassBox. Sounds just what I need.

The book I'm waiting for is
"Designing, Building and Testing Your Own Speaker System:
With Projects" David B. Weems; Paperback

Peter Scott

In article , Peter Scott
wrote:

"Wally" wrote in message
...

[snip]

In a similar way, a heavily clipped output stage can be said to be
producing loads of HF harmonics, but it isn't really

[snip]

What you've said is interesting and makes sense.

I'm afraid it didn't make much sense to me given the parts I snipped this
time, nor did I agree with what Wally wrote. :-)

Clipping after the amplifying stage does not require the amplifier to
produce the frequencies.

Alas, the output stage of an amplifier *is* part of the amplifier, not
'after' it. Nor is it invariable that clipping only occurs at the output
stage. Depends on the design, etc. Also, the components created by
clipping *won't* always be at higher frequencies than those in the input.
Clipping 'musical' waveforms can lead to unwanted components at all sorts
of frequencies, many of which are lower than, or comparable with, those in
the input waveform.

Hence you would need an 'adaptive' filter that always knew what components
to pass and what to reject, right across the audio band. The usual
arrangement for this is - an amplifier that is not clipping in the first
place. :-)

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html
Barbirolli Soc. http://www.st-and.demon.co.uk/JBSoc/JBSoc.html

"Jim Lesurf" wrote in message
...
In article , Peter Scott
wrote:

I'm afraid it didn't make much sense to me given the parts I snipped this
time, nor did I agree with what Wally wrote. :-)

Clipping after the amplifying stage does not require the amplifier to
produce the frequencies.

Alas, the output stage of an amplifier *is* part of the amplifier, not
'after' it. Nor is it invariable that clipping only occurs at the output
stage. Depends on the design, etc. Also, the components created by
clipping *won't* always be at higher frequencies than those in the input.
Clipping 'musical' waveforms can lead to unwanted components at all sorts
of frequencies, many of which are lower than, or comparable with, those in
the input waveform.

Yes the new waveform might well have new low frequency components
needed to synthesise it. But I come back the point about the hf. Surely
if the clipping is caused by effects in the output transistors, that drive
them
into non-linearity, it is in effect after the amp stages? Otherwise how
could
components be generated that lie outside the range of the amp. Or is it
that fourier doesn't come into it. The fact that you need this range of
frequencies to synthesise the wave doesn't mean that the amp has to produce
them. All that is needed is a slew rate that is at least equal to the
steepest
slope on the wave. As you can tell I'm struggling a bit with this but I do
want to understand it. I had exactly the same problem with discrete
cosine transformation when learning about jpeg and mp3.

Hence you would need an 'adaptive' filter that always knew what components
to pass and what to reject, right across the audio band. The usual
arrangement for this is - an amplifier that is not clipping in the first
place. :-)

Definitely. Hence I am now buying power amps with oodles of power. Max
Headroom lives!

Peter Scott

In message , Peter Scott
writes
Otherwise how
could
components be generated that lie outside the range of the amp.

It is extremely likely that the overall NFB loop includes the output
stage of the amp. Therefore clipping in the output stage is just about
the worst place for it to occur. The limiting of the NFB signal during
clipping will force the earlier gain stages to handle gross overload
signals. In many amps, this internal overload has the additional effect
of lengthening the time for the amp to recover once the input overdrive
signal has gone away.
--
Chris Morriss

"Chris Morriss" wrote in message
...

It is extremely likely that the overall NFB loop includes the output
stage of the amp. Therefore clipping in the output stage is just about
the worst place for it to occur. The limiting of the NFB signal during
clipping will force the earlier gain stages to handle gross overload
signals. In many amps, this internal overload has the additional effect
of lengthening the time for the amp to recover once the input overdrive
signal has gone away.
--
Chris Morriss

So back to the original question. Is it the case that to produce a waveform
that includes hf, eg supersonic, components an amplifier must be capable of
producing those frequencies? Or does the notion of fourier synthesis not
apply here? Or is it that even rapidly rising wave edges are not rising
rapidly enough to need such hf?

Peter Scott

In message , Peter Scott
writes

"Chris Morriss" wrote in message
...

It is extremely likely that the overall NFB loop includes the output
stage of the amp. Therefore clipping in the output stage is just about
the worst place for it to occur. The limiting of the NFB signal during
clipping will force the earlier gain stages to handle gross overload
signals. In many amps, this internal overload has the additional effect
of lengthening the time for the amp to recover once the input overdrive
signal has gone away.
--
Chris Morriss

So back to the original question. Is it the case that to produce a waveform
that includes hf, eg supersonic, components an amplifier must be capable of
producing those frequencies? Or does the notion of fourier synthesis not
apply here? Or is it that even rapidly rising wave edges are not rising
rapidly enough to need such hf?

Peter Scott



I'll do a quick SPICE sim of the harmonics produced by clipping a 1kHz
sine wave at say 90% of its full level and look at the spectrum. I'll
let you know.
--
Chris Morriss

In message , Peter Scott
writes

"Chris Morriss" wrote in message
...

It is extremely likely that the overall NFB loop includes the output
stage of the amp. Therefore clipping in the output stage is just about
the worst place for it to occur. The limiting of the NFB signal during
clipping will force the earlier gain stages to handle gross overload
signals. In many amps, this internal overload has the additional effect
of lengthening the time for the amp to recover once the input overdrive
signal has gone away.
--
Chris Morriss

So back to the original question. Is it the case that to produce a waveform
that includes hf, eg supersonic, components an amplifier must be capable of
producing those frequencies? Or does the notion of fourier synthesis not
apply here? Or is it that even rapidly rising wave edges are not rising
rapidly enough to need such hf?

Peter Scott



I SPICE simulated a 20V pk-pk sine wave of 1kHz, being clipped to 18V
pk-pk by a couple of Schottky diodes and a pair of voltage sources. I
added a series resistor to give the effect of an amp 'soft-clipping'.

THD (for the first 16 harmonics) was 0.77%.

3rd of 0.37%
5th of 0.385%
7th of 0.362%
9th of 0.30%
11th of 0.23%
13th of 0.145%
15th of 0.0835

So even with only a small amount of fairly soft clipping, the harmonics
are not starting to drop off quickly until above the 11th, although the
amount of energy in them is quite small.
--
Chris Morriss

In article , Peter Scott
wrote:

"Jim Lesurf" wrote in message
...
[snip]

But I come back the point about the hf. Surely if the clipping is
caused by effects in the output transistors, that drive them into
non-linearity, it is in effect after the amp stages?

If you are observing components at the amplifier's output terminals when
driving a normal passive linear load then the signals you see are all
getting their non-zero power from the amplifier to that point. Consider in
this context two questions:

1) How can any part of the amplifier be "after" the amplifier? (This sounds
like we are heading towards a Zen Koan if we aren't careful. ;- )

2) The load is not generating the power observed, so if it isn't being
provided by the amp, what would be meant by saying it comes from "after the
amp stages"?

Otherwise how could components be generated that lie outside the range
of the amp.

This depends what definition you are using for the "range of the amp".

When operating as a unit, the overall amplifier circuit will have a given
frequency response into the load. However individual gain devices in the
amp may well have gain 1 over a wider range. Also, when you push devices
into nonlinear behaviour by clipping/saturating/switching them, this may
create changes in level which occur in shorter timescales (and hence at
higher frequencies) than the range over which the amp has a gain 1 in
normal, unclipped use. Is this what you have in mind?

However in my experience most SS amps have devices with gain up to
frequencies well above 20kHz (typically to MHz region or higher). Hence I
doubt this is a problem as such for audio. The audible problems will tend
to be at much lower frequencies.


Or is it that fourier doesn't come into it.

Fourier Tranformation and Fourier Theory simply provides a method for
looking at the same information in terms of either the time waveform (time
domain) or frequency spectrum (frequency domain). You can use this as you
prefer, provided the tranformation is performed correctly over an
appropriate waveform interval, etc. So it can come into it or not depending
on what you are trying to explain/understand, etc.


The fact that you need this range of frequencies to synthesise the wave
doesn't mean that the amp has to produce them.

If they appear in the clipped waveform, then they have power. Thus they
will have come from somewhere. If they are at the ouput terminals of the
amp, and the load is essentially passive and linear, the question becomes,
where else are they coming from?

The 'sharpness' of the clipping will depend on the behaviour of the
amp/devices. Hence you can't assume perfectly sharp clipping (which would
potentially produce components up to an arbitrarily high frequency).

All that is needed is a slew rate that is at least equal to the steepest
slope on the wave.

The slew rate isn't really the issue, I think. It is the sharpness of the
'corners' produced as the signal enters/leaves clipping. i.e. the
derivatives of the shape if considering the time domain waveform.

'Perfect' or idealised clipping would involve a discontinuous change in the
slope of the waveform at the instants when clipping starts/ends. Such
discontinuities would imply an infinite bandwidth for the output. In
reality, this doesn't happen, though. All real devices have finite
bandwidths and finite response times.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html
Barbirolli Soc. http://www.st-and.demon.co.uk/JBSoc/JBSoc.html

"Chris Morriss" wrote in message
...
In message , Peter Scott
writes




I SPICE simulated a 20V pk-pk sine wave of 1kHz, being clipped to 18V
pk-pk by a couple of Schottky diodes and a pair of voltage sources. I
added a series resistor to give the effect of an amp 'soft-clipping'.

THD (for the first 16 harmonics) was 0.77%.

3rd of 0.37%
5th of 0.385%
7th of 0.362%
9th of 0.30%
11th of 0.23%
13th of 0.145%
15th of 0.0835

So even with only a small amount of fairly soft clipping, the harmonics
are not starting to drop off quickly until above the 11th, although the
amount of energy in them is quite small.

Interesting. Of course the sim will model a perfectly shaped wave
and perhaps just a little more rounding would reduce the higher harmonics
even more? Looks like SPICE is even better than when I last used.
The modelling you've done is very clever. Thanks

Peter Scott

In message , Peter Scott
writes

"Chris Morriss" wrote in message
...
In message , Peter Scott
writes




I SPICE simulated a 20V pk-pk sine wave of 1kHz, being clipped to 18V
pk-pk by a couple of Schottky diodes and a pair of voltage sources. I
added a series resistor to give the effect of an amp 'soft-clipping'.

THD (for the first 16 harmonics) was 0.77%.

3rd of 0.37%
5th of 0.385%
7th of 0.362%
9th of 0.30%
11th of 0.23%
13th of 0.145%
15th of 0.0835

So even with only a small amount of fairly soft clipping, the harmonics
are not starting to drop off quickly until above the 11th, although the
amount of energy in them is quite small.

Interesting. Of course the sim will model a perfectly shaped wave
and perhaps just a little more rounding would reduce the higher harmonics
even more? Looks like SPICE is even better than when I last used.
The modelling you've done is very clever. Thanks

Peter Scott


No it's bog-standard in modern versions of Spice. The 3F5 engine can
get confused when doing distortion analysis of the output of class-d
amps though. I don't know the reason why.

I did use a perfect sine wave as the source. I think that in reality
many power-amps don't clip cleanly. The delays due to the internal
saturations that can occur when the NFB loop is inoperative due to
output stage clipping can give a nasty fast glitch on the edge as the
o/p stage comes out of clipping. I bet Jim Lesurf has got lots of info
about this sort of nasty.
--
Chris Morriss