New page on Squares waves and amplifier performance
 

In article , David Looser
wrote:
"Don Pearce" wrote in message
...
On Mon, 11 Jan 2010 18:19:50 -0000, "David Looser"
wrote:


It may have been "unheard of", but it existed all the same. Any
amplifier, with or without feedback, can exhibit slew-rate limiting if
the bandwidth is insufficient to cope with the rate of rise or fall
of the input signal.

Excessive bandwidth is the cause of the problem.

You mean excessive bandwith of the input signal I presume. That's
another way of saying insufficient bandwidth in the amplifier.

Define your terms. :-)

One of the points of the webpage which started this thread is that a test
waveform generator may deliver a far wider bandwidth for the test signal
than the actual domestic sources which the amp is designed for.

So saying the amp has "insufficient bandwidth" has to be accompanied by
specifying "for the purpose of...".

In terms of curing/avoiding the problem the simple approach is to fit an RC
passive filter to the input of the amp. This is then chosen to *suit the
design and the intended use*. i.e. for domestic audio, not for driving lab
squarewaves into test loads. So the amp bandwidth should then be fine for
the intended use, but not for all other purposes.

Indeed, excessive bandwidth of the incoming signal. But as Jim showed a
simple passive filter on the input to the amplifier solves it.

Not quite "solves". In principle the damn user can always go and find an
even more evil load which has more capacitance or some other awful RF
behaviour. Hence having a passive output network and worrying about
stability, etc, is also sensible in this context.

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

New page on Squares waves and amplifier performance
 

In article , Ian Iveson
wrote:

Engineers weren't daft in the days of valve amps.

Rather a sweeping assertion. You mean *none* of them ever did anything that
lead to amplifier imperfections? If so, then I'm afraid that many tests -
including in the 1955 article I've referred to - show otherwise.

Had slew rate limiting been a problem, it would have been recognised.

Again, you do seem here to be making assertions of faith in a rather
sweeping manner.

Yes there are various forms and causes of current limiting, which may
appear similar to slew rate limiting, but that's not what they are.

I'm afraid that seems rather unclear to me.

I agree that current limiting is not slew rate limiting. But you'd need to
explain what you mean beyond that.

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

New page on Squares waves and amplifier performance
 

In article , Ian Iveson
wrote:
David Looser wrote:

Indeed, excessive bandwidth of the incoming signal.

Excessive slew rate. There really is no other adequate way of saying
it. Not excessive bandwidth, or amplitude, or combination of the two,
but slew rate, pure and simple.

Although as IIUC you indicated earlier, the input slew rate that does
become 'excessive' will also depend on other factors like signal amplitude
if - for example - the output load also has resistance as so that takes
some of the output current.



But as Jim showed a simple passive filter on the input to the
amplifier solves it.

Not if it's bad enough so that it can occur at audio frequencies, or
frequencies otherwise necessary for the satisfactory operation of the
amp. Then the cause of the limiting must be dealt with.

Well, the fix could prevent the slew limiting, but at the unacceptable
sacrifice of spoiling the performance in another way (i.e. rolling away the
HF to an audible extent). So, yes, the design would need altering in some
other way to avoid exchanging one problem for another.

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

New page on Squares waves and amplifier performance
 

In article , David Looser
wrote:
"Don Pearce" wrote

No, it is part of the amplifier - a very necessary part. Otherwise it
would be part of the record deck, the microphone, the CD player etc -

It doesn't have to be part of anything.

and it would need to be variable to cope with all the different
amplifiers that might be connected.

No, because it's only ever used with the the one amplifier. *Physically*
it may be part of the amp, but as it's outside the NFB loop it's not
part of the amp in a functional sense.

I must admit that I regard signal conditioning at input and output to be a
part of the design of the amp. Without this you can't define what tasks the
amp is being expected to perform, etc. Although your argument with David
here does seem to me to be approaching 'how many angels can dance on the
head of a pin'. :-)

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

New page on Squares waves and amplifier performance
 

In article , David Looser
wrote:

The theory behind this applies to more than Hi-Fi, more than audio. You
cannot say where and how band-limiting filtering may be applied in every
design. The filtering is there so that the slew-rate of the signal does
not cause problems to the amplifier; I take it we can agree on that?
Arguing about whether this filter is regarded as part of the amplifier,
or not, seems to me to be equivalent to arguing about how many angels
can dance on the head of a pin.

Snap. 8-]

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

New page on Squares waves and amplifier performance
 

In article , David Looser
wrote:
"Ian Iveson" wrote

In the meantime I suggest you check out how a cathode follower works,
and in particular how its output impedance is defined.


Think of it like this. The valve can source current into the load
capacitance, but not sink current from it. So the only discharge path
for the charge stored in that capacitance is via the cathode resistor,
and hence the rate of drop of the cathode voltage cannot exceed that
determined by the time constant of those two components.

There is a distinction which I *think* Ian may have had in mind.

The 'fully conducting' state of the gain device will tend to approximate to
a current source. The 'fully off' state leaves the bias resistor to supply
the output.

One state acts as a current source attached to any load capacitance. Thus
producing a defined rate of change of the voltage.

The other generates a time constant, so the rate of change then varies
exponentially.

So you can argue that one is 'slew rate limiting' in a way that the other
is not.

However in practice both effects represent the max rate of change the amp
can deliver under the circumstances and produce a period where any small
details of the input may be blocked or ignored. So the difference is in
practice rather cosmetic. Hence personally I'd regard both as being due to
current limiting into a capacitance, both produce distortion, both need to
be avoided by means like those discussed. Can't say I'm too fussed if
people want to call them both slew limiting or not since in practice the
observed difference is that actual limiting rate has different behaviours,
but is limited in both cases according to the circumstances of the design.

So rather than argue about the distinction, design to avoid both of them.
:-)

However my real concern here is similar to the one I raised a while ago wrt
loading due to cable+speaker combinations. If no-one is checking, how can
we know what *all* modern designs actually are doing in practice?

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html

New page on Squares waves and amplifier performance
 

"Jim Lesurf" wrote

There is a distinction which I *think* Ian may have had in mind.

The 'fully conducting' state of the gain device will tend to approximate
to
a current source. The 'fully off' state leaves the bias resistor to supply
the output.

One state acts as a current source attached to any load capacitance. Thus
producing a defined rate of change of the voltage.

The other generates a time constant, so the rate of change then varies
exponentially.

So you can argue that one is 'slew rate limiting' in a way that the other
is not.

However in practice both effects represent the max rate of change the amp
can deliver under the circumstances and produce a period where any small
details of the input may be blocked or ignored. So the difference is in
practice rather cosmetic. Hence personally I'd regard both as being due to
current limiting into a capacitance, both produce distortion, both need to
be avoided by means like those discussed. Can't say I'm too fussed if
people want to call them both slew limiting or not since in practice the
observed difference is that actual limiting rate has different behaviours,
but is limited in both cases according to the circumstances of the design.

So rather than argue about the distinction, design to avoid both of them.
:-)

The reason for my original reply to Ian was to counter his assertion that
negative going slew-rate limiting in CFs applies to high amplitude signals
only. In frequency response terms a CF can cope with higher frequencies at
low amplitude than high, because the slew-rate for low-amplitude signals is
less than for high amplitude for a given frequency, but in terms of slew
rate the limitation is independent of amplitude. As far as negative going
signals is concerned that rate is determined by the ability of the cathode
load resistor to sink current out of the load capacitance. It would be the
same if the cathode load were a constant-current sink, except that now it
would be a true current-limited effect rather than being exponential. In any
practical standard CF the negative going slew-rate limit will be worse,
probably a lot worse, than the positive going slew-rate limit.

As I also tried to point out that a CF is inherently unsuitable as driver
for supplying fast slew-rate signals into highly capacitive loads even
though the low output impedance might suggest otherwise. Designing to avoid
slew-rate limiting in these cases probably means using something other than
a standard CF.

David.

New page on Squares waves and amplifier performance
 

"Phil Allison"

Bandwidth and rise times are small signal parameters -
but " slew rate limiting " is a LARGE signal phenomenon.

All audio amplifiers have slew rate limits, usually quoted
in V/uS - often different rates apply for positive and
negative going output voltages.

The usual cause is a current source internal to the amp's
topology being driven to its limit in trying to charge and
discharge internal capacitances.

As originally coined, "slew rate limiting" applied to op
amps. Where gain and feedback are both very large, dv/dt can
directly translate into an internal current requirement,
regardless of amplitude. Consequently the slew rate limit
will be true for any amplitude. That's why quoted slew rates
make no mention of how big the signal is.

However, an *infinitesimally* small signal can be considered
to have no significant slew, and hence no significant slew
rate. That's why slew rate limiting is not a consideration
of small signal analysis.

Large signal analysis concerns itself with any signal that
is not infinitesimally small.

Ian

New page on Squares waves and amplifier performance
 

Jim Lesurf wrote:

Engineers weren't daft in the days of valve amps.

Rather a sweeping assertion. You mean *none* of them ever
did anything that
lead to amplifier imperfections? If so, then I'm afraid
that many tests -
including in the 1955 article I've referred to - show
otherwise.

The assertion is qualified in the next sentence.

Had slew rate limiting been a problem, it would have been
recognised.

See, that's what I meant by not daft. Not all daft.
Collectively not daft.

Again, you do seem here to be making assertions of faith
in a rather
sweeping manner.

Hardly. The symptoms of slew rate limiting are not hard to
spot. Very many very competent engineers applied themselves
to valve circuits, including the cathode follower. Do you
imagine that it's limitations, and especially one so plainly
obvious, would remain unknown? Check out RDH4, for example,
and tell me they didn't know that a clipping triode in a
cathode follower should be avoided.

An op amp circuit has a slew rate limit. It can be ascribed
a value expressed as a number. The same cannot be said of a
cathode follower.

Or do you believe it can? So let's take a weedy AX7 CF
driving 10nF, set up as per a datasheet example, with 220k
cathode resistor, and a bias of 1V at 1.2mA. Can you tell me
what it's slew rate limit is, expressed as a single number?

The reason you could not is that, as long as the input
signal is appreciably smaller than that 1V bias voltage, the
triode will never be driven into cut-off or saturation,
regardless of frequency and hence regardless of slew rate.
The output impedance will always be of the order of 650
ohms, rather than the 220k it becomes if the valve turns
off.

Drive it with a signal appreciable greater than the bias
voltage and at some point, as frequency is increased and the
effective feedback proportion diminishes, the effective Vgk
signal input will exceed the bias voltage. Assuming the
valve is set up for optimum class A operation, it will be
starting to cut off at one extreme and pass grid current at
the other. The transformation from 650 ohms to 220k happens
fairly suddenly as the triode cuts off, and is from an
voltage source to an approximate constant current source.
But it's quite a distant approximation.

A triode makes a poor CCS even when optimised for that
purpose, because of a shortage of gain.

Yes there are various forms and causes of current
limiting, which may
appear similar to slew rate limiting, but that's not what
they are.

I'm afraid that seems rather unclear to me.

Exactly. The simple definition of slew rate limiting *is
very* clear. A definition that includes consideration of
amplitude or frequency *is not at all* clear.

I agree that current limiting is not slew rate limiting.
But you'd need to
explain what you mean beyond that.

Take that AX7 CF and 10nF as an example. It won't fail with
a 5V input signal as long as frequency is low. It won't fail
with a 20kHz signal as long as the signal amplitude is low,
but it will fail with a combination of 5V and 20kHz. Because
it only fails with a combination of high frequency and high
amplitude, it can appear to be slew rate limited, because
slew rate also increases with amplitude and frequency.

However, a signal of less than 1V would not be limited no
matter how high its slew rate, although it might be
attenuated by the filter. So it's not the slew rate that
causes the limit, but rather the combination of amplitude
and frequency.

Ian

New page on Squares waves and amplifier performance
 

"Ian Iveson needs to be shot in the head "

As originally coined, "slew rate limiting" applied to op amps.


** But the topic here is audio amps.

You ****wit, context shifting, autistic pommy ****.




.... Phil