New page on Squares waves and amplifier performance
 

"David Looser"
"Trevor Wilson"

DolbyT.

What's DolbyT?


** Thomas Dolby:

http://en.wikipedia.org/wiki/Thomas_Dolby

" She blinded me with science .... "

Shame how in TW's case, the blinding is due to a far more prosaic cause.




.... Phil

New page on Squares waves and amplifier performance
 

David Looser wrote:
"bcoombes" bcoombes@orangedotnet wrote in message
...
Bill Taylor wrote:
On Sun, 10 Jan 2010 10:22:12 +0000, bcoombes bcoombes@orangedotnet
wrote:

The cassette deck might have sliced everything above 8k off, but Dolby
B didn't.
But on these decks it did because with the Dolby off there was response up
to at least 12k whereas with it on it didn't just remove the hiss it
removed everything..as per my not properly implemented comment.


What tolerance was there on the response up to12k?, and at what rate did the
response fall off above 8k with Dolby on? What level, relative to Dolby
level, did you make your measurements at?


Hardly necessary since what was missing was/is clearly audible. As it happens I
still have a Panasonic RQ-X20D cassette player which I've just dug out of
storage and had a listen to to confirm the effect and it's totally obvious, with
the Dolby on some of the 'music' is just plain MISSING. My original 8k figure
was a [mis]remembered estimate, it actuallity it must be lower than that.
ISTR that Panasonic were established to be one of the manufacturers that took a
while to get the *new* Dolby technology right but the defect was quite common
for a while and was often commented on in reviews of cassette decks of the time.

New page on Squares waves and amplifier performance
 

David Looser wrote:
"bcoombes" bcoombes@orangedotnet wrote in message
...
Brian Gaff wrote:
Now, don't get all hot about it.
I have a dbx recorder or two here, and one can actually record bandwidth
limited squarewaves at higher levels. The one artefact you tend to see of
cours, is down to the finite time the processor takes to do things. You
tend to get level overshoots and undershoots and an obvious worsening of
the noise performance on louder recordings
I never really understood why everyone went to Dolby,
I remember when Dolby B first started appearing on cheaper Japanese
cassette players. It solved the hiss problem by simply slicing off
everything above 8k. Of course these days that wouldn't bother
me...unfortunately :(

Except that that's not what Dolby B does [when properly implemented].

Fixed your post..see my other reply above.

New page on Squares waves and amplifier performance
 

Jim Lesurf wrote:

However the passage of time has changed both of those
factors! Although I
do still wonder about some modern 'high end' designs. What
peak current and
slew rates can they provide, and is it adequate for real
use with music
into speakers? In particular I wonder about some valve
designs. At what
current do they limit, and what happens when they do?...
Alas, if no one
measures this you can't tell, or often even guess with any
reliability.

It could be that a square wave test isn't the best way of
identifying the kinds of limiting that you mention, and
never was.

When valve amps were the norm, true "slew rate limiting" was
unheard of, AFAIK. Not enough gain or feedback. Books of the
time carried pages of square wave test results, which can
become comically abstruse once you've got an unmatched phase
splitter with two different phase errors, or a saturating
transformer.

One effect that I have seen referred to as slew rate
limiting can happen when a cathode follower with inadequate
bias current fails to keep up with the demands of a
capacitive load. Like real slew rate limiting, it is an
effect of feedback, and results when the triode approaches
turn-off only when the signal is both high amplitude and
high frequency...which is when the slew rate is highest.
However, for small amplitudes, the same CF wouldn't suffer
from the same problem even if the input slew rate were the
same. A square wave test could miss the problem unless it
was full amplitude, and then it might be strewn with other
debris. A positive pulse wouldn't show it either, although a
negative-going one would.

In practice, other tests would be used to identify most
problems. The most common use of a square wave was to check
for a well-damped transient response. For that the square
wave must rise significantly faster than the frequency of
the dominant pole, and hold steady enough so the consequent
ringing can be observed. For a power amp, that frequency was
likely to be of the order of 50kHz. This test makes a nice
picture that anyone can understand directly...a steep rise,
overshoot, three declining wriggles, then flat. Lots of
things do that when you hit them, everyone sees it every
day.

Stability was pretty much the big issue then, hence the need
for that particular test, but is it still? What other
problems were square waves used to illustrate then, that
they could also be best used for now?

Once again, it could be that the issues of concern have
changed such that other tests are more appropriate.

If you wish to re-introduce some stringent analytical
testing to magazine reviews, then you need to pick the most
directly illustrative test for each effect you wish to
portray. Re-introducing the square wave, willy nilly, might
not prove popular or particularly instructive.

What I would most like to see, now as then, is a clear set
of graphs showing distortion across the claimed frequency
range at small, medium and full power levels, into several
representative loads (how many 3D plots would it take to
cover those 4 dimensions?), and a Bode diagram. Some
confirmation that EMC and safety standards have really been
met would be reassuring.

It's easy for amps to look good when they're well within
their comfort zones, so they do need to be stretched to the
limits of their performance and operating conditions, so I'm
with you on that point. These days, it should be possible to
traverse those limits systematically, rather than sample
them conveniently as they seem to do.

Ian

New page on Squares waves and amplifier performance
 

bcoombes wrote:
David Looser wrote:
"bcoombes" bcoombes@orangedotnet wrote in message
...
Bill Taylor wrote:
On Sun, 10 Jan 2010 10:22:12 +0000, bcoombes bcoombes@orangedotnet
wrote:

The cassette deck might have sliced everything above 8k off, but Dolby
B didn't.
But on these decks it did because with the Dolby off there was
response up to at least 12k whereas with it on it didn't just remove
the hiss it removed everything..as per my not properly implemented
comment.


What tolerance was there on the response up to12k?, and at what rate
did the response fall off above 8k with Dolby on? What level, relative
to Dolby level, did you make your measurements at?


Hardly necessary since what was missing was/is clearly audible. As it
happens I still have a Panasonic RQ-X20D cassette player which I've just
dug out of storage and had a listen to to confirm the effect and it's
totally obvious, with the Dolby on some of the 'music' is just plain
MISSING. My original 8k figure was a [mis]remembered estimate, it
actuallity it must be lower than that.
ISTR that Panasonic were established to be one of the manufacturers that
took a while to get the *new* Dolby technology right but the defect was
quite common for a while and was often commented on in reviews of
cassette decks of the time.

On the above subject here's some technical stuff on Dolby NR implementation
problems. No doubt many of the early and 'cheap' cassette decks *incorporated*
them but it seems it may also be a problem with the recordings. Any comments
from some of the people here who obviously know about this stuff gratefully
received since I'm more interested in learning than 'being right'.

1. Pumping:
Incorrect selection of the control path bandwidth external components can result
in an audible increase in noise as the input level changes. This is most likely
to be heard on solo instruments or on speech. Sometimes the S/N rate is too poor
and masking will not be completely effective - i.e., when the bandwidth is wide
enough to pass the program material, the increase in noise is audible. Cutting
down on the pumping will also affect the program material to some extent and
judgement as to which is preferable is required. Sometimes a shorter decay time
constant in the detector circuit will help, especially for a source which always
shows these characteristics, but for better program material a return to the
recommended detector characteristics is imperative.
2. High Frequency Loss:
This can be caused by an improper control path gain setting—perhaps deliberate
because of the source S/N ratio as described above—or incorrect values for the
audio path filter capacitors. Capacitors larger than the recommended values will
scale the operating bandwidth lower, causing lower -3 dB corner frequencies for
a given control path signal. Return to the correct capacitor values and the
appropriate control path gain setting will always ensure that the h.f. content
of the signal source is preserved.

New page on Squares waves and amplifier performance
 

In article , Ian Iveson
wrote:
Jim Lesurf wrote:

However the passage of time has changed both of those factors!
Although I do still wonder about some modern 'high end' designs. What
peak current and slew rates can they provide, and is it adequate for
real use with music into speakers? In particular I wonder about some
valve designs. At what current do they limit, and what happens when
they do?... Alas, if no one measures this you can't tell, or often
even guess with any reliability.

It could be that a square wave test isn't the best way of identifying
the kinds of limiting that you mention, and never was.

Well, the 1955 article that Mike just sent me a copy of does show that it
might well have been common! However since people now often don't check
I have no idea how revealing it might turn out to be nowdays. It is
easy enough to deal with this by good design. But I am less than sure
that *all* modern designs *are* 'good' in respects like this TBH.

When valve amps were the norm, true "slew rate limiting" was unheard
of, AFAIK. Not enough gain or feedback.

I appreciate that it was largely "unheard of" in the sense that many people
hadn't heard of it or knew what it was. :-)

However you don't need either high gain or overall feedback for slew
limiting to occur. So if a design isn't made in a way that ensures
it doesn't happen then the absence of feeback or high gain isn't a
guarantee. So far as I know, if the gain stages are not
preceeded by a passive LPF and you can connect arbitrary loads to the
output the slew limiting may be possible, regardless of having no feedback
back or low voltage gain.



One effect that I have seen referred to as slew rate limiting can
happen when a cathode follower with inadequate bias current fails to
keep up with the demands of a capacitive load. Like real slew rate
limiting,

Erm.. what you describe above does seems like slew rate limiting to me,
so I don't know why you say it is "like" slew rate limiting. In what
was is it *not* slew rate limiting?

it is an effect of feedback,


Again, your reason for saying that isn't clear. cf below.


and results when the triode
approaches turn-off only when the signal is both high amplitude and
high frequency...which is when the slew rate is highest.

Your approach above seems different to mine. Mine is that the stage has a
limited output current ability. So if you attach a large enough capacitance
and try to slew the voltage fast enough you get current limiting determined
slew rate limiting of the output voltage. Not sure why feedback would be
regarded as the cause of that. Perhaps you can explain if I have not
understood your point.

So far as I know the above can arise even with a single gain stage with
no feeback. The only requirement is a load capacitance that is non
zero, a current limiting mechanism in the circuit, and an input that
has too high a rate of change for the stage to then handle without
the problem arising.

The current limiting mechanism can arise in various ways. Obvious
examples being the saturation/max current the gain device can pass.
You then get essentially a current source attached to a capacitance.

However, for small amplitudes, the same CF wouldn't suffer from the same
problem even if the input slew rate were the same.

I'm not clear why you think that, I'm afraid.

I appreciate that real 'squarewaves' tend to have finite bandwidth
and so using a given generator you tend to find that the maximum slew
rate of the generator output scales with the waveform amplitude.

But if the mechanism of the problem in the stage is that it has finite
current capacity, and is connected to a load capacitance, then that
mechanism will be the same. Is your argument that with small signals
the current required for the load resistance is smaller, so more is
available for the capacitance? If so, yes, I'd agree that makes
sense.

Either way, I agree you'd need to check for this with suitably large
waveform amplitudes as smaller test signals may well not provoke
a problem which larger signals would show.


A square wave test
could miss the problem unless it was full amplitude, and then it might
be strewn with other debris. A positive pulse wouldn't show it either,
although a negative-going one would.

I'd assume the voltage amplitude required would vary with the load
capacitance and the current limit value for the part of the system which
was involved.

In practice, other tests would be used to identify most problems. The
most common use of a square wave was to check for a well-damped
transient response. For that the square wave must rise significantly
faster than the frequency of the dominant pole, and hold steady enough
so the consequent ringing can be observed. For a power amp, that
frequency was likely to be of the order of 50kHz. This test makes a
nice picture that anyone can understand directly...a steep rise,
overshoot, three declining wriggles, then flat. Lots of things do that
when you hit them, everyone sees it every day.

Yes. I think that was why later on some reviewers checking SS amps with an
output inductor assumed the ringing with a capacitative load was something
other than the passive LC resonance of the output inductor with the load
capacitance... and then though it was something to do with the stability
margin of the amp.

Stability was pretty much the big issue then, hence the need for that
particular test, but is it still? What other problems were square waves
used to illustrate then, that they could also be best used for now?

Once again, it could be that the issues of concern have changed such
that other tests are more appropriate.

For well designed amps I'd agree. However many of the 'high end' designs
these days tend to be valve types with output transformers and valve stages
that could current limit in quite complex ways. As per the results in the
1955 article I do wonder what some of the 'new' designs would do if square
wave tested into loads other that a kindly matched resistor load. :-)

If you wish to re-introduce some stringent analytical testing to
magazine reviews, then you need to pick the most directly illustrative
test for each effect you wish to portray. Re-introducing the square
wave, willy nilly, might not prove popular or particularly instructive.

Matter of horses for courses, yes. However the problem here is that in some
cases you don't know what a test will show until you do it and see.

What I would most like to see, now as then, is a clear set of graphs
showing distortion across the claimed frequency range at small, medium
and full power levels, into several representative loads (how many 3D
plots would it take to cover those 4 dimensions?), and a Bode diagram.
Some confirmation that EMC and safety standards have really been met
would be reassuring.

I'd also like measurements to determine the in-loop output impedance and
stability behaviour.

It's easy for amps to look good when they're well within their comfort
zones, so they do need to be stretched to the limits of their
performance and operating conditions, so I'm with you on that point.
These days, it should be possible to traverse those limits
systematically, rather than sample them conveniently as they seem to do.

A number of people have proposed various tests. However as with squarewaves
I always tend to end up feeling you need a variety of these on tap so you
can check for the unexpected! :-)

I presume many amps would be fine. But the importance of the tests would be
to know which ones do 'pass ok' and which get caught out in ways that might
crop up with real music into real speakers.

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
 

"bcoombes" bcoombes@orangedotnet wrote

Hardly necessary since what was missing was/is clearly audible. As it
happens I still have a Panasonic RQ-X20D cassette player which I've just
dug out of storage and had a listen to to confirm the effect and it's
totally obvious, with the Dolby on some of the 'music' is just plain
MISSING. My original 8k figure was a [mis]remembered estimate, it
actuallity it must be lower than that.
ISTR that Panasonic were established to be one of the manufacturers that
took a while to get the *new* Dolby technology right but the defect was
quite common for a while and was often commented on in reviews of cassette
decks of the time.

What you were hearing was dynamic expansion of the HF. The HF isn't
"missing" exactly, but it has been pushed down in level by up to 10dB,
depending on it's frequency and on it's original level. The audible effect
is unpleasant IMO, probably subjectively worse than the simple HF roll-off
you took it to be.

I long ago gave up using Dolby NR on cassettes; I have a fairly up-market
Aiwa with auto bias and eq adjustment, but even with that I preferred to
sound with Dolby off. A little bit of tape hiss is pretty innocuous, far
less subjectively disturbing than hearing NR systems at work.

David.

New page on Squares waves and amplifier performance
 

"bcoombes" bcoombes@orangedotnet wrote

.. It solved the hiss problem by simply slicing off
everything above 8k. Of course these days that wouldn't bother
me...unfortunately :(

Except that that's not what Dolby B does [when properly implemented].

Fixed your post..see my other reply above.

Sorry, your "fix" is wrong. Even when improperly implemented Dolby doesn't
do that. In fact there was usually very little wrong with the implementation
of Dolby in cheap cassette decks, there was a standard circuit which most
manufacturer's faithfully followed. The problem was that to work properly
Dolby needs a reasonably good recorder between the encode and decode
operations, and the actual recorders in these cheap cassette decks wasn't up
to the job.

David.

New page on Squares waves and amplifier performance
 

"Ian Iveson" wrote in message
...
Jim Lesurf wrote:


When valve amps were the norm, true "slew rate limiting" was unheard of,
AFAIK. Not enough gain or feedback.

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.

One effect that I have seen referred to as slew rate limiting can happen
when a cathode follower with inadequate bias current fails to keep up with
the demands of a capacitive load. Like real slew rate limiting, it is an
effect of feedback, and results when the triode approaches turn-off only
when the signal is both high amplitude and high frequency...which is when
the slew rate is highest. However, for small amplitudes, the same CF
wouldn't suffer from the same problem even if the input slew rate were the
same.

Nope, CF failure can occur even with small amplitude signals. The cathode
voltage can only fall at the rate determined by the time constant of the
cathode resistor and load capacitance. If the input falls faster than this
the output will not follow regardless of the signal amplitude. CF failure
can be a problem with video signals; to drive a high-capacitance load with a
CF may require a load resistor so low in value as to represent a real
problem. In the original TV transmitter at Alexandra Palace the cathode
resistor of the CF at the output of the modulation amplifier required
water-cooling as it dissipated over a kilowatt. Since the valve was a DH
type the filament supply for it came from a motor-generator set mounted on
tall insulators to minimise it's capacitance to earth! More recent TV
transmitter design has used White cathode-followers or other forms of
push-pull drive for this function to reduce the power dissipation. The
transmitter I am currently working on uses 4 PL38s in a White
cathode-follower.

David.

New page on Squares waves and amplifier performance
 

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. Slew rate limiting is
in fact plain ordinary limiting (sawing the tops off a sine wave) but
in the current domain when feeding a capacitor. Because of the
differentiation it looks in the voltage domain like a straight slop.
Excessive bandwidth permits large fast signals that will show limiting
of this kind.

d