Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

Jim Lesurf wrote:

In article , Patrick Turner
wrote:

[big snip]

I did make some remarks here earlier about the non flat and peaked
response the open loop response of a typical SS amplifier, one which I
am completing as we type.

When measuring an SS amp for open loop response, it is necessary to
shunt the AC signal NFB with a large value cap from the NFB port of the
amp to 0V, so the DC voltage at the output can still be fed back to
maintain CD stability at least, necessary because passive trimming of
the DC gain is impossible because in fact the gain is around 250,000
times, or over 100 dBv.

One point I'd like to raise here is that we need to take care with quoting
values for voltage gain as the actual value may well depend upon the output
load. This is particularly the case when considing open loop values as the
system then lacks the feddback network that would control the output level
and tend to stop the voltage gain varying with o/p load. I'm not clear what
value of o/p load you are assuming or using for the values you quote in
your posting.

One must assume that gain is for the rated load unless otherwise stated.

Open loop gain in amps with emitter follower outputs does not vary much with
load value since the input impedance to the output stage is so high
regardless of the load at the output, if the amp has been designed that way.
Crown use darlington triples in their output stages to achieve
the high impedance so the gain of the VAS can be at maximum.




A side-point here is to be wary of quoting gains in "dB's" in this context
as they can be misleading or confusing unless we take the impedances
through the system (and the load) into account.

Gain can be assumed to be voltage gain, and expressed in dB.

Its 20 x log of the number of times voltage is amplified.

All engineers know this, and routinely talk and think in dB.



So when a cap is used to shunt the DC, the cap does have some impedance,
even if it is 300 uF, and so the response does not measure flat below
the open loop HF pole, which is around 1 kHz.

I assume that a large cap will also exhibit resistance and inductance as
well, so this would need to be taken into account if you are trying to
asses the 'true' open loop gain via measurements with a cap.

The caps esr is negligible at LF. The impedance of a 300 uF cap
at LF is mainly just a function of its capacitance value.



Curious as to why your amp starts its open loop rolloff as low as 1 kHz.

Because of the way the FB port is shunted to make the tests.

When the amp has its normal FB applied, the error signal is substantially flat
from 5 Hz to 2 kHz, and then the error signal has rise to about 13k, then a
roll off,
and this tells you that the open loop gain rolls off at 2 kHz...

This was after some tweaking of the gain character...

I have not time to write you a book on basic amplifier engineering.
Please don't take all of what i say as gospel whch can be applied universally.



SS amps are usually all direct coupled, which means their response
*should* extend down to 0.0 Hz, but in fact this is impossible, unless
we have infinitely low impedance of the PS, which is almost possible if
we used good active regulation, or about 50 truck batteries..

Afraid I still am puzzled by the above comment. Having a high gain at dc
should not mean the amp must have rails with zero impedance so far as I can
see. Can you explain what you mean here? FWIW I have used instrumentation
power amps that go down to dc and these didn't need what you say, nor can I
see a theoretical reason for it in the practical cases I have in mind.
Hence I am wondering if I've understood what you are saying here.

In theory, direct coupled amplifiers are regarded as being able to go down to
DC.

In most amps there is the usual
two resistor FB R network, say 33 k and 1 k, with the junction taken to the
FB port of the diff pair, and then between the 1k and 0V there is about 100 uF.

so that at DC, *all* the output voltage is fed back to the FB port.
But the input resistance has an effect on the FB application.

So if you apply +1 volt DC to the input, you will get +1 volt DC at the output
and very close to
+1V applied to the FB port.

The rail caps at DC are an open circuit.
But the PS rectifiers create a supply impedance of a finite value.
Usually its quite low, so rails don't move much even when VLF signals are
created at the output.




[snip]

The output stage is emitter follower, and has about 40 dB of NFB applied
there. At 400 Hz, and when global NFB is applied, the open loop gain is
reduced from 262,000 to 33, a reduction of nearly 8,000 times. Since thd
at 34vrms of output is around 0.005%, we could have expected thd to be
8,000 times worse with no NFB, and that would mean it would be 40%.

When i did test the amp with no NFB, thd sure wasn't 40%, more like 5%.
so the use of NFB in SS amps does not always seem to follow the laws of
gain/feedback equations like it generally does in tube amps using far
less NFB.

There are various possible reasons for that, and they may not have much to
do with choosing SS devices.

Two possibilities.

1) That some of the 'open loop' and test gains you quoted are voltage gains
and hence may be subject to the comments I made above about being loading
dependent and not taking current gains or impedances fully into account.

Loadings were taken into account.
Gain measurements were made with FB connected, and at an F where
the error signal is lowest at the output of the diff pair.



2) Some distortion mechanisms may be 'outwith the loop' so aren't affected.
An obvious example here is that in some AB designs you might find that some
of the rail current variations induce error voltages at the ground
reference point used for feedback. This then can inject distortion by a
route that the feeback loop can't really deal with.

I am not sure about this in my case.
I tried various 0V wiring set ups, and nothing changed the measured thd.
The was so low, that the acumulated errors could easily outwit me.

I got the thd as low as I want for this amp, and its stable,
no ring on square waves in the error signal, so all is well.



I suspect you will have noticed that when looking at low-levelTHD figures,
particularly at higher frequencies, you can sometimes adjust the reading by
moving around wires or components and inducing various effects similar to
(2). On some cases this gives almost a 'null' in some circumstances, but
the low value this gives can be highly misleading.

Perhaps the fuse at the output really is contributing all that thd; my
oscillator which makes about 0.002% at the 1v level. And the wiring of
this re-engineered amp isn't state of the art, and so star earthing not
entirely optimal,

Ah. Perhaps (2) above may be relevant here.

but then 0.005% at 144 watts isn't too bad a result.

The snag is that to check for things like 'nulled' distortion effects we
really need to try things like intermod and asymmetric signals.
Unfortunately, these can be harder to do well than thd...

That said, it becomes open to question if this matters when the distortion
levels are genuinely low.

Maybe I can wheel out the 1 kHz LC filter with 30 dB
attenuation at 3 kHz, and that may improove the thd measurememnts, along
with placing a shunt across the fuse, or using a 40 amp auto fuse, which
should have less effect than the 10 amp one I have their now. The fuse
is outside the NFB loop.

What current is required with these fuses to make them blow in, say, less
than 100 milliseconds? I don't know about currently pun! available fuses,
but I found in the past that you had to use surprisingly low values to get
them to blow quickly enough to protect bipolars.

I am not worried, I have 5 output transistors each side of the PP circuit.




[snip]

I may appear to not be willing to discuss every widget and twadget in
the thread; I have to earn a living, and I am time poor.

I appreciate that. I am now retired, but still find I don't have enough
time to do everything I intend/plan. :-)

But I hope I have been of use to those actually building something,
rather than spending time all night picking holes in people's arguments
and building nothing, observing nothing, measuring nothing, and
understanding less.

No idea who you are referring to here. :-)

Can't speak for anyone else. But I hope I have been asking questions and
making comments with the aim of improving *all* our understanding and
aiding better design.

FWIW I have built, measured, etc, quite a few things in my time, so I
assume you are not referring to me above. ;-

But what are you building now?



And BTW, I do have a website, and one due for re-vegetation after all
the browsers have passed over the last 4 years, and I will get around to
it, along with my tax return, and all the other work.

There is little at my website regarding solid state design and build
methodology. The master at that appears to be Douglas Self, from whom I
have learnt a lot. Ben Duncan is another who inspired me to consider the
considerables. Fritz Langford-Smith taught me more about tube craft than
anyone else. he wrote the Radiotron Designer's Handbook, along with his
merry team of helpers.

FWIW I also have a high regard for Doug Self and Langford-Smith. Alas, I
can't say the same for Ben Duncan as I have found too many misleading,
ambiguous, out-of-context, or inaccurate things in his published work to be
comfortable with a lot of what he has written.[1] I regret to say that many
textbooks contain things which should be treated with some caution.

Slainte,

Jim

[1] If interested: For a specific example, go to the 'Analog and Audio'
section of the "Scot's Guide" and have a loop at the linked page on
"Current dependent phase shifts in cable?"

I will have a peek later.

Patrick Turner.



--
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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

In article , Patrick Turner
wrote:

I have since removed the fuses, and will never ever use clipped or
pressure clamped fuses again. I have since rewired the output to
include a board with soldered links of fuse wire. The metalic
interaction of the ends of the pressure clamped fuse causes quite
some thd, as would any connection when not perfect.

That solves the distortion problem, but may not deal with any
medium-term dynamic variations in fuse resistance produced by the
music. It also is a bit awkward for the user if they have to change a
fuse. :-)

The soldered in fuse link seems to be quite blameless, since it is but a
piece of plated copper wire. The distortion measurements indicate it has
no large effects worth a worry.

At what frequency and power did you do the distortion measurements?

The paper by Greiner that Arny has pointed out in another thread shows
fuses giving quite high levels of distortions under some quite plausible
conditions.



Mind you, I was happy enough to put the fuses on the boards so the
user had to undo the lid to replace them. Did this to give the PSU
caps time to discharge before the user could touch the fuses.
c100,000microF caps up at over 80V can make the user jump. :-)

All you need with 100,000 uF at 80v is 1k5 at 10 watts as a bleed
resistor across the cap. The time constant is 150 seconds, no? so rails
would be down by the time you opened the case.

Yes. This is the approach I employed.

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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

In article , Patrick Turner
wrote:


Jim Lesurf wrote:
[snip]
One point I'd like to raise here is that we need to take care with
quoting values for voltage gain as the actual value may well depend
upon the output load. This is particularly the case when considing
open loop values as the system then lacks the feddback network that
would control the output level and tend to stop the voltage gain
varying with o/p load. I'm not clear what value of o/p load you are
assuming or using for the values you quote in your posting.

One must assume that gain is for the rated load unless otherwise stated.

Well what one *must* assume can sometimes be incorrect if the other person
did not share your assumptions. :-)

Open loop gain in amps with emitter follower outputs does not vary much
with load value since the input impedance to the output stage is so high
regardless of the load at the output, if the amp has been designed that
way. Crown use darlington triples in their output stages to achieve the
high impedance so the gain of the VAS can be at maximum.

Again, the problem here is establishing things like the meaning of the
term "open loop gain". It seems implicit in what you write above that you
are thinking of the voltage gain. The actual power gain will be load
dependent. And some distortion mechanisms, etc, will depend upon that.

It is perhaps also worth bearing in mind that although maybe "Crown use
darlington triples" this may not be universal practice for all designs.
Hence making assumptions like this in general statements about all
amplifiers might be unreliable. (Nor do I know if even Crown used
triples in *all* their designs.) Hence my interest here is in the
reliability of what you say in terms of being applicable in 'all'
cases, or just being a comment which applies in some cases...





A side-point here is to be wary of quoting gains in "dB's" in this
context as they can be misleading or confusing unless we take the
impedances through the system (and the load) into account.

Gain can be assumed to be voltage gain, and expressed in dB.

Again what you regard as "can be assumed" I'd tend to phrase more like "may
be guessed to be".

Its 20 x log of the number of times voltage is amplified.

All engineers know this, and routinely talk and think in dB.

I know that many do. The problem is that it is only strictly correct where
the impedances involved are the same. Since this often isn't the case, the
voltage ratio result quoted in dB is ambiguous and can be misleading.



So when a cap is used to shunt the DC, the cap does have some
impedance, even if it is 300 uF, and so the response does not
measure flat below the open loop HF pole, which is around 1 kHz.

I assume that a large cap will also exhibit resistance and inductance
as well, so this would need to be taken into account if you are trying
to asses the 'true' open loop gain via measurements with a cap.

The caps esr is negligible at LF. The impedance of a 300 uF cap at LF is
mainly just a function of its capacitance value.

TBH my main interest here was actually in the overall response - at HF
as well as LF.

I am not quite as certain of the above as you seem to be. However I'd agree
with you in many circumstances. I'd be interested to know, though, the
leakage current values of the caps involved.
[snip]

I have not time to write you a book on basic amplifier engineering.
Please don't take all of what i say as gospel whch can be applied
universally.

I don't. :-) That is why I was previously trying to question some of the
statements that you made in the form of 'universal' statements in past
postings.

I am am very happy to agree that many of the things you say are correct
in specific cases. However I am much less sure of some of the more
sweeping statements you have made.




SS amps are usually all direct coupled, which means their response
*should* extend down to 0.0 Hz, but in fact this is impossible,
unless we have infinitely low impedance of the PS, which is almost
possible if we used good active regulation, or about 50 truck
batteries..

Afraid I still am puzzled by the above comment. Having a high gain at
dc should not mean the amp must have rails with zero impedance so far
as I can see. Can you explain what you mean here? FWIW I have used
instrumentation power amps that go down to dc and these didn't need
what you say, nor can I see a theoretical reason for it in the
practical cases I have in mind. Hence I am wondering if I've
understood what you are saying here.

In theory, direct coupled amplifiers are regarded as being able to go
down to DC.

In most amps there is the usual two resistor FB R network, say 33 k and
1 k, with the junction taken to the FB port of the diff pair, and then
between the 1k and 0V there is about 100 uF.

so that at DC, *all* the output voltage is fed back to the FB port. But
the input resistance has an effect on the FB application.

So if you apply +1 volt DC to the input, you will get +1 volt DC at the
output and very close to +1V applied to the FB port.

Agreed.

The rail caps at DC are an open circuit. But the PS rectifiers create a
supply impedance of a finite value. Usually its quite low, so rails
don't move much even when VLF signals are created at the output.

The above does not seem to explain your previous statement about the rails
having to have an infinitely low impedance for the response of the amp to
extend to d.c. Hence it does not seem to answer the question I was asking.

So far as I can see, what you have now written above does not require what
you previously claimed.
[snip]



2) Some distortion mechanisms may be 'outwith the loop' so aren't
affected. An obvious example here is that in some AB designs you might
find that some of the rail current variations induce error voltages at
the ground reference point used for feedback. This then can inject
distortion by a route that the feeback loop can't really deal with.

I am not sure about this in my case. I tried various 0V wiring set ups,
and nothing changed the measured thd. The was so low, that the
acumulated errors could easily outwit me.

I got the thd as low as I want for this amp, and its stable, no ring on
square waves in the error signal, so all is well.

That's fair enough. I was suggesting some possible mechanisms for what
you reported. However I have no idea of the actual cause. My point
was that it is not inevitable that the distortion will always be
reduce in the same way with feedback as this depends upon the details
of how the distortion is arising, etc.

[snip]


FWIW I have built, measured, etc, quite a few things in my time, so I
assume you are not referring to me above. ;-

But what are you building now?

These days, mostly my garden. :-)

However, as with yourself, I assume that some of my recollections,
experiences, etc, may still sometimes prove useful.

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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

Jim Lesurf wrote:

In article , Patrick Turner
wrote:

I have since removed the fuses, and will never ever use clipped or
pressure clamped fuses again. I have since rewired the output to
include a board with soldered links of fuse wire. The metalic
interaction of the ends of the pressure clamped fuse causes quite
some thd, as would any connection when not perfect.

That solves the distortion problem, but may not deal with any
medium-term dynamic variations in fuse resistance produced by the
music. It also is a bit awkward for the user if they have to change a
fuse. :-)

The soldered in fuse link seems to be quite blameless, since it is but a
piece of plated copper wire. The distortion measurements indicate it has
no large effects worth a worry.

At what frequency and power did you do the distortion measurements?

1 kHz.



The paper by Greiner that Arny has pointed out in another thread shows
fuses giving quite high levels of distortions under some quite plausible
conditions.

The thermal dynamic changes and resistance changes would be negligible
with a soldered in peice of copper wire.

Patrick Turner.

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

Jim Lesurf wrote:

In article , Patrick Turner
wrote:

Jim Lesurf wrote:
[snip]
One point I'd like to raise here is that we need to take care with
quoting values for voltage gain as the actual value may well depend
upon the output load. This is particularly the case when considing
open loop values as the system then lacks the feddback network that
would control the output level and tend to stop the voltage gain
varying with o/p load. I'm not clear what value of o/p load you are
assuming or using for the values you quote in your posting.

One must assume that gain is for the rated load unless otherwise stated.

Well what one *must* assume can sometimes be incorrect if the other person
did not share your assumptions. :-)

Well why would I quote open loop gain without a load, or for 1 ohm?



Open loop gain in amps with emitter follower outputs does not vary much
with load value since the input impedance to the output stage is so high
regardless of the load at the output, if the amp has been designed that
way. Crown use darlington triples in their output stages to achieve the
high impedance so the gain of the VAS can be at maximum.

Again, the problem here is establishing things like the meaning of the
term "open loop gain". It seems implicit in what you write above that you
are thinking of the voltage gain.

Of course.

The actual power gain will be load
dependent. And some distortion mechanisms, etc, will depend upon that.

To get maximum voltage gain for the voltage gain FB equations to work,
we must keep load impedances seen by voltage amp stages of the amp high,
and yes, it inherently means having a huge current gain between the
input of the output stage and its output.
So if you have 40V at 10A for a 10 ohm load, and
the input to the EF output is 40.5V, and the current swing to the
bases of the driver bjts is 0.2mA, then the current gain is
10 / 0.0002 = 50,000, and Rin to the bases is 40.5 / 0.0002 = 202,500 ohms.
this would most likely only be possible using a darlingtom triple, or the
an EF + gain transistor driving the EF outputs.

With mosfets or valves we don't need to worry about this issue.



It is perhaps also worth bearing in mind that although maybe "Crown use
darlington triples" this may not be universal practice for all designs.

Probably not.
Most makers have several toplogies, and many are possible.


Hence making assumptions like this in general statements about all
amplifiers might be unreliable. (Nor do I know if even Crown used
triples in *all* their designs.) Hence my interest here is in the
reliability of what you say in terms of being applicable in 'all'
cases, or just being a comment which applies in some cases...

I would stipulate that in all cases where the very high impedance of the
collector output of a drive amp stage
is used to drive an output stage whose input impedance can vary with the output
load, and with the
unmatched hfe of npn and pnp devices, it is wise to buffer the VAS with enough
unity voltage gain high current gain devices to make the output stage look like
a
high impedance load to maximise the gain of the VAS, and allow it to be fairly
linear,
so that NFB application is easy and effective.

One does not want non linear current loads to affect the voltage output
from a current output VAS.





A side-point here is to be wary of quoting gains in "dB's" in this
context as they can be misleading or confusing unless we take the
impedances through the system (and the load) into account.

Gain can be assumed to be voltage gain, and expressed in dB.

Again what you regard as "can be assumed" I'd tend to phrase more like "may
be guessed to be".

Its 20 x log of the number of times voltage is amplified.

All engineers know this, and routinely talk and think in dB.

I know that many do. The problem is that it is only strictly correct where
the impedances involved are the same. Since this often isn't the case, the
voltage ratio result quoted in dB is ambiguous and can be misleading.

Not really.
Voltage gain is gain, regardless of impedances...





So when a cap is used to shunt the DC, the cap does have some
impedance, even if it is 300 uF, and so the response does not
measure flat below the open loop HF pole, which is around 1 kHz.

I assume that a large cap will also exhibit resistance and inductance
as well, so this would need to be taken into account if you are trying
to asses the 'true' open loop gain via measurements with a cap.

The caps esr is negligible at LF. The impedance of a 300 uF cap at LF is
mainly just a function of its capacitance value.

TBH my main interest here was actually in the overall response - at HF
as well as LF.

The ESR has negligible effect on the closed loop bandwidth of the completed amp

between 5 Hz to 68 kHz.


I am not quite as certain of the above as you seem to be. However I'd agree
with you in many circumstances. I'd be interested to know, though, the
leakage current values of the caps involved.

The cap used for the signal grounding of the FB networl is a BP cap, and it has

about 0.3v dc across it.

I have never had problems with leakage.


[snip]

I have not time to write you a book on basic amplifier engineering.
Please don't take all of what i say as gospel whch can be applied
universally.

I don't. :-) That is why I was previously trying to question some of the
statements that you made in the form of 'universal' statements in past
postings.

I am am very happy to agree that many of the things you say are correct
in specific cases. However I am much less sure of some of the more
sweeping statements you have made.

Learn what to ignore, or at least what to be suspicious of.





SS amps are usually all direct coupled, which means their response
*should* extend down to 0.0 Hz, but in fact this is impossible,
unless we have infinitely low impedance of the PS, which is almost
possible if we used good active regulation, or about 50 truck
batteries..

Afraid I still am puzzled by the above comment. Having a high gain at
dc should not mean the amp must have rails with zero impedance so far
as I can see. Can you explain what you mean here? FWIW I have used
instrumentation power amps that go down to dc and these didn't need
what you say, nor can I see a theoretical reason for it in the
practical cases I have in mind. Hence I am wondering if I've
understood what you are saying here.

In theory, direct coupled amplifiers are regarded as being able to go
down to DC.

In most amps there is the usual two resistor FB R network, say 33 k and
1 k, with the junction taken to the FB port of the diff pair, and then
between the 1k and 0V there is about 100 uF.

so that at DC, *all* the output voltage is fed back to the FB port. But
the input resistance has an effect on the FB application.

So if you apply +1 volt DC to the input, you will get +1 volt DC at the
output and very close to +1V applied to the FB port.

Agreed.

The rail caps at DC are an open circuit. But the PS rectifiers create a
supply impedance of a finite value. Usually its quite low, so rails
don't move much even when VLF signals are created at the output.

The above does not seem to explain your previous statement about the rails
having to have an infinitely low impedance for the response of the amp to
extend to d.c. Hence it does not seem to answer the question I was asking.

So far as I can see, what you have now written above does not require what
you previously claimed.

The power supply seen by the amp at 1 Hz at full power will have a finite
impedance.
Its a complex model.
Its not as if the impedance is infinitely small.
A 10,000 uF rail cap has 16 ohms of reactance at 1 Hz, so it may as well not be

present in the PS, except that id smothes the supply of DC which is pulsed into
the cap
as the 100 Hz rate....

You don't need me to work out the equivalent circuit for an amp at 1 Hz, full
power,
I am quite sure you'll work it out yourself.



[snip]



2) Some distortion mechanisms may be 'outwith the loop' so aren't
affected. An obvious example here is that in some AB designs you might
find that some of the rail current variations induce error voltages at
the ground reference point used for feedback. This then can inject
distortion by a route that the feeback loop can't really deal with.

I am not sure about this in my case. I tried various 0V wiring set ups,
and nothing changed the measured thd. The was so low, that the
acumulated errors could easily outwit me.

I got the thd as low as I want for this amp, and its stable, no ring on
square waves in the error signal, so all is well.

That's fair enough. I was suggesting some possible mechanisms for what
you reported. However I have no idea of the actual cause. My point
was that it is not inevitable that the distortion will always be
reduce in the same way with feedback as this depends upon the details
of how the distortion is arising, etc.

Exactly.

With SS amps using such high amounts of NFB, the very tiniest of
parasitic currents can cause thd to leap above 0.005% easily, and one has to
decide whether its worth a week of fiddling around to reduce the thd to 0.001%
at 180 watts.....



[snip]


FWIW I have built, measured, etc, quite a few things in my time, so I
assume you are not referring to me above. ;-

But what are you building now?

These days, mostly my garden. :-)

However, as with yourself, I assume that some of my recollections,
experiences, etc, may still sometimes prove useful.

Well, I don't have too many flame burns.....

Thanks for the interest.

Patrick Turner.



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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

In article , Patrick Turner
wrote:


Jim Lesurf wrote:

In article , Patrick Turner
wrote:


The soldered in fuse link seems to be quite blameless, since it is
but a piece of plated copper wire. The distortion measurements
indicate it has no large effects worth a worry.

At what frequency and power did you do the distortion measurements?

1 kHz.

It may be worth your while to also do measurements at 20Hz or use LF/HF
intermod. The thermal effects of fuses on distortion are likely to be
greater at LF than at 1kHz due to the time constants involved.



The paper by Greiner that Arny has pointed out in another thread shows
fuses giving quite high levels of distortions under some quite
plausible conditions.

The thermal dynamic changes and resistance changes would be negligible
with a soldered in peice of copper wire.

I am not at all sure you are correct. The working method of a metal fuse is
that it becomes heated until it ceases to hold together. This means that at
or above the rated current the temperature of the metal has to rise
sufficiently for this to occur. All normal metals have similar behaviour in
that their resistivity rises with temperature. I don't know of any reason
why a soldering in piece of copper wire would not follow the same physical
laws as other types of metal link arranged to fail at a similar current
level.

No idea what the relevant values would be for the fuses you are using, but
unless you do some LF measurements at high currents I would not assume the
effects are "negligable".

As I mentioned in an earlier posting, I was surprised by how large the
effects were that were reported by Greiner. e.g. he quotes 1 percent THD
for 5A fuses with a 4 Ohm load in one set of conditions. Higher values for
other conditions/fuses...

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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

In article , Patrick Turner
wrote:


Jim Lesurf wrote:

In article , Patrick Turner
wrote:



One must assume that gain is for the rated load unless otherwise
stated.

Well what one *must* assume can sometimes be incorrect if the other
person did not share your assumptions. :-)

Well why would I quote open loop gain without a load, or for 1 ohm?

The problem is with your use of the word *must*. My problem is that you
also seem to regularly use dB values for 'gain' and mean voltage ratios not
power gain. This undermines my confidence that the values quoted always
mean what you seem to imply or assume. :-)

[snip]

The actual power gain will be load dependent. And some distortion
mechanisms, etc, will depend upon that.

To get maximum voltage gain for the voltage gain FB equations to work,
we must keep load impedances seen by voltage amp stages of the amp high,
and yes, it inherently means having a huge current gain between the
input of the output stage and its output. So if you have 40V at 10A for
a 10 ohm load, and the input to the EF output is 40.5V, and the current
swing to the bases of the driver bjts is 0.2mA, then the current gain is
10 / 0.0002 = 50,000, and Rin to the bases is 40.5 / 0.0002 = 202,500
ohms. this would most likely only be possible using a darlingtom triple,
or the an EF + gain transistor driving the EF outputs.

With mosfets or valves we don't need to worry about this issue.

Again, you omit a qualifier like "often" or "in many designs". The
difficulty hear stems from your earlier comment about a situation where you
have a large open-loop gain and then find that the distortion did not fall
as you expected when you applied feedback. One of the points I was making
is that using 'dB' for voltage gains was one reason why the results of
applying feedback may not be as you expect. This being an example of why
expressing voltage gains in dB can be miselading unless other information
is taken into account.

[snip]

Hence making assumptions like this in general statements about all
amplifiers might be unreliable. (Nor do I know if even Crown used
triples in *all* their designs.) Hence my interest here is in the
reliability of what you say in terms of being applicable in 'all'
cases, or just being a comment which applies in some cases...

I would stipulate that in all cases where the very high impedance of the
collector output of a drive amp stage is used to drive an output stage
whose input impedance can vary with the output load, and with the
unmatched hfe of npn and pnp devices, it is wise to buffer the VAS with
enough unity voltage gain high current gain devices to make the output
stage look like a high impedance load to maximise the gain of the VAS,
and allow it to be fairly linear, so that NFB application is easy and
effective.

I'd agree that is a perfectly reasonably approach to *choose* to take.

However I am not sure it is universal practice. Nor am I sure that it is a
necessary design constraignt, and hence a better design might be possible
without it. Nor does it always guarantee that attempting to do as you
stipulate really will avoid the kinds of effects I have mentioned.

One does not want non linear current loads to affect the voltage output
from a current output VAS.

Well the primary concern is that the current load should not affect the
output voltage when the amplifier is in its intended working situation and
arrangement. (i.e. with the chosen feedback applied). This may or may not
be equivalent to what you say to a greater or lesser extent. Depends upon
the design.

[snip]

Gain can be assumed to be voltage gain, and expressed in dB.

Again what you regard as "can be assumed" I'd tend to phrase more like
"may be guessed to be".

Its 20 x log of the number of times voltage is amplified.

All engineers know this, and routinely talk and think in dB.

I know that many do. The problem is that it is only strictly correct
where the impedances involved are the same. Since this often isn't the
case, the voltage ratio result quoted in dB is ambiguous and can be
misleading.

Not really. Voltage gain is gain, regardless of impedances...

I am afraid I do not agree. :-)

Consider as a reductio argument the humble signal transformer. ;-

The transformer may have a turns ratio that gives it a nominal voltage step
up of, say, a factor of 10. However regarding this as a 'gain of 20dB' is a
very dubious description. Putting such a transformer into a black box may
in some circumstances make it look like a differential amp with a "gain of
20dB". But I doubt this result would be "regardless of impedance". Nor do I
think the results of trying to apply feedback would be as you'd expect if
you assumed the "amplifier" had a gain of 20 dB.

Yes, it is quite common for engineers to specify voltage ratios in dB, and
sometimes call the results "gain". However: yes, this also means that the
results they get when they fail to consider the actual *power* gain can be
unexpected (to them).

This is the point of keeping it clear in mind that actual gain in dB is a
power or energy ratio.


I assume that a large cap will also exhibit resistance and
inductance as well, so this would need to be taken into account if
you are trying to asses the 'true' open loop gain via measurements
with a cap.

The caps esr is negligible at LF. The impedance of a 300 uF cap at
LF is mainly just a function of its capacitance value.

TBH my main interest here was actually in the overall response - at HF
as well as LF.

The ESR has negligible effect on the closed loop bandwidth of the
completed amp

between 5 Hz to 68 kHz.

But note that my comment was w.r.t. to its possible effect when trying to
measure the *open loop* response. The point here being that this is another
area where - if using the method you describe - the results of applying
feedback may not always be as you expect as your 'open loop' results may
not translate as you expect into the *actual* open loop behaviour of the
amplifier.


I am not quite as certain of the above as you seem to be. However I'd
agree with you in many circumstances. I'd be interested to know,
though, the leakage current values of the caps involved.

The cap used for the signal grounding of the FB networl is a BP cap, and
it has

about 0.3v dc across it.

I have never had problems with leakage.

In use, I'd say that was quite reasonable. However I was pointing out the
possibility that when doing open-loop measurements near d.c. the leakage of
the cap might affect the observed open loop gain.



I am am very happy to agree that many of the things you say are
correct in specific cases. However I am much less sure of some of the
more sweeping statements you have made.

Learn what to ignore, or at least what to be suspicious of.

Indeed. FWIW I and my (ex-) research group worked on devising, building,
and using measurement instrumentation for the NPL. Indeed, they still use
our designs for some of their work on setting National Standards for
complex impedance and noise in the mm-wave region. This means I spent a
fair while worring about what affects precision measurements of various
kinds.





SS amps are usually all direct coupled, which means their
response *should* extend down to 0.0 Hz, but in fact this is
impossible, unless we have infinitely low impedance of the PS,
which is almost possible if we used good active regulation, or
about 50 truck batteries..

Afraid I still am puzzled by the above comment. Having a high gain
at dc should not mean the amp must have rails with zero impedance
so far as I can see. Can you explain what you mean here? FWIW I
have used instrumentation power amps that go down to dc and these
didn't need what you say, nor can I see a theoretical reason for
it in the practical cases I have in mind. Hence I am wondering if
I've understood what you are saying here.

In theory, direct coupled amplifiers are regarded as being able to
go down to DC.

In most amps there is the usual two resistor FB R network, say 33 k
and 1 k, with the junction taken to the FB port of the diff pair,
and then between the 1k and 0V there is about 100 uF.

so that at DC, *all* the output voltage is fed back to the FB port.
But the input resistance has an effect on the FB application.

So if you apply +1 volt DC to the input, you will get +1 volt DC at
the output and very close to +1V applied to the FB port.

Agreed.

The rail caps at DC are an open circuit. But the PS rectifiers
create a supply impedance of a finite value. Usually its quite low,
so rails don't move much even when VLF signals are created at the
output.

The above does not seem to explain your previous statement about the
rails having to have an infinitely low impedance for the response of
the amp to extend to d.c. Hence it does not seem to answer the
question I was asking.

So far as I can see, what you have now written above does not require
what you previously claimed.

The power supply seen by the amp at 1 Hz at full power will have a
finite impedance. Its a complex model. Its not as if the impedance is
infinitely small. A 10,000 uF rail cap has 16 ohms of reactance at 1 Hz,
so it may as well not be

present in the PS, except that id smothes the supply of DC which is
pulsed into the cap as the 100 Hz rate....

You don't need me to work out the equivalent circuit for an amp at 1 Hz,
full power, I am quite sure you'll work it out yourself.

What I am still unable to work out, I'm afraid, is how your statements
above show that you previous assertion that "infinitely low impedance of
the PS" is actually required for d.c. output.

I can see that the capacitors used have to be large enough to ensure that
the ripple should be small enough not to bother the amplifier with the
current being drawn. Your comments seem consistent with this, but not with
your assertion that "infinitely low impedance" is required.

If anyone else *does* see how your comments explain this assertion I'd be
interested to know. But at present it looks to me as if your assertion is
simply incorrect, or does not describe what you were actually wishing to
say at the time...

[snip]

With SS amps using such high amounts of NFB, the very tiniest of
parasitic currents can cause thd to leap above 0.005% easily, and one
has to decide whether its worth a week of fiddling around to reduce the
thd to 0.001% at 180 watts.....

This I would certainly agree with. Used to have a pair of plastic pliers to
allow me to wiggle wiring around whilst the amp was doing 200Wpc and watch
the effects on the THD readings. :-) Like youself, I ended up concluding
that was pretty meaningless. However I've seen the same in low feedback
designs, and don't know of any reasons why it would be confined to SS
designs...

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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

In message , Jim Lesurf
writes
In article , Patrick Turner
wrote:


Jim Lesurf wrote:

In article , Patrick Turner
wrote:


The soldered in fuse link seems to be quite blameless, since it is
but a piece of plated copper wire. The distortion measurements
indicate it has no large effects worth a worry.

At what frequency and power did you do the distortion measurements?

1 kHz.

It may be worth your while to also do measurements at 20Hz or use LF/HF
intermod. The thermal effects of fuses on distortion are likely to be
greater at LF than at 1kHz due to the time constants involved.



The paper by Greiner that Arny has pointed out in another thread shows
fuses giving quite high levels of distortions under some quite
plausible conditions.

The thermal dynamic changes and resistance changes would be negligible
with a soldered in peice of copper wire.

I am not at all sure you are correct. The working method of a metal fuse is
that it becomes heated until it ceases to hold together. This means that at
or above the rated current the temperature of the metal has to rise
sufficiently for this to occur. All normal metals have similar behaviour in
that their resistivity rises with temperature. I don't know of any reason
why a soldering in piece of copper wire would not follow the same physical
laws as other types of metal link arranged to fail at a similar current
level.

No idea what the relevant values would be for the fuses you are using, but
unless you do some LF measurements at high currents I would not assume the
effects are "negligable".

As I mentioned in an earlier posting, I was surprised by how large the
effects were that were reported by Greiner. e.g. he quotes 1 percent THD
for 5A fuses with a 4 Ohm load in one set of conditions. Higher values for
other conditions/fuses...

Slainte,

Jim


Many years ago, in my first job at British Aerospace in Bristol, one of
my colleagues has patented a 'water fuse'. Basically just a piece of
fuse wire surrounded by water. The fuse had a much sharper blowing
action than normal, as it was cooled by the water until it had heated up
so much that the water in contact with the wire turned to steam, so the
wire then lost its cooling and blew very quickly.

Never seen the principle used in the real world though.

--
Chris Morriss

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

Jim Lesurf wrote:

In article , Patrick Turner
wrote:

Jim Lesurf wrote:

In article , Patrick Turner
wrote:


The soldered in fuse link seems to be quite blameless, since it is
but a piece of plated copper wire. The distortion measurements
indicate it has no large effects worth a worry.

At what frequency and power did you do the distortion measurements?

1 kHz.

It may be worth your while to also do measurements at 20Hz or use LF/HF
intermod. The thermal effects of fuses on distortion are likely to be
greater at LF than at 1kHz due to the time constants involved.

But I know that gain / feedback quantities are high OK at 20 Hz,
so there is little need for testing at 20 Hz.





The paper by Greiner that Arny has pointed out in another thread shows
fuses giving quite high levels of distortions under some quite
plausible conditions.

The thermal dynamic changes and resistance changes would be negligible
with a soldered in peice of copper wire.

I am not at all sure you are correct. The working method of a metal fuse is
that it becomes heated until it ceases to hold together. This means that at
or above the rated current the temperature of the metal has to rise
sufficiently for this to occur. All normal metals have similar behaviour in
that their resistivity rises with temperature. I don't know of any reason
why a soldering in piece of copper wire would not follow the same physical
laws as other types of metal link arranged to fail at a similar current
level.

Sure the fuse link may get warm, but not much during normal operation.




No idea what the relevant values would be for the fuses you are using, but
unless you do some LF measurements at high currents I would not assume the
effects are "negligable".

I have never had cause to do much about it.



As I mentioned in an earlier posting, I was surprised by how large the
effects were that were reported by Greiner. e.g. he quotes 1 percent THD
for 5A fuses with a 4 Ohm load in one set of conditions. Higher values for
other conditions/fuses...

Often it isn't the fuse wire itself, its the connection that causes the
bothers.

Patrick Turner.



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

Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's
 

On Tue, 07 Dec 2004 20:57:12 +1100, Patrick Turner
wrote:

It may be worth your while to also do measurements at 20Hz or use LF/HF
intermod. The thermal effects of fuses on distortion are likely to be
greater at LF than at 1kHz due to the time constants involved.

But I know that gain / feedback quantities are high OK at 20 Hz,
so there is little need for testing at 20 Hz.

Is the fuse inside the feedback loop then? That would be an unusual
topology since a blown fuse would immediately put the amplifier into
open loop mode, and cause massive signals in the output stage.

d

Pearce Consulting
http://www.pearce.uk.com