| Author |
Message |
Ruud Broens
Guest
|
 Posted:
Sun Dec 19, 2004 8:44 pm Post subject:
Re: NFB loop question |
|
|
"Ian Iveson" <IanIveson.home@blueyonder.co.uk> wrote in message
news:GB6xd.35249$Rf4.5905@fe2.news.blueyonder.co.uk...
: "Ruud Broens" <broensr@wanadoo.nl> wrote
:
: > You still want your driver transfer function to be as close as
: > possible
: > to the power stage transfer function as you can get :)
:
: Quite what does "close" mean here, Ruud? And why?
Good questions here and in other parts of this thread ! :-)
Let's for a moment ignore the actual loading with a loudspeaker
and assume a purely resistive, passive load.
The situation i was referring to:
an inverting driver stage driving an SE power stage.
If you plot a Vout-Vin graph for all frequencies from say 10 Hz
to 50 kHz for the intended input range for the power stage
- and that includes the OT, obviously -
you would have, scaling aside, hopefully some y=x 'like'
function graphs.
the driver should, for each of these, have the inverse shape
and amplitude transfer function, to get a lineair 2 stage amplification.
NFB will now reduce low to very low distortion and lower Ro.
Rudy
: For a pp stage, if the two are equal then the even harmonics are
: cancelled completely. If the characteristics of the valves are
: different, then some even H may remain. Odd H are summed, but let's
: assume that we're not concerned about them at the moment.
:
: Can the part of the transfer function that gives rise to the even
: harmonics be isolated, for the purpose of comparing odd candidates
: for pp operation? Does "close" only concern that part of the
: function?
:
: For cascaded stages, the situation is more complicated, because we
: are multiplying rather than summing. Here, if the functions are
: equal wrt even H, then is it cancelled? I don't think so. Is it
: possible to get closer to cancellation if the functions are
: different wrt even H?
:
: What you really need is the inverse function I suppose...except you
: would have no gain. So part of the function needs inverting, part
: not. Tricky.
:
: Proper wondering, don't know.
:
: cheers, Ian
:
:
:
:
|
|
| Back to top |
|
 |
John Stewart
Guest
|
| Posted:
Mon Dec 20, 2004 8:35 pm Post subject:
Re: NFB loop question |
|
|
Tom Schlangen wrote:
| Quote: | Gentlemen,
consider a simple two staged SE amp, w/o gNFB. As we know, it
might even be desirable to get a certain amount of 2H from the
driver to cancel out some of the 2H produced by the output
stage.
Now consider a global NFB loop (voltage feedback) added
around the stages.
Do we still want the driver to produce a certain amount of 2H,
or should the driver now act as clean as possible?
Tom
--
When in doubt, use brute force.
- Ken Thompson
|
Why sweat it? Others have looked at this possibility in detail.
I've posted at ABSE something that makes a lot of sense by Ivan
Johnston.
Was to have been included in Sound Practices magazine, Issue 17 that
never saw the light of day. Already sent this to Patrick some time ago
but I guess he has lost it. Or perhaps some of his comments are based
on this very good paper.
There are 3 1/3 pages, about 800K.
Sending direct to Patrick & Tom. Hope I got your address correct, Tom.
Cheers, John Stewart
PS- I don't believe in predistorting at all. Many who do
manage to conveniently ignore higher order effects. JLS |
|
| Back to top |
|
|
Ruud Broens
Guest
|
| Posted:
Tue Dec 21, 2004 2:39 am Post subject:
Re: NFB loop question |
|
|
"John Stewart" <jh.stewart@sympatico.ca> wrote in message
news:41C6F15B.6A099B45@sympatico.ca...
: Tom Schlangen wrote:
:
: > Gentlemen,
: >
: > consider a simple two staged SE amp, w/o gNFB. As we know, it
: > might even be desirable to get a certain amount of 2H from the
: > driver to cancel out some of the 2H produced by the output
: > stage.
: >
: > Now consider a global NFB loop (voltage feedback) added
: > around the stages.
: >
: > Do we still want the driver to produce a certain amount of 2H,
: > or should the driver now act as clean as possible?
: >
: > Tom
: >
: > --
: > When in doubt, use brute force.
: > - Ken Thompson
:
: Why sweat it? Others have looked at this possibility in detail.
: I've posted at ABSE something that makes a lot of sense by Ivan
: Johnston.
: Was to have been included in Sound Practices magazine, Issue 17 that
: never saw the light of day. Already sent this to Patrick some time ago
: but I guess he has lost it. Or perhaps some of his comments are based
: on this very good paper.
:
: There are 3 1/3 pages, about 800K.
:
: Sending direct to Patrick & Tom. Hope I got your address correct, Tom.
:
: Cheers, John Stewart
:
: PS- I don't believe in predistorting at all. Many who do
: manage to conveniently ignore higher order effects. JLS
:
Interesting but not convincing:
if you'd have, say a driver with y=sqrt(x) transfer
followed by a power stage with y=x^2
why would there be *any* distortion in the composite transfer function ?
(that it is pretty difficult to actually achieve an exact inverse transfer
f.
is another matter :)
Rudy |
|
| Back to top |
|
|
John Stewart
Guest
|
| Posted:
Tue Dec 21, 2004 2:57 am Post subject:
Re: NFB loop question |
|
|
Ruud Broens wrote:
| Quote: | "John Stewart" <jh.stewart@sympatico.ca> wrote in message
news:41C6F15B.6A099B45@sympatico.ca...
: Tom Schlangen wrote:
:
: > Gentlemen,
:
: > consider a simple two staged SE amp, w/o gNFB. As we know, it
: > might even be desirable to get a certain amount of 2H from the
: > driver to cancel out some of the 2H produced by the output
: > stage.
:
: > Now consider a global NFB loop (voltage feedback) added
: > around the stages.
:
: > Do we still want the driver to produce a certain amount of 2H,
: > or should the driver now act as clean as possible?
:
: > Tom
:
: > --
: > When in doubt, use brute force.
: > - Ken Thompson
:
: Why sweat it? Others have looked at this possibility in detail.
: I've posted at ABSE something that makes a lot of sense by Ivan
: Johnston.
: Was to have been included in Sound Practices magazine, Issue 17 that
: never saw the light of day. Already sent this to Patrick some time ago
: but I guess he has lost it. Or perhaps some of his comments are based
: on this very good paper.
:
: There are 3 1/3 pages, about 800K.
:
: Sending direct to Patrick & Tom. Hope I got your address correct, Tom.
:
: Cheers, John Stewart
:
: PS- I don't believe in predistorting at all. Many who do
: manage to conveniently ignore higher order effects. JLS
:
Interesting but not convincing:
if you'd have, say a driver with y=sqrt(x) transfer
followed by a power stage with y=x^2
why would there be *any* distortion in the composite transfer function ?
(that it is pretty difficult to actually achieve an exact inverse transfer
f.
is another matter :)
Rudy
|
If the transfer equation was simply quadratic then the sun would
shine & the birds would sing. However, it is cubic or worse.
Now all the gremlins show up to feast on the singing birds!!!!!!
Cheers, John |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Tue Dec 21, 2004 7:24 am Post subject:
Re: NFB loop question |
|
|
Ruud Broens wrote:
| Quote: | "John Stewart" <jh.stewart@sympatico.ca> wrote in message
news:41C6F15B.6A099B45@sympatico.ca...
: Tom Schlangen wrote:
:
: > Gentlemen,
:
: > consider a simple two staged SE amp, w/o gNFB. As we know, it
: > might even be desirable to get a certain amount of 2H from the
: > driver to cancel out some of the 2H produced by the output
: > stage.
:
: > Now consider a global NFB loop (voltage feedback) added
: > around the stages.
:
: > Do we still want the driver to produce a certain amount of 2H,
: > or should the driver now act as clean as possible?
:
: > Tom
:
: > --
: > When in doubt, use brute force.
: > - Ken Thompson
:
: Why sweat it? Others have looked at this possibility in detail.
: I've posted at ABSE something that makes a lot of sense by Ivan
: Johnston.
: Was to have been included in Sound Practices magazine, Issue 17 that
: never saw the light of day. Already sent this to Patrick some time ago
: but I guess he has lost it. Or perhaps some of his comments are based
: on this very good paper.
:
: There are 3 1/3 pages, about 800K.
:
: Sending direct to Patrick & Tom. Hope I got your address correct, Tom.
:
: Cheers, John Stewart
:
: PS- I don't believe in predistorting at all. Many who do
: manage to conveniently ignore higher order effects. JLS
:
Interesting but not convincing:
if you'd have, say a driver with y=sqrt(x) transfer
followed by a power stage with y=x^2
why would there be *any* distortion in the composite transfer function ?
(that it is pretty difficult to actually achieve an exact inverse transfer
f.
is another matter :)
Rudy
|
But I think a truth escapes you.
regardless of whether the the two series stages have equal transfer curves,
the
thd of the first will always be distorted by the second, so there is
distortion of the distortion,
and lots of IMD products, and the clever amoung you can work it all out
mathematically if you want to.
Its impossible to have a pair of distortionless stages.
The article JS posted shows the IMD THD spectra to be more complex
where there is a deliberate attempt to make V1 distort more than usual so
2H cancels that of the output V2, which always naturally produces much more
thd than V1.
So don't do this, simple.
Its better to linerarize the output stage with local NFB to get thd low so the
natural thd of the driver
will do some useful cancelling of the even order harmonics.
In thwe article tenedered by JS the majority of what extra complexity exists
in the THD spectra appears to be below
the -80 dB levels, or less than 0.01%, and les than the noise floor.
The article says this extra harmonic "dirt" at such a low level has a poor
subjective effect
on the percieved sound.
Maybe it would, if it was at all audible.
0.01% of 2.83 volts at a speaker to make 1 watt is 0.283 mV.
The speakers I have sound quite silent at 3 metres away with 0.283mV of
signal of any kind.
So a linear input and driver stage will always produce a simpler mix of
harmonics,
at the output of the output tube but there will be a larger quantity of them
without any cancelling.
Do two wrongs make a right?
It depends, it does.....
Patrick Turner. |
|
| Back to top |
|
|
Ian Iveson
Guest
|
| Posted:
Tue Dec 21, 2004 8:33 am Post subject:
Re: NFB loop question |
|
|
"Ruud Broens" <broensr@wanadoo.nl> wrote
| Quote: | Interesting but not convincing:
if you'd have, say a driver with y=sqrt(x) transfer
followed by a power stage with y=x^2
why would there be *any* distortion in the composite transfer
function ?
(that it is pretty difficult to actually achieve an exact inverse
transfer
f.
is another matter :)
|
OK if you want a gain of 1.
Not at all obvious to the Fourier-fixated, either. Someone might
argue that the harmonics in the driver would be distorted by the
power stage, leaving residual distortion.
In your example, every term in the driver series would result in a
set of terms in the power stage, such that the cancellation would
ripple down the whole series, leaving an infinitessimally small
remainder of an infinitely high harmonic.
Cancelled is the wrong term to use BTW, because it suggests
subtraction rather than the actual multiplication, but hopefully you
can follow my drift.
The chances of getting all that right with real valves is remote, I
would guess :(
Hence Turner Misapprehension #46 will appear true for real valves,
and ppl who prefer one mix of distortion to another will continue to
tune cascaded stages for the sound they like.
I have come across ppl moaning about harsh sound from feedback, and
even from PP, but not about this kind of thing. I wonder why not?
cheers, Ian |
|
| Back to top |
|
|
Stewart Pinkerton
Guest
|
| Posted:
Tue Dec 21, 2004 5:58 pm Post subject:
Re: NFB loop question |
|
|
On Tue, 21 Dec 2004 15:11:07 GMT, "Ian Iveson"
<IanIveson.home@blueyonder.co.uk> wrote:
| Quote: | "Ruud Broens" <broensr@wanadoo.nl> wrote
Of course it's difficult to actually implement, no doubts there :)
but the point is rather, don't a priori close an avenue of
exploration.
BTW what is interesting in the article John posted, something to
think about for the 'EE' minds, our cognitive abilities DO
include
perception below noise-floor...
|
The EEs already know this. That's one reason why dithering is an
*essential part* of any digital system - it retains signal detail well
below the wideband noise floor.
| Quote: | We are much more sensitive to coherent sound. A fixed tone such as a
PS harmonic can be heard at lower levels than noise. We also pick
out what our minds are alert to. Distortion is coherent, and some
forms may be alarming.
|
Indeed so, and the established limit for this is some 15-20dB below
the wideband noise floor (it has to do with the bandwidth of our
multiple aural receptors, much like the 'bins' in a FFT analyser).
Also, the absolute threshold of hearing is well established at 0dB SPL
(indeed, that's *why* it's 0dB!). If we have a system which plays at
an average level of say 85-90dB SPL (pretty darned loud!), and we have
a noise floor 80dB below this, with no distortion artifacts above the
noise floor (any decent 60-watt SS amp, and a few good valve ones), it
follows that there will be *zero* audible distortion from the
amplifier at normal replay levels.
It also follows that whatever the SET boys are hearing, it has nothing
to do with a 'cleaner' signal......................
--
Stewart Pinkerton | Music is Art - Audio is Engineering |
|
| Back to top |
|
|
Ruud Broens
Guest
|
| Posted:
Tue Dec 21, 2004 7:25 pm Post subject:
Re: NFB loop question |
|
|
"Ian Iveson" <IanIveson.home@blueyonder.co.uk> wrote in message
news:2MNxd.2773$DF3.1627@fe3.news.blueyonder.co.uk...
: "Ruud Broens" <broensr@wanadoo.nl> wrote
:
: > Interesting but not convincing:
: > if you'd have, say a driver with y=sqrt(x) transfer
: > followed by a power stage with y=x^2
: > why would there be *any* distortion in the composite transfer
: > function ?
: > (that it is pretty difficult to actually achieve an exact inverse
: > transfer
: > f.
: > is another matter :)
:
: OK if you want a gain of 1.
:
: Not at all obvious to the Fourier-fixated, either. Someone might
: argue that the harmonics in the driver would be distorted by the
: power stage, leaving residual distortion.
:
: In your example, every term in the driver series would result in a
: set of terms in the power stage, such that the cancellation would
: ripple down the whole series, leaving an infinitessimally small
: remainder of an infinitely high harmonic.
:
: Cancelled is the wrong term to use BTW, because it suggests
: subtraction rather than the actual multiplication, but hopefully you
: can follow my drift.
:
: The chances of getting all that right with real valves is remote, I
: would guess :(
:
: Hence Turner Misapprehension #46 will appear true for real valves,
: and ppl who prefer one mix of distortion to another will continue to
: tune cascaded stages for the sound they like.
:
: I have come across ppl moaning about harsh sound from feedback, and
: even from PP, but not about this kind of thing. I wonder why not?
:
: cheers, Ian
:
Of course it's difficult to actually implement, no doubts there :)
but the point is rather, don't a priori close an avenue of exploration.
BTW what is interesting in the article John posted, something to
think about for the 'EE' minds, our cognitive abilities DO include
perception below noise-floor...
Rudy |
|
| Back to top |
|
|
Ian Iveson
Guest
|
| Posted:
Tue Dec 21, 2004 8:11 pm Post subject:
Re: NFB loop question |
|
|
"Ruud Broens" <broensr@wanadoo.nl> wrote
| Quote: | Of course it's difficult to actually implement, no doubts there :)
but the point is rather, don't a priori close an avenue of
exploration.
BTW what is interesting in the article John posted, something to
think about for the 'EE' minds, our cognitive abilities DO
include
perception below noise-floor...
Rudy
We are much more sensitive to coherent sound. A fixed tone such as a |
PS harmonic can be heard at lower levels than noise. We also pick
out what our minds are alert to. Distortion is coherent, and some
forms may be alarming.
Just as there are many well-known visual illusions (which line is
longest, are these lines parallel, how many prongs has this fork
got, candlestick or two faces, etc), our ears can also be fooled.
In some ways I think audio systems are less about fidelity to
signal, and more about preserving the illusion. The trick of
"cancellation" by successive stages may not work very well, but the
strategy may still compare well soundwise with other methods of
reducing distortion. There could be all sorts of reasons.
cheers, Ian |
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Tue Dec 21, 2004 8:25 pm Post subject:
Re: NFB loop question |
|
|
Ian Iveson wrote:
| Quote: | "Ruud Broens" <broensr@wanadoo.nl> wrote
Interesting but not convincing:
if you'd have, say a driver with y=sqrt(x) transfer
followed by a power stage with y=x^2
why would there be *any* distortion in the composite transfer
function ?
(that it is pretty difficult to actually achieve an exact inverse
transfer
f.
is another matter :)
OK if you want a gain of 1.
Not at all obvious to the Fourier-fixated, either. Someone might
argue that the harmonics in the driver would be distorted by the
power stage, leaving residual distortion.
In your example, every term in the driver series would result in a
set of terms in the power stage, such that the cancellation would
ripple down the whole series, leaving an infinitessimally small
remainder of an infinitely high harmonic.
Cancelled is the wrong term to use BTW, because it suggests
subtraction rather than the actual multiplication, but hopefully you
can follow my drift.
The chances of getting all that right with real valves is remote, I
would guess :(
Hence Turner Misapprehension #46 will appear true for real valves,
and ppl who prefer one mix of distortion to another will continue to
tune cascaded stages for the sound they like.
I have come across ppl moaning about harsh sound from feedback, and
even from PP, but not about this kind of thing. I wonder why not?
|
I don't know.
If you have a 3 stage SE amp like the SEUL amps with 13E1 where I use
screen FB from the ande circuit to linearize the appalling SE tetrode
transfer curve,
and then apply 16 dB of global NFB, then one still ends up with
0.5% of 2H at 20 watts and a whole potpourrie of
lesser amplitude harmonics which are so small as not to change the
thd measurement much from the 0.5%.
PP class AB amps have the same low level complex thd close to clipping
except that
the dominant H is 3H, and maybe 0.2%, with perhaps up to 0.2% of 2H
and then all the rest at low levels.
At 2 watts from either 20 watt amp the thd/imd is not a big problem.
In my latest SE35 watt amps the thd of the output stage is reduced to
under
2% mainly 2H at near clipping with the CFB windings, and then the CFB
makes the output stage less sensitive
than UL or tetrode mode, so a largish 46vrms drive is needed, and 12BH7
was set up with no attempt to load it adversly with a low load value or
bias it
adversely to enchorage 2H production, and along with it other harmonics
that spoil the outcome,
and the result is that the rate of increase of 2H and the amount of 2H
produced
about matches that of the output stage, thus giving a huge reduction in
2H
at the output into 5 ohms.
When viewed on the CRO, the thd is mainly 3H and other products
which are normally not easily visible when mixed with a large amount of
2H
which is inherent in most SE amps.
Thus I get under 0.1% thd at 10 watts into any load between 4 and 12
ohms,
with declining THD below this power as one approaches 0.0 watts.
I did have a client with a cheap asian made SE amp with a lone EL34 with
a load of
2.5k and Ea of +450v, and only 12 dB of global NFB.
Its wasn't a good performer, and the owner said it was no improvement
over his crummy old SS amp.
It had that sand papery sound of poor tube sound, not as top end harsh
as bad SS,
but not good.
The loading was wrong, the Ea and Ia were wrong, the OPT like a toy,
and the FB too small, and so not every SE amp is all that wonderful.
The measurements were dreadful.
So a bit of 2H voltage cancelling intelligently implemented
isn't too bad in a basically well concieved amp.
And when its all said and done, a well engineered SE triode
class A amp with or without NFB can be a delight to the ears
with or without cancelling if the power ceiling is adequate.
Its hard to make a bad sounding triode amp.
So if one cannot manage a lot of 2H cancelling between the driver and
output,
then its certainly isn't a disaster for the music.
Patrick Turner.
|
|
| Back to top |
|
|
Patrick Turner
Guest
|
| Posted:
Tue Dec 21, 2004 8:41 pm Post subject:
Re: NFB loop question |
|
|
Ruud Broens wrote:
| Quote: | "Ian Iveson" <IanIveson.home@blueyonder.co.uk> wrote in message
news:2MNxd.2773$DF3.1627@fe3.news.blueyonder.co.uk...
: "Ruud Broens" <broensr@wanadoo.nl> wrote
:
: > Interesting but not convincing:
: > if you'd have, say a driver with y=sqrt(x) transfer
: > followed by a power stage with y=x^2
: > why would there be *any* distortion in the composite transfer
: > function ?
: > (that it is pretty difficult to actually achieve an exact inverse
: > transfer
: > f.
: > is another matter :)
:
: OK if you want a gain of 1.
:
: Not at all obvious to the Fourier-fixated, either. Someone might
: argue that the harmonics in the driver would be distorted by the
: power stage, leaving residual distortion.
:
: In your example, every term in the driver series would result in a
: set of terms in the power stage, such that the cancellation would
: ripple down the whole series, leaving an infinitessimally small
: remainder of an infinitely high harmonic.
:
: Cancelled is the wrong term to use BTW, because it suggests
: subtraction rather than the actual multiplication, but hopefully you
: can follow my drift.
:
: The chances of getting all that right with real valves is remote, I
: would guess :(
:
: Hence Turner Misapprehension #46 will appear true for real valves,
: and ppl who prefer one mix of distortion to another will continue to
: tune cascaded stages for the sound they like.
:
: I have come across ppl moaning about harsh sound from feedback, and
: even from PP, but not about this kind of thing. I wonder why not?
:
: cheers, Ian
:
Of course it's difficult to actually implement, no doubts there :)
but the point is rather, don't a priori close an avenue of exploration.
BTW what is interesting in the article John posted, something to
think about for the 'EE' minds, our cognitive abilities DO include
perception below noise-floor...
Rudy
|
There are those who would maintain that SS amps which might measure at
0.005% thd at 10 watts, mostly 3H, with all other artifacts at
below 0.001% as being somewhat tiring, lifeless, or harsh to listen to.
Similarly another crew, or even the same ppl would condemn any practice
of trying to rid an amp of its even order harmonics by way of **voltage**
cancelling
as I have spelled out in numerous posts, or by **current** cancelling as
spelled out
by everyone since 1920 with regard to PP amps.
I listen to all these claims and opinions, and yes, its all very well for
everyone to have such opinions.
But I don't have any power of judgement over whatever folks think,
and I cannot verify other folks' perceptions, because I have not their ears,
nor
their brains, and amidst such good music most of the time I have no time for
the
ABX comparisons; its all mostly a fine vintage.
Its my own view that SE amps don't suffer from being constructed
to give low thd, or thd as low as a decent PP amp, and with as low an amount
of
NFB as possible and still get a damping factor of at least 8.
I very much like class A PP amps, as well as SE amps which are always class
A.
Patrick Turner. |
|
| Back to top |
|
|
Ruud Broens
Guest
|
| Posted:
Tue Dec 21, 2004 11:07 pm Post subject:
Re: NFB loop question |
|
|
"Stewart Pinkerton" <patent3@dircon.co.uk> wrote in message
news:i9ogs0hml3spoccf156qc705hc0mk0sh0u@4ax.com...
: Indeed so, and the established limit for this is some 15-20dB below
: the wideband noise floor (it has to do with the bandwidth of our
: multiple aural receptors, much like the 'bins' in a FFT analyser).
: Also, the absolute threshold of hearing is well established at 0dB SPL
: (indeed, that's *why* it's 0dB!). If we have a system which plays at
: an average level of say 85-90dB SPL (pretty darned loud!), and we have
: a noise floor 80dB below this, with no distortion artifacts above the
: noise floor (any decent 60-watt SS amp, and a few good valve ones), it
: follows that there will be *zero* audible distortion from the
: amplifier at normal replay levels.
:
: It also follows that whatever the SET boys are hearing, it has nothing
: to do with a 'cleaner' signal......................
: --
:
: Stewart Pinkerton | Music is Art - Audio is Engineering
Of course, threshold=threshold = minimal energy required.
But 90 db SPL being loud, well, that's subjective, isn't it ? :)
(You wouldn't believe the SPL's i've come across
at certain parties, hehe:)
Rudy |
|
| Back to top |
|
|
Tim Williams
Guest
|
| Posted:
Wed Dec 22, 2004 12:48 am Post subject:
Re: NFB loop question |
|
|
"Ian Iveson" <IanIveson.home@blueyonder.co.uk> wrote in message
news:f2Xxd.4887$DF3.1342@fe3.news.blueyonder.co.uk...
| Quote: | In some ways I think audio systems are less about fidelity to
signal, and more about preserving the illusion.
|
Hence why Mr. Bose is doing so well.
Tim (ducking and running)
--
"I've got more trophies than Wayne Gretsky and the Pope combined!"
- Homer Simpson
Website @ http://webpages.charter.net/dawill/tmoranwms |
|
| Back to top |
|
|
Ruud Broens
Guest
|
| Posted:
Fri Dec 24, 2004 8:23 pm Post subject:
Re: NFB loop question |
|
|
"Stewart Pinkerton" <patent3@dircon.co.uk> wrote in message
news:i9ogs0hml3spoccf156qc705hc0mk0sh0u@4ax.com...
: The EEs already know this. That's one reason why dithering is an
: *essential part* of any digital system - it retains signal detail well
: below the wideband noise floor.
:
: >We are much more sensitive to coherent sound. A fixed tone such as a
: >PS harmonic can be heard at lower levels than noise. We also pick
: >out what our minds are alert to. Distortion is coherent, and some
: >forms may be alarming.
:
: Indeed so, and the established limit for this is some 15-20dB below
: the wideband noise floor (it has to do with the bandwidth of our
: multiple aural receptors, much like the 'bins' in a FFT analyser).
: Also, the absolute threshold of hearing is well established at 0dB SPL
: (indeed, that's *why* it's 0dB!). If we have a system which plays at
: an average level of say 85-90dB SPL (pretty darned loud!), and we have
: a noise floor 80dB below this, with no distortion artifacts above the
: noise floor (any decent 60-watt SS amp, and a few good valve ones), it
: follows that there will be *zero* audible distortion from the
: amplifier at normal replay levels.
:
: It also follows that whatever the SET boys are hearing, it has nothing
: to do with a 'cleaner' signal......................
: --
:
: Stewart Pinkerton | Music is Art - Audio is Engineering
Hm, not convinced, i'm afraid :)
why would a truely down to the lsb linear DA + amp
pumping out 20 or more bit wide @ 192 kHz
*need* dithering, Stewart ?
happy travels,
Rudy
Oh, btw,
even those signals that -by themselves- are below the
energy level required for detection
*can* be perceived.
How ? |
|
| Back to top |
|
|
Stewart Pinkerton
Guest
|
| Posted:
Tue Dec 28, 2004 6:30 pm Post subject:
Re: NFB loop question |
|
|
On Fri, 24 Dec 2004 16:23:16 +0100, "Ruud Broens" <broensr@wanadoo.nl>
wrote:
| Quote: | "Stewart Pinkerton" <patent3@dircon.co.uk> wrote in message
news:i9ogs0hml3spoccf156qc705hc0mk0sh0u@4ax.com...
: The EEs already know this. That's one reason why dithering is an
: *essential part* of any digital system - it retains signal detail well
: below the wideband noise floor.
:
: >We are much more sensitive to coherent sound. A fixed tone such as a
: >PS harmonic can be heard at lower levels than noise. We also pick
: >out what our minds are alert to. Distortion is coherent, and some
: >forms may be alarming.
:
: Indeed so, and the established limit for this is some 15-20dB below
: the wideband noise floor (it has to do with the bandwidth of our
: multiple aural receptors, much like the 'bins' in a FFT analyser).
: Also, the absolute threshold of hearing is well established at 0dB SPL
: (indeed, that's *why* it's 0dB!). If we have a system which plays at
: an average level of say 85-90dB SPL (pretty darned loud!), and we have
: a noise floor 80dB below this, with no distortion artifacts above the
: noise floor (any decent 60-watt SS amp, and a few good valve ones), it
: follows that there will be *zero* audible distortion from the
: amplifier at normal replay levels.
:
: It also follows that whatever the SET boys are hearing, it has nothing
: to do with a 'cleaner' signal......................
Hm, not convinced, i'm afraid :)
|
No problem, the truth will remain, regardless of your beliefs! :-)
| Quote: | why would a truely down to the lsb linear DA + amp
pumping out 20 or more bit wide @ 192 kHz
*need* dithering, Stewart ?
|
It wouldn't, because DACs don't use dither.
*All* ADC systems need dithering, otherwise they are broken. The
sampling rate is irrelevant, and one could certainly argue that a true
20-bit system is *self* dithering due to analogue noise levels, but
that's a different matter.
| Quote: | Oh, btw,
even those signals that -by themselves- are below the
energy level required for detection
*can* be perceived.
How ?
|
Because they are only below the *wideband* noise floor. So long as
they are above the narrow-band noise floor that equates to the comb
receptor bandwidth of the frequency to which we're listening, we can
hear them. We are also able to dig a little deeper when the signal is
periodic, as we can average out the noise and pick up a signal even a
little below this narrow-band noise floor. Old-time Morse Code radio
operators were reputed to be the best at this, along with submarine
'phones'.
--
Stewart Pinkerton | Music is Art - Audio is Engineering |
|
| Back to top |
|
|
|
|
|
|
Profile
Log in to check your private messages
Log in