AndrewT said:
So, what is so magical about 0.8uS? Why is it better than 1uS, or 5uS? Small timeconstants increase greatly the risk of increased input noise, interference signals getting in, etc. So, if you want to go to (much) higher input bandwidth, you need to have some pretty good reasons to take that extra noise and interefence as a consequence.
Jan Didden
"and the usual range is 0.3uS to 1.5uS fitted to the front end"
Yes, I think this is done to have reasonably flat frequency response up to 20 Khz or so. It's a part of commercial specification.
"Higher sampling freq does NOT produce higher freq music!"
I guess it does a bit ,but better not over do it.
I think another main benefit of the higher sampling rate is in time domain, not only frequency response. Previously high frequency sinewave can only exist in some time slot in the sampling period, now they have more "timing" slot/resolution.
for a given semiconductor technology, stretching the frequency range might degrade other parameter, so this should be taken into consideration.
now let's not threadjack
Yes, I think this is done to have reasonably flat frequency response up to 20 Khz or so. It's a part of commercial specification.
"Higher sampling freq does NOT produce higher freq music!"
I guess it does a bit ,but better not over do it.
I think another main benefit of the higher sampling rate is in time domain, not only frequency response. Previously high frequency sinewave can only exist in some time slot in the sampling period, now they have more "timing" slot/resolution.
for a given semiconductor technology, stretching the frequency range might degrade other parameter, so this should be taken into consideration.
now let's not threadjack
I don't quite see this as thread-jack; as is often the case peripheral matters can influence the main issue substantially in the absence of proper definition. Frequency (audible) influence or not did come up regarding the presence of the series output coil.
That was my intuition when things got bogged down in semantics re "continuous sine waves". It has by now hopefully been properly defined by several previous contributions. There had been several tests way back where sharp filtering >20 KHz resulted in a more acceptable CD rendition, showing that there were artifacts there that could create listener fatigue. Digital devices have hopefully cleaned up their act since then, still.
I would repeat again that the danger of anything above 20 KHz intermodulationg, convoluting, overdriving and generally discombobulating in a region where the amplifier is no longer kosher is to me unacceptable - I chicken out! (Unless anybody has a fixation with designing audio amplifiers with a constant amplitude/zero phase shift characteristic into the MHz region, to boast that their products can handle anything. We already have those; we call them radios. ) Like the question: Why not a phase correction coil in the output circuit, I would also ask: Why not cut above 20 - 25 KHz and get rid of whatever there might be lurching in the harmonic distortion spikes ready to upset matters?
That was my intuition when things got bogged down in semantics re "continuous sine waves". It has by now hopefully been properly defined by several previous contributions. There had been several tests way back where sharp filtering >20 KHz resulted in a more acceptable CD rendition, showing that there were artifacts there that could create listener fatigue. Digital devices have hopefully cleaned up their act since then, still.
I would repeat again that the danger of anything above 20 KHz intermodulationg, convoluting, overdriving and generally discombobulating in a region where the amplifier is no longer kosher is to me unacceptable - I chicken out! (Unless anybody has a fixation with designing audio amplifiers with a constant amplitude/zero phase shift characteristic into the MHz region, to boast that their products can handle anything. We already have those; we call them radios.
) Like the question: Why not a phase correction coil in the output circuit, I would also ask: Why not cut above 20 - 25 KHz and get rid of whatever there might be lurching in the harmonic distortion spikes ready to upset matters?Graham,
Your post #366 (pardon for not quoting all here).
I did try to indicate earlier that I am keeping all of that in mind, and that in revisiting previous design investigation I tried to implement what you posted some time ago in this regard. I did some minor changes in feedback topology as a result of that, after also having implemented the ideas put forward by John Ellis on PLIL a few years ago.
Having done that as best as I know how, I am then careful not to "swipe the dam from under the duck" (not saying that you do). We are still in audio, and I sometimes gets the impression that a lot of effort is going into matters not of prime and proven importance to the pure reproduction of audio signals.
Regards
Your post #366 (pardon for not quoting all here).
I did try to indicate earlier that I am keeping all of that in mind, and that in revisiting previous design investigation I tried to implement what you posted some time ago in this regard. I did some minor changes in feedback topology as a result of that, after also having implemented the ideas put forward by John Ellis on PLIL a few years ago.
Having done that as best as I know how, I am then careful not to "swipe the dam from under the duck" (not saying that you do). We are still in audio, and I sometimes gets the impression that a lot of effort is going into matters not of prime and proven importance to the pure reproduction of audio signals.
Regards
jcx said:
John, napping? Obviously you've never met or spoken with him?
I can attest, not napping...
But we can accuse him of other things...
jcx, I am unaware of any dbt where there are any controls on the equipment that was used to do the presentation of the "stimuli"! Are you?
By this I mean even basic, simple before and after parameter testing and verfication of those things that we can agree are "testable"; THD, IM (vs. SPL) polar response, frequency response(!!), waterfall and related FFT derived system responses, room characteristics, etc. Without those factors being controlled it is difficult to give credence to the dbt(s) applicability beyond the actual test situation. Eh?
This has been a serious concern of mine for a long time - one usually side-stepped and avoided/ignored for reasons that escape me...
Curl is still a troublemaker.
_-_-bear
The thought that I tried to convey to Curl on the phone, and in what i posted (not that I have any special insight or wisdom) is simply that the output coil is generally audible.
I use an output coil, as most of us do, in most of my amps. (My DC coupled SE Mosfet amp (predates Nelson Pass' by a little bit - I think he did a DC coupled one... not sure) had no output inductor. The square waves out of that one could die for... beautiful looking!) But, I found that the 'nature' of the coil - if it is used - makes an audible difference. In other words, all coils are not equal.
This goes along the lines of what some have found with passive high level speaker crossover components - going with litz or ribbon wound coils. Seems they can and may make a difference! Clearly, the lower the distortion of your speaker, the more important this becomes.
Put another way, for most people IT DON'T MATTER AT ALL!!! .
Why it matters seems to have been hit upon in the posts that have been made - I do not think it is any single factor, although one or two may dominate. It would be nice to reduce it to some set of measurable parameters...
In my phone convo with Curl (and you can call me up too, if you want...) we touched upon the history of power amps. Back in the day, high current, high voltage output devices were not so robust, nor were amp circuits all that stable. The prospect of looking into a very low Z at some high freq outside the audio band meant the prospect of the magic smoke emerging from the devices. As we all know.
I'm wondering if todays output stages that use stacks of very high current, high voltage transistors really care at all if they see a very very low Z at high freq, especially given that the signal that is likely to appear there is transient? Of course, there is the case of the large scale high level parasitic being shoved into the front end of the amp... but even then, would it fry?? How about if the load looked scary capacitively reactive? As long as the phase angle didn't shift enough to bring the amp into positive feedback, would it really do anything much other than get hot??
Honestly, I do not know the answer, since I haven't had the chance to attempt a distructive test of a well made, big monster modern power amp under those conditions...
I know that my own BEAR Labs Symphony No.1 has survived operation at high listening levels into a dead short across the load... for a short period of time (a surprising LONG short period of time), and that it is stable into more than 2ufd with anything in the audio band... and maybe a bit out... but I honestly never tried the ULTIMATE full power blast into a big cap at or near 200kHz or so... or if I did, I don't remember, which means it survived ok.
Thinking out loud.
Btw, I am an ignoramus.
Curl is still stirring up trouble...
_-_-bear
I use an output coil, as most of us do, in most of my amps. (My DC coupled SE Mosfet amp (predates Nelson Pass' by a little bit - I think he did a DC coupled one... not sure) had no output inductor. The square waves out of that one could die for... beautiful looking!) But, I found that the 'nature' of the coil - if it is used - makes an audible difference. In other words, all coils are not equal.
This goes along the lines of what some have found with passive high level speaker crossover components - going with litz or ribbon wound coils. Seems they can and may make a difference! Clearly, the lower the distortion of your speaker, the more important this becomes.
Put another way, for most people IT DON'T MATTER AT ALL!!! .
Why it matters seems to have been hit upon in the posts that have been made - I do not think it is any single factor, although one or two may dominate. It would be nice to reduce it to some set of measurable parameters...
In my phone convo with Curl (and you can call me up too, if you want...) we touched upon the history of power amps. Back in the day, high current, high voltage output devices were not so robust, nor were amp circuits all that stable. The prospect of looking into a very low Z at some high freq outside the audio band meant the prospect of the magic smoke emerging from the devices. As we all know.
I'm wondering if todays output stages that use stacks of very high current, high voltage transistors really care at all if they see a very very low Z at high freq, especially given that the signal that is likely to appear there is transient? Of course, there is the case of the large scale high level parasitic being shoved into the front end of the amp... but even then, would it fry?? How about if the load looked scary capacitively reactive? As long as the phase angle didn't shift enough to bring the amp into positive feedback, would it really do anything much other than get hot??
Honestly, I do not know the answer, since I haven't had the chance to attempt a distructive test of a well made, big monster modern power amp under those conditions...
I know that my own BEAR Labs Symphony No.1 has survived operation at high listening levels into a dead short across the load... for a short period of time (a surprising LONG short period of time), and that it is stable into more than 2ufd with anything in the audio band... and maybe a bit out... but I honestly never tried the ULTIMATE full power blast into a big cap at or near 200kHz or so... or if I did, I don't remember, which means it survived ok.
Thinking out loud.
Btw, I am an ignoramus.
Curl is still stirring up trouble...
_-_-bear
Hi John Curl
Perhaps you might have noticed the parallel thread on output coils.
I agree with your comment that ringing is a possible issue, but for capacitive loads. I do not know of conventional speaker crossover networks which place capacitors directly across the load terminals: they usually have damping resistors etc in series or inductors in series or parallel. Is it the case that your audible results were made using an electrostatic loudspeaker? The (simulated) differences for 20 kHz signals for resistive loads and inductors of 3 uH were pretty small. That is not to say zero, but what a real load would do is something else... a bit like your transistor modelling. I'd say that the SPICE level 3 does quite a good job considering all the different mechanisms at work inside the real device.. . but one deficiency in most implementations is the lack of a proper BVceo description. VBIC has superceded it to improve things ... and maybe we need a "speaker model" which is better!
cheers
John
Perhaps you might have noticed the parallel thread on output coils.
I agree with your comment that ringing is a possible issue, but for capacitive loads. I do not know of conventional speaker crossover networks which place capacitors directly across the load terminals: they usually have damping resistors etc in series or inductors in series or parallel. Is it the case that your audible results were made using an electrostatic loudspeaker? The (simulated) differences for 20 kHz signals for resistive loads and inductors of 3 uH were pretty small. That is not to say zero, but what a real load would do is something else... a bit like your transistor modelling. I'd say that the SPICE level 3 does quite a good job considering all the different mechanisms at work inside the real device.. . but one deficiency in most implementations is the lack of a proper BVceo description. VBIC has superceded it to improve things ... and maybe we need a "speaker model" which is better!
cheers
John
bear said:
Bear,
Doug Self has investigated coil matters in one of his installments in EW (and possibly in one of his books but I don't have them here right now) and found out that the orientation of the coil wrt the other parts of the circuit made a big difference - he could measure non-linearities depending on the geometry of the circuit elements and the coil orientation. This is a matter of EM fields from the coil influencing oter parts of the system.
It may well be that what is being sometimes reported as coil audibility could be related to this. It is measureable and explainable - two great advantages of this phenomenon! It also points to the solution - get the layout right.
Jan Didden
Hartono said:
Hartono,
I'm not sure I understand what you want to say here. I'm sure you know that with a 44.1kHz sampling system 20kHz signals can be sampled and reconstituted unchanged, because of the (red-book mandatory) anti-aliasing post filter. So, higher sampling rate is NOT necessary because otherwise we would loose 20kHz signal or something.
I say again: with 44.1kHz sampling you get correct 20kHz signal component reconstitution, even if it is counter-intuitive. We think that having only 2 or 3 samples for a wave cannot exactly reconstuitute the wave, but it is really true nevertheless. The key is the DAC postfiltering.
Jan Didden
bear said:
Fwiw, the converse is true:
It is amazing how many "scientific" and "engineering" types are deaf?
_-_-bear
This is not fair. I won't accept justifications for or against output coils based on wrong technical reasons.
Neither based on audibility as long as no one has been capable of at least describe the audible arctifacts introduced.
Give me rational technical arguments or a hint of audible efects one can hold on to pursue focused research, and I will accept whichever outcome results regarding coils. Obvious as may sound, yet not that frequently occuring.
Rodolfo
Hi Jan,
I'm not saying that 20 Khz can not be reproduced.
I'm saying if for example we take sample 20 Khz sine into digital
with higher sampling rate, we have more "slot" where we can put the data in time.
with 40Khz sampling we only have 2 slots, and occupy 2 sampling period.
at sampling of 160Khz we can have the 20khz signal in slightly different time slot, and occupy 8 sampling period.
I hope this make sense.
I'm not saying that 20 Khz can not be reproduced.
I'm saying if for example we take sample 20 Khz sine into digital
with higher sampling rate, we have more "slot" where we can put the data in time.
with 40Khz sampling we only have 2 slots, and occupy 2 sampling period.
at sampling of 160Khz we can have the 20khz signal in slightly different time slot, and occupy 8 sampling period.
I hope this make sense.
- Status
- Not open for further replies.