Hi,
Well, well well. Another round of Mythbusters? Well, actually, I don't really care, so I won't bother to cover this in detail, honest.
Really, so you mean using two DAC's in parallel reduces SNR by 3dB?
Are you sure you really understand how the PCM1704 works?
Really, I guess it is time that I take Bob Adam's advice to Scott Wurcer...
You don't? As it so happens I see one differential current output DAC for left channel and another for the right channel. Makes two DAC's in my milmaid 1 + 1 = 2 calculation, but WTFDIK.
More to the point, I even know quite a few examples.
It highly depends on the DAC used and many other factors. But yes, you can do that and I have documented doing that for Audio DAC's over a decade ago.
Ciao T
Well, well well. Another round of Mythbusters? Well, actually, I don't really care, so I won't bother to cover this in detail, honest.
Really, so you mean using two DAC's in parallel reduces SNR by 3dB?
Are you sure you really understand how the PCM1704 works?
Really, I guess it is time that I take Bob Adam's advice to Scott Wurcer...
You don't? As it so happens I see one differential current output DAC for left channel and another for the right channel. Makes two DAC's in my milmaid 1 + 1 = 2 calculation, but WTFDIK.
More to the point, I even know quite a few examples.
It highly depends on the DAC used and many other factors. But yes, you can do that and I have documented doing that for Audio DAC's over a decade ago.
Ciao T
Hi,
Lovely. I knew I could not be the only ho could see the potential in this principle...
So, maybe I'll try it.
Edit, sanken still make it...
http://www.sanken-mic.com/en/product/product.cfm/3.1000700
Ciao T
Lovely. I knew I could not be the only ho could see the potential in this principle...
So, maybe I'll try it.
Edit, sanken still make it...
http://www.sanken-mic.com/en/product/product.cfm/3.1000700
Ciao T
Last edited:
You started by claiming that two analog signals in parallel (balanced) reduces S/R by 3 dB. Now you doubt my claim that two DACs in parallel would suffer the same 3 dB S/R reduction? Is there some magic that spares us from the usual rules, or are you claiming that the DACs are noise free, or are you claiming that their noise is perfectly matched and therefore cancels out?
I only know what the data sheet documents. They mention "dual balanced current segments" and "away from zero with small steps in both directions," both of which seem to describe a balanced pair of DACs. Also, the block diagram on page one clearly shows the current output from each DAC being combined before it is presented on a single output pin. What is wrong with my suggestion that each DAC be brought out to a separate pin where they can be combined or not by the circuit designer?
Are you sure you really understand how the PCM1704 works? It would be vastly more helpful if you tried to explain what you think you know, rather than tossing out nonproductive challenges.
I'm talking about "per channel" - of course a stereo DAC has 2 independent DACs internally, but they're each working on a separate channel. My apologies for not holding your hand through all of this discussion, but you first act like you don't need any help understanding this sort of thing and then you act confused by a simple omission that should be plainly obvious in the context of the conversation.
Are you trying to dither this thread by injecting "fuzzy" noise?
Hi,
I did not claim anything. I merely observed the facts. If your electronic knowledge is so limited that you cannot handle basic noise calculations I am happy to demnsotrate for you how it is done.
Of course I do, since it runs counter to the laws of physics as they are currently understood. If you have new ones maybe you publish in "Nature", rather than here?
But this magic seems to apply where you are, so instead of suffering the unavoidable 3dB noise penalty (based on the "all else being equal)" dictum of course) thay have a 6dB noise advantage and paralleling two DAC's decreases
SNR by 3dB, instead of increasing it, as the laws of physics demand.
Quite frankly these are some of the most bizarre claims in a long time and so far off base, I'm gobsmacked...
Not that any of this particulary matters in the case of the PCM1704.
This sufficies, but I do happen to know a bit more.
Yes.
First, you are the ones who knows everything better anyway, so you explain it, secondly, if you need EE remedials, my hourly rate is a reasonable 50 Euro.
As the PCM1704 is a single channel DAC and you did not elaborate your application (which could have been non-audio) you may have found the second channel redundant and hence the Arda Tech DAC not suited for your project on BOM cost. That was all.
Well, let's see? Who is making claims that have no basis in physics as they are currently understood and constantly uses counter factuals in his arguments here?
So I really think I am best off with taking Bob Adam's advise to Scott Wurcer for myself as well.
Qui ignorabat, ignorabitur
Ciao T
I did not claim anything. I merely observed the facts. If your electronic knowledge is so limited that you cannot handle basic noise calculations I am happy to demnsotrate for you how it is done.
Of course I do, since it runs counter to the laws of physics as they are currently understood. If you have new ones maybe you publish in "Nature", rather than here?
But this magic seems to apply where you are, so instead of suffering the unavoidable 3dB noise penalty (based on the "all else being equal)" dictum of course) thay have a 6dB noise advantage and paralleling two DAC's decreases
SNR by 3dB, instead of increasing it, as the laws of physics demand.
Quite frankly these are some of the most bizarre claims in a long time and so far off base, I'm gobsmacked...
Not that any of this particulary matters in the case of the PCM1704.
This sufficies, but I do happen to know a bit more.
Yes.
First, you are the ones who knows everything better anyway, so you explain it, secondly, if you need EE remedials, my hourly rate is a reasonable 50 Euro.
As the PCM1704 is a single channel DAC and you did not elaborate your application (which could have been non-audio) you may have found the second channel redundant and hence the Arda Tech DAC not suited for your project on BOM cost. That was all.
Well, let's see? Who is making claims that have no basis in physics as they are currently understood and constantly uses counter factuals in his arguments here?
So I really think I am best off with taking Bob Adam's advise to Scott Wurcer for myself as well.
Qui ignorabat, ignorabitur
Ciao T
Last edited:
Hi John,
With age comes wisdom.
I see you posses much of it and are very tolerant to others flaws, a virtue I am rather lacking, I must admit.
I, meanwhile, feel the getting older, but wisdom escapes me.
Plus, a severe failing of mine that I am keenly aware of but have yet to change, I have a low tolerance of those who criticise my knowledge and understanding of advanced electronic concepts (quite rudely at that) and then reveal themselves to be sorely lacking in very basic electronic knowledge (never mind more advanced concepts), plus I am rather allergic to Merda Taurorum.
I was actually very tempted to take recourse to more robust anglo-saxon vocabulary here, but I shall contain myself to avoid winning another all expenses paid holiday in Sin-Bin-City, where the grass ain't green and the girls ain't pretty.
So, being the wise and tolerant person you are John, I am sure you can forgive my foibles and youthful, if unwise exuberance.
Caelum, non animum, mutant, qui trans mare currunt.
Ciao T
With age comes wisdom.
I see you posses much of it and are very tolerant to others flaws, a virtue I am rather lacking, I must admit.
I, meanwhile, feel the getting older, but wisdom escapes me.
Plus, a severe failing of mine that I am keenly aware of but have yet to change, I have a low tolerance of those who criticise my knowledge and understanding of advanced electronic concepts (quite rudely at that) and then reveal themselves to be sorely lacking in very basic electronic knowledge (never mind more advanced concepts), plus I am rather allergic to Merda Taurorum.
I was actually very tempted to take recourse to more robust anglo-saxon vocabulary here, but I shall contain myself to avoid winning another all expenses paid holiday in Sin-Bin-City, where the grass ain't green and the girls ain't pretty.
So, being the wise and tolerant person you are John, I am sure you can forgive my foibles and youthful, if unwise exuberance.
Caelum, non animum, mutant, qui trans mare currunt.
Ciao T
Thorsten, I just want to keep you contributing. You have a much more detailed knowledge of digital and its weaknesses. I want to learn more, not get the debate left one sided.
I have known about digital's 'weaknesses' just from listening tests, first starting in 1968.
I was ever hopeful that it would improve in the early 70's with delay lines that we used for large concerts, but it didn't. Then a Philips Research Lab demo in 1974, with 14 bits and a 50K sample rate was pretty good, and I hoped for the future. In 1977 or so, Ampex started to make delay lines for analog record preview. We had a terrific fight over 50K vs 100K clock rates, and they settled on 50K, although we could hear the difference. Then Stockham came out with his digital recorder, and we competed head to head with it for the final mixdown of the Tusk recording. Stockham won out over me, but I then predicted that from what I heard, that this recording would fail in the marketplace, because of the digital addition. (The analog masters were just fine, I heard them repeatedly, myself in the studio in 1979.)
Digital, for me, has always been problematic, sonically. Perhaps, I just seem to be extra sensitive to it, intrinsically. So, now that 30+ years have passed, I still have problems with digital, both live vs recorded (as we do every year at CES), standard CD's, digital radio, TV, etc. I do feel that a basic 'rethink' of digital may be in order, rather than relying on today's specs and 'rationalizations' so that hi end digital can become COMPLETELY transparent.
I have known about digital's 'weaknesses' just from listening tests, first starting in 1968.
I was ever hopeful that it would improve in the early 70's with delay lines that we used for large concerts, but it didn't. Then a Philips Research Lab demo in 1974, with 14 bits and a 50K sample rate was pretty good, and I hoped for the future. In 1977 or so, Ampex started to make delay lines for analog record preview. We had a terrific fight over 50K vs 100K clock rates, and they settled on 50K, although we could hear the difference. Then Stockham came out with his digital recorder, and we competed head to head with it for the final mixdown of the Tusk recording. Stockham won out over me, but I then predicted that from what I heard, that this recording would fail in the marketplace, because of the digital addition. (The analog masters were just fine, I heard them repeatedly, myself in the studio in 1979.)
Digital, for me, has always been problematic, sonically. Perhaps, I just seem to be extra sensitive to it, intrinsically. So, now that 30+ years have passed, I still have problems with digital, both live vs recorded (as we do every year at CES), standard CD's, digital radio, TV, etc. I do feel that a basic 'rethink' of digital may be in order, rather than relying on today's specs and 'rationalizations' so that hi end digital can become COMPLETELY transparent.
Last edited:
I would like to see similar measurement for S-D A/D converter. I have never seen something similar, maybe I am just only not aware of it and there might be many who have already done it. Anyway, it is not as simple to make such measurement as to type on PC keyboard in a high-end pub.
I am with Thorsten here - not so much interested in long-run averaged FFT captured on periodic signals. Once again, rising transients are very, probably most important for human hearing and such signals should be investigated. Music is in fact a stream of transients, rather than 1 minute lasting 1 or 2 sinuses, that may be easily averaged.
Joachim,
This seems less of a "2-Way" mike and more of a "Multi-Mike", as it combines an Omni with a figure 8 and lets you mix the signals yourself to blend the directivity as you like.
Still quite interesting though. Not sure I quite would have selected this precise combo...
Ciao T
This seems less of a "2-Way" mike and more of a "Multi-Mike", as it combines an Omni with a figure 8 and lets you mix the signals yourself to blend the directivity as you like.
Still quite interesting though. Not sure I quite would have selected this precise combo...
Ciao T
Any signal path has some amount of noise along the way. If you mix two basically identical signal paths, then the noise will increase. If the noise is Gaussian, then the increase in noise by combining two signal paths will be 3 dB. This ignores the signal content, though.
With balanced signal paths you have two basically identical signals of opposite polarity. When combined, one signal is inverted such that the sum is double the original signal. That gives the signal a boost of 6 dB in a balanced system. It also happens to cancel out any common mode noise. The cancellation of common mode noise is not limited to 6 dB, in fact it can be an almost infinite improvement.
You have to look at both signal and noise. If the signal is increased 6 dB by sending it twice, but the noise is only increased by 3 dB, then the net gain is 3 dB better S/N. If you add common mode noise to the consideration, then S/N can be improved by much more than 3 dB, but only if you don't get the topology wrong. Unfortunately, the non-common Gaussian noise from each signal path does not cancel just because one signal path is inverted, but fortunately the combination of two Gaussian sources is not 6 dB but only 3 dB, no matter the polarity.
If there is a flaw in any of the above, I would appreciate someone actually spending a sentence or two explaining what's wrong. I have no pride, I don't mind being wrong. I'm here to learn, so if I'm missing something then I can only gain by being corrected. I'm not asking for a remedial course, because I think I provided that above. If I made a mistake, then kindly take the time to provide at least one detail or two that can be confirmed. I regret that I may be feeding the troll, here, but perhaps a discussion based on communication instead of posturing will benefit someone.
If you want to claim that balanced suffers 3 dB worse noise, then show how that is not overwhelmed by the 6 dB increase in signal, not to mention the cancellation of common mode noise.
If you want to claim that a pair of DACs do not follow the same math and physics, then please show how. Why is the Gaussian noise from each DAC not combined to be 3 dB more than a single DAC?
My understanding would be that a pair of DACs converting the same data would combine (mix) to a 6 dB greater signal, but any Gaussian noise produced by these DACs would only increase by 3 dB.
Basically, "a pair of DACs" or "a balanced system" has exactly the same math going on when it comes to signals and Gaussian noise. Note that I said basically, not perfectly identical.
The distinction between the two is that parallel DACs cannot cancel common mode noise unless you digitally invert the data going into one DAC and invert the output before combining them. That's certainly one valid topology. But the hacker topology of soldering one DAC chip on top of another will not accomplish this. In fact, any common mode noise in that topology, i.e., any noise that is generated similarly in both identical DACs, would be 6 dB higher when mixed. Thus, some aspects of noise are not improved by simple doubling of DACs, but rather get worse.
As for the PCM1704, I explained that the data sheet uses phrasing like "two DACs are combined in a complementary arrangement," "utilizes dual balanced current segments," and "steps away from zero with small steps in both directions," which all strongly imply to me that the whole thing is balanced. If anyone has information to the contrary, then please share that information. Real information, please, not more posturing.
With balanced signal paths you have two basically identical signals of opposite polarity. When combined, one signal is inverted such that the sum is double the original signal. That gives the signal a boost of 6 dB in a balanced system. It also happens to cancel out any common mode noise. The cancellation of common mode noise is not limited to 6 dB, in fact it can be an almost infinite improvement.
You have to look at both signal and noise. If the signal is increased 6 dB by sending it twice, but the noise is only increased by 3 dB, then the net gain is 3 dB better S/N. If you add common mode noise to the consideration, then S/N can be improved by much more than 3 dB, but only if you don't get the topology wrong. Unfortunately, the non-common Gaussian noise from each signal path does not cancel just because one signal path is inverted, but fortunately the combination of two Gaussian sources is not 6 dB but only 3 dB, no matter the polarity.
If there is a flaw in any of the above, I would appreciate someone actually spending a sentence or two explaining what's wrong. I have no pride, I don't mind being wrong. I'm here to learn, so if I'm missing something then I can only gain by being corrected. I'm not asking for a remedial course, because I think I provided that above. If I made a mistake, then kindly take the time to provide at least one detail or two that can be confirmed. I regret that I may be feeding the troll, here, but perhaps a discussion based on communication instead of posturing will benefit someone.
If you want to claim that balanced suffers 3 dB worse noise, then show how that is not overwhelmed by the 6 dB increase in signal, not to mention the cancellation of common mode noise.
If you want to claim that a pair of DACs do not follow the same math and physics, then please show how. Why is the Gaussian noise from each DAC not combined to be 3 dB more than a single DAC?
My understanding would be that a pair of DACs converting the same data would combine (mix) to a 6 dB greater signal, but any Gaussian noise produced by these DACs would only increase by 3 dB.
Basically, "a pair of DACs" or "a balanced system" has exactly the same math going on when it comes to signals and Gaussian noise. Note that I said basically, not perfectly identical.
The distinction between the two is that parallel DACs cannot cancel common mode noise unless you digitally invert the data going into one DAC and invert the output before combining them. That's certainly one valid topology. But the hacker topology of soldering one DAC chip on top of another will not accomplish this. In fact, any common mode noise in that topology, i.e., any noise that is generated similarly in both identical DACs, would be 6 dB higher when mixed. Thus, some aspects of noise are not improved by simple doubling of DACs, but rather get worse.
As for the PCM1704, I explained that the data sheet uses phrasing like "two DACs are combined in a complementary arrangement," "utilizes dual balanced current segments," and "steps away from zero with small steps in both directions," which all strongly imply to me that the whole thing is balanced. If anyone has information to the contrary, then please share that information. Real information, please, not more posturing.
I made none of the claims that you attribute to me. That would explain your confusion.
Last edited:
There are lots of dual capsule mikes, usually the outputs are mixed to vary pattern. The ones I showed you can do this. Download the Neumann FET-47 schematic and you can see. Paralleling small capsules lowers the noise as long as you keep the noise decorellated just as paralleling amplifiers. The best example I can not speak about.
I cookup a fairly dramatic experiment that anyone can do at home for my article.
Last edited:
PMA look at it on a sample by sample basis if you want after subtracting the desired output signal. Examine the statistics it is white noise in the simplest case. Dither works on any signal that does not violate the sampling theorem. You can not put illegal computer generated "transients" through a DAC and make any conclusions. You guys are hung up on the idea that just the right length FFT would show "tones", etc. It ain't so. The noise behaves just as Rice showed in 1948.
Last edited:
Hi,
While your veritable vichyssoise of verbiage veers most verbose, once reduced to the point I can see where your confusion comes in. You actually are comparing apples to oranges. In fact, you clearly do not even understand how balanced connections and systems operate.
Please do yourself the favour and actually fully work out two examples where ALL ELSE IS EQUAL (that is two DAC's, one set parallel for SE the other balanced, two I/V converters for balanced, one for SE) and do yourself the favour to not just wave hands and make bizarre claims that repeat urban myth, but actually calculate the noise in detail from first principles.
Do we have to do this? Honest. Do you enjoy making me waste my time?
I really have no desire to spend the time to pick out your various statements, show that you made them and then illustrate just how wrong they, especially given that nothing more than having passed EE202 is needed to immediately see the fallacies...
As said, if you need remedials, you know my rate, PM me...
Otherwise I'm done with this.
Ciao T
While your veritable vichyssoise of verbiage veers most verbose, once reduced to the point I can see where your confusion comes in. You actually are comparing apples to oranges. In fact, you clearly do not even understand how balanced connections and systems operate.
Please do yourself the favour and actually fully work out two examples where ALL ELSE IS EQUAL (that is two DAC's, one set parallel for SE the other balanced, two I/V converters for balanced, one for SE) and do yourself the favour to not just wave hands and make bizarre claims that repeat urban myth, but actually calculate the noise in detail from first principles.
Do we have to do this? Honest. Do you enjoy making me waste my time?
I really have no desire to spend the time to pick out your various statements, show that you made them and then illustrate just how wrong they, especially given that nothing more than having passed EE202 is needed to immediately see the fallacies...
As said, if you need remedials, you know my rate, PM me...
Otherwise I'm done with this.
Ciao T
A good idea to introduce vertical phase problems where you really do not want them from a mike, for sure. AKG tried-it too with two concentric dynamic caps (D202).
This could work only with 1X 2" cap in between 2X 1" paralleled ones ( d'Appolito).
BTW, i'm realy upset by the arrogance of some posters, here, even where they do not understand the random aspect of the noise of the analog output stages of DACs.( not for you, Joachim ).
Last edited:
John,
Let me say I have developed my own understanding based in part on theory, in part on listening, in part on deductive analysis and in part on inductive analysis. Much of it actually comes from possibly knowing TOO MUCH about analogue systems and the human hearing...
I would hardly claim my view is RIGHT in any absolute sense or should be adopted unquestioningly by all. But I do like to raise some issues, like the one I like call "fuzzy distortion" to allow people to think about these issues from a different angle, to provide so to speak a different perspective.
Of course, to benefit from this it is essential to THINK, instead to just quote the orthodox position (not that I do not know, I just disagree).
Ciao T
Let me say I have developed my own understanding based in part on theory, in part on listening, in part on deductive analysis and in part on inductive analysis. Much of it actually comes from possibly knowing TOO MUCH about analogue systems and the human hearing...
I would hardly claim my view is RIGHT in any absolute sense or should be adopted unquestioningly by all. But I do like to raise some issues, like the one I like call "fuzzy distortion" to allow people to think about these issues from a different angle, to provide so to speak a different perspective.
Of course, to benefit from this it is essential to THINK, instead to just quote the orthodox position (not that I do not know, I just disagree).
Ciao T
- Status
- Not open for further replies.