As your own measurements are showing, square to sine conversion degrades the phase noise floor by 10-20dB. Of course, you may chose to ignore this degradation and to claim that close in phase noise is everything in audio.
Close-in phase noise should also be degraded by the sine to square conversion, although at a lesser degree, and depending on the phase deviation; the reference I already posted and this one:
http://people.mpi-inf.mpg.de/~adogan/pubs/IFCS2018_comparator_noise.pdf
are clearly proving this fact.
Your claim is that it is not affected, at all. I apologize for not trusting your measurements in this respect, this would be an extraordinary result that needs an extraordinary explanation to be trustful. Perhaps you are using a novel circuit that optimizes the close in phase noise of the conversion process, beyond to what's measurable with a timepod. If so, we could discuss this if you care to post the schematic you are using.
Last edited:
I don't think you formulated the question right, but since I think I understand where you are coming from, the closest answer is the Parseval theorem applied to the Fourier series (where the squares are actually proportional to the signal(s) (and in particular noise) power).
Last edited:
That would produce a weird waveform rather than a jittered square wave. Normally the n-th harmonic is modulated the same as the fundamental, but with an n times larger phase deviation.
It would not only measure fine, it also could be used for fine tasks.
That is the idea behind the Oliver Collins paper. Low pass filter or even
band pass filter the fundamental And THEN amplify the edge speed,
maybe repeat several times.
The harmonics of the incoming signal do not carry useful information,
nor do their sidebands. All that counts are the zero crossings of the
fundamental.
The harmonics only collect dirt over the increased bandwidth. It's
just that logic chips do not like sine inputs. They only like 0 and 1.
Somewhere in the archives of the time nuts list is a spreadsheet to
calculate the optimum low pass filter edges and optimum number
of stages.
Having an additional high pass may remove some 1/f noise.
If you search a process that modulates only a harmonic:
Assume you have a 5 MHz carrier with harmonics and a
low pass filter that happens to have the corner at 15 MHz,
then imagine the Bode plot: at 5 MHz the phase is completely
flat, 10% component variations in the filter won't have any influence
at 5 MHz.
But at 15 MHz, the phase will flip n*180° over a MHz or two.
10 % variations on the L or C will be a disaster for the phase
of this harmonic.
Think of a C like a varicap with unclean DC across it or a
ferrite coil in a mag field.
That's why I like notch filters with high Q to remove the
close-in sub/harmonics. They leave the fundamental alone.
< [time-nuts] Understanding Oliver Collins Paper "Design of Low Jitter Hard Limiters" >
< Sci-Hub | The design of low jitter hard limiters. IEEE Transactions on Communications, 44(5), 601–608 | 10.1109/26.494304 >
< Bruce's: Zero Crossing Detectors - KO4BB >
Last edited:
You missed my point Mark. There is a distrust of plastic toslink amongst many due to reading that it's bad, nothing to do with intrinsic problems. Glass ST looked nicer and might have had a jitter advantage in the old days, but died out.
You seem to have totally and utterly misread or misunderstood what I wrote. Either that or you have a persecution complex.
Bruce Griffiths wrote:
"The problem of optimal zero crossing detector design was essentially
solved by Oliver Collins in the 1990's.
Essentially a series of cascaded limiter stages with appropriate gain
and bandwidth distribution are used.
With a 10MHz 1V rms signal only 2-3 stages suffices.
However unless you need fs jitter less complex zero crossing detectors
should suffice.
1) a comparator (or line receiver) based design should achieve sub 10ps
jitter.
2) AC coupling to the input of a CMOS (AC04, AHC04 LVC04) should achieve
a jitter of 1ps or less
3) A simple differential pair with AC coupled emitters (reduces
asymmetry due to component tolerances ) is capable of sub ps jitter."
Spreadsheet attached.
The second way is just our AC04 sine to square converter, so no "extraordinary result", just what expected.
The third way (cascaded) is just what we are working on as the ultimate sine to square converter.
Last edited:
Router stuff- I had a long discussion with an audiophile/network wizard to try to reason some way this could matter. There are two aspects that could impact things.
First, the clock you are fiddling with controls the packet rate, not the wire data rate.
Second, using two routers will reduce the number of data packets that are not relevant to your player. The processor in the first need to talk to all the live ports but the second only has your device in its mac table so only those packets would be forwarded. (I'm not a network guru so I my have butchered this.) So there is some validity IF the player doesn't have good isolation between the ethernet interface and the rest of the box.
Also, all the RJ45 ports are galvanically isolated so a simple ferrite on the cable should reduce the remaining coupling. No need for a lot of other stuff.
Network audio should be the easiest to do a good blind AB test since levels etc. will be "perfect" every time.
First, the clock you are fiddling with controls the packet rate, not the wire data rate.
Second, using two routers will reduce the number of data packets that are not relevant to your player. The processor in the first need to talk to all the live ports but the second only has your device in its mac table so only those packets would be forwarded. (I'm not a network guru so I my have butchered this.) So there is some validity IF the player doesn't have good isolation between the ethernet interface and the rest of the box.
Also, all the RJ45 ports are galvanically isolated so a simple ferrite on the cable should reduce the remaining coupling. No need for a lot of other stuff.
Network audio should be the easiest to do a good blind AB test since levels etc. will be "perfect" every time.
Ah, this is your DIY partner Roberto A spreadsheet, available in many places on the Internet (ko4bb, time-nuts, etc...), essentially implementing Oliver Collins theory and method. Great stuff indeed (and no, I am not sarcastic)!
Why don't you ask him to drop by so we could have an educated discussion about, hopefully without any high end audio
Last edited:
Yes, the spreadsheet was created by my co-developer Roberto with the help of Bruce Griffiths who is an authority in this matter.
Although some here believe that we do things randomly, in reality any of our design is preceded by months or years of study.
I'm sorry but Roberto does not like this forum, he also does not like to share our designs because as I have said several times we are designing for ourselves, basically we are designing our whole audio system following the way we believe.
Me and Roberto we do not agree on this point and the compromise was to share our projects but not the schematics.
I respect him and therefore I find it right to respect his will.
However you could not have with him the discussion you name "without any smelly high end audio noise", because on this matter (importance of timing in digital to analog conversion) he is much more fundamentalist than me, although our thougth is very similar on the whole audio chain.
So we will keep designing our audio devices based on our beliefs to get the best performance as possible following what we believe is the correct approach.
In a few words for us audio is not just measurements, in the end it is always listening.
And we care less than zero if anyone considers our approach "snake oil".
Everyone is free to think whatever he wants, it's not our problem.
We respect the opinions of others even if we do not agree them, but in any case we will follow our beliefs even if others believe them to be wrong.
And we don't do audio research, we're just hobbyists, so we wouldn't even have the time as well as the skills.
As we have done so far we will continue to make the appropriate measurements, but always following our own approach and not that of others that we do not agree.
For example we will use an upconverter to measure the jitter at the output of the DAC rather than an audio analyzer, sorry but we will never perform the Jtest.
And as long as the members of this forum are interested in our projects we will keep sharing them with the audio community.
Although some here believe that we do things randomly, in reality any of our design is preceded by months or years of study.
I'm sorry but Roberto does not like this forum, he also does not like to share our designs because as I have said several times we are designing for ourselves, basically we are designing our whole audio system following the way we believe.
Me and Roberto we do not agree on this point and the compromise was to share our projects but not the schematics.
I respect him and therefore I find it right to respect his will.
However you could not have with him the discussion you name "without any smelly high end audio noise", because on this matter (importance of timing in digital to analog conversion) he is much more fundamentalist than me, although our thougth is very similar on the whole audio chain.
So we will keep designing our audio devices based on our beliefs to get the best performance as possible following what we believe is the correct approach.
In a few words for us audio is not just measurements, in the end it is always listening.
And we care less than zero if anyone considers our approach "snake oil".
Everyone is free to think whatever he wants, it's not our problem.
We respect the opinions of others even if we do not agree them, but in any case we will follow our beliefs even if others believe them to be wrong.
And we don't do audio research, we're just hobbyists, so we wouldn't even have the time as well as the skills.
As we have done so far we will continue to make the appropriate measurements, but always following our own approach and not that of others that we do not agree.
For example we will use an upconverter to measure the jitter at the output of the DAC rather than an audio analyzer, sorry but we will never perform the Jtest.
And as long as the members of this forum are interested in our projects we will keep sharing them with the audio community.
Understood that the waveform would not be exactly square, and its shape would be changing dynamically. The question for me is how much deviation from perfectly square might there be and how might we measure resulting time modulation of zero crossing (or whatever the detection threshold of the dac clock input may be).
Perhaps consider that if, say, the 3rd harmonic, were to be phase modulated at 5Hz, wouldn't close-in phase noise of the approximately square clock pulses be affected? (Not saying this is something I would expect to happen, but what if it did to some subtle degree?)
Last edited:
Hi Doede,
thanks for review.
And I could measure the phase noise of the Neutron 3 clock if someone would send me one.
Andrea
The clock could take a rout among Germany and Italy
Thanks, I look forward to measure it.
If you mean at the input of the DAC, so LRCK, BCK or MCK depending on the DAC, it's enough easy: with a phase noise analyzer since zero crossing means phase.
The problem is go down to audio frequencies with a phase noise analyzer.
"...with a phase noise analyzer since zero crossing means phase"
Understood, but I deliberately tried to think up a way that a phase noise analyzer could give a different result if the waveform was filtered before measurement as verses if the waveform had not be filtered. In my mind at least, I think I did describe such a condition. However, I'm not sure if it could be a problem in reality.
Understood, but I deliberately tried to think up a way that a phase noise analyzer could give a different result if the waveform was filtered before measurement as verses if the waveform had not be filtered. In my mind at least, I think I did describe such a condition. However, I'm not sure if it could be a problem in reality.
From the measurements we have made it would seem not
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator
The first 4 harmonics are not attenuated by the low pass filter and the plot of the non-filtered waveform is exactly superimposed to the one of the filtered waveform.
Zero crossing is not alterated so the phase noise is not modulated.
The Well Tempered Master Clock - Building a low phase noise/jitter crystal oscillator
The first 4 harmonics are not attenuated by the low pass filter and the plot of the non-filtered waveform is exactly superimposed to the one of the filtered waveform.
Zero crossing is not alterated so the phase noise is not modulated.
My hypothetical example was with a 24Mhz audio clock, so the 3rd harmonic would be up around 72MHz. Wouldn't that get attenuated by the phase measurement input filter?
If we started with a 5Mhz clock then I would have had to specify a higher order harmonic as being phase modulated so as to put it above the input low pass filter cutoff (that is, I would have to do it in order to create a suitable hypothetical question).
If we started with a 5Mhz clock then I would have had to specify a higher order harmonic as being phase modulated so as to put it above the input low pass filter cutoff (that is, I would have to do it in order to create a suitable hypothetical question).
Last edited:
- Status
- Not open for further replies.