XXHighEnd

Today I am going to show you how jitter looks like on a capable oscilloscope and explain a few things which are not so commonly known. This already starts with the way (or the means used) how jitter is observed and of which I found that this should be done through analogue (scope) means.

Side note : When done through digital means it is so much subjective to the "programs" used to convert the digital signal(s) to the analogue screen for observation that as many mistakes can be made as that, say, XXHighEnd could improve SQ over the years by means of all the (infinite) interpretations needed to do it really well.

In addition the (ADC) sampling rate of the digital scope has to be so high in order to interpret the resolution of today's "femto second" oscillators, that this actually does not exist. Mind you, no matter it's said that it does. So, if today's highest sampling rate of digital scopes is 160GS/s (giga samples per second) then the interpretable bandwidth is to be considered around 1/3 of that, thus something like 53GHz and the remainder is a matter of "time". I mean, when the scope can measure 53GHz "cycles per second" then the resolution of that comes down to 1/53,000,000,000 of a second which is equal to 0.00000000001887 seconds =  19ps (pico seconds). This is still 1000 times less for time resolution compared to what's needed (think 19fs (femto seconds)) while the analyzers concerned still "can do it".
But with so many tricks (already at the ADC side to achieve the high sampling rate) that in my personal opinion nothing much can be trusted. Or at least will vary vastly depending on the analyzer used.
Also thinking about the oscillators in the (ADC) analyzer itself and how they would need to be rated ... It just can not exist against the money such analyser costs (which is multi 100K to begin with).

Remember, all does not prevent the digital analyzers to come up with jitter data after all but regarding reliability - it is just not worth it. Remember : my opinion and judgement.

With an analogue scope the only problem is that it does not exist. :swoon:. Old technology. So, maybe analogue scopes are still being made in the lower bandwidth range (like under 100MHz) but at the moment when digital broke through (say 15 years ago) and maybe one manufacturer was working on higher bandwidth analogue scopes, it was considered that digital was the way to go and (high tech) and development of analogue scopes stopped.
And mind you, if I observed correctly, the analogue extreme high bandwidth scopes is so complex to design/manufacture that actually two such (1GHz) scopes existed, the Tektronix 7104 from last century's 70's and the Iwatsu TS-81000 from the year 2002.

For comparison, what an 1GHz analogue scope can do compares with what a digital scope of 300GS/s would be able to do for the time domain theoretically (theoretically for reasons described above).

Side note again : It is not so that digital analyzers are totally incapable because they can't meet the necessary specs regarding the time domain; digital analyzers can do more because the time domain can be derived from the frequency domain (which is what "digital" can do), but as said, it is full of tricks and such analyzers cost a large house anyway).

This is how an analogue scope "works" and outperforms the digital scope without hassle (I mean, what you see is just real) :

The 1GHz we are talking about is not about 1 giga samples per second or something - this is about the number of "sweeps" (say scans) the scope can perform per time unit. This is defined as 1GHz for the whole (horizontal) screen or more literally : 10 divisions per ns (nano second).

1GHz = 1/1.000.000.000 = 0.000000001 second = 1ns.

In other words : when the whole screen is written in 1ns, one division is written in 1ns/10 = 100ps (pico second).
or :
One division contains 10GHz of time resolution (10 x 1GHz).

In practice this is 200ps per division (don't ask me why but it can be related to nyquist stuff after all and there is math behind this which is somewhat beyond me) with the further notice that at least the time cursor has a resolution of 1/60'th division. Now 200ps/60 = 3.33ps and so the time resolution which can be observed is 3.33ps.

Also notice that for the means of observing the time domain the screen's resolution is important, which in this case is 800 pixels for the (horizontal) time domain (480 vertical). Thus, 80 pixels are available per division which is sufficient for the 60 time cursor steps to begin with but which is also sufficient for seeing the full resolution by eye. Thus, the full resolution which can be observed is 200ps / 80 = 2.5ps.

2.5ps is thus 80 times 200ps is thus 80 times 5GHz which can be derived from the above. This means that 400GHz of time resolution is available for observement.
This 400GHz can still be compared to the 53GHz mentioned in the beginning of this post.

To keep in mind : this is only about observing the time domain or performing math directly on that (which lattter a digital scope can do and which an analogue scope can not do).

Now let's see what we can do with this, with the notice that I never saw pictures of what I will present anywhere else. This is obvious because chances are very small that people in the same field use such an analog 1GHz scope.
(and yes, I got myself a 1GHz Iwatsu :heat:)

To keep in mind : Such a scope is not an "analyzer" as such and analysis goes by means of the eye.

---

Let's first look at an "analysis" of jitter distribution (as how it is called) of a digital scope (courtesy Rohde & Schwarz) :

(http://www.stordiau.nl/Phasure NOS1/Jitter Jistogram 01.png)
Watch the more horizontal beam which is the "analysed" representation of the signal of an oscillator and how the more light blue colour denotes more intensified jitter. Read this as : The light blue colour denotes the most occurrence of the signal. This is how the mid of that beam is the most intensified because it is the signal itself and the deviations (jitter) of that signal are less in occurrence so less intensified.

If you look at the plot in dark blue at the bottom of the screen you see how the deviation of the signal can be plotted and you see two major side beams. The right hand one is the highest and you can see this as the secundary light blue beam in the horizontal main beam. The left hand beam in the plot at the bottom is lower in intensity and this represents the dark blue beam in the horizontal beam.
Lastly, you can see a second side beam in the left hand side and this is the more orange colour in the horizontal beam.

Hard to read is the 51.2ns of deviation to the perceived optimal signal which shows on top of the right hand side beam at the bottom. This means that there's a frequently occurring jitter of 51.2ns towards one side of the optimal signal; when the left hand side beam would also show 51.2ns (but this is not readable) it would mean that peak to peak jitter is 102.4ns for this more intensified (more occurring) deviation. Mind you, only for these more intensified parts because you can see there is jitter beyond that - and it extends to the yellow (least intensified) beams in the horizontal beam. With some visual interpretation you can see that this spans around 400ns just because the distance is around twice that to the light blue and dark blue beams shown in the horizontal beam.

Got that ?
Then let's now look at the analogue scope and what we can make of that :

(http://www.stordiau.nl/Phasure NOS1/C0000001- 1KHz.PNG)
This shows an 1KHz signal and you actually see exactly the same. One difference of course - it is one colour only and the more occurring deviation (jitter) shows by the more intensified beams;
You can see that the optimal signal - now simply and "honestly" derived from the most occurring signal without any math - is in the middle and that at equal spacing more occurring jitter shows; as in the (first) example picture you can see that the distribution to the left and to the right of the signal is not equal; to the right there's more intensified and less spread deviation and to the left it is more spread (wide) and shows (thus) more vague. With some experience you could see that the deviation to the right and to the left in itself is equal but the deviation to the left is more "wobbling" in itself.

Notice that "to the right" means "late" and "to the left" means "early". So we have jitter that makes our samples come out later than intended and also more early than intended. And, the distribution of that is not equal.
Mind this word "distribution" because in the end a lot is about that (for Sound Quality).

If you look to the last picture and now concentrate on the middle of the beam (the intended signal) then you can see there too a more intensified and less intensified part. So, don't think this is just the "phosphor" of the screen dying out or something - it is just jitter again and, there too, more deviated to the left than to the right and again the early samples are more "distributed" than the late ones.

Lastly for this picture, notice that the parts between the middle beam and the outside ones are completely black ...

(http://www.stordiau.nl/Phasure NOS1/C0000002-Normal music.PNG)
But not so here !
What happened ?

While the previous picture shows a 1KHz test signal, the last one shows normal music playing.
Aha.
What insiders right away will tell us, is that this is thus "data dependent" jitter. Well, don't believe everything they say (and decide whether you believe me or not) but as we already know there is more going on than data impeded jitter as it formally exists (without explaining what that exactly is at this time) - and this is i2s dependent jitter. This too I won't explain plus it has been sorted out by "us ourselves" (with  Nick in the base).
Is it really i2s impeded jitter ? maybe maybe not, but it is dependent on things beyond known (formally). Look :

(http://www.stordiau.nl/Phasure NOS1/C0000003-144dBS attenuated.PNG)
This time again music is playing, but now it has been attenuated by 144dBFS. In't that interesting ...

If you compare this picture with the previous one you will be able to see how "volume" plainly creates jitter. All you need to do for that is seeing that the main beam almost has vanished at that almost all has been spread into deviation.
But not so with the last picture !

Side note : Can you already see how "jitter analysis" (from reviews or whatever other means) need context never provided (because not recognized) ? This is how I love this ... Something new again. Something new to work with.

(http://www.stordiau.nl/Phasure NOS1/C0000004-120dBFS attenuated.PNG)
Here you see the same but now attenuated to -120dBFS.
See ? it already starts to happen.

---

In the big picture we must see this in the context of a software player (by PC) and how it can influence jitter. Now we must be careful for the interpretation of the remainder in this post because it states that this *always* happens. However, this is empirical finding in itself and it had been derived only from so many people's various D/A converters with plainly nobody ever telling that XXHighEnd's influence is not there. Thus the other way around, it always does and this means that for the remainder of this post what I say/claim is true in general for everybody and everywhere (DACs).
Still this does not prove 100%, so please keep this in mind.

---

What you saw so far has been taken from your fine Phasure NOS1 D/A converter as you have it almost all (USB version I mean). Thus what I showed will show form your NOS1 just the same. Now notice : This has been taken from the 22.xxxMhz oscillator output and the oscillator is rated at 200fs. But look at the picture again and see the two vertical yellow ("cursor") lines and see that really 210ps of peak to peak jitter comes from it. :thankyou:
yeah, well, what to say.

Lousy after all ? maybe, but that doesn't tell at all that other DACs do better. It is all a matter of how the interpretation is done and all I personally can tell is that whatever the Stereophile's etc. come up with is by far superseded by me hence the NOS1 if I apply the same type of measurements. Apparently that doesn't tell much or too much is vaguened under way. And/but mind you, any first real jitter measurement of a DAC manufacturer himself I still have to see. IOW, showing oscillator measurements doesn't tell a thing at all. I hope this is clear now; only when this is done in the physical DAC (like I just showed) it shows reality.
An of course no Stereophile is going to put measurement probes on oscillator legs.

On to something new again (for the world and as far as I can tell) :

(http://www.stordiau.nl/Phasure NOS1/C0000012-Not streaming I think.PNG)
What you see here is the top of the oscillator's wave cycle (but with amount of 1 giga scans per second again) and you can clearly recognize that what's shown there has to resemble 100% what was shown before. The difference ? this is the "level" of the output voltage and not the time domain as such. Now, with the notice that the horizontal scale is 50mv per division, the "peak to peak" level variation shows something like 100mV (on a total of 3.3V peak-peak);

It is not said that the same happens to our audio samples (I did not check that) but it can be expected. Thus, this is again the oscillator output and "something" influences not only the time stability but also the level stability (output voltage stability). And since the "jitter" distribution looks the same, its source will be the same (whatever that source is).

Back to XXHighEnd influence ...
... Notice that the maximum output level of the PCM1704 chip in amperage is 1.2mA (which is converted to voltage and amplified). You are entitled to see 1.2mA as totally nothing to begin with, but notice that this is the maximum output or in other words what it outputs at -0dBFS (no digital attenuation). Can you now please divide 1.2mA by 16,777,216 to learn the minimum output at 24 bit attenuation or 144dBFS ? And also divide that by 1,048576 which implies 20 bits which is equal to -120dBFS ?

... and look back at above pictures which comared the -144dBFS and -120dBFS and that it makes a difference in jitter ?
And that whatever your calculators came up with for super low amperage is THUS so easily influenceable by all other unpronounceable tiny current draw ?

So that is how it works, and it will also work for level influence. Thus we don't deal with jitter in the time domain only, but also with jitter in the level domain (you might call this frequency domain, but not sure). Or in other words : when two samples with exact same calculated level are presented to the D/A conversion both will vary because "something" influences it.
Generally this "something" can be called your DAC and all what happens in there in order to convert the digital samples to analogue output. And that is A LOT.

But let's try to improve on this; what we can measure can be improved;

(http://www.stordiau.nl/Phasure NOS1/C0000013.BMP)
Aha. Well, between what you saw from the previous pictures and this one reads like a 1 minute change but take it that it really was two years. Even more actually because some know I have been working on it forever.
Anyway now it shows 13ps;
Watch out : This 13ps is measured peak to peak while general "one number" jitter figures are expressed in RMS and this also depends on the distribution; we can't do the real math from looking at a beam like this but think that for the RMS figure very roughly the 13 can be divided by 8 (so RMS jitter would be something like 1.625ps). Remember, "very roughly" and it can be higher or lower.

But is it 13ps or is it actually "analyzer limit" ? It is the latter. How ? well, it is just the beam of the signal and although I am able to minimize the beam more than you see, it comes down to showing "no beam" if less width of it is to be shown (those familiar with scopes will readily understand).
So I can only show the width of the beam and as long as no more intensified parts show (which it does not), it shows "no jitter".

Of course there is jitter (because no oscillator exists without jitter) but at least it will be under 13ps. And, if all was done 100% right it will be the 200fs rating of the oscillator.

Back to the beginning of this post and how digital analyzers may be able to dig out more (for various types of jitter I mean), let's say that for myself I don't need to have dug out more once nothing "is there" anyway. As said, in the end this is not true because down at the femto second level similar will be seen, but whether that is useful - I don't think so (also see the end of this post).


For our fun look here :

(http://www.stordiau.nl/Phasure NOS1/C0000033.BMP)
The scope is a so-called storage oscilloscope which means that it can store the once passed wave forms and keep on showing that to the screen. You can see this by the somewhat different colour of blue when used. So, this picture as well as the previous one used this feature and it was "all open". This means that it forever keeps on showing the once captured output which in digital terms is named "phosphor technology" and which resembles the after glowing of old Tektronix oscilloscopes.
This feature comes handy when analyzing jitter by this visual means because a "once occurring" deviation will be captured and showed, no matter it is not occurring any more.
So what happened to the last situation shown ?

I hopped up and down on the floor;
While jitter will be under 13ps inherently, I implied a 36ps of jitter which was incurred for by vibration; you can't see it here, but the one time the beam would jump to the right, the other to the left and after a few minutes of exercise on the floor this came out of it.
So be careful of what sound pressure (or speaker vibration) itself can do - not much different from the vast importance for turntables regarding this.

But didn't I put up a couple of posts on a nice Friday evening on how damping of the DAC (protecting it from vibrations) is important ? Right.
To understand this better you need to read to the end so you can see how even more low the jitter will be than this "analyzer limit" BUT that hopping on the floor is still capable to make 36ps of it. Relatively this only emphasizes the importance of good damping (read : should jitter be down to e.g. 200fs while vibration of the floor makes that 36ps then the difference is a lot.

(http://www.stordiau.nl/Phasure NOS1/C0000015.png)
Here you can see that the "jitter" on the voltage output level of the oscillator has vanished as well.
Please notice that this is not 100% comparable to the other pictures because the beam is intensified too much, plus this shot was taken with PNG (compression) which you can clearly see, while the others have been taken with lossless BMP. But the general idea is clear - no any side bands visible anywhere any more.

---

What technically happened is the 100% isolation from the (USB) interface. This has been accomplished by a means that is *not* detrimental to jitter and which is the big trick. There's not even long term jitter involved like how a PLL would imply that (so just saying); I am pretty sure that whatever I did has not been done before ...
Secondary (but a prerequisite), the lowest possible supply noise has been applied to all parts everywhere. Think 10's of mV before (which also is very good) to less than 2uV (RMS !) now.

(http://www.stordiau.nl/Phasure NOS1/C0000034.BMP)
As a bonus, here's the eye diagram at a sampling rate of 352800 and with the (phosphor) intensity full open for 15 minutes. It looks the very best (Google for pictures of eye diagrams).
Notice : In itself the eye diagram only tells about the headroom available for errors in the D/A conversion. So this can be seen as when there's so much jitter that there's no opening (this is the part between the crossing lines), there are no continuous "signal edges" to let trigger the output clocking upon (always).

---

Questions ? shoot. Recognize I sprouted cr*p somewhere ? say it.
But indeed, there is no audible XXHighEnd influence any more ...

Peter

Peter,

This is revolutionary stuff...congratulations...these new mods make the NOS1 _IMMUNE_ to the computer. 

Wow.

Wow.

Wow.

Not only that but there are several parts of this that I am going to have to re-read to fully comprehend but I believe you just showed a mechanism by which a solid state device is susceptible to vibration.  Again wow!

It's too late here now, but I will re-read this in the morning.

Super stuff!

Regards,

Anthony

Peter,

I will not pretend I understand everything, but I never really thought that vibrations would affect the NOS1 like this (and, I guess, other components). Wow !

Am I ok to interpret that at these levels (such low level voltage), anything in the power will also affect jitter ? And that new power supply (the thread that leads to this one) will improve on these artifacts ?

Maybe I am not using the right terms (like "artifacts"), but I will need to re-read this a few times...

This isolation that you brought (floating on water) is not so funny anymore...

Finally, when you mention that XXHE will not affect the sound anymore, you mean that the dials and other adjustments will no more be needed ? Provided that the power supply is upgraded and the isolation taken care of ?

Again, thanks for sharing this !!!!! It's a major eye opener !

Regards,

Alain

Hey Alain,


Generally yes but with the notice that I didn't take the effort to really prove that. But also take note that what can be done today with ultra low noise supplies was not possible even two years ago so it needs some rethinking (for everybody). What I "just" did was applying the "technology" as far as I could while I knew I was now out of the "USB noise" anyway. So mind you please, that is the major issue (or was) and it is obvious that we all knew that (with Nick's/Paul's attempts and with me telling about it even before the NOS1-USB was out). That is just USB ...
So with that out of the way as major bottle neck - and which is totally unrelated to the low noise supplies - the next bottle neck should be that noise now.


So yes of course. An (active) oscillator is directly driven by a supply and it's (clock) output signal is related to that. So if the supply varies ("ripples") it's output will do too (although there will be a certain amount of "rejection" - also dependent on the type of oscillator).


Exactly. Now, it may look like some "hobby" will disappear (wasn't that introduced by me myself ?) but to be the very most honest - am I glad that this hobby is out of the way. I think most of us will think that one settings is better for such but not better for so while the other setting implies the other way around. But also think of this angle :

It is you who knows foremost (not even any one else I think) how much work has been involved with getting the last device out of the PC. Well, while I did not get round to really achieve that (while knowing it is possible for sure) I had better enjoyed 6 months or so of vacation. It is just the wrong thing to approach.
Easy said of course when that "other" (this) wouldn't jump to my mind how to do it.
All has been the horse behind the wagon ...

At least from day one I meantioned that I hunted for this (actually elininating my own software).

And oh, I don't think I told how "easy" all suddenly sounds. Looks like LP for that matter.

Best regards,
Peter

Hi Peter,

Wow, this sounds great. XXHighend having no effect on SQ anymore also means that we just can use a 'normal' pc. The only requirement is that it has enough CPU and memory power to do the job?
No special psu needed for the pc? No minimised OS?

And offcourse I need to order the upgraded NOS1!

Regards, Arjan

Hi Arjan,

Yes, all correct. It is only that the more cpu cores and cpu speed, the more fast tracks will load.
Of course this has been the case always, but this sort of went against us when someone like me told (or found) that the lower cpu speed (like my 430 MHz) sounded better. Thus, this was so (IMHO) but is now out of the way.

Laptops ? haha, also no problem any more. But again mind the inherently needed speed to conveniently control the lot.

Kind regards,
Peter

Hey Peter, like many people reading your post, I'll admit to not understanding all the technicalities. But even so, it's obvious that what you've done here is beyond superb.

I describe you to my wife jokingly as 'a genius'. I'm seriously beginning to think the joke's on me.

Mani.

Hi Peter,

I am Chris/hochopeper from SNA, Anthony mentioned this post to me and I enjoyed reading, thank you for sharing details of your experiment. I first read this post in detail while watching the Catalunya F1 GP so I may have missed some of the details in your post so I'll give it a read (and possibly a 3rd read for good measure) before I respond in detail.

One thing that I thought while reading was: Have you been able to take a freq domain measurement of the DAC output for any of these scenarios? If yes, did it simply show nothing of interest?



Hi Anthony,

There has been some discussion in low noise regulator design datasheets that I've read that piezoelectric effect of some capacitor types can impact noise performance in a circuit. These capacitors are the same ones that are absolutely vital in their performance for the clock to perform well. I've read papers/discussion where different decoupling designs give dramatic improvements to clocks as well as jitter on the output of CPLD/FPGA processors.


Regards,
Chris

Thanks for the reply Peter. I, like Mani, often talk about you as a "genius", some other times like my "guru". This, all in good spirit of course.

And yes, USB plus the elimination of the final device... But one thing at a time right ? The achievement you have again reached is something to have a couple of beers about !

As for the adjustements that would be obsolete (I still have some troubles with this, but it is not intended on the fact it will "cut" in the hobby), if "that was good for this", and "this was good for that", there lie some explanations that may be answered someday... I would not remove all of them now, well not the "Minimize OS" ? Unless this brings too much assle with this breakthrough of yours... But of course, this is your "child", and it has grown a lot since :)

Kind regards,

Alain

Hi there Chris,

Well, I was putting up the posts during that same "event" and I must admit that these days this is easy to do with those stupidly "not sounding" F1 cars which are slow in the mean time. So for us it looks like something of the past, meanwhile giving an additional 2 hours of "normal Sunday". The only thing beneficial from it as how it is now is that Vettel stopped winning ...

But Chris, very nice to see you here, especially because I was of course indirectly referring to you as "Aussies". But I think you guessed that.

Regards,
Peter

Chris,

Of course each time I apply something "major" I am excited to try the THD and hope for the better, but no, that never happens.

I think we can well say that the, say, main process in D/A conversion is implying too much noise to begin with. Well, I now call that noise, but maybe I should call it "smoothing factors". So for example, when a D/A chip has an SNR of 120dB then it is rather useless to improve on THD when the noise figures way under that 120dB improved (like I told about 1dB regarding that). Or, in the line of what I also said : when the system output noise shows at -140dB (with good PSSR and all) it is kind of useless to improve the supply noise further (it has been rejected sufiiciently already).

So yes, against all odds I observe the THD, but that won't change a bit.

While I don't think you were referring to noise as such (while this is your actual subject - I know), all changes when we think Jitter. Well, I tried to explain that regarding the impact of noise, and this is quite obvious. However, when you *now* think how that would improve on THD ... never. But I think I have an explanation for that too :

We can think two directions :
1. No analogue filter behind the D/A process so the stepping distortion will be there (depending on the measure of upsampling);
2. There is an analogue filter and depending on at what frequency it hooks in and what frequency we observe, the stepping can be out of the way.

Either direction - in my view - implies a multi-1000 whatever factor compared to what jitter implies. Thus, the stepping is a relative severe distortion (and directly related to THD) while the response of a filtering capacitor is not much different (and at least totally out of our control).
IOW, such distortion so much overwhelms the minute distortion jitter will imply that we will never be able to see that in THD figures.
Or : with a 1 second recharging capacitor jitter can be seen (within those 1 second boundaries) but *that* capacitor won't filter the stepping distortion. So see ? dead end.

This may be a too brief "explanation", depending on how one is read into these matters.

Peter

Hi there Alain,

What I had in mind to add somewhere (but I did not) is that there's indeed something in me too like "but hey, I haven't been dealing with this for ages without reason ? come on !" ... and thus I too use MinOS, Unattended and or BUT I do still utilize the various (now) explicit SQ settings because ...

Well, because they influence the OS for speed. Just think MinOS as the best example - all is just x times faster.

I know, this is a (now) strange angle, but obviously (?) it has been an angle which has been there anyway; just compare the Q5=1 from 0.9z-9 and how that did NOT allow for very low SFS while -with nothing changed to Q5 itself- this for 1.186 allows for ultra low SFS (like up to 0.07 for my 430MHz).
All IOW, it may not be used for SQ any more, but it still can be used for responsiveness, track loading and whatever real time stuff later.

I think this is a tad too much :offtopic: for in here, but we have a running topic for it elsewhere.
And no, it is not offtopic of course for those who are explicitly against all this hassle (which I can tell you are most).

Peter

Hey Peter, your traces look very similar to those that Keith Johnson showed in his talk at RMAF2010. I've watched the video of the talk a few times now, and I've found it very helpful. (But you have to persevere.)

Mani

Doh Mani !
I was searching the internet for this video then found it by looking for the title I recognized and then found Phasure as the second link and then ... saw that you linked to it previously and that I never watched it. Grrr ...

Going to watch it now !
Peter

Peter hi,

This is an absolutely facinating post  :)

Partly because its a really pragmatic approach your taking using and analogue scope to determine the spread of jitter and tune it qualitatively.

Also because its great to see such a focused approach being applied to the clock time domain in the NOS1.

Reading your post above and your other resent post on the NOS1 upgrade, if I have understood them correctly you have applied a number of really important changes to the audio clock setup since the upgrades to my NOS1 were applied a couple of months ago. This is moth watering  :)

Although I don’t have anything line the scope resolution you have with your TS8100, amplitude jitter and audio clock edge smearing and smearing of eye diagrams can be seen hear to some degree even on my modest scope.

The tuned results you are getting get in your last three graphs above look great, is the trace being taken from the directly from the bit clock feed to the DAC chips ?

Again what a fascinating and great post.

Kind regards,

Nick.

Ps I took a look at the spec for the TS8100, what a great bit of kit, I can now add scope envy to the FFT analyser envy I already had haha !

Hey Nick,

Yeah, nice eh ?


Envy ? sheer obsession ! Well, I think you know how long I have been trying to find something suitable (which is all related to $$$$$ of course). And then to think that I sort of blindly ordered this one from some obscure looking Korean company expecting a brick in the box ... But it went all right.
TS-81000 btw.


Nah Nick, I don't think so. But the problem is, once the sampling speed is high enough this starts to look like that. But all what happens is that the sampling itself (and/or the math behind it) is jittery and that's what you see. Just do the math on what sampling speed it takes for analogue already to enable the visibility of 210ps of jitter. And when you are as far as that (will give "buyable" results) then know that the bandwidth (still in order) is too low to get any decent upright rise out if it (like you see in my screen shots something towards 90 degrees). So instead you'll have a triangle sine like flat shape and that won't allow time domain jitter to be seen.
Make that a digital sampling scope and not even 40GHz will bring you close.
But ok.


Well yes. And for other people's information : that few months ago I was as far as finally trying the isolation itself just to see what it would bring with small PCBs for that which I had (let) made and which were in the house for maybe 6 months already. So these were only there to test it, not to really apply it. So, they worked which wasn't all *that* exciting (but with my expectation that there could be a ground-potential-looping problem just because of the potential difference on either side which had to go somewhere (radiation !) but that problem didn't happen and actually it sounded quite OK as well. But as I told Nick, now we could be guaranteed to have jitter just because the isolation implies that. But it was just a trial in itself ...

With the boards and wires all over the place with a next attempt because I got that "other" idea of how to possibly apply this to the existing NOS1 while nothing was intended for that, that worked in one go and all what remained what now designing something that could be piggy-bagged to the existing NOS1; a challenge in itself because of the available space from all directions. But in the end it all worked out.
Right at this moment the only thing left is the wiring between USB and DAC in a nice decent fashion.
Oh, might someone know a source for 2.5-3.5 mm internal diameter cotton braided sleeve ... (and I mean one which ships and delivers within 2-3 days instead of possible weeks from the US).


Not sure whether you really intended to refer to the last three graphs, but yes. Or almost yes, because that couldn't be done directly for the best result without additional measurement wires. Point : The best result will not look better (scope limit). So if you look closely you see that I sneaked in a Word Clock shot (watch the frequencies) and this is all in advance of where the real stuff happens - but still good enough and thus knowing that the scope is not able to show better anyway.


On this one and especially for you (because you will be able to envision it from your situation as of now) :
The oscillators now are on one of these new PCBs ...

Hey Nick, as you will understand you were one of the great motivators for this. I mean, no matter what we all could find or try to reason for this USB sh*t, in my opinion this was not exactly under our control, no matter we could audibly improve on it. So that's why. Or maybe that's why with some priority now, because of course I have always been thinking about THE solution.

Thank you ...
Peter

Not necessary anymore !
Just found it close by.

Thanks.
Peter

Hey Peter,

I have to say (to echo Nick) all the work you have done on this is super impressive. From an engineering perspective it is almost unheard of these days to experience this level of clarity, vision and execution.

Everything is digital nowadays and there is little proper engineering done - just bolt a few IC's together without any proper understanding of what is going on tends to be the normal approach. The old days of proper engineering (i.e. in the days of analogue) are largely a thing of the past. Except here of course!!!

Cheers

Paul

Thank you so much for the kind words Paul.
But aren't we all working on the same case ? It is only that some might have more $$ to begin with and investments *are* needed you know. And lots of time. Well, you guys just give me that time, helped mucho to set the priorities for it, motivate and ... well the $$ isn't the problem accidentally. It just has extended my own hobby to a level I could never have imagined a few years back. And I can honestly tell you, the real thing happens when we can turn that into sharing. But this is what audio is about and thus I can tell *once again* that this only can all happen when we do this together.

What to say more ...
Peter