Ringing needs inductance but with small package and short trace lengths inductances can be kept small. At these clock frequencies (i.e. above self-resonant frequency) acrylic caps do next to nothing and just take up PCB space. The internal decoupling capacitors of larger clock oscillators are typically ceramic (even class II) so the benefits of external non-ceramic decoupling capacitors are slim. Probably better not to use external decoupling capacitors for those oscillators as trace lengths (and inductance) can become excessive. But decoupling capacitors are used not only on clock oscillators but also on ICs where such issues can be mitigated with proper layout.
Here are some pictures showing what larger package oscillators contain:
https://www.diyaudio.com/community/threads/inside-a-cchd-925-output-dead.396310/post-7280163
https://www.audiosciencereview.com/forum/index.php?threads/e1da-cosmos-adc.27038/post-1996953
This may also be of interest:
https://www.ti.com/lit/an/sloa069/sloa069.pdf
Here are some pictures showing what larger package oscillators contain:
https://www.diyaudio.com/community/threads/inside-a-cchd-925-output-dead.396310/post-7280163
https://www.audiosciencereview.com/forum/index.php?threads/e1da-cosmos-adc.27038/post-1996953
This may also be of interest:
https://www.ti.com/lit/an/sloa069/sloa069.pdf
There is capacitance between the PCB surface fill, the ground plane underneath, and the power plane. It doesn't have be lumped where you think it does if the layout supports doing it another way in a particular design case. ThorstenL has already talked about some those concepts in another thread. The bottom line is that the clock board works, and it works quite well for its intended purpose.
Again, if you think you can do better then please share your own design with the group. Otherwise your armchair quarterbacking is not helpful.
With that, I will stop here on this subject.
Again, if you think you can do better then please share your own design with the group. Otherwise your armchair quarterbacking is not helpful.
With that, I will stop here on this subject.
I have not commented on your design but just suggested other components.
My approach is quite different as I have clock buffers on my USB-I2S, ADC and PCM2DSD boards and reclocking on DAC boards (where needed). They all accept external clocks if needed. I have a separate clock board if I need to override the onboard clocks. That has similar power supply as this: https://www.diyaudio.com/community/attachments/final-mclk-reg-png.950839/
My approach is quite different as I have clock buffers on my USB-I2S, ADC and PCM2DSD boards and reclocking on DAC boards (where needed). They all accept external clocks if needed. I have a separate clock board if I need to override the onboard clocks. That has similar power supply as this: https://www.diyaudio.com/community/attachments/final-mclk-reg-png.950839/
The optimum bypass capacitor value/type thing with clock oscillators has been going on for years. I could post pics of clock adapters made by Twisted Pear, Iancanada, etc., that can accept various kinds of bypass caps for experimental optimization. IME they all sound like crap (and most of the voltage regulators didn't sound too good either but that's a different topic). The unpublished work I did with different bypass caps for clocks and buffers started out exactly according to textbook dogma. However, something appears to be going on that is not well explained by the most common engineering models. Moving away in increments from textbook caps to .22uf MU caps shows a clear improvement in multiple dimensions of music reproduction (e.g. soundstage depth, vocal dynamics, etc.). However, things are not even that simple. MU caps take time to stabilize -- less than a day for most of the settling to occur under actual operating conditions. The sound goes through some interesting phases as the caps settle in over a few or several hours (say, overnight). From a scientific perspective I would very much like to know exactly what is going on inside the caps during that period of time. From a practical engineering perspective, there is no need to keep repeating the same experiments over and over, expecting a different result. Thus no need to keep trying small ceramic caps. Its been done enough times already. IMHO, same thing with ferrites and clocking for audio reproduction (whether or not the problem for audio is exactly due to "encagement"). All the above having been said, if somebody has new idea to try that might work better than MU caps, then I will be more than happy to listen to their version of the clock board.
With that, I hope we can leave this sub-topic. Its been talked through enough already.
With that, I hope we can leave this sub-topic. Its been talked through enough already.
Last edited:
New version of Clock Board with PLL is in process at JLCPCB. Thought I might take a moment to show a simple block diagram of what the board does:
In case it wasn't clear before that's the pretty much the basic idea. Start with a couple of clocks (originally 45/49MHz). The USB board controls the relays that switch clock frequencies, and so on. The details of adapting the board to also work with 22/24MHz (or even 11/12MHz) clocks complicates things a bit. There is now also a PLL to multiply clock frequencies as needed, and there are a couple of spare inverters for other contingencies.
In case it wasn't clear before that's the pretty much the basic idea. Start with a couple of clocks (originally 45/49MHz). The USB board controls the relays that switch clock frequencies, and so on. The details of adapting the board to also work with 22/24MHz (or even 11/12MHz) clocks complicates things a bit. There is now also a PLL to multiply clock frequencies as needed, and there are a couple of spare inverters for other contingencies.
Hi everyone,
thanks Markw4 for this nice work ! Very interesting.
In a few posts, someone mentioned "phase measurement", but I saw no data !?
Nobody uses a TinyPFA ?
Here => https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.Homepage
LiteVNA, TinySA, TinyPFA are the 3 cheapo (around 100Eur) measurement devices that any DIYer must have (IMO) !
I can't find the comparative that was made on eevblog a while ago... but these devices are dead cheap AND as accurate as xkEur "lab-grade" devices (when frequency below 300MHz if I remember ; then, above, their noise floor goes up but they are still top-notch devices).
Other point : I read some post about capacitors (some can ring etc...) => a LiteVNA is a must-have tool to check any unstability / noise etc created by a component or a combination of components (passive filters eg).
Cherry on the cake : each of these devices have a PC interface to get much more out of them :
Hope it helps (or it doesn't help because everyone here already knows all that
)
thanks Markw4 for this nice work ! Very interesting.
In a few posts, someone mentioned "phase measurement", but I saw no data !?
Nobody uses a TinyPFA ?
Here => https://www.tinydevices.org/wiki/pmwiki.php?n=TinyPFA.Homepage
LiteVNA, TinySA, TinyPFA are the 3 cheapo (around 100Eur) measurement devices that any DIYer must have (IMO) !
I can't find the comparative that was made on eevblog a while ago... but these devices are dead cheap AND as accurate as xkEur "lab-grade" devices (when frequency below 300MHz if I remember ; then, above, their noise floor goes up but they are still top-notch devices).
Other point : I read some post about capacitors (some can ring etc...) => a LiteVNA is a must-have tool to check any unstability / noise etc created by a component or a combination of components (passive filters eg).
Cherry on the cake : each of these devices have a PC interface to get much more out of them :
- LiteVNA : I use nanovnaapp to run the measurement, and save the s2p file ; and then I use the app of the LibreVNA (LibreVNAgui) to display the s2p file, do some comparison btw measurements etc... (this GUI is awesome, IMO ; I bought a LibreVNA => expensive & crap vs a LiteVNA which is the best in that category)
- TinySA : it has its app, too.
- TinyPFA works with TimeLab (see link above). We can't use all the features of TimeLab but the avaiiable features are far enough (ADEV, freq "stability", phase "stability" etc...). The graphs given by TimeLab can be saved as screenshot (basic), but we also can save the measurements data (frequency & phase) as csv file to do other calculations using Excel... Amazing device this TinyPFA because what it can measure accuratly, no other instruments can do it (an oscilloscope can't measure the tiny unstability of oscillators, with a spectrum analyzer the lower RBW value is always to big to give a relevant measure... etc)
Hope it helps (or it doesn't help because everyone here already knows all that
)We are talking about "phase noise," which is one of the signal-correlated noise sources in dacs. The equipment to measure clock phase-noise with high precision is quite expensive (have plots for some of my clocks though). However it is possible to look at spectral line noise skirts as can be seen at: https://www.diyaudio.com/community/threads/phase-noise-in-ds-dacs.387862/post-7063038
Hi rfbrw,
no !
we can do a measurement of a DUT connected to ONLY 1 port of the TinyPFA !
in that case, the "reference" is the PLL of the TinyPFA, BUT
it's way enough vs precision required in there !hi markw4,
"phase noise" is phase-noise... whatever the device that host a oscillator.
well..
please go back to the TinyPFA web page to understand the way it works & how accurate it is...
your link is fine, but these spectrums (from spectrum analyzer), thus, the scale is way higher that TinyPFA can achieve !
As I said above, a spectrum analyser if limited by its RBW...
Honestly, the cheapo&fasto way => buy a TinyPFA, and you'll realize that...
- although it looks like a cheapo nanovna (100Eur)
- it's the must-have tool... because it'll give you crazy "fine" measurements of you DUTs
@phile1,
Looks to me like you misunderstand what we are interested in. Are saying that nonvna is a State Of the Art instrument (SOA) that is as good or better than anything Microchip, Holzworth, HP, Keysight, Tek, etc., has ever made or knows how to make?
The reason I ask is that some of the clocks we use are SOA ultra-low phase-noise. There are few instruments made in the world that can accurately measure the clocks, and the instruments cost over $10k each. Some people have spent $50k on a rack full of instruments to measure such clocks.
So you see, it would be silly to believe a $100 instrument could be good enough to measure the clocks as accurately as we want. Especially so if using a PLL for a reference, because PLLs are notoriously jittery. It makes no sense at all.
Besides, its not clear if nonvna could produce a measurement anything at all like this even for a very poor clock:
Do you even understand what this graph shows?
Looks to me like you misunderstand what we are interested in. Are saying that nonvna is a State Of the Art instrument (SOA) that is as good or better than anything Microchip, Holzworth, HP, Keysight, Tek, etc., has ever made or knows how to make?
The reason I ask is that some of the clocks we use are SOA ultra-low phase-noise. There are few instruments made in the world that can accurately measure the clocks, and the instruments cost over $10k each. Some people have spent $50k on a rack full of instruments to measure such clocks.
So you see, it would be silly to believe a $100 instrument could be good enough to measure the clocks as accurately as we want. Especially so if using a PLL for a reference, because PLLs are notoriously jittery. It makes no sense at all.
Besides, its not clear if nonvna could produce a measurement anything at all like this even for a very poor clock:
Do you even understand what this graph shows?
Last edited:
Mark, you are right. It does sound like squaring the circle. OK why start a discussion? I will buy one and connect it to my Andrea Mori Clocks, the Accusilicons and the included simple clocks from the FiFoPi and see what happens and if any correlation between "quality" can be found. Yesterday, I was doing some tests with my new DAC setup and in fact I felt I "needed" your board here - So please keep going and when you are "ready" I will be among the first to build it
update: Amazon has them and delivery will be today before 10PM
- So please keep going and when you are "ready" I will be among the first to build it update: Amazon has them and delivery will be today before 10PM
The 20R resistors were experimental at one point. LMK1C1104PWR has a rated output impedance of 50R, which is higher than most other buffers and CMOS devices. In theory it would be a match to a 50-ohm transmission line with a 0R series damping resistor. I tried to explain about series damping resistor values in the Clock Board Notes file I posted a few posts down after posting the KiCad files. They can affect dac sound. Sometimes higher resistor values than theoretical matching would suggest, end up sounding better in dacs. For that reason some people use 100R as a standard series damping value in dacs. Possibly that's because more resistance reduces risetime a little and thus may reduce some very high frequency RF harmonics from coupling noise into other circuitry, or maybe reducing coupling of RF noise through a dac chip's substrate.
For Marcel's dac as I have shown in at least one pic, I like the dac sound of the dac best with more than 0R series damping on the reclocker board master clock line. Also, changing the subject a little, I will say that with the overall setup I have shown for Marcel's dac, it is sounding exceptionally good in SE output mode. Definitely better than I had the Acko version of Marcel's dac sounding previously. A large part of the reason for the improvement was removal of the ferrites and C-II ceramic caps from the reclocker board. Replaced the caps with Rubycon and Cornell-Doublier SMD film caps, gave it a day of running for them to settle, and then found sound which I like better. In particular the ferrites seem to impart an artificially pronounced metallic sound to things like cymbals, and brass wind instruments. Its an effect that doesn't sound like real instruments, at least not to me and on my system. In addition, found similar ferrite and cap side-effects and remedies for the Squaring Boards I use.
For Marcel's dac as I have shown in at least one pic, I like the dac sound of the dac best with more than 0R series damping on the reclocker board master clock line. Also, changing the subject a little, I will say that with the overall setup I have shown for Marcel's dac, it is sounding exceptionally good in SE output mode. Definitely better than I had the Acko version of Marcel's dac sounding previously. A large part of the reason for the improvement was removal of the ferrites and C-II ceramic caps from the reclocker board. Replaced the caps with Rubycon and Cornell-Doublier SMD film caps, gave it a day of running for them to settle, and then found sound which I like better. In particular the ferrites seem to impart an artificially pronounced metallic sound to things like cymbals, and brass wind instruments. Its an effect that doesn't sound like real instruments, at least not to me and on my system. In addition, found similar ferrite and cap side-effects and remedies for the Squaring Boards I use.
Last edited:
BTW, found a new u.fl tool which in principle works similar to the Sparkfun tool, but IMHO is overall better (a better fit to the u.fl connector, and a better handle):
https://rfsuperstore.com/products/u-fl-insertion-extraction-hand-tool
https://rfsuperstore.com/products/u-fl-insertion-extraction-hand-tool
Hi Markw4,
no brainer about all what you said... that's clear & anyone knows that !
This graph you show is from the manufacturer (or anyone else).
But my concern is about => YOUR PCB OUTPUT !
as clear & simple as that ?
I wrote "YOUR PCB OUTPUT" in "BIG" just kidding
cause what matters here is =>
given any "common people" / basic DIYer,
=> how can I get any idea of my own PCB output with a cheapo & accurate measurement device (cheapo is better, accurate is required
of course )It I am going to measure the output of my board, it can only be done with the same equipment used to measure the SOA clocks (given that buffer close-in phase noise should be about as good as SOA clocks). I could send a board to someone with the necessary equipment if I thought it necessary.
OTOH, a device that is primarily spec'ed for comparing the frequency two clocks is useless. One clock needs to be at least 10 times better than the other clock so the better clock can act as a reference. If both clocks are SOA, or if they are otherwise about the same, there is no way to tell which one is right and which one is wrong. You only see a difference. A PLL-based measurement would also be useless.
In addition, I looked at nonvna website and saw nothing about using it for close-in phase noise measurements at all. It was spec'ed for frequency comparison of two clocks, and for phase mesurments of two clocks of the EXACT same frequency. Well, audio clocks don't have to be the exact right frequency. They don't necessarily need to have terrific Allen Deviation either. But IMHO they do need to have very low close-in phase noise for best results.
That said, IIRC there is a new-ish Chinese instrument that works something like a TimePod. IIRC the new instrument goes for about $900-$1k, somewhere around there. However its not as accurate as a TimePod for close-in phase noise at SOA levels. Again, so far as I know there no $100 solution to the measurement problem. You keep claiming there is, but you haven't shown a link to any information that would confirm that nonvna can do what you seem to be round-about inferring. Also, you never explain exactly what its supposed to be able to do that we should want. IOW, you are waving your arms around and making a lot of noise without being specific in a technical sense.
On the issue of the average diy'er needing to measure clocks, again we need get specific about exactly what you are claiming they should be measuring, how to do such a measurement, and why it is needed. I mean, we could measure the weight and the volumetric displacement of a clock or of a board, but why do that if what we are concerned about is the sound? You need to explain how what nonvna can measure relates to dac sound. Specifically, and technically.
OTOH, a device that is primarily spec'ed for comparing the frequency two clocks is useless. One clock needs to be at least 10 times better than the other clock so the better clock can act as a reference. If both clocks are SOA, or if they are otherwise about the same, there is no way to tell which one is right and which one is wrong. You only see a difference. A PLL-based measurement would also be useless.
In addition, I looked at nonvna website and saw nothing about using it for close-in phase noise measurements at all. It was spec'ed for frequency comparison of two clocks, and for phase mesurments of two clocks of the EXACT same frequency. Well, audio clocks don't have to be the exact right frequency. They don't necessarily need to have terrific Allen Deviation either. But IMHO they do need to have very low close-in phase noise for best results.
That said, IIRC there is a new-ish Chinese instrument that works something like a TimePod. IIRC the new instrument goes for about $900-$1k, somewhere around there. However its not as accurate as a TimePod for close-in phase noise at SOA levels. Again, so far as I know there no $100 solution to the measurement problem. You keep claiming there is, but you haven't shown a link to any information that would confirm that nonvna can do what you seem to be round-about inferring. Also, you never explain exactly what its supposed to be able to do that we should want. IOW, you are waving your arms around and making a lot of noise without being specific in a technical sense.
On the issue of the average diy'er needing to measure clocks, again we need get specific about exactly what you are claiming they should be measuring, how to do such a measurement, and why it is needed. I mean, we could measure the weight and the volumetric displacement of a clock or of a board, but why do that if what we are concerned about is the sound? You need to explain how what nonvna can measure relates to dac sound. Specifically, and technically.
Last edited:
Cool ddac !Mark, you are right. It does sound like squaring the circle. OK why start a discussion? I will buy one and connect it to my Andrea Mori Clocks, the Accusilicons and the included simple clocks from the FiFoPi and see what happens and if any correlation between "quality" can be found. Yesterday, I was doing some tests with my new DAC setup and in fact I felt I "needed" your board here
update: Amazon has them and delivery will be today before 10PM
Nice buy ! => plenty of nice vids on youtube to understand all that stuff is required
Honestly, I can't understand how a lot of people, here & there, talks about designs without feedback from any "basic" measurement.
Given that we are in 2024 and any DIYer can ask as Christmas gift a LiteVNA / TinySA Ultra / TinyPFA (all in once, or only one and LiteVNA is the basic of these 3 high-end devices)....
Simple question to Markw4 (cause I admit I'm lost there) :
how do you manage to get feedbacks from your designs ?
1. is it based on measurements ? on the tiny parts that can have an effect on SQ
2. is it an "overall" SQ feedback based on your own hifi setup ?
Dear Markw4,
I just quote this tiny piece,..
so what, you have no clue about a VNA can do ?
no brainer => many info via... the WEB ! LOL !
then, when you read some tech tech about "VNA" => you buy a NanoVNA or a LiteVNA (I've them all).
Oups !
My point here is not to "force" anyone to buy "any" devices stated above
My Point is to FORCE anyone
to understand how VNA & so on work !then...
if someone gots any idea thru his foggy brain how these VNA work (like me, a few month ago) => you buy a "cheapo" LiteVNA and play with it (many features & samples required !!!!!!!! but fun is there !!!!!)
On the contrary, I know what they do. They are two-port network analyzers. They can inject a frequency and measure reflected and transmitted power, including phase, and including spectral analysis of signals appearing at the input ports. From those measurements a number of things can be calculated. In fact there is an old, obsolete HP VNA favored by audio diy'ers because its good down to something like 10Hz. Most VNAs don't work below 9kHz or higher. Tiny VNA gets pretty noisy down at 1MHz and below. Also, it can be easily blown up by a too large of an input signal (isn't it limited to somewhere around 1V maximum input?).
What I want to know from you is exactly how nonvna is good for measuring close-in phase noise down to .1Hz and -170dBc, or exactly whatever you are claiming diy'ers need to be using it for in dacs? Please feel free to get specific.
Last edited: