I suppose it depends on your expectations. If a PLL reference is enough to meet yours then that is all to the good.in that case, the "reference" is the PLL of the TinyPFA, BUTit's way enough vs precision required in there !
The TinySA arrived, but as Mark mentioned, these devices can be easily killed by signals above 10dBm. So I will have to wait for an ordered -30dB attenuator before I connect it to the Andrea Mori clocks (18dBm). I did test the device with my signal generator yesterday evening and I can already say there is no way this will produce anything useful for close in phase noise (needed for DACs). The device's own phase noise is according to manual: -90dB/Hz at 100kHz offset and -120dB/Hz at 1MHz offset... No mentioning of 10Hz or 100Hz...
but maybe, maybe, it wil show differences between clocks, which might point into a direction of quality difference. But you will not have useful absolute data. I will try and come back with some results, whatever these will turn out to be.
for the little money it costs, is a nice addition to my other test equipment. But definitively not more.
in the meantime, this is a nice white paper from the experts: see attached pdf
make sure to read this: 2.5 Challenges/limitation of the spectrum analyzer method
but maybe, maybe, it wil show differences between clocks, which might point into a direction of quality difference. But you will not have useful absolute data. I will try and come back with some results, whatever these will turn out to be.
for the little money it costs, is a nice addition to my other test equipment. But definitively not more.
in the meantime, this is a nice white paper from the experts: see attached pdf
make sure to read this: 2.5 Challenges/limitation of the spectrum analyzer method
Regarding VNAs, I wouldn't say they are necessarily useless for dac work, but a multi-meter and a dual channel, 100MHz scope (or better) are far more important to have first. A VNA with a set of field probes can be useful for things like helping to sniff out radiated EMI/RFI noise, for example. That said, a scope probe ground lead clipped to the probe tip can sometimes be used as well.
Tested the TinySA today. Long story short, if you use the phase noise function and the clock is better than the tiny, you only measure the tiny… I compared the sine wave from the Andrea Mori clocks and compared it with a sine wave from my signal generator with 120dB phase noise at 100k. so not kind of high end…. I could see no difference at all. The values are always the limitation of Tiny, which is crap.
In normal FFT mode, I used highest possible resolution and points, no difference visible.
Still a very handy tool, to check HF clocks if all is fine. You see the frequency, harmonics and dBm level, which is nice. I will not send it back to Amazon…
So we are still stuck with the dilemma, that real test equipment is out of reach for most of us. At least for me ….
For me case closed…
In normal FFT mode, I used highest possible resolution and points, no difference visible.
Still a very handy tool, to check HF clocks if all is fine. You see the frequency, harmonics and dBm level, which is nice. I will not send it back to Amazon…
So we are still stuck with the dilemma, that real test equipment is out of reach for most of us. At least for me ….
For me case closed…
hi Markw4,
(please reply straight if this post is not clear because we have a party running... and... I'll have to read this post 10x b4 sending it to be sure it is OK
happy saturday night fever to anyone ! ; in such a "state", the proper adjective is french is "pompette", a state where you're drunk but "the smart way", not at all "booze" like)So,
- to answer to your question => no link between NanoVNA / phase noise measurement. Of course => back to youtube & the many vids that show how a "vna" works... & CAN help you !
- a very nice & right video anyone must watch on youtube is from "IMSAI guy" about how he got to electronics, like me given both we don't have an EE degree, but we "can learn" like anyone (sorry I can't find the link to this very kind & cool video from IMSAI Guy...
But overall, my thought about the last posts is : "WTF, in audio stuff, they level down to a BUNCH of components",
BUT they have no clue about what's going into this BUNCH of components !
So, LiteVNA (the best VNA) / TinyPFA (the best & UFO-like in its category) => are the "Must-have tool of any diyer".
Given their price... BUY & PLAY with them.... & LEARN from your "BIG PCB main ouput" / tiny-basic output PCB disign
The key stuff is, how IMSAI Guy said in his vid => "DO !"
do tiny part of circuits => run measurements thru VNA / TinyPFA etc and draw you conclusion vs this PCB vs that one...
Sorry but I'm realizing I'm doing blabla about basic stuff there ....
ok, happy eve to anyone !
Clock board v2.1 (with PLL) is starting to warm up:
Found two old NDK SDA 22/24MHz clocks on adapters in a drawer. Area outlined in red is the PLL area. Its programmed by two solder jumpers. The contingency inverters had to be moved over to the yellow outlined area. Added about 2.5mm to the length of the board. PLL is correctly outputting 45/49MHz. It is (optionally) fed through JP14 as can be seen in the yellow outlined area. Also, right now the PLL is solder-jumpered to run from the Auxiliary Power regulator. In addition, the auxiliary regulator now has a 47R load resistor to ground. Another change is the clock dividers can be bypassed with solder jumpers (which they are in this case).
Will give it overnight to settle-in some, then try it with a dac tomorrow.
Found two old NDK SDA 22/24MHz clocks on adapters in a drawer. Area outlined in red is the PLL area. Its programmed by two solder jumpers. The contingency inverters had to be moved over to the yellow outlined area. Added about 2.5mm to the length of the board. PLL is correctly outputting 45/49MHz. It is (optionally) fed through JP14 as can be seen in the yellow outlined area. Also, right now the PLL is solder-jumpered to run from the Auxiliary Power regulator. In addition, the auxiliary regulator now has a 47R load resistor to ground. Another change is the clock dividers can be bypassed with solder jumpers (which they are in this case).
Will give it overnight to settle-in some, then try it with a dac tomorrow.
Last edited:
Oops! Soldered the wrong relays on there
Correct relays should be here in a few days.
Other than that, it worked in the dac. However, my impression was the NDK SDA clocks sound rather rough at least as compared to much better clocks. Of course, maybe some more running time would help. Those particular clocks hadn't been run in years
Correct relays should be here in a few days.
Other than that, it worked in the dac. However, my impression was the NDK SDA clocks sound rather rough at least as compared to much better clocks. Of course, maybe some more running time would help. Those particular clocks hadn't been run in years
Last edited:
Clock Board with PLL is now operational. Fitted with 22/24MHz Crystek clocks at the moment (with .22uf Rubycon 805 SMD film caps soldered to the underside of the clock adapter boards):
Sounds better than NDK SDA to me. Always been my experience in the past too, at least of if Crystek are fitted with the bypass caps they like.
The USB board is now being driven at 45/49MHz by the PLL output.
Also, looking at the PLL output with a scope, looks like it is roughly in quadrature with the PLL input at 22/24MHz. However, reclocker board is handling the time shift fine. That said, if there were to be some timing problem, the solution could be one possible use for the contingency inverters at the end of the board; they could be used to introduce a small delay and or inversion if needed.
With that, I think this thing is working as I said I would do for the forum. Probably will upload supporting files later to today or tomorrow morning. As this is diy project with the full KiCad files, there will probably not a be a formal BOM. The board can be configured in different ways with some different parts values depending on intended use. The Clock Board Notes document should be enough to provide any needed information. If not, please feel free to ask.
The next topic should probably be what to do about a reclocker board (for those folks who don't feel comfortable doing their own). However, it may be that not any particular reclocker board design will be well suited to every dac.
Regarding the performance of Marcel's dac with the setup as it is configured here (except using an Iancanada clock instead of Crystek), it was starting to sound better than my Andrea Mori dac setup. Which is to say Marcel's dac can sound very good indeed with no output stage other than DC blocking into a line level transformer.
Would also note that while waiting for the correct relays to arrive, the past few days were spent modifying my Andrea dac. Its now back in 1st place but it took some doing. I have emailed Andrea to advise him of the changes.
Sounds better than NDK SDA to me. Always been my experience in the past too, at least of if Crystek are fitted with the bypass caps they like.
The USB board is now being driven at 45/49MHz by the PLL output.
Also, looking at the PLL output with a scope, looks like it is roughly in quadrature with the PLL input at 22/24MHz. However, reclocker board is handling the time shift fine. That said, if there were to be some timing problem, the solution could be one possible use for the contingency inverters at the end of the board; they could be used to introduce a small delay and or inversion if needed.
With that, I think this thing is working as I said I would do for the forum. Probably will upload supporting files later to today or tomorrow morning. As this is diy project with the full KiCad files, there will probably not a be a formal BOM. The board can be configured in different ways with some different parts values depending on intended use. The Clock Board Notes document should be enough to provide any needed information. If not, please feel free to ask.
The next topic should probably be what to do about a reclocker board (for those folks who don't feel comfortable doing their own). However, it may be that not any particular reclocker board design will be well suited to every dac.
Regarding the performance of Marcel's dac with the setup as it is configured here (except using an Iancanada clock instead of Crystek), it was starting to sound better than my Andrea Mori dac setup. Which is to say Marcel's dac can sound very good indeed with no output stage other than DC blocking into a line level transformer.
Would also note that while waiting for the correct relays to arrive, the past few days were spent modifying my Andrea dac. Its now back in 1st place but it took some doing. I have emailed Andrea to advise him of the changes.
Last edited:
Those Hakko soldering vices look marvellous !
Can you post a better picture of them Mark ?
I can't seem to find anything with a google search so assuming they are 'historic'
A picture would help me to appreciate how to DIY some.
Thanks.
I believe these are the same thing: https://www.amazon.com/Hakko-C1390C-Omnivise-Holder-Fixture/dp/B000B61IHW/ref=asc_df_B000B61IHW/
They sell for a higher price if advertised differently:
https://www.amazon.com/Hakko-C1390C-Holder-Stand/dp/B00N3M1G46/ref=sr_1_318941380&s=hi&sprefix=hakko+soldering+vic,tools,155&sr=1-3
Really good for their intended purpose though. Highly recommended.
Last edited by a moderator:
Hi,
That's interresting the acrylic decoupling cap had not given the expected results with the NDK SA. I noticed not only the acrylic caps were good with the CDHH 957 but also it was sensible according the capacitance choice and also according the setup for the tonal balance you hear at the end through the loudspeakers. I finallly used it not only for decoupling HF but also as decoupling little local power bank with 1 Uf (in my set up it gave an audible difference VS only 0.1 uF)
I wonder how your board could give with accusilon you sligthy prefer over the Chrystek at feeding the JLSOUNDS board via the pins header if I remember correctly that possibility in its datasheet ? Could be interresting for us poorest diyers.
Btw thank you to share
That's interresting the acrylic decoupling cap had not given the expected results with the NDK SA. I noticed not only the acrylic caps were good with the CDHH 957 but also it was sensible according the capacitance choice and also according the setup for the tonal balance you hear at the end through the loudspeakers. I finallly used it not only for decoupling HF but also as decoupling little local power bank with 1 Uf (in my set up it gave an audible difference VS only 0.1 uF)
I wonder how your board could give with accusilon you sligthy prefer over the Chrystek at feeding the JLSOUNDS board via the pins header if I remember correctly that possibility in its datasheet ? Could be interresting for us poorest diyers.
Btw thank you to share
Hi diyiggy,
Once again its thanks to you for finding the caps that made Crystek CCHD 957 sound good. It later turned out the same caps also made the clock buffers sound good. In fact, to some extent it seems the more caps peppered around the board the better. OTOH, IME these caps in larger sizes don't seem very good for analog audio.
Regarding Accusilicon, I haven't directly compared them to Crystek. The Accusilicon clocks are 45/49MHz, whereas the Crystek clocks I have running now are 22/24MHz. The different clocks are on different boards which are setup differently. Thats why I say it depends on what you want to do with a board because it has some influence on how you configure it.
Regarding NDK SDA, I have never had good results with them going back years. I got some selected ones from Jocko Homo (RIP). He was baffled as to why I found they didn't sound as good as other clocks.
Once again its thanks to you for finding the caps that made Crystek CCHD 957 sound good. It later turned out the same caps also made the clock buffers sound good. In fact, to some extent it seems the more caps peppered around the board the better. OTOH, IME these caps in larger sizes don't seem very good for analog audio.
Regarding Accusilicon, I haven't directly compared them to Crystek. The Accusilicon clocks are 45/49MHz, whereas the Crystek clocks I have running now are 22/24MHz. The different clocks are on different boards which are setup differently. Thats why I say it depends on what you want to do with a board because it has some influence on how you configure it.
Regarding NDK SDA, I have never had good results with them going back years. I got some selected ones from Jocko Homo (RIP). He was baffled as to why I found they didn't sound as good as other clocks.
Last edited:
thanks for feedback Markw4 ; (and thanks for the thanks)
Tempted by Marcel Dac but not sure I readed enough to understand the USB digital front end needed, the power supply needs and the average total cost of it (BOM). I surely PM you one day or another in the couple of months to ask you.
I remember also Jocko sorted a lot also crystal from Citizen iirc... What a good job he made ; (yeah RIP Jocko)
Edit : btw, wonder how the good priced single channel AD8597 (and good enough over the AD797) could perform over the modern NJM 5534 iteration and else choices of Operational amps in the RTZ thread, but that is off topic here. Certainly is worthing a benchmark over te 1602. While since since all the talks with the double channel 4562 with its equal sota PSRR and CMNR number... with so much suceptibility with EMR/EMI ? I use AD8597 with good results as a buffer but very do not know if having the quality for filter conf. ?!
Tempted by Marcel Dac but not sure I readed enough to understand the USB digital front end needed, the power supply needs and the average total cost of it (BOM). I surely PM you one day or another in the couple of months to ask you.
I remember also Jocko sorted a lot also crystal from Citizen iirc... What a good job he made ; (yeah RIP Jocko)
Edit : btw, wonder how the good priced single channel AD8597 (and good enough over the AD797) could perform over the modern NJM 5534 iteration and else choices of Operational amps in the RTZ thread, but that is off topic here. Certainly is worthing a benchmark over te 1602. While since since all the talks with the double channel 4562 with its equal sota PSRR and CMNR number... with so much suceptibility with EMR/EMI ? I use AD8597 with good results as a buffer but very do not know if having the quality for filter conf. ?!
Last edited: