gk2013
100+ Head-Fier
- Joined
- Oct 26, 2015
- Posts
- 242
- Likes
- 137
Default is PLL off; No matter which one you choose, the jiiter performance just can't compare to a femto clock.
This is absolutely true! You just hit the nail on the head.
Default is PLL off; No matter which one you choose, the jiiter performance just can't compare to a femto clock.
(1) J1 {PLLEN}, J2 {ATT0}
Master clock
98 Mhz - [On, On], [Off, Off] --- directly use TCXO 98.304MHz clock, clock jitter would be determined by 98.304 XO's quality
196 Mhz (PLL1) - [On, Off] --- use Cyclone's PLL function to multiply the input clock by 2, clock jitter would be determined by PLL jitter performance (200ps max according to Cyclone's spec)
294 Mhz (PLL2) - [Off, On]--- use Cyclone's PLL function to multiply the input clock by 3, clock jitter would be determined by PLL jitter performance (200ps max according to Cyclone's spec)
Master clock is used to drive the low pass filter and generate BCLK, WCLK and DATA for 1704s.
refer to "http://www.audio-gd.com/Pro/NOS/NOS7/NOS7EN_Use.htm"
In mode 1, you can use either PLL, in mode 2 you can only use PLL1 cause the FIR filter take too much computation power.
Default is PLL off; No matter which one you choose, the jiiter performance just can't compare to a femto clock.
About PLL, please refer to "http://bbs.audio-gd.com/dispbbs.asp?boardid=2&Id=30056&page=2"
It's audio gd's Chinese forum, sorry it is in Chinese.
About Bypass, Audio-gd won't recommend anyone use it; By email exchange with Kingwa, he just told me all functions will be disabled and after doing some testing it's easy to find out the rest.
Actually if you disconnect the 98Mhz XO or just cut off its power, the DSP-1 will automatically enter BYPASS mode.
That makes me wonder about the master-7 in-phase processing: is the i2s clock preserved when pll is turned off (i assumed so) or rounded up to some internal dsp clock period? If it is rounded up already, I will not loose the benefit of having a direct and exact femto clocks by going to the nos7. Same question is valid for the f-1, does it round-up the crystek or use it as is (or a sub-frequency)?
In Master 7, Master clock (MCLK) is taken from external source --- in the case of F-1, 22.5792 or 24.576Mhz depending on the input sampling rate; Then depending on PLLEN setting, it might be regenerated using Cyclone's PLL function; Now there are two possibilities, (1) Only use MCLK to drive oversampling/dithering function and pass F-1's BCLK and WCLK to 1794s. (2) Also use MCLK to derive BLCK and WCLK. In case (1) you won't experience any SQ difference by changing PLLEN setting. So obviously it's case (2) which is implemented.
There is no "rounding up" since DSP-1 doesn't have any internal clock other than the input MCLK, and if PLLEN is turned off, the output BCLK and WCLK would be a delayed version of F-1 BLCK and WCLK with worse jitter; In short the i2s clock will be preserved (but deteriorated a bit).
For F-1, it's used as it is:
384000*2*32 = 24576000
352800*2*32 = 22579200
Some words about femto clock: CCHD 575 might perform according to its spec, once it pass through F-1's cpld its jitter will get worse a little bit; the more logic gates it pass the worse the jitter. Remember the i2s signal also have to pass through (1) 74VHC125 - channel selection (2) Cyclone II to reach 1704s. We don't have official Singxer released test report, but quite sure it won't be as good as you might hope.
Instead of applying external power to the dc-dc converters of the clean side on the f-1, i would rather replace them. Simpler setup and i am worried of grounding problems. How easy to solder?
I assume you have read my posts "http://www.head-fi.org/t/803111/xmos-xu208-usb-bridges-the-latest-gen-has-arrived/3094"
There are two 3.3V ldo on the clean side, one for the XOs, the other for cpld. The one for XOs require the cleanest power. You certainly can remove the two ldo and use two external uV grade lps to drive the XOs and cpld directly, just make sure use shielded cables. Single digit uV noise level is not so easy to maintain at a distance over 20/30 cm. Grounding should not be a problem as long as you don't form a ground loop which will pick up EM waves. As for myself, my 0.8uVrms LT3042 just arrived yesterday, now I'm waiting for the circuit boards.
Soldering should be easy; Removing the two ldo on the other hand might be a little tricky if you haven't done this before. So please refer to "http://www.head-fi.org/t/803111/xmos-xu208-usb-bridges-the-latest-gen-has-arrived/3108".
Are you planning on feeding the LT3042 with 5V taken from the M7 board?
I assume you have read my posts "http://www.head-fi.org/t/803111/xmos-xu208-usb-bridges-the-latest-gen-has-arrived/3094"
There are two 3.3V ldo on the clean side, one for the XOs, the other for cpld. The one for XOs require the cleanest power. You certainly can remove the two ldo and use two external uV grade lps to drive the XOs and cpld directly, just make sure use shielded cables. Single digit uV noise level is not so easy to maintain at a distance over 20/30 cm. Grounding should not be a problem as long as you don't form a ground loop which will pick up EM waves. As for myself, my 0.8uVrms LT3042 just arrived yesterday, now I'm waiting for the circuit boards.
Soldering should be easy; Removing the two ldo on the other hand might be a little tricky if you haven't done this before. So please refer to "http://www.head-fi.org/t/803111/xmos-xu208-usb-bridges-the-latest-gen-has-arrived/3108".
Panel mounting a DC jack would be ideal, say replacing the RJ45 connector. I have not found something suitable yet.
BTW did you find out what the I2S header cable connector is? I was considering making my own cable but then could not find this connector, and also remembered metallic compatibility might be problem with copper/silver on the in plated pins. Do you think this would be a problem in this application?
I plan to replace the RJ45 connector too; what I have in mind is something like this "http://www.digikey.com/product-detail/en/cui-inc/PJ-005A/CP-5-ND/165838", and handcraft a metal plate with three holes - one for the socket, two for screw. Or I can remove RJ45 socket and components from the original RJ45 connector board and install something like "http://www.digikey.com/product-detail/en/cui-inc/PJ-002A/CP-002A-ND/96962" on the board.
I bought the 5 pin 2.54mm pitch connector (housing and pins separately, similar to "http://www.grtelectronic.com/e_productshow/?209-XHS-HA-TJC3-5pin-straight-angle-wafer-connector-209.html#image_3") from local electronic store. Then use Neotech OCC 7N 24awg single thread wire and WBT 4% silver solder to connect to F-1 board's 20 pin through holes directly. If you have trouble locate the connector locally, you might consider reusing the connector+wire from the original RJ45 connector board.
"metallic compatibility" --- Sorry I don't get it. If you means "Seebeck effect", the generated offset voltage is just couple uV for 10 degree temperature difference between copper and silver, and also consider I2S is digital signal, so it should be fine. As to if signal will distort across different metal interface, i have no idea.
I was concerned over electrolytic or galvanic reaction between different metal in contacts or in crimp connection (not something I understand), but I'm not sure how important this is, just I heard that previously some hifi companies tried using OFC wire with molex in amplifiers and ran into problems.
Hopefully local electronics supplier has something similar to that I2S connector also. Did you find normal crimp tool works?
In that case a silver plated occ wire might give you a peace of mind. Or solder the wire to the pin in addition to crimping like I did.
Don't have crimp tool so cant be sure; I use needle nose pliers instead. Strip about 2mm of the wire and fix the pin to wire's insulation layer with its longer arms first, then crimp the exposed wire with the shorter arms.
Thanks!
That is consistent with what i hear. The better the i2s signal, the better the sound, therefore it is a matter of clock performance.
I am very happy with how my m7 sound with audirvana raising depth to 24 bits using >1M samples time window (advanced settings). I use as hardware: jitterbug -> intona -> schiit wyrd -> f-1 -> 3-inch cat6a cable -> m7. All usb cable are very good, the last one using separate runs for signal and power.
Instead of applying external power as substitute to the dc-dc converters of the clean side on the f-1, i would rather replace them. Simpler setup and i am worried of grounding problems. How easy to solder?
Did you put an RJ45 socket on the F1 or did you hard wide a cable on it?