Dustin on the radio through Sabre, of course
Ross, could you elaborate on the reference system(s) that you heard? Amps, speakers, power conditioners, room treatment, the works. We wanna know what toys the boys at ESS get to play with...and to help frame a reference for us to compare our experiences with.
Straight ESS demo board, or something snazzier? Do you know what filter settings, modulator bits settings, etc?
WMS
krgaunt said:
Ross, could you elaborate on the reference system(s) that you heard? Amps, speakers, power conditioners, room treatment, the works. We wanna know what toys the boys at ESS get to play with...and to help frame a reference for us to compare our experiences with.
Straight ESS demo board, or something snazzier? Do you know what filter settings, modulator bits settings, etc?
WMS
Wavelength said:
Hi Gordon,
Looking forward to seeing what you measure. I have done measurements in my lab, and we hired an independent 3rd party to do the measurements, and we really wanted him to to give the straight scoop. I know its might seem fishy saying the we (ESS) hired someone to verifiy our results since there is a vested interest, since we paid him, but that was one of the terms of the contranct, the data was as is and it is what it is. Im really looking forward to more measurements made by others in the hope to just kee lning more. As a design Engineer, you feel no work is ever finished, but then the tapeout date comes......
Thanks
Dustin
Wobbly, what frequency is the Tentlabs clock for the CD Pro?
The Sabre DAC will work best with a clock 50Mhz to 80Mhz. You may find it best to clock the Sabre at 2X or 4X the rate of the transport.
Thus you would need to generate a low jitter clock for the Sabre and divide by 4 for the transport. I believe Tentlabs has a schematic for a clock divider circuit on their website.
The Sabre DAC will work best with a clock 50Mhz to 80Mhz. You may find it best to clock the Sabre at 2X or 4X the rate of the transport.
Thus you would need to generate a low jitter clock for the Sabre and divide by 4 for the transport. I believe Tentlabs has a schematic for a clock divider circuit on their website.
The Tent XO3 for the CD PRO is 8.4672MHz,I am amazed and somewhat bemused by the fact there is no real "standard" involved with the bitclock in I2S.
Here we have ostesibly the best DAC money can buy and the best red book transport as well, but they cant be used together with a simple "plug and play" approach.
I would never have immagined this, and would have been tearing my hair out for weeks trying to find my wiring "error" as the damn system wouldnt work.
Again my experience is severely limited in this area so any help given to implement this scenario is gratefully accepted.
Here we have ostesibly the best DAC money can buy and the best red book transport as well, but they cant be used together with a simple "plug and play" approach.
I would never have immagined this, and would have been tearing my hair out for weeks trying to find my wiring "error" as the damn system wouldnt work.
Again my experience is severely limited in this area so any help given to implement this scenario is gratefully accepted.
Reply from Guido - use an 8414 to recieve reclocked Spdiff from CD PRO. XO3 running at 33.8688 with divide by 4 ( 8.4672) to run CD PRO and divide by 3 ( 11.2896) to make 8414 in slave mode, output 64fs data.
Same clock to run Sabre.
Well, maybe a clever workaround but really defeats the whole idea of deleting Spdiff and running ( supposedly better) I2S.
And like im going to build THAT in an afternoon with no backup help.
Still cant believe this totaly unforseen balls up.
Same clock to run Sabre.
Well, maybe a clever workaround but really defeats the whole idea of deleting Spdiff and running ( supposedly better) I2S.
And like im going to build THAT in an afternoon with no backup help.
Still cant believe this totaly unforseen balls up.
wobbly said:
It appears you are making things unnecessarily complicated. Just
use a local dedicated, low jitter clock for Sabre and let it do it's
thing, which is reject jitter from whatever source you feed it. Spend
your money on Sabres XO and PS first. This XO should be right at
the DAC.
cheers
Terry
wobbly said:
As above:
I would suggest feed the Sabre Spdif and spend your money on the
local XO and it's PS.
I'd be surprised if there was any advantage running I2S over SPDIF.
The internal ASRC has a jitter rejection CF around 10x lower than other
previous products such as AD1896 and Anagram etc.
It would appear to be getting to the point where other things (than
jitter rejection) will make bigger differences... maybe I am wrong.
Has anyone here directly compared the two on Sabre, all else being
equal?
T
Terry Demol said:
Absolutely.
Along with about at least half dozen clocks so far, and several I/V stages.
My experience is this regarding I2S vs SPDIF. If you have a good SPDIF source there is absolutely no difference. Use the SPDIF if your can.
If you have a less than ideal SPDIF source you will need to open up the DPPL bandwidth to some degree to keep it locked which (very very slightly) decreases jitter reduction.
The Sabre has some (very minor) design flaws regarding SPDIF (such as not detecting pre-emphasis, and a longer setting time), those can be mitigated or removed by using a separate receiver.
Cheers!
Russ
Russ White said:
Absolutely.
Along with about at least half dozen clocks so far, and several I/V stages.
My experience is this regarding I2S vs SPDIF. If you have a good SPDIF source there is absolutely no difference. Use the SPDIF if your can.
If you have a less than ideal SPDIF source you will need to open up the DPPL bandwidth to some degree to keep it locked which (very very slightly) decreases jitter reduction.
The Sabre has some (very minor) design flaws regarding SPDIF (such as not detecting pre-emphasis, and a longer setting time), those can be mitigated or removed by using a separate receiver.
Cheers!
Russ
Russ,
I don't think I have ever known a CD or dig source to need
de-emphasis.
WRT settling time, that is purely a product of the vary low CF
DPLL. Does it ever actually not lock or is the issue purely long lock
time?
Also, did you hear any difference between the different XO's you
tried? If so, which one was the best and what are it's specs.
cheers
Terry
Now I have been beaten into submission and are forced to use Spdiff ( already reclocked with Tent and PS) I suppose im now looking at a reclock and PS for the Sabre. Would be interested of course in your comments Russ.
PS - Guido Tent also added that his dac is 48fs to suit the Philips and reiterated my dismay at various manufacturers arbitrarily deciding to use other formats for I2S - any particular reason the Sabre is 62fs???
PS - Guido Tent also added that his dac is 48fs to suit the Philips and reiterated my dismay at various manufacturers arbitrarily deciding to use other formats for I2S - any particular reason the Sabre is 62fs???
Hey Folks,
My experience with different clocks has been that the well known clocks you can get easily for ~$5 to ~$50 pretty well sound identical - which is to say absolutely fantastic. At least I have not been able to note any difference. Call me deaf if you like, but I am fairly confident I am hearing things just fine.
I found that even the lower end clocks from Crystek and Abracon sound fantastic in the buffalo. I found no audible difference between the $50 clocks and the $8 clocks. I did not feel like buying anything more expensive than $50. If someone wants to send me a ultra mega super cool audiophile approved clock to test, I will. Otherwise, I am already completely satisfied with the sound coming from the buffalo with the clock Brian and I chose for it.
For those of you who just have to try clock swapping for yourselves, the clock header is there to pipe in your own clock should you desire to try one, and the on board clock is also easily disabled.
The only clock I would probably not use is a cheapo crystal. But, honestly I did not give them a fair test.
I also played with dedicated CCS/shunt(LM4051 + EF) regulator for the clock, and could not detect any audible change, but this is to be expected as the clock's supply is the digital supply, and it is bypassed well. Also the clock and the rest of the digital side is very effectively isolated from the analog supply with a low pass filter.
On the buffalo the analog supply has it own pre-regulator when using the LCDPS and the digital supply is used simply as a voltage reference with a low pass filter into the analog VREG which is a dual opamp.
For anyone who really wants try it, the VREF low pass filter can be made sub sonic by soldering a 100uf low ESR cap across C6. I have done this, and it certainly works, but I can't say I noticed any change except that it takes a bit for VA to reach its target 3.3V. I think the Fc is quite low enough as it is without going to this extreme.
The more experience I have with the Sabre chip the more I believe that the output stage is the most significant key to performance, as small changes there can and do yield evident and consistent differences. So that is where I am dedicating most of my time these days.
Cheers!
Russ
My experience with different clocks has been that the well known clocks you can get easily for ~$5 to ~$50 pretty well sound identical - which is to say absolutely fantastic. At least I have not been able to note any difference. Call me deaf if you like, but I am fairly confident I am hearing things just fine.
I found that even the lower end clocks from Crystek and Abracon sound fantastic in the buffalo. I found no audible difference between the $50 clocks and the $8 clocks.
I did not feel like buying anything more expensive than $50. If someone wants to send me a ultra mega super cool audiophile approved clock to test, I will. Otherwise, I am already completely satisfied with the sound coming from the buffalo with the clock Brian and I chose for it. For those of you who just have to try clock swapping for yourselves, the clock header is there to pipe in your own clock should you desire to try one, and the on board clock is also easily disabled.
The only clock I would probably not use is a cheapo crystal. But, honestly I did not give them a fair test.
I also played with dedicated CCS/shunt(LM4051 + EF) regulator for the clock, and could not detect any audible change, but this is to be expected as the clock's supply is the digital supply, and it is bypassed well. Also the clock and the rest of the digital side is very effectively isolated from the analog supply with a low pass filter.
On the buffalo the analog supply has it own pre-regulator when using the LCDPS and the digital supply is used simply as a voltage reference with a low pass filter into the analog VREG which is a dual opamp.
For anyone who really wants try it, the VREF low pass filter can be made sub sonic by soldering a 100uf low ESR cap across C6. I have done this, and it certainly works, but I can't say I noticed any change except that it takes a bit for VA to reach its target 3.3V.
I think the Fc is quite low enough as it is without going to this extreme.The more experience I have with the Sabre chip the more I believe that the output stage is the most significant key to performance, as small changes there can and do yield evident and consistent differences. So that is where I am dedicating most of my time these days.
Cheers!
Russ