|
Home
/ FAQ
/ News Classifieds / Events |
Audio Asylum Thread Printer |
Get a view of an entire thread on one page |
84.138.120.32
In Reply to: RE: But does the BUFFALO chip upsample as well? posted by rickmcinnis@dogwoodfabrics.com on May 29, 2009 at 19:53:32
Rick,yes, I've had this confirmed by the designers of the Buffalo32S, the oversampling FIR filter is also clearly mentioned in the data sheet (although they don't really disclose how the upsampling is done and to which frequency), the ESS9018 chip has what they call the Time Domain Jitter Eliminator which basically is upsampling and reclocking, as with your Burr Brown a technique that is commonly used. The ESS9018 is highly praised for exactly this feature, it is said to be very efficient and to do things very well, but I agree with you, I have yet to hear a solution where a hardware upsampler does a better job (or at least not worse) than a software upsampler like SRC or SoX.
And yes, the chip is designed with the ability of turning this feature off, it requires setting a few registers via i2c, but the firmware used by the Buffalo32S designers does not offer that functionality, you need a custom firmware for that. Surely the Buffalos are a very fine piece of audio design and manufacture, but I believe the chip can do even better. Since like you I do not want the DAC to do the upsampling, plus I do not think that our cMP2 environment produces too much jitter anymore that needs attenuating (the measurements cics had done in March last year were already impressive, and so much improvements have been made since then), I will build my own DAC later this year with two ESS9018 chips in dual mono configuration (recommended by ESS for best audiophile performance) and compare - at least that's the plan ;-)
Best,
Robert
Edits: 05/30/09 05/31/09
Robert,
I'm planning same thing.
I was reluctant to select Sabre unit over TI 1704 for the same exact reason you mentionned, ie internal upsampling / downsampling.
This technique allows the widest compatibility with the different sampling rates sound format availlable (everything from 32k up to 192k)
Although this is common practice in Silicon SRC and good to rmove jitter of common source, here we have a machine and a software that provide us with minimum jitter at the PCI Sound Card output which mean (if well implemented) at the DAC Input.
I'm thinking for my dac implementation to try to achieve as low jitter as possible all the way down to the analog which means the use of a Master Clock close to the DAC chip.
The problem here if one wants to avoid too much complexity implies to select a single as universal sampling rate (178K4 or 192K).
Should you select 178k4 it would be perfect for original files sampled at 178K4 and close to perfect with 88k2 and 44k1 providing that you select the right settings in Sox (that i suspect may be full system dependant).
So the question is what will be the standard sampling rates (if any) for Hi res files pick the right clock for that standard and accept imperfection for other sampling rates.
Now that you explained us that Sabre can be forced not to use its internal SRC, it comes back to interrest to me because of its ability to work with 32 bits resolution.
So, if you move on with this project and if we can share energy, you can count on me.
Best,
Jean
Jean,absolutely great to hear you're heading down the same path - or rather up :-)
Exactly, the first assumption I'm making is that we in our cMP2 environment have eliminated or reduced jitter to such an extent that the downside of on-chip jitter reduction techniques, e.g. hardware upsampling and reclocking, more or less clearly outweigh the benefits of doing the upsampling in software and closely coupling the DAC via i2s - in my expectation this is superior, but we'll have to test and see. IMHO the ESS9018 is currently the best DAC chip around, but need to properly re-evaluate.
Secondly, I'm also with you that it is key to use the best master clock available on the DAC. As far as I can see, that is NewClassD's Neutron Star (http://www.newclassd.com/index.php?page=36), Lars has done an outstanding job with this.
Thirdly, the question of selecting one "standard" sampling rate for everything and optimizing for this is easy for me I thought - since all of my playback is through cPlay, I intended to use 24/192 (or rather its 32bit incarnation through ASIO2) through SRC or SoX exclusively. Your comment on 176k4 though suggests that this probably might be the best choice (another assumption I make for me is that most important for me is data derived from red Book CDs, thus 16/44.1, that's what I focus on), probably wil go for that, need to get more info on this, and then choose clock and settings etc. accordingly.
And lastly but most important for me, I'll be absolutely glad to sync with you and join forces once I get started with this part of the project - look forward to this very much! Please PM me (if possible) so I have your coordinates.
Best,
Robert
Edits: 05/31/09 05/31/09
Hi Robert,
Glad to hear from you.
I've selected 176k4 as the best for CD material for the exact same reason mentionned earlier.
Up to now i could only test on my reference system (i'm hifi dealer) with 44k1 and 88k2, to be honest i didn't even try 96k.
But i suspect that upcoming hi res material will be 192k ...
Reagarding clock i was also interrested in Audiocom stuff. Also praised as the best one in town.
The next important thing is power supply and IV conversion.
I will be playng soon with different PSU for evaluation.
My problem with Sabre is how to get it, as i don't want to use Buffalo, buffalo one not any more availlable and at least no more support, buffalo two is too expensive if you take into account that i consider changing IV and PSU. Second big problem Board design.
That's why i am on the conservative approach thinking to use TI 1704.
Best regards,
Jean
Jean,you're probably right that over time there might be a shift to 192kHz material - for the time being for me this still is almost irrelevant since by far most of my music ("old" one I have as well as "new" one I buy) is 16/44.1 because it comes on CD. Let's see how SRC vs SoX as part of cPlay develop, I'd very much be interested in cics' opinion on what sample rate he sees best for Red Book material (176 vs 192) and for what reason (SNR? etc.)
Audiocom is certainly good - need to ask Lars why he's better :-) My knowledge in clock technology and architecture needs to be expanded and updated before I will decide on which to choose.
Well, power supply - you certainly have seen that I fancy battery plus proper capacitance - can only recommend that, need some extremist spirit though ;-)
Buffalo32S IMHO is not expensive at USD 469 at all, given the top-of-the-hill solution you get with it. I'm currently setting it up as my reference, then quite likely will go with 2 x ESS9018 evaluation boards as a next step (not a bargain at 2 x > EUR900), and then take the experience of those two to get to some "final end-all-search" DAC, who knows ;-)
Cheers
Robert
Edits: 05/31/09 05/31/09
I know we will all look forward to your implementation.
Bye,
Rick McInnis
Post a Followup:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2009, Audio Asylum, all rights reserved ]
[ Go Green ]
[ Take5 ]