Topic: Zoltrix Transfers and their Bit Accuracy

[rustoleum]

I picked up a RatShack Digital Signal format convertor to test for bit accuracy and stumbled upon what seems to be an even bigger issue I wanted some input on. 

As an initial test I took a version of I am Hydrogen (nice quick tune) and put it onto DAT tape with silence on either side of it.   Using WaveLab4, I transferred the H2 via Core-Sound 7-pin -> Monster Digi Cable -> Zoltrix Nightingale (coax) -> WaveLab4 using a Sony TCD-D100 for playback.

I then did the same H2 again using Core-Sound 7 pin -> Monster Digi Cable -> Rat Shack Convertor -> toslink cable -> Zoltrix (toslink) -> WaveLab4

WaveLab4 was configured to "Auto-start if sound detected" and "Auto-stop if silence" so each WAV file should begin and end on the same sample making for an easy comparison of data.

After multiple tries, I was never able to get the 2 WAVs to match up using WaveLab's "File Comparer" tool and was all set to blame it on the convertor and call it  day.  However, as a Sanity check I decided to run the DAT tape through the coax again and compare against my original transfer.  To my surprise,  the files did not match.  After 5 more transfers from each (toslink and coax) I have not gotten a single file that will accurately match up with any of the others.  Each file size is ever so slightly different, WaveLab reports millions of differences and zooming in to individual samples in WaveLab shows that   the wavs are indeed different. 

To be completely sure I used EAC's WAV compare tool and came up with the similar results reporting minute differences in WAV length and missing and/or different samples between the two.  WTF!?!?  I though the Zoltrix Nightingale was a Bit-Accurate sound card? 

Any thoughts on this?  Is this something flawed in my testing?  Has anyone done similar testing of the zoltrix and found it to be accurate? 

On one hand, it means that the Rat Shack convertor may indeed be bit-accurate, but because the zoltrix is flaking out my tests have been inconclusive... on the other hand, WTF is going on with the zoltrix?

The only other thing I can think of is the DAT deck is flawed in someway. I'm going to test this theory out by running MiniMe -> USB into my Vaio and MiniMe -> Coax -> Zoltrix into my Desktop PC to take the D100 out of the equation.  I won't be able to get to this until tomorrow, so if anyone has any other thoughts on the whole situation in the meantime, I'd love to hear them.


Rusty


Coax Transfer Results as seen through a cmd window:

09/14/2004  03:35p          31,419,088    H2_coax_1084.wav
09/14/2004  03:40p          31,420,096    H2_coax_2085.wav
09/14/2004  03:46p          31,419,216    H2_coax_3086.wav
09/14/2004  03:58p          31,419,744    H2_coax_4087.wav
09/14/2004  04:07p          31,418,928    H2_coax_5088.wav

[dklein]

I tested my Nightingale and verified bit accuracy, though I only used the optical in.  I also found it kind of picky as far as other stuff going on when transferring.  I got in the habit of doing every transfer twice, then doing a paste invert - if it didn't null out, I did another transfer until I got 2 that nulled.

As far as your methodology goes, I wouldn't rely on auto-start and stop to match files precisely.  I always did this by zooming into a single sample and matching the start samples, then copy, paste invert (or mix the 2 in multitrack with one inverted).  Most software that compares files doesn't have the smarts to look ahead/behind and see if it matches better if you toss out the first few samples (did that make sense?)

Now, why aren't your DAT transfers matching each other?  I found that when the Nightingale 'glitched', it was easy to see using the paste invert approach.  You'd get a null up to a point and then all of a sudden the music would start again.  This could be happening.  Then there's the whole issue of varying s/pdif voltage levels (standards that aren't always followed)- optical connections should negate that problem.

I think that some folks also had trouble with the Nightingale doing 48k.  I always ran 44.1 so I'm not sure about that one.  I have an ftp server if you want to upload a few copies I can take a look.  pm me for login info if you're interested.

Which model of nightingale do you have?

[rustoleum]

Thanks for the reply!  I thought of the problem of using auto-start, too and have begun doing some transfers recording silence on either end and cutting to the start and end samples manually.  It's a little more work, but I'm optimistic that this will clear things up on the DAT -> coax side of things.  Only then will I feel confident enough to try testing the Rat Shack convertor again.

These were 44.1 K samples I was transferring and I believe that it's a Pro 6 Nightingale, although I'm not 100% sure as it's an oem card that I got new without a box, manual, etc.  It's based on the CMI8738 chipset and is a PCI card.

I appreciate the offer to look at some samples.  I'm going to keep playing around with it today while at work and will report back.  May need to take you up on that offer if I can't get some accurate comparisons.  Thanks and +T

-Rusty

[rustoleum]

ok... now we're getting somewhere.

After several unsuccessful attempts to match up my start and end samples of "I am Hydrogen" I decided to have WavLab generate some Sine curves to use for this test to make it easier to identify the samples.  I generated a 5 second curve that spans ~44 Hz all the way up to 20,000 Hz and copied it over 11 more times to create a 1 minute sample.  I fed the final test signal to my DAT deck via the Coax-digiout and fed it back in 2 times via the coax digi-in.

As expected, the samples were very easy to match up and the wav file comparison tools in both EAC and WaveLab have reported that the two samples are identical.  Looks like the zoltrix is behaving after all.

Now the cool part:  I fed the same signal in through the Rat Shack convertor and into the zoltrix optical in and the WaveLab reported it was identical to the coax transfers... Holy Shit, this Rat Shack thing may be bit-perfect afterall!

Going to run some longer tests (probably hour long) and vary the sine curves more to give the test more room to screw up, but so far I'm pretty optimistic about this Rat Shack thing.

Rusty