Is HiMD optical in bit accurate?

[poorlyconditioned]

Let me say that my preliminary experiments indicate that the digital input is

*NOT* bit accurate!

I recorded a "click track" generated by Audacity (audacity.sourceforge.net).

I then played it out of an Edirol UA5 optical input into the MD (MZ-NHF800).

I then captured from MD to computer using USB, and captured to WAV by

playing the track from SonicStage and recording using TotalRecorder.

The original and final copies were compared in Audacity.

The result is that the file broken, both in levels and in timing. It appears

to apply some variable gain to the clicks, probably due to AGC. It also

appears to mess up the timing, there are dropped samples and slight time

variations.

I tried manual level controls, setting the level to various things, like 20,

25, 30 (out of 30). I'm investigating this now, but I am not optimistic.

I am really disappointed. Recording with digital input should produce an

exact copy. This is bad, both for listening (eg., transfering CDs to unit in

realtime) and field recording (eg., an outboard A/D).

[aeriyn]

If you didn't record in linear PCM on the device, your test is useless, since the ATRAC DSP is compressing the data and altering the waveform.

If you did record PCM... than... this is definitely not good.

[bhangraman]

I'd say there are too many variables in the equation to start pinning it on Hi-MD at this time... especially with a USB sound card too (I'm not referring to the deficiencies of the card but with the PC with USB). Have you tried an initial loopback test using Total Recorder in the chain?

[jonahn]

Sorry to re-open an old thread... but has this been tested any further?

I've done some tests of my own and I'm also finding that PCM recording from optical input is not bit accurate - can anyone confirm or deny this?

[jadeclaw]

The problem is that almost all soundcards are not bit accurate.On Samedayaudio one customer reported problems with his UA5. Clicks and other problems. Another one claimed, that the UA5 is not bit true.

Expect the fault not in the NHF800, instead expect your soundcard at fault here. Especially when the soundcard resamples everything up to 48kHz only to have it downsampled to 44.1kHz by the recorder.

Some points to observe: When the specifications lists only 48kHz for playback, then you have a resampler. That is about 70% of all USB-Audio and about 90% of all AC97 compliant soundcards including almost all onboard sound thingies. Second, when you can adjust the volume of the digital out, then that means goodbye bit accuracy. My soundcard loses about 0.5dB on the digital out in Windows. In Linux, digital out is fixed, I will test it next week for accuracy.

My experience is, that some soundcards perfom better with Linux. Seems to be a driver issue...

[jonahn]

I wasn't using a soundcard :wink:

From what I can tell, it looks like the unit is doing sample rate conversion on the digital input, even if the sample frequency is already 44.1kHz...

[reefbeef]

Jadeclaw,

Have you tried the latest Intel 'HD Audio'? Is it better in this regard?

[dex Otaku]

Interesting bit here: [no pun intended]

RandomHajile2 pointed this article on minidisc.org out to me while we were discussing something else:

http://www.minidisc.org/mj_ja3es.html

Assuming that HiMD's processing is a descendant of the ATRAC3.5 chips [a fairly reliable assumption] .. scroll way down, right near the bottom, and you'll get the news - YES, MD -always- resamples the digital input. Or, more accurately, it reclocks it to help eliminate jitter.

Which means that it's next to impossible for anything to be recorded with excat bit-accuracy from a digital source, because the resampling will always impose some [probably extremely minor] shifts in timing, and probably be interpolating samples when jitter actually does occur.

Or at least, that's what I got out of it. Still - I'm making the assumption that what applied to the 20-bit I/O pro deck in 1995 has made it into the portables of 2004. Which still seems a fair bet, to me.

[KCT]

I had encuntered similar issues when using ROLAND optical interface with Windows XP, it turned out

working just fine when using Windows 95/98. So I believe the MS operation system was at fault not

the MD unit or Sound Card.

Cheers

KCT

[guymrob]

The problem may be due to the sample rate converter. The sample rate converter reformat the incoming signals even though the sample frequency is 44.1kKHz, it does not bypass the sample rate converter so bit to bit recording is not possible in PCM mode. Try recording a HDCD encoded CD via optical digital to HI-MD in PCM mode. Then download the file and burn it to CD-R. Play it on Windows media player 10 and observe if the HDCD logo appears. If it doesn't it means information is lost.

The second factor is the digital gain level control. This will digitally add or substract the digital data if the the gain control is not set to neutral (centre). In Hi-MD there's no neutral in manual setting. However, The Onkyo MD-133 may able to do bit to bit recording accuracy coz it may have a 'Direct digital pass circuit' which bypass the sample rate converter if the frequency is 44.1kHz. (Anyone can confirm this???)

As for jitter, sample rate converter is notoriously bad, it induces high rate of jitter into signal and may affect the sound quality.

Edited September 17, 2006 by guymrob

[The Low Volta]

  guymrob said:

Try recording a HDCD encoded CD via optical digital to HI-MD in PCM mode. Then download the file and burn it to CD-R. Play it on Windows media player 10 and observe if the HDCD logo appears. If it doesn't it means information is lost.

indeed, but that could also be the fault of the download proces and reconversion from PCM to wav... therefore this is no proof of non-bit-perfect recording but only of a non-bit-perfect 'full recording AND downloading'-cycle