Computer interface measurement help required

[ddaudio]

I'm starting to digitize my vinyl collection, and to do so I'm using an old Sony Minidisc recorder for ADC (you can hit the record button with no minidisc in the unit and it does ADC (16/44.1 with no ATRAC compression), connected to the tape out of my amplifier and then its optical output goes into an old Foxconn mini PC which has an optical input onboard. I'm using Linux/ALSA to capture from the iec958 input to wav which I then transfer to a modern laptop for editing in Audacity.

I thought I'd check the transparency of my chain (not including the ADC bit) by comparing a 1khz test tone:
1. Flac directly loaded into REW
2. LMS --> Squeezebox Touch optical output --> Foxconn --> ALSA record --> REW
3. LMS --> Squeezebox Touch optical output --> Minidisc recorder optical in/out --> Foxconn --> ALSA record --> REW

The good news is that 2 and 3 are virtually identical, so the MD recorder in the loop is transparent. 1 seems better but in both cases the noise is 120 below peak except at high frequencies. However, see images below: I was hoping for something that looked a bit more like the graphs in Amir's reviews - can I adjust the FFT parameters in REW somehow?

Flac loaded into REW (left channel):

Squeezebox --> Sony MD --> Foxconn --> Alsa record (left channel):

 

[boxerfan88]

Erm … any reason why you don’t just get a high quality audio interface (eg. Motu or Topping) and record directly into Audacity?

 

[Blumlein 88]

Under the gear symbol in REW are settings where you can change size of the FFT. I think Amir uses 32k FFTs.

 

[ddaudio]

Erm … any reason why you don’t just get a high quality audio interface (eg. Motu or Topping) and record directly into Audacity?

Well, I could use Audacity on the minicomputer for the capture, it makes no difference to the resulting audio. But I'd rather do the post-processing on a more powerful machine, so a simple ALSA capture is all that's needed.

As to a new audio interface, it's all money. And I like reusing stuff I have hanging around, if possible. Would an external USB device be better than an internal soundcard with built in s/pdif input? Would I notice the difference using more modern ADC when recording old vinyl? I don't know, which is partly why I opened this discussion (together with wanting some help with REW).

 

[DVDdoug]

Would I notice the difference using more modern ADC when recording old vinyl?

No. The analog-side is the weak link... The record, cartridge, and phono preamp.

 

[Sokel]

The purple chart smells like a mic calibration file is used.
Delete it if you are about to measure electrically.

 

[ddaudio]

No. The analog-side is the weak link... The record, cartridge, and phono preamp.

That's what I thought.

 

[ddaudio]

The purple chart smells like a mic calibration file is used.
Delete it if you are about to measure electrically.

No mic calibration file used or even exists on my system.

 

[ddaudio]

Here's a better graph: there was obviously something wrong with my original 1khz sine wave file, which was downloaded from Focal! This is with an REW-generated sine wave: the red is the original, the blue is after passing through my capture chain as described above. I'm definitely happy with this, the noise floor of my vinyl is going to be way above those artifacts.

 

[Blumlein 88]

Here's a better graph: there was obviously something wrong with my original 1khz sine wave file, which was downloaded from Focal! This is with an REW-generated sine wave: the red is the original, the blue is after passing through my capture chain as described above. I'm definitely happy with this, the noise floor of my vinyl is going to be way above those artifacts.

View attachment 435188

Was this at 32k FFT sizing?

Maybe good to do a short bit with the needle in the lead in groove of an Lp before music starts.

 

[ddaudio]

Was this at 32k FFT sizing?

Maybe good to do a short bit with the needle in the lead in groove of an Lp before music starts.

No, that was just opening the files in REW. Here is a graph showing the results of 32k FFT, Hann window on the above files plus the run-out groove of one of my LPs.
Red: original 1khz tone wav
Blue: 1khz wav after passing through the capture chain
Lilac: run-out groove captured via Sony MDS-JE510 ADC monitor

 

[boxerfan88]

Is the runout groove chart after/before RIAA correction?

 

[ddaudio]

Is the runout groove chart after/before RIAA correction?

After. Sounds good when live monitoring after AD - DA. Like most of the records I want to digitize, it's second hand, of unknown provenance, cleaned to the best of my ability. Maybe I should run the same test with one of my few new, pristine LPs.

 

[Blumlein 88]

I think that shows pretty conclusively that your recording chain has fidelity well beyond your source material on LPs. So record away without worry. Only at the highest frequencies (13 khz and above) might noise of the recording chain enter into it. It will be masked by everything else and should still capture any signal up in the higher frequencies when music is playing.

 

[ddaudio]

I think that shows pretty conclusively that your recording chain has fidelity well beyond your source material on LPs. So record away without worry. Only at the highest frequencies (13 khz and above) might noise of the recording chain enter into it. It will be masked by everything else and should still capture any signal up in the higher frequencies when music is playing.

Especially since my poor old ears can't hear anything above 12 khz, anyway!

 

[ddaudio]

Just for fun, here's the graph again with an extra capture - this time the 1 khz sine test file through:

LMS -> Squeezelite -> SMSL PO100 USB to optical converter -> Cambridge Audio 840C DAC -> analog to Yamaha receiver -> MD recorder ADC -> optical to Alsa/Audacity.

Something is definitely chucking in a 50hz mains artifact.

 

[ddaudio]

This is interesting: I thought the rising noise above 13 khz was due to the nearly 30 year old ADC technology in the Sony recorder. But no, using sox in a terminal instead of Audacity makes it go away, maybe because the capture device has only 2MB memory and an Intel Atom processor, so struggles running the window manager/Audacity gui.

 

[Blumlein 88]

Buying one of those RCA to SPDIF pocket ADCs from miniDSP for $55 might be money well spent if you are doing a large number of vinyl disks. You could record straight to the modern laptop.

 

[ddaudio]

Buying one of those RCA to SPDIF pocket ADCs from miniDSP for $55 might be money well spent if you are doing a large number of vinyl disks. You could record straight to the modern laptop.

I'd need USB input to my laptop, so a bit more expensive, I think. I can control it all from my laptop anyway - ssh or remote X session. Just need to power on the mini pc at the same time as I switch the turntable on

 

[AnalogSteph]

This is interesting: I thought the rising noise above 13 khz was due to the nearly 30 year old ADC technology in the Sony recorder. But no, using sox in a terminal instead of Audacity makes it go away, maybe because the capture device has only 2MB memory and an Intel Atom processor, so struggles running the window manager/Audacity gui.

I'd suspect that Audacity is (needlessly) applying dither. It shouldn't be doing that when recording in 16 bit and exporting as such but you never know... change quality settings and you should find out in short order. I think I've previously run across an issue like that in older versions.

BTW, Intel Atom suggests netbook-grade hardware so I presume that's 2 GiB worth of memory and not just 2 megs (which would be more befitting of a 286 or 386). In which case it would be more well-appointed than the machine I used to make lots and lots of radio recordings using Audacity 2.0.6 15-20 years ago.

On another note, it seems you can only record straight to 44.1 with the MDS-JE510... here's to hoping that the ADC anti-alias filter is sufficiently bombproof, not to mention anything about its passband ripple (a measurement of frequency response would be interesting, noting that you're looking for variations sub-+/-0.1 dB). Odds are it's a fair bit better than the Aztech ISA card (probably with a CS4248 codec) that I did my first vinyl rips with way back when, but possibly not quite as good as the WM8770-based jobs I relied on for years (although that one didn't exactly have good filter ultimate rejection at 65 dB, so oversampling would be recommended).

It would appear that measurements for the ADC sections of MD decks are a bit hard to come by. We roughly know what to expect in terms of dynamic range (probably low-mid 90s dB(A)), and distortion should be low enough either way, but nobody ever tests filter ripple and out-of-band rejection.