Lossy or 32Khz DAT

[corsair]

I have a recording whose source is supposed to be going
DPA 4061 > Sony TCD-D8.

However, after conducting some test with both AnalFreq,
and Spectral Analysis with EAC. I'm not so sure
if this is mp3 source or just a case of 32KHz DAT compression...

[range_hood]

From what I can see here:

EACs spectral view is bad.
The frequency response goes up to about 16-17kHz, so this could be 32kHz DAT (theoretically frequency response is half of the sampling rate).
But it also could be mp3, from what we have.

Get Cool edit, take spectral shots of about 2 seconds and post it in png format or at least gif.

Spotting Lossy Sources--links inside (http://www.thetradersden.org/forums/showthread.php?t=4288)

[Five]

agreed, hard to see from these pics. still could go either way

[corsair]

Ok... here's what I did after downloading and installing CoolEdit 2k.

I zoomed in onto the wav, till its only ~ 2secs, and did a spectral view.
Correct me if I'm doing it wrong :)

[range_hood]

Could you post just the spectral pictures so you do not have to resize the picture? I think one channel is enough on this one.

If you want you can upload a 10 second sample on http://www.yousendit.com/ for further investigation. Just fill in your e-mail address and post the link to the file.

[corsair]

Something seems to be not right with my spectral picture...
Why is it only in greyscale, while on the rest of website, I see multi colors?
Can you enlighten me on the settings to change in CE Pro?

I'll get home later and re-do the image and upload a sample to you. :)
Thanks alot!

[corsair]

Ok.. I've redone the spectral views..
I've included 2 samples... each is from different track.

Thanks!

[corsair]

Link to the sample:

http://s24.yousendit.com/d.aspx?id=1D7CRAAKLBJ4G3DJ2YRIM28U7O

[range_hood]

I´m not that experienced in 32 kHz Dats, so I just try.

I´ve found this in my collection:
nirvana1988-12-21
ANA(x) > DAT(1)@32khz > CDR(5) > EAC > WAV > FLAC > shntool > flac


Converted this to 320kbps mp3 for comparison it almost looks the same in coloured spectral analysis.

http://img150.imageshack.us/img150/6185/0132khzdatcolor9jk.png




Switching to b/w mode and some sort of difference is noticable. Look at the artificial looking cut at 16 kHz (mp3 pic).

http://img150.imageshack.us/img150/9427/0232khzdatbw9zf.png




On the last one the y-axis is also zoomed. In the mp3 pic you can see this tiny, little blocks around the cutoff. This is an indicator for a lossy encoded audiofile - this time.

http://img150.imageshack.us/img150/5121/0332khzdatbwzoomed1qr.png


.

[corsair]

This is how mine looked like...after zooming in up close...

Seems more like a 32khz DAT after all. :D

[corsair]

I decided to encode my wav file to mp3 320Khz. Wow...the initial SA, when its NOT black & white looks almost the same. I did the same zoom in, B&W.. then I saw the difference.. :)

This thread makes a pretty good documentation for 32khz dATs heheh,
or are we doing it the wrong way :p

[U2Lynne]

This is great stuff. I know Five wants to write a document all about SA and FA sometime so I'm sure he just loves all this information. (And, the way you have laid it out makes it really easy to read and see all the differences.)

[corsair]

There is another point I want to add too. Its that, at first glance, it might not be obvious. U might need to move along the wav line and observe carefully. On a 32khz dat source, the surface should not have sudden trough or peaks... while on a mp3 source, the surface is not as smooth in a straight line, with chunks here and there at times.

More illustration.

[thornhill]

ive got some screenshots of some stuff im 99% sure is sourced from 32k dat. lineage is CSB's > Sony PCM 1 > Fostex D5 > Tascam cdrw5000 > CD. theres a couple digipops in the first SA next to the yellow cursor so disregard those (unless of course you think its something of significance).

[range_hood]

thanks for the input thornhill. the nirvana source didnt´t really have that much frequency response in the higher frequencies.

[range_hood]

could you please provide a 10 secs. sound sample?

[ssamadhi97]

yea, this stuff is 32kHz sourced indeed

take a look at these

http://www.thetradersden.org/forums/attachment.php?attachmentid=6574

http://www.thetradersden.org/forums/attachment.php?attachmentid=6589

and note how the spectrum close above 16kHz (fs/2) looks like a flipped image of the spectrum close below 16kHz. that's a typical sign of upsampling (whose presence depends on the resampling implementation used - it might be more or less pronounced in other 32kHz sourced recordings).

[BoldCaptain]

I went back to the original DAT in question, and yes indeed, this was recorded at 32k. I dont know what to say-I am so embarrassed. My apologies to Thornhill, who seeded this for me, and anyone else offended by the 32k recording.

[corsair]

Just curious. So is my original assessment of the recording that I've asked in the first place is indeed 32Khz DAT as well?

[thornhill]

I went back to the original DAT in question, and yes indeed, this was recorded at 32k. I dont know what to say-I am so embarrassed. My apologies to Thornhill, who seeded this for me, and anyone else offended by the 32k recording.

32k aint exactly a shit in the eardrums. its a quite welcome break from all the damn mp3 sourced and minidisc b.s. ive gotten in the past from other sources, and for anyone who recieved this recording in the seed, im sure they can attest.

[Five]

Just curious. So is my original assessment of the recording that I've asked in the first place is indeed 32Khz DAT as well?
that's how it looks to me

thanks everyone for the great discussion here

[range_hood]

[...] and note how the spectrum close above 16kHz (fs/2) looks like a flipped image of the spectrum close below 16kHz. that's a typical sign of upsampling (whose presence depends on the resampling implementation used - it might be more or less pronounced in other 32kHz sourced recordings).
This is great info. Thanks ssamadhi97.

Taken from corsairs sample:

[ssamadhi97]

This is great info. Thanks ssamadhi97.
np.

I should have some more blatant examples somewhere, "for great illustration". Stuff like 32kHz DAT upsampled to 48kHz and back to 44.1kHz on the fly during transfer through a SBLive!, for example. (Thanks a lot, Creative. Please die.) I shall dig for some good pics once I'm back home.

[corsair]

I've got 2 recordings where the source more or less swears its never on mp3, but rather on 32KHz DAT. Unlike the waveform posted earlier in another thread, its rather different. And at 16Khz, there is mirror image flip at that point to signify upsampling.

I'll attached images of the SA of the 1st of 2 recodings...

[corsair]

The attachments here will be the 2nd of the 2 recordings. :)