Musings on digitizing vinyl

[manisandher]

I want to share my experiences in case anyone else is thinking of doing this.

My little 'studio' is now all set up (see below) and I have to say that early attempts at digitizing vinyl suggest that the effort has been worthwhile. It's probably gross distortion, but there's something about the sound of vinyl that is really, really nice. I'll put some 24/192 and or DSD files up for anyone who's interested at some point.

I'll be using this thread as a bit of a blog, as I'm sure there's still a lot for me to learn. But for now, I'll just list the equipment that's gone into the 'studio':

- sealed and sound-insulated room with semi-Faraday cage (builder's wire mesh used instead of copper plating) and integrated air purifier and ventilation system
- PurePower PP2000 (used in battery mode for turntable only)
- 5KVA balanced transformer (for linear PSs)
- 5KVA balanced transformer (for SMPSs)
- Loricraft vacuum record cleaning machine
- Technics SP10 MKII turntable
- Transfi T3Pro arm
- London Decca Reference cartridge
- Ray Samuels F-117 phono stage
- Pacific Microsonics Model Two (for PCM 24/192)
- Weiss AFI1 (with Paul Hynes linear PS)
- Korg MR1000 (with Paul Hynes linear PS) (for DSD128)
- Tascam DV-RA1000HD
- Berning Siegfried 300B
- AKG K-1000
- Sony PS3 (for SACD rips)

As far as possible, all power supplies sit outside the studio.

More posts later...

Mani.

[boleary]

I'll put some 24/192 and or DSD files up for anyone who's interested at some point.

Can't wait!

The "studio" looks very impressive; hope the results exceed expectations.

[PeterSt]

Nice Mani !

As far as possible, all power supplies sit outside the studio.

And no speakers in there !
Well, you will understand ...

Peter

[CoenP]

That's one impressive setup!

I am looking forward to your postings of the results.

Any reason not to do the phono eq in the digital domain? (I.e.http://www.stereophile.com/phonopreamps/pure_vinyl_lp_recording_amp_editing_software/index.html)

Regards, Coen

[manisandher]

Haha, this ties in nicely with what I'm doing:

Going through the various paces of streaming, CD spinning and finally vinyl in quick succession had to admit that streaming with XXHighEnd has far more in common with vinyl than CD even when the latter exploits a memory-based transport. The traits we so enjoy with this streaming source like speed and scale are far more present with vinyl than polycarbonate.

http://www.6moons.com/audioreviews/vaessen2/1.html

I've just made three recordings of the 'Dead Can Dance' track, 'Radharc' in 16/44.1, 24/88.2 and 24/176.4 on the PMII. I've compared my files to the 16/44.1 CD rip (same master by the sound of it), and here's what I'm hearing when I play back on the NOS1 with all files upsampled to 705.6 in XX:

Firstly, the CD obviously has a more stable sound, with massive channel separation. But Lisa Gerrard's voice is 'thick' and flat - dull. I saw DCD live in October and can definitely say that she doesn't sound like this.

The 16/44.1 file has a remarkably 'similar' sound to the CD rip, albeit with much diminished channel separation, and approx. 1dB lower volume. IMO it has more rhythm though.

Surprisingly, the 24/88.2 file doesn't sound that much different to the 16/44.1 file. A little more 'holisitc', but a similar sound.

The 24/176.4 file however is fundamentally different, and soooo much closer to how the vinyl sounds. Lisa's voice is now 'delicate'. There's a sweetness and liveliness to the sound that the other files just don't have.

IMO, only the 24/176.4 file played back through XX/NOS1 really sounds identical to the vinyl. This is my subjective opinion, but there is one thing I know for sure: Only at 24/176.4 (and 24/192) is the PMII a truly non-oversampling ADC.

My hypothesis is that any oversampling in the ADC and/or the DAC kills the sound. (It seems the guys are ESS, who produce the Sabre DAC, are onto this too: RMAF 11: Noise Shaping Sigma Delta Based Dacs, Martin Mallison, CTO, ESS Technology).

I'm happy to share these files for anyone to try themselves but am having difficult setting privileges in 'Rapidshare'. Does anyone know of another (easier) way to share them?

Mani.

[PeterSt]

Put them to FileMail. That gives 10 downloads. I will be the first.
Haha.

Otherwise DropBox may do the job - today without (e.g. me) needing to install DropBox. May, because it needs I link in here to it to really work. I never tried that myself, but I think I received "links" from others I can use like this.

Btw, for me no other than those 176.4 will be necessary. But then I would be judging the sound to absolute merits, and I think I already know what I will say ("bweh"  ).
But then this is not all about me you know ...

Thanks for sharing Mani,
Peter

[manisandher]

Peter, I'm just about to catch a plane but will sort this out at the other end.

Mani.

[manisandher]

I hope these links work...

CD rip: http://rapidshare.com/files/3980571067/Dead%20Can%20Dance%20-%2012%20-%20Radharc.wav

Digitized Vinyl Files

16/44.1: http://rapidshare.com/files/2382222736/12.%20Radharc%2016_44.1.wav

24/88.2: http://rapidshare.com/files/1295849148/12.%20Radharc%2024_88.2.wav

24/176.4: http://rapidshare.com/files/375223364/12.%20Radharc%2024_176.4.wav

Mani

[juanpmar]

Sorry Mani but the links don´t work.
Maybe using FileMail as Peter suggested https://www.filemail.com/

Juan

[JohanZ]

Sorry Mani but the links don´t work.

Yesterday it was working ok! Maybe a limitation on the number of download's
Johan

[PeterSt]

Something like that. "File owner's public traffic exhausted".

Edit : I think there's a daily limit of 1GB on it. And Johan consumed that. Haha.

Maybe if you put it up as MP3 it works better ?

[manisandher]

Yep, daily limit of 1GB, so just 5 24/176.4 files plus a few of the others.

Mani.

[manisandher]

The 16/44.1 file has a remarkably 'similar' sound to the CD rip... IMO it has more rhythm though.

Here are three different files:
- my 16/44.1 vinyl recording
- an SACD rip (using a PS3 and then converted to 16/44.1 PCM)
- a CD rip

Can you guess which is which?

Mani

[CoenP]

The first one looks clipped to me, the last one compressed. The first one obviously having much less dr and level, the last one having the most harmonics.

Pffew, this does not add up. What source could possibly be clipped and low in level at the same time? A bad lp rip? I hold you in higher regard!

Lets give it a shot:
1: cd
2: sacd
3: lp

Regards, Coen

[manisandher]

Thanks Coen. Anyone else want to have a shot?

Mani.

[ed linssen]

Thanks Coen. Anyone else want to have a shot?

Mani.

Hi Mani,

 a)I hope the bottom one to be the LP
 b)Midlle one the rip
 c)Upper one the conversion asf. At least the left(upper) channel is compressed and somehow limited.

Ed

[GerardA]

a: cd     peaks cut off
b: lp      highest irregular peaks
c: sacd  in between a and b

What do I win?

[christoffe]

B = LP

The others would be a guess.

[PeterSt]

1 = SACD
2 = LP
3 = CD

I said it before : This means of DR measure isn't worth much (I can easily tweak it for the clear worse, while the DR number goes sky high).

So ...

3 is the good master version of a CD.
1 is taken from the bad master version of a CD which also exists.
2 is quite noisy and thus your LP. And no, I don't derive the noisy from the obvious.

The DR from 2 (LP) looks right to me (despite what I said above) and this is because no precautions were taken to compress it somewhat like has been done in 3 and more in 1 (this is not necessary for LP).

Warning :
1 (my SACD) and 3 (my CD) can just as well be turned around when the masters used are turned around. Then this would come from it :

1 = CD
2 = LP
3 = SACD

with the explanation that the only thing happened to 3 is that it was expanded from the same master as used for 1.

And if that were not all, the "master itself" as we could see it in 3 can already be an expanded one of the master used for 1.
What am I saying ?
That both 1 and 3 can be CD which I can easily show you. Ok, I will, see below.

So the question : which one is the SACD ?
haha
Moral : it tells nothing because it depends on the master.

[manisandher]

The 'winners' are... GerardA, Christoffe and of course Peter!

The order was:

File 1 = CD
File 2 = LP
File 3 = SACD

What's interesting for me is that these were all derived from the original 1990 digital master, which from the spectra (not shown) is quite clearly 44.1 (and almost definitely 16 bits). Why the massive compression on the CD? I have no idea. There seems to be a little bit of compression on the SACD, whereas the LP doesn't seem to have as much (or any).

FWIW, I prefer the sound of the vinyl. I know it's from the same digital master as the CD/SACD, but to me it's simply more satisfying to listen to. Is this due to the lesser compression? Or perhaps a euphonic effect caused by the inherent distortion in all LP replay? I don't know. But what I do know is that even XX and the NOS1 can't 'polish the turd' they've put on the CD.

As so many people have been saying, we're stuck with what the mastering engineer did and no format is going to be able to improve on that.

Mani

[christoffe]

The digitized LP files (EMT 950, EMT XDS, KORG MR 2000s, Audiogate software;  192KHz/24bits (superior to 96/24), )
are totally different in SQ to CD files  via XXH.

Instruments on CD files are well seperated, less soundstage, cymbals are more harsh , but I like it.

LP files smoother SQ  and nice soundstage

Christoffe

[PeterSt]

No no no, I can't be a winner because I supplied two solutions. But to be clear, the screenshot of the two files I provided were both from existing CDs. Both exactly resemble the 1 and 3 provided by you, Mani.

To work it out further you could look into the SACD hi-res rip you will also have. Does it contain frequency above 22.05KHz ? I'm afraid it will not. And when not, both are from the same digital source which shouldn't be "hires" in the first place. The SACD is just expanded somewhat and is the same as my second picture (which btw is an MFSL).

Not that it is all *that* important, but this is how our own rips from vinyl can work out for the better. Not that I will believe that more than 20KHz can be obtained for it, but for the remainder ... yes.

Thank you Mani. These things are always fun.
Peter

[manisandher]

To work it out further you could look into the SACD hi-res rip you will also have. Does it contain frequency above 22.05KHz ? I'm afraid it will not.

Nope, no signal over 15KHz or so. It does make you question the wisdom of putting a 16/44.1 master onto SACD - but I guess your equipment appreciates the HF workout.

Mani.

EDIT: for completeness, I've added the vinyl 24/176.4 recording spectrum.

[AlainGr]

If I understand correctly, the first graph shows that this SACD originated from a redbook right ?

But the graph of the LP ? Can someone explain why there is no dip ?

Thanks.

Alain

[PeterSt]

Mani, no. I know that "analysing" this sh*t is difficult, but your LP goes to 15KHz just the same. The remainder is sheer noise. Sorry about that, but this is how I see it.

Set the scaling to liniar and use any windowing you like, but no "hires" behaves like this.
The top of the caret is where the music ends. The remainder is "anomaly".
Besides that, I don't think it can exist from LP, although I don't know the price of your cartridge. But still ...

Peter

[PeterSt]

But the graph of the LP ? Can someone explain why there is no dip ?

Alain,

With some imagination you can see the dip just the same in my screenshot from my previous post. But this is coincidence. Well, unless the LP was taken from the LP of course. Hahaha, and you know what, this can even be true (DSD recording - ehm assumed that this existed in 1990). But too far sought anyway.

The dip in the SACD is from "noise shaping". Think like "the 1-bit quantization noise is moved to beyond the audio band".
In this case Mani will have done that himself. So, he determined how steep the "audio band" ends and where the (inaudible (??)) noise starts. Mani may have used Audio Gate for that which IMO is no good anyway. But alas. Mani will tell it himself if it suits him.

If we are not careful, the whole lot comes from a "vinyl-rip" in the first place.

What's interesting for me is that these were all derived from the original 1990 digital master.

Maybe. But then not the best.

[manisandher]

Mani, no. I know that "analysing" this sh*t is difficult, but your LP goes to 15KHz just the same. The remainder is sheer noise. Sorry about that, but this is how I see it.

Hey Peter, I agree 100%. What do you think I meant by "no signal over 15KHz"? I was talking about the original master, and therefore all of the versions too.

EDIT: On the graph that Peter showed, it's clear that there is output from the cartridge all the way to 30KHz, but no signal on the LP. So it's just noise from 15KHz to 30KHz at about -84dB.

Oh and the cartridge cost me about the same as a NOS1!

Mani.

[PeterSt]

Hey Peter, I agree 100%. What do you think I meant by "no signal over 15KHz"? I was talking about the original master, and therefore all of the versions too.

Ah, I missed that (interpretation). Ok.

Peter

[manisandher]

The dip in the SACD is from "noise shaping". Think like "the 1-bit quantization noise is moved to beyond the audio band".

Yep, that's what it is.

In this case Mani will have done that himself. So, he determined how steep the "audio band" ends and where the (inaudible (??)) noise starts. Mani may have used Audio Gate for that which IMO is no good anyway. But alas. Mani will tell it himself if it suits him.

I just told Audiogate not to apply any filtering, and that's all. The dip from 15KHz to 22KHz on the SACD is identical to the dip on the CD. However, after 22KHz, the noise builds on the SACD due to the noise-shaping.

What's interesting for me is that these were all derived from the original 1990 digital master.

Maybe. But then not the best.

Yep, certainly not the best. But it's the reason I chose this track... to show that the mastering is the most important factor, before format.

Mani.

[PeterSt]

When we are at it anyway, I listened to the CD-rip provided by Mani and the 44.1 + 176.4 Vinyl-rip. Btw, all output to 705.6.

CD - Good.

44.1 Vinyl - poor. No channel separation. Highs in the left channel too faint.

176.4 Vinyl - Good. Slightly better than the CD, audible in the lower end of the drum (at the start). Little lower, little longer lasting.

Peter

[manisandher]

It's not really how I'm hearing things. I have a slight preference for the 16/44.1 vinyl recording over the CD rip. Oh, I forgot to mention: you need to disengage HDCD in XX before listeing back to the 16/44.1 recording - hdcd.exe isn't very clever and thinks HDCD peak extension has been used, when it hasn't.

In any event, the 24/176.4 recording is much better than the CD rip... at my place at least. Maybe you need to get some better amps... haha.

Mani.

[christoffe]

Mani, no. I know that "analysing" this sh*t is difficult, but your LP goes to 15KHz just the same. The remainder is sheer noise. Sorry about that, but this is how I see it.

Peter

The main difference is the "amount of information" you hear.
The amount of steps of a sinus curve at files with 16/44,1 are smaller and the height bigger than .............. .
You hear more refined vibrations of the guitar string at 24/176,4 than ................ .
So, the amount of  steps is higher and the height smaller at 24/176,4 at the same sinus curve.

[PeterSt]

Hoachim,

The height is smaller ?
Can you elaborate somewhat ?

Regards,
Peter

[christoffe]

Please see the link

http://www2.leuphana.de/medienkulturwiki/medienkulturwiki2/images/4/45/Digitalisierung.png

I'm not so bright to insert the image into this post. Copy and paste does'nt work, even if I press "Insert Image" before.


Joachim

[PeterSt]

Here you go. "Save target as".

[PeterSt]

But now you must be so bright to elaborate.

[christoffe]

But now you must be so bright to elaborate.

As a sample,

over one second
the amount of steps ( 0's and 1's) are increasing with the sample rate, you get more informations (overtones, please see http://www.music.vt.edu/musicdictionary/texto/overtone.html) about the same impulse.

Joachim

[PeterSt]

Ok, you had your chance.

So, the amount of  steps is higher and the height smaller at 24/176,4 at the same sinus curve.

Incorrect. The relation between 16/44.1 and 24/176.4 is :

16bits:24bits = 1:256 (= your height)
44.1:176.4 = 1:4

So, height is 256 times more, while sample rate is only 4 times more.

If this is not what your subject is, then ... please elaborate.
haha

Peter

[CoenP]

Here you go. "Save target as".

What a bad example pic. There is even a phaseshift between analog and digital domain, never seen that before.

This is a totally useless pic to make any other point than that the drawer didn't understand sampling either (probably a marketeer).

Just a warning: Digital theory is highly abstract and mathematical. Not much if anything in there that is inutitive or will explain "better" sound. Implementation of it in electronics is more of an experimental art than science, though Peter is on to some principles by now! I think the field of psychoacoustics will provide for better insghts in our sound perception of various formats and recording/playback means. Much generalisations in there and little attention for the "high-end audio" however.

regards, Coen

[christoffe]

Ok, you had your chance.

So, the amount of  steps is higher and the height smaller at 24/176,4 at the same sinus curve.

Incorrect. The relation between 16/44.1 and 24/176.4 is :

16bits:24bits = 1:256 (= your height)
44.1:176.4 = 1:4

So, height is 256 times more, while sample rate is only 4 times more.

If this is not what your subject is, then ... please elaborate.
haha

Peter

He, he

now I'm trapped!!!

In general my intention is to say, coming back to my first quote, that it is not sheer noise, there are more overtones on the LP.

Joachim

P.S. Is there a better program than Audio Gate on the market?

[PeterSt]

there are more overtones on the LP.

If that were so, it would show as frequencies. It takes a bit to learn "reading" such graphs, but here there is nothing. Just fairly random noise (beyond 15Khz I mean).

P.S. Is there a better program than Audio Gate on the market?

Obviously I challenged for the question, but although I don't know, there will be undoubtedly. Well, "others" and I don't know about "better". But you can look at it as downsampling. really dozens of programs exist for it, and they all behave differently (notice that DSD-PCM actually is a form of downsampling or "decimating"). The one is worse than the other and possibly none is good. Too many variables involved.
And then there's the subjectiveness of theories. So, take Arc Prediction for "filtering" (generally called "upsampling"). Here too dozens and dozens exist. Now go find one better behaving than Arc Prediction. You might find it theoretically, but personally I don't think you will. So here it is the technique used which is subjective to my ideas. And it is practice (or objective) once it is clear that it works for the better and best.

Do you see ? this is how I can say that e.g. Audio Gate isn't worth much. Too few parameters and what can be applied gives a result at least I don't like.

There's much more to say, but this is all in the DSD realm and already that us subject to "is that better ?" in the first place.
It's a large subject Joachim. One thing : it determines SQ, just like Arc Prediction does that for your NOS1 ...

Regards,
Peter

[christoffe]

The digitized LP files (EMT 950, EMT XDS, KORG MR 2000s, Audiogate software;  192KHz/24bits (superior to 96/24), )
are totally different in SQ to CD files  via XXH.

Instruments on CD files are well seperated, less soundstage, cymbals are more harsh , but I like it.

LP files smoother SQ  and nice soundstage

Christoffe

Disengage HDCD during replay of digitized LP files (192KHz/24bits) is mandatory. (see also Mani's note before)

[PeterSt]

I don't think there is an HDCD flag in 24/192(176.4) ...

[manisandher]

Disengage HDCD during replay of digitized LP files (192KHz/24bits) is mandatory. (see also Mani's note before)

I don't think there is an HDCD flag in 24/192(176.4) ...

Actually, there is! If you have an Alpha, an Esoteric D70 or a PMII, the HDCD light will come on with 24/192(176.4) material...

... provided these files have been recorded on a PMII, as no other ADC will encode HDCD at any rates.

BUT the HDCD decoding in XX should not detect the HDCD flag at these rates. So having it set to on/off in XX should make no difference.

Mani.