Vinyl to digital - best ADCs?

[Petrus]

The younger generation here apparently does not know about the test Ivor Tiefenbrun, the founder and owner of Linn Products, participated in in the eighties. He was strongly anti-digital, which is no wonder as he is the designer of Linn Sondeck turntable and the father of modern vinyl based hi-fi.

He agreed to a test where the sound from his Linn LP-12 turntable was either passed pure to the amps, or AD/DA converted with a Sony PCM-F1 16-bit converters.

He could not hear the difference in a blind test, even though he was (and still is) revered as The Guru.

The funny part: As the converters were fed from consumer system, the peak levels reached only -20 dBFS, which means Ivor was listening a 13 bit system...

I would say this alone is quite good proof that 16/44 is good enough for vinyl transcription. Or maybe even 13 bits.

Or maybe the hearing of younger generation is wastly better than Ivor's. They should be able make a fortune in hi-fi.

The story is here as told by the person who arranged it: Boston Audio Society - ABX Testing article

[Lemonsqueezer]

Originally Posted by Petrus

The younger generation here apparently does not know about the test Ivor Tiefenbrun, the founder and owner of Linn Products, participated in in the eighties. He was strongly anti-digital, which is no wonder as he is the designer of Linn Sondeck turntable and the father of modern vinyl based hi-fi.

He agreed to a test where the sound from his Linn LP-12 turntable was either passed pure to the amps, or AD/DA converted with a Sony PCM-F1 16-bit converters.

He could not hear the difference in a blind test, even though he was (and still is) revered as The Guru.

I would say this alone is quite good proof that 16/44 is good enough for vinyl transcription.

Or the hearing of younger generation is wastly better than Ivor's. They should be able make a fortune in hi-fi.

The story is here as told by the person who arranged it: Boston Audio Society - ABX Testing article

Interesting will read that.

I would agree for many cases 16/44 is fine. However I did tell the difference between in 16/44 and 24/96 on my own system in ABX test so I obviously chose what worked for me in my set up.

[dtf]

Originally Posted by Plush

I am pleased to answer your other question. It is very easy to determine the bit rate of an analog source. One can do this for an LP, for FM radio broadcast, for cassette tape etc. One could even do it for one of your mics if it was tested for its dynamic range capabilities.

The way you do it is to determine the dynamic range of the analog source. A LP might have at best a 75dB dynamic range, more usually in the range of 50-65 dB dynamic range. The 75 dB figure corresponds to just over 12 bits, so this is where I get my figure of it being a 12 bit source.

Dear Plush,

"bit rate" is something different than "bit depth". If we want to have a good discussion about the topic we should try to use the same definitions. Before, you already mixed up dBu with dBV and used a wrong definition of 0 VU.

The bit depth is the amplitude resolution of a sampled analogue signal, measured in [bit]. The dynamic range is linked to this - the more bits generate a larger theoretical dynamic range which is rarely achieved exactly in practical application.

The bit rate is the amount of data per time, usually measured in [bit/s].

From measuring the dynamic range of a system one can not deduce the bit depth of a digital signal. This would imply that the quantization steps are equally large, independent of the bit depth, which is not the case.

Best,
Dirk

[mixedupsteve]

Originally Posted by bde

the chip i posted is used for a/d in korg mr2000s.

I also emailed dh labs and they said they can make custom cables with rca in and xlr out- so that questions is solved more or less)

If your interested in thoughts about the chip rather the unit as a whole you might ask on the Geekslutz forum.

[Martin Kantola]

Originally Posted by Petrus

I would say this alone is quite good proof that 16/44 is good enough for vinyl transcription. Or maybe even 13 bits.

Your little story is not proof, sorry. As someone who is old enough to have used the F1, I easily can imagine it was an amazing experience when first heard. No noise, no wow and flutter. But as time passed, our ears became more perceptive to the shortcomings of early digital audio. 16 bit digital was nowhere near mature with the F1.

"Younger generation...."

Martin

[johnnyc]

Originally Posted by Petrus

The funny part: As the converters were fed from consumer system, the peak levels reached only -20 dBFS, which means Ivor was listening a 13 bit system...

This not exactly true. A 16 bit system at -20dBFS will not have the same resolution/step size and quantization noise as a native 13 bit system. They are not equivalent.

[Petrus]

Sorry, it would be a bit less than 13 bits...

Besides, if you think different bit depth systems have varying step sizes, you are wrong. In PCM the sample represents the voltage, in binary math each added digit doubles the maximum representable value. Thus adding one bit doubles the voltage value, which in turn means 6.02 dB higher volume. No matter if it is 12, 16, 20 or 24 bit system.

In decimal system: original number is 4, we add a zero at the end we get 40 (ten times the value)
In binary system: original number is 100 (4 in decimal), we add zero and get 1000 (8 in decimal, or doubled)

[didier.brest]

It's equivalent within 2 dB.

[The Listener]

Ok, here are two short mono clips (in 24/88 it exceeds 6mb if I would make a longer clip - and that is the attachement limit here) recorded from Eela audio EA924 broadcasting turntable with XLR outputs, directly through TC K48, no processing on either of the clips, one was recorded in "CD" resolution, the other in 24/88,2 - the resolution which I think might be appropriate to digitize anyone's vinyl collection and still have some feeling like listening to the original LP... There is just more high end information, nicer/cleaner/firmer bass, better depth... subtle of course... but quite audible even in those mono files.
If someone is an audiophile, having 300$/0.5m cables, who knows what kind of exotic full range speakers, etc., he probably wants to squeeze the last "bit" out of his transfer, no?

Merry Christmas!

[Daniel Stark]

Shouldn't it be 24/44.1 and 16/44.1?

[Lemonsqueezer]

Originally Posted by Daniel Stark

Shouldn't it be 24/44.1 and 16/44.1?

why?

[Martin Kantola]

There's clearly energy in the samples over 22kHz, so I agree that the sampling rate should be the same if we're only talking bit depth. You might want to re-check the levels too, and the samples seem to have slightly different EQ. Haven't listened yet.

Martin

[johnnyc]

Originally Posted by Petrus

Besides, if you think different bit depth systems have varying step sizes, you are wrong. In PCM the sample represents the voltage, in binary math each added digit doubles the maximum representable value. Thus adding one bit doubles the voltage value, which in turn means 6.02 dB higher volume. No matter if it is 12, 16, 20 or 24 bit system.

Do you really believe this? You might want to read up a bit more on some of the basics. Bit depth is precisely what determines step size/resolution.

By what you are saying a 24 bit converter would be 48db louder than a 16 bit???!!!?? Suppose on your converter at 16bit 0dBFS equals 10V, then at 24bit 0dBFS would need to 2,560V, greater than 2kV, really??? That would need to be some pretty fancy electronics.

[The Listener]

Any difference must be caused by the difference in conversion resolution, all else was the same, no eq, no gain change. I contributed this mini test for showing if it makes any difference to use higher resolution than 16/44,1, not just bit debth. I chose 24/88,2, anyone can listen and hear if vinyl transfer sounds any different in "CD" and higher resolutions.

Originally Posted by Martin Kantola

There's clearly energy in the samples over 22kHz, so I agree that the sampling rate should be the same if we're only talking bit depth. You might want to re-check the levels too, and the samples seem to have slightly different EQ. Haven't listened yet.

Martin

[Petrus]

Originally Posted by johnnyc

Do you really believe this? You might want to read up a bit more on some of the basics. Bit depth is precisely what determines step size/resolution.

By what you are saying a 24 bit converter would be 48db louder than a 16 bit???!!!?? Suppose on your converter at 16bit 0dBFS equals 10V, then at 24bit 0dBFS would need to 2,560V, greater than 2kV, really??? That would need to be some pretty fancy electronics.

Step size (if we are talking the effect of one bit) is always double voltage, no matter how many bits are used for sampling. When you set the recording level you determine where each voltage level sits. Basically you try to aim the peak as high as you dare to minimize noise. Have you heard about setting the recording levels depending on the inpul levels?

We all (should) know that the values are used in a way that the extra bits afforded by 24 bits compared to 16 extend towards the low, quiet levels, not useless loud levels at imagined kiloVolt ranges. Many people do use one or two bits for peak safety (setting levels slightly lower than they would with 16 bit system).

It is also possible that we mean different things here. Resolution is the smallest possible step, yes. The thing is that in real life random noise drowns everything below certain level. If the signal has only noise below certain level, like vinyl below -60 dB or so, using more bits achieves nothing, as they are only random numbers. In a 16 bit recording of 60 dB S/N signal at least the last 5 bits are random, in 24 bits there are 13 random, useless bits. These random numbers are spread evenly on the signal, the added "resolution" is totally random, thus nonexistant.

[Martin Kantola]

Originally Posted by Petrus

The thing is that in real life random noise drowns everything below certain level. If the signal has only noise below certain level, like vinyl below -60 dB or so, using more bits achieves nothing, as they are only random numbers. In a 16 bit recording of 60 dB S/N signal at least the last 5 bits are random, in 24 bits there are 13 random, useless bits.

Really? So you don't hear any music in the attached file?



Martin

[dtf]

Hi Petrus,

a lot of what you write is wrong, and explains why there is confusion.

Originally Posted by Petrus

Step size (if we are talking the effect of one bit) is always double voltage, no matter how many bits are used for sampling.

No, the step size is the constant voltage difference between two quantization values. The amplitude resolution is is not logarithmical! In a 16 bit system all 65535 steps are of equal size.

The effect of one bit is something different.

Originally Posted by Petrus

We all (should) know that the values are used in a way that the extra bits afforded by 24 bits compared to 16 extend towards the low, quiet levels, not useless loud levels at imagined kiloVolt ranges. Many people do use one or two bits for peak safety (setting levels slightly lower than they would with 16 bit system).

No, the only thing that is extended is the theoretical dynamic range - in other words: the noise floor will be lower. As said before, step size will decrease in a higher bit system, therefore the amplitude resolution will increase. 0 dBfs stays the same, as you write correctly.

More bits were never required for better noise values! They are advocated for more accurate signal reproduction, effects such as aliasing etc.

Originally Posted by Petrus

It is also possible that we mean different things here. Resolution is the smallest possible step, yes. The thing is that in real life random noise drowns everything below certain level. If the signal has only noise below certain level, like vinyl below -60 dB or so, using more bits achieves nothing, as they are only random numbers.

Resolution is not the smallest step - it is the amount of steps within a defined bandwith. Again, there is no extension towards lower signal levels!

Best,
Dirk

[johnnyc]

Originally Posted by Martin Kantola

Really? So you don't hear any music in the attached file?



Martin

Thanks for posting this. I did this too but I think your example is even better.

Can we as an audio community finally get rid of the silly notion that we can determine bit depth soley based on dynamic range and S/N ratios?? That argument might be valid from a systems/communications link perspective but clearly falls short in regards to audio as this example clearly indicates.

[Martin Kantola]

One of the main problems as I see it is defining what's noise and what's signal. Here's another set of noisy files for your consideration. Try playing them both together.

Martin

[JonesH]

Originally Posted by dtf

No, the only thing that is extended is the theoretical dynamic range - in other words: the noise floor will be lower. As said before, step size will decrease in a higher bit system, therefore the amplitude resolution will increase. 0 dBfs stays the same, as you write correctly.

More bits were never required for better noise values! They are advocated for more accurate signal reproduction, effects such as aliasing etc.

I was under the impression that increased resolution leads to less quantization distortion which when properly dithered amounts to less noise in the conversion... Am I mistaken (or just over-simplified)?

[johnnyc]

Originally Posted by JonesH

I was under the impression that increased resolution leads to less quantization distortion which when properly dithered amounts to less noise in the conversion... Am I mistaken (or just over-simplified)?

You are correct, more bits equals more resolution which means lower quantization noise/distortion, lower noise floor, etc. Bit depth really has nothing to do with aliasing. Perhaps dtf was speaking in terms of total bit rate (sample rate and bit depth)?

I'm not quite sure what you mean about the dither part though, dither is used when lowering the bit depth (say 24 to 16 bit as opposed to simple truncation). When you dither you actually add noise to the system, but you do it to retain low level information and it sounds more "musical" than simple truncation.

[JonesH]

Originally Posted by johnnyc

I'm not quite sure what you mean about the dither part though, dither is used when lowering the bit depth (say 24 to 16 bit as opposed to simple truncation). When you dither you actually add noise to the system, but you do it to retain low level information and it sounds more "musical" than simple truncation.

Yes, but dither is also used when doing A/D conversion for the same purpose as before; increasing the dynamic range of the conversion. It might of course differ from designer to designer, since we're working with so low levels one might assume that the thermic noise from the circuits self-dither to around -120/-130 dB.

[EV676]

So tonight I'm reading this thread while transfering vinyl to digital!
I'm not to worried about bit depth or conversion quality. The turntable is a Technics 1200, the cartridge is an Ortofon (not one of their DJ models) and I'm using the mono 1mil stylus for it. The recorder is a Zoom H4, set for 16/44.1. The source material, a number of big band albums from the likes of Lawrence Welk and Freddy Martin made in the 50's.

They're for my 80 year old uncle. He got an I-pod for Christmas and asked me if I could somehow get his favorite music into his new player.

I'll take the SD card, load it into my computer at work, call up Bias Peak, sequence the recordings burn him CD's and make the m4a files for his
I-pod. When I'm done he'll be perfectly happy and I, well, I guess this is one of the things you do to make your uncle smile.

Plus there's something disturbingly charming about the mid-1950's incarnation of the Champagne music of Lawrence Welk and the artwork of the album covers!

Great thread and yes, I did learn a lot from the discussion. Thanks!

[Lemonsqueezer]

^^^

Nice to hear you did that for your uncle.

Its funny, I am sure he won't care for the bit depth, yet will still be very happy just to hear the music.

[twentyhertz]

conversion questions aside, If someone has found a really good tool for processing the raw audio into songs with the necessary artist/album/track info I'd love to hear it. I have transferred a few of my cherished LPs but taking the raw audio out of soundforge, splitting the files and then naming everything made it not worth it for me, most of the time. Most of the time If I buy an LP that doesn't come with a CD I just live with playing it on the turntable only, or get the CD from the library and rip it to FLAC