Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Archiving vinyl once only 96/24? (Read 20275 times) previous topic - next topic
0 Members and 2 Guests are viewing this topic.

Archiving vinyl once only 96/24?

Hi All,

I wonder if you can help.
I have recently registered with HA but have been reading the forums for months.
Very interesting and informative.

Being a recovering so called 'audiophile (fool)' I think I have at last found some sense in the real world!
But thats a whole other story for a different thread!

I have a collection of some 2000 LPs that I want to archive to digital.
Now I release that the topics of 44/16 v 96/24 has been widely discussed but just to be clear and please
correct me if Im wrong on any of these points;

- Set the phono amp output to produce a max peak volume just below ADC clipping to maximise ADC coverage?

- 16 bits is more than enough, 24 bits does not produce a more detailed rip?

- Would 24 bits needed if I want to use declicking functions and rumble filters?

- If I always set the ADC input correctly normalisation should not be necessary?

- If tracks do need normalising to say -3db what does 24 bit over 16 bit prodivde?

- The noise floor is not lower with 24 v 16 bits, I dont think I understand the concept properly yet?

- 96K rather than 44K is worth doing for capture as it pushes any anti aliasing well out of the audio band.?

- There is no point in having a final cleaned corrected file greater than 44/16 as there is no audio benefit?

(BTW Im using a great software package called Vinyl Studio!)

Thanks for any comments,
Matthew

Archiving vinyl once only 96/24?

Reply #1
- If tracks do need normalising to say -3db what does 24 bit over 16 bit prodivde?


You can be more wrong in where you set the input gain and still recover fully. 16 bits gives you at least 20 dB more range than you need.

Quote
- Would 24 bits needed if I want to use declicking functions and rumble filters?


If your filters use something more than 16 bit, you could keep it in that precision for the duration of your editing, even if recorded at 16 bit.

Archiving vinyl once only 96/24?

Reply #2
Btw I didn't want to prejudice any answers but I already have had
a lot of excellent results with a number of albums capturing at 44/16.
96/24 may sound 'slightly better/different' but it's not massive whatever it is!

Be interested to know the formal reasons....

Archiving vinyl once only 96/24?

Reply #3
If you are by some means certain of what the peak value will be, then setting the level to get the maximum non-clipping signal into the ADC is generally fine (with some reservations mentioned below) but it seems to me the only way to accomplish that is to record first at a significantly lower level just to find that peak. Using a reasonable mid-level setting to just record and be done would thus save a lot of time and stylus wear and will produce a recording that is of no lower quality.

24 bits  will not capture more detail, but there is an advantage to recording at a higher bit depth. It is unlikely to ever be audible, but it makes a real difference in the data if you do any post-recording processing -- unless the software you are using doesn't work above 16 bit. The quantization errors at 24 bit are smaller by a factor of 256 over 16 bit. Some editors may automatically convert to a higher bit depth when loading, or may convert the data for each operation if the file is 16 bit. In the former case there is still an advantage in having the greater bit depth to begin with if you end up saving and reloading the file a number of times before being finished and in the later there is some additional quantization noise added with every operation.

24 bits is not "needed" but the technical advantage is as stated above.

Normalization is never needed but is a matter of taste. If you capture the maximum value possible for your ADC, producing some peaks at or near 0dBfs, and then process, some operations will raise the level above 0dBfs, thus clipping integer sample values. If you are working in floating point, you can recover by amplifying a negative amount afterwards, but 24 bit integer will not help you there. It is easier to capture at a lower level and normalize to what you want at the end of processing. Being afraid of normalizing is irrational.

Any post-recording processing at a higher, vs a lower, bit depth makes for smaller quantization errors.

The ADC noise at 24 bit is generally lower than at 16 bit, but if you are using any decent ADC, the LP surface noise is so much higher that the ADC noise is totally irrelevant.

Recording at a higher sample rate will move the inevitable aliasing to higher frequencies but "well worth doing" is technically incorrect as you will never be able to hear the difference unless perchance the ADC is quite poor. In that case you can't depend on a higher sample rate giving you even that much. Test with RMAA to see what you have.

No one has been able to demonstrate a benefit to listening to a higher sample rate recording. There have been a few isolated incidents of someone claiming to be able to detect a difference with one or two sample tracks but these seem to have been equipment deficiencies as they have not been verified as being generally audible. Also, if such differences are ever audible, they are only audible during very close A/B comparisons. This cannot be done while actually listening to music.


Archiving vinyl once only 96/24?

Reply #4
On clicking preview, I see AndyH has replied. We very much agree, though he has more experience of vinyl ripping than me.

- Set the phono amp output to produce a max peak volume just below ADC clipping to maximise ADC coverage?


Not necessary to cut it so close and risk clipping (which would require re-capturing the audio at a lower setting). Far better to set it to 6 to 10 dB below ADC clipping just in case of an unexpected over.

The effective 14 or 15 bits left from a 16-bit sampling depth is ample to capture down to the noise floor of vinyl and you can adjust the volume later (e.g. ReplayGain scan will be more consistent than normalization). Any gain adjustment will be calculated at high bit depth and properly dithered down to 16-bit (you don't have to worry about rounding errors) and will add negligible noise compared to vinyl noise, so don't worry.

Quote
- The noise floor is not lower with 24 v 16 bits, I dont think I understand the concept properly yet?


The noise introduced by the digital resolution is far below the noise inherent in vinyl or audio tapes, so it will not add an amount that's remotely audible over the noise you're faithfully capturing.

Quote
- 96K rather than 44K is worth doing for capture as it pushes any anti aliasing well out of the audio band.?


Any reasonably modern sound card should anti-alias perfectly well at 44.1kHz. One or two old ones might work best at 48 kHz. Other than choosing a rate that's compatible with your software, 44.1 or 48 kHz should be essentially the same quality.

Quote
- There is no point in having a final cleaned corrected file greater than 44/16 as there is no audio benefit?


True, thought 48/16 is not unreasonable either, but will occupy a bit more disk space. Some computer audio mixers run at 48kHz internally, so you'd save them from resampling (reduced processor load) but both 44.1 and 48 kHz would be audibly identical.
Dynamic – the artist formerly known as DickD

Archiving vinyl once only 96/24?

Reply #5
I would record at 48k/24 bit, then after processing save at 48k/16 bit as FLAC.  You can of course save higher, but its largely pointless unless you plan to further process the material (and even then probably unnecessary).

Archiving vinyl once only 96/24?

Reply #6
My 1st advice is always...  Buy the CD, if it's avaliable!    With that many recordings, I assume cost is a big factor, so you might consider used CDs.  Or, you might want to consider CDs or downloads only for recordings that are in poor condition.

You don't NEED 24/96, but if your hardware supports it there's no harm in using a high resolution format.  24/96 seems to be the current "studio standard".  (But of course,that's  way-overkill for digitizing "scratchy" vinyl.)  You may also want to consider archive storage space.  24/96 requires about 3 times the space as 16/44.1.    (By using FLAC or ALAC lossless compression you can disc space, but the 3:1 ratio still holds.)  My small collection of digitized vinyl is archived on audio CD (16/44.1).

You might want to consider archiving the original digitized copy as well as the de-clicked copy.  You can sometimes get artifacts from DSP, and I suspect that you are not going to carefully audition 2000 cleaned-up recordings.  So, it might be useful having the ability to start-over without re-recording.  (I normally keep an un-processed original only when I've done a LOT of EQ & noise reduction.)

Quote
- Set the phono amp output to produce a max peak volume just below ADC clipping to maximise ADC coverage?
Right... I usually shoot for between -3 and -6 dB.  After digitizing, I check the peaks and if anything hits 0dB I assume it's clipped and I start-over at a lower level. 

The idea is to get the best signal-to-noise ratio as possible.    You can improve the theoretical signal-to-noise ratio by increasing the signal while the analog noise in the soundcard and the quantization noise remains constant.  But of course, most of the noise is in your analog signal and that noise goes-up when you boost the signal, so it's not that critical unless you have a really bad soundcard.

Quote
- 16 bits is more than enough, 24 bits does not produce a more detailed rip??
Right.  Two reasons...  Our ears are not that good.  And with analog vinyl, there is so much analog noise that you cannot make use of all that digital resolution or digital dynamic range.

Noise is basically randomness added/subtracted to the digital value.  If I say the digital value is 1000, but there is noise of +/- 100, that means the real/original value could be anywhere between 900 and 1100.    It doesn't help at all if I add resolution and get 1000.000 or 1000.001.  The noise/uncertainty makes that resolution/precision useless. 

Quote
- Would 24 bits needed if I want to use declicking functions and rumble filters??
Some filters & effects may work better at higher resolution.    But since there is no useful information in those extra bits, the DSP programmer could have easily increased the resolution (bit depth and/or sample rate) for temporary processing, and then convert it back.    In fact, most audio software works internally at 32-bit floating-point.    Some DSP effect algorithms may increase the sample rate temporarily, but it's not common.

Quote
- If I always set the ADC input correctly normalisation should not be necessary??
it's not always "necessary", but it can be helpful and there's no harm in normalizing (after all other processing).  You can't predict the peaks, so you need headroom when recording.  But, you don't need that headroom in your finished digital audio file, and you just might want to make it louder.    (It's best to normalize the album as a whole, so that quiet songs remain relatively quiet and loud songs remain loud, etc.)


Quote
... - If tracks do need normalising to say -3db what does 24 bit over 16 bit prodivde??...

... - The noise floor is not lower with 24 v 16 bits, I dont think I understand the concept properly yet??
More bits results in lower quantization noise.    But at 16-bits the quantization noise is at -93 or -96dB...  That's way below any other noise in the recording.  If you want to know how quiet that is, try reducing the level by to -80 or -90 dB in your audio editor.    Play the original file as loud as you like, and then listen to the -80/-90dB copy (without touching the volume control).  That should also demonstrate that you are not loosing any "detail" at 16-bits.  If you want to know what quantization noise sounds like, make an 8-bit file.

Quote
- 96K rather than 44K is worth doing for capture as it pushes any anti aliasing well out of the audio band.??
I've never heard any problems with aliasing or with anti-aliasing filters.

Quote
- There is no point in having a final cleaned corrected file greater than 44/16 as there is no audio benefit??
True.

Quote
(BTW Im using a great software package called Vinyl Studio!)?
I haven't used Vinyl Studio.    For recording, your choice of software doesn't matter.    The recording software basically just sends the ADC output-data to a file on your hard drive.  But for cleaning-up clicks & pops, some software works better than others and there are trade-offs between removing defects and adding artifacts.    (I use Wave Repair in the manual mode, but with 2000 records I don't recommend that for you.)





Archiving vinyl once only 96/24?

Reply #7
Thanks for the answers.

Im curious though as to why digital interfaces provide 96k (or even 192) if 44/48 is good enough?
Is it a manufacturing ploy to get people to buy new stuff?

I assumed that a 24 bit v 16 bit sample of a vinyl signal peak adjusted prior to the adc at the phono
stage to give a -3db peak would actually contain more info or am i missing something.
So then consequently replayed through a 24 bit dac would have more detail?

Dont shout as im learning about this stuff!

Thanks
Matthew

Archiving vinyl once only 96/24?

Reply #8
Sorry DVDdoug your reply beat my next question whilst i was typing so ignore my last post about 24 bits.
Great answers!

Archiving vinyl once only 96/24?

Reply #9
Im curious though as to why digital interfaces provide 96k (or even 192) if 44/48 is good enough?
Is it a manufacturing ploy to get people to buy new stuff?
It’s not inherently bad to support high resolutions. It’s inherently bad when people start asserting without evidence that such resolutions are audibly superior. And yes, some companies use their (trivial to implement) support for high resolutions as a dishonest selling point.

Quote
I assumed that a 24 bit v 16 bit sample of a vinyl signal peak adjusted prior to the adc at the phono
stage to give a -3db peak would actually contain more info or am i missing something.
So then consequently replayed through a 24 bit dac would have more detail?
It will naturally have inaudibly lower quantisation noise introduced by the ADC and may additionally be able to preserve minuscule noise way below any human’s hearing (below –96 dB) from the original vinyl but most likely due to its various shortcomings than to anything correlated with the actual original signal. That’s about it.

Archiving vinyl once only 96/24?

Reply #10

Ive just realised something fundamental!

Why are hires audio downloads now becoming popular and often with glowing reviews when 44/16 already provides the correctly engineered and ultimate solution?
Some reviews have compared them to say a previous CD release for the same recording and found them to be substantially better, Im thinking of some of the Linn recordings etc.

Im sure a lot of people have asked that here on the forum?
(Ive probably missed something!)

 

Archiving vinyl once only 96/24?

Reply #11
Either placebo effect and marketers exploiting it, or dishonest tactics via supplying a better master to the high-res files. It’s as simple as that.

Archiving vinyl once only 96/24?

Reply #12
- 16 bits is more than enough, 24 bits does not produce a more detailed rip?

- Would 24 bits needed if I want to use declicking functions and rumble filters?


Likely not, but if 24 is supported, then you can use that and have much more margin of error. You can always peak-normalize and dither down to 16 in the final step. But if you perform RIAA correction in software, that curve is quite a few bits from top to bottom.


Under certain conditions, you may care whether the final format should be 48 or 44.1. For example, my DAC does not seamlessly switch between the two (an inevitable consequence of its buffer design), so I get dropouts (and a minor scratch sound) if I make a playlist with alternating samplerates. Since I have lots of CD rips, I would be biased towards 44.1.

Archiving vinyl once only 96/24?

Reply #13
Why are hires audio downloads now becoming popular and often with glowing reviews when 44/16 already provides the correctly engineered and ultimate solution?

Somehow "ultimate" feels like too emphatic a word here; "adequate" seems enough particularly in reference to recording LPs.  Let's not close the book just yet; I'm sure there were people who believed the Edison wax cylinder to be the ultimate solution.

...dishonest tactics via supplying a better master to the high-res files.

It doesn't even need to be "better" (whatever that may mean to you), just perceptually different:
"I paid more for it" -> "It's high-res" -> "It sounds different to me" (or "It looks different to me") -> "It must be better"

Archiving vinyl once only 96/24?

Reply #14
Why are hires audio downloads now becoming popular and often with glowing reviews when 44/16 already provides the correctly engineered and ultimate solution?


Just because it's enough resolution and can be correctly engineered doesn't mean it is correctly engineered. 

It's been alleged that the same albums on CD and SACD have been deliberately degraded on CD just to show there is a difference and you should buy SACD (more profit margin and copy protected)

Archiving vinyl once only 96/24?

Reply #15
I assumed that a 24 bit v 16 bit sample of a vinyl signal peak [...] would actually contain more info [...] So then consequently replayed through a 24 bit dac would have more detail


24-bit audio played through a 16-bit pipe shouldn't sound any different, unless the recording has > 100 dB of dynamic range, the difference between a jet engine and a butterfly. Your ears would be very unhappy if you had the volume cranked loud enough to that entire range, rather than just the range where the music is.

You probably noticed that when you've got the recording level set high enough for -3 dB peaks, the silent parts of the record, i.e. the whoosh of vinyl rushing past the stylus, hover around -50 or -60 dB. That's the "noise floor" and there's nothing to listen to below that; certainly all the music is way above that, even in the fade-outs. So in theory, you need maybe 11 bits, as far as dynamic range is concerned.

Increased bit depth also captures the input amplitude with greater precision, which is what you're wondering about. This precision has to be really low before the quantization noise it creates is audible and distracting (if undithered). At 8 bits, it can be rather unpleasant. At 16 bits, it's not even close to being audible.

In other words, the bit depth needs to be high enough to capture the dynamic range your input requires (about 10 or 11 bits for vinyl), and it needs to be high enough to get the quantization noise below audibility (maybe 14 bits, if undithered, I think). Beyond that, no one can tell the difference. If it makes you feel better, and you have gobs of disk space, sure, record at 24-bit. But you're not getting increased fidelity in any of the sounds that you can hear.

Archiving vinyl once only 96/24?

Reply #16
You've had lots of good answers. I have just a few brief comments.

I have a collection of some 2000 LPs that I want to archive to digital.

That's a lot of LPs to transfer. And you shouldn't underestimate how much effort is required.
As DVDDoug has said, it's easier to just buy the CDs, but I will just point out these possible reasons why a vinyl transfer may be preferable:
1. Some CD releases have been remixed or edited and the music simply isn't the same.
2. Some CD reissues have unacceptable amounts of dynamic range compression and/or clipping, which means the vinyl can actually sound better (despite its inherent technical limitations).
3. Perhaps (like me) you just enjoy the transfer process. Unlikely, but it takes all sorts.

- Set the phono amp output to produce a max peak volume just below ADC clipping to maximise ADC coverage?

I usually pick what looks like the loudest section of an album by visual inspection of the grooves and just play that, setting the recording level to peak at -3dB. This should give you sufficient headroom for unexpected peaks. I've transferred hundreds of LPs this way and honestly haven't ever had an over using this strategy.

That said, with modern ADCs, aiming for peak levels close to -3dB is probably pointless. My approach is a legacy of years of analogue tape usage and the fact I started doing digital transfers back in the early 90's with 16 bit converters that had fairly high noise floors, so it was necessary back then. These days it's probably easier to just peak around -10dB or so and normalise later.

- Would 24 bits needed if I want to use declicking functions and rumble filters?

There is a theoretical advantage to using greater bit depths. As others have explained, the accumulation of rounding errors will increase the quantisation noise. However in practice it really doesn't matter. The noise floor of vinyl (about -60dB on a good day) is so high that you'd have to do dozens (possibly hundreds) of passes before the accumulation of 16 bit quantisation noise would become audible against the inherent vinyl noise. You can probably even get away without dithering.

On the other hand, there is a *practical* disadvantage to working at 24 bit. This is simply that there are some useful tools that only work on 16 bit files.

(BTW Im using a great software package called Vinyl Studio!)

I always intended to check out Vinyl Studio, but never got around to it. It has a very good reputation.

Archiving vinyl once only 96/24?

Reply #17
...BTW Im using a great software package called Vinyl Studio!...


I too can vouch for VinylStudio as a very good all-in-one solution.  I also use iZotope RX2 for post-processing, but VinylStudio can do pretty much everything you want to do when transferring vinyl. While both programs are available on Windows and Macintosh, iZotope comes with a steep learning-curve and a high price tag. VinylStudio is pretty easy and costs little.

I can't think of any reason not to record at 24/96 if you have the means to. (Edit: unless, as cliveb said, you are using tools that only work at 16 bit.) You don't have to but why not? If you have the space and your soundcard supports it you can always just archive them as you wish and VinylStudio can make you a good Redbook copy on demand. If it was more involved than that then I think it would make more sense to finalize as 16/44.1 but it's really not. I personally never resample/dither my own needledrops (unless it's going on my iPhone or something) because I have a DVD-A player and usually have spindles of DVD-Rs whereas I rarely have a lot of CD-Rs to spare. That's really not the norm for most people so YMMV. Most people don't own a SACD/DVD-A player, nor do they really need one.

BTW: I don't know if you're working from a laptop or not, but it's not necessary to keep all your files on your primary HDD for VinylStudio. Even 16/44.1 files can accumulate and eat up valuable disc-space on a laptop. I personally save them to an external HDD and import/export the files as needed.
The Loudness War is over. Now it's a hopeless occupation.

Archiving vinyl once only 96/24?

Reply #18
These days it's probably easier to just peak around -10dB or so and normalise later.


I'll take this as an opportunity to offer some advice on peak-level normalization when ripping vinyl, as I'd hate for him to figure this out halfway through his 2000-record rip...

Vinyl albums, whether single disc or multi-LP sets, are usually mastered such that all sides will be played at one volume setting. Thus when recording, you'll get proper relative volume, song to song, if you just leave the input gain at one setting, whatever's appropriate for the loudest song on the whole album. Likewise, any normalization or ReplayGain 'album gain' scanning ideally should be done across the entire album, not on a per track, per-disc, or per-side basis. (Though if you get the ReplayGain scanning wrong, it's easy to fix... normalization is not.)

Singles, including 12-inch singles & EPs, on the other hand, are generally mastered for maximum loudness on each side, and that level will vary depending on several factors, such as duration and bass levels. For example, on a 12", often the A-side will be one 7-minute song cut especially loud, and the B-side will be a good 4-6 dB lower because it's maybe 15 minutes long, not because the songs are supposed to be that much quieter than the song on the A-side. In this situation, for proper relative volume levels, you might want to adjust the input gain per-side. If you don't, there's no harm in recording both sides at the same level (that's what I usually do), but just be sure to do your normalization or ReplayGain 'album gain' scanning on a per-side basis.

Of course, there will be exceptions. I've seen it all.

Archiving vinyl once only 96/24?

Reply #19
16/44.1 is more than good enough, and has maximum compatibility.

On rare occasions, wear-free LPs played on great set-ups can deliver greater (all-be-it inaudible) frequency response than 44.1kHz can sample (i.e. there's real content above 22kHz). Far more commonly, information from LP playback that's above 22kHz is pure distortion, but occasionally it's something real. For those of us who might want to pull out the digital copy to check the provenance of a given recording, sampling at 96kHz might be useful to preserve this information. It's also conceivable that some future audio restoration technique might benefit from capturing at the higher sample rate. It's also just conceivable that a large proportion of audio reproduction decades into the future will use higher sample rates, and one day, probably after we're all dead, upsampling from 44.1kHz will be more of a pain than having native 24/96 or DSD or whatever.

Of all the decisions you'll have to make regarding LP transfer, sample rate is by far the least important, because 44.1kHz more than good enough (I said that already, didn't I?) and anything better can always be down-converted to 44.1kHz for use on CDs, players than only support 44.1kHz, etc. Whereas many many other factors have an audible impact on the results. e.g. condition of LP, choice of turntable, and (potentially) bad decisions regarding denoising + EQ.

Sanity hints: use FLAC for storage, add correct metadata (tags) ASAP, and ReplayGain album mode for levels.

Unless you're at home all day with nothing else to do, 2000 LPs will take you years. Start with a couple you don't really care about to test your process on, and then start again with the ones you're most likely to listen to, working downwards in preference. Then if you never finish, at least you'll have the most important ones copied.

Cheers,
David.

Archiving vinyl once only 96/24?

Reply #20
Great Advice!!

Ive learnt more from this forum than any other source especially manufacturers websites!

The quality of replies is very erudite for any digital newbie.
Hopefully anyone similar to myself looking for answers on the basics would find this thread extremely useful.

VinylStudio is a great package and now offers IIR filters for any user audio requirements. The developer Paul is extremely responsive and strives to keep it the leading Vinyl ripping tool folowing user input. Its nearly an order of magnitude cheaper than the package I was looking at Im sure you know which that is...

Its also sparked my interest in digital audio electronics!
I will be getting my dusty copy of Horowitz and Hill off the bookshelf.

In conclusion then from what Ive gathered, as far as hires downloads etc are concerned it is a marketing ploy and what we really need is a better quality audio master in 44/16 rather than some heavily compressed rubbish that sometimes ends up on the final CD?

Another question maybe for another forum is;
How important is Jitter in effecting things in audio? a lot of DAC/ADC make a big deal about jitter is it really that much of a problem and consequently charge the necessary money?

Matthew

Archiving vinyl once only 96/24?

Reply #21
Another question maybe for another forum is;
How important is Jitter in effecting things in audio? a lot of DAC/ADC make a big deal about jitter is it really that much of a problem and consequently charge the necessary money?

This has been discussed numerous times, which you will find when you search. Also, it is bad form to ask an unrelated question in someone else's thread, and you are much less likely to get the answer that you are looking for.

Archiving vinyl once only 96/24?

Reply #22
Whilst I agree that another discussion on jitter is not needed, especially since there have been several in recent memory, it only takes a few seconds to see this is not “someone else's thread”.

Archiving vinyl once only 96/24?

Reply #23
In the context of this discussion, any amount of jitter you might find in the digital portion of your system will be completely swamped by that which is inherent in vinyl by several orders of magnitude.

http://www.hydrogenaudio.org/forums/index....st&p=790112

Archiving vinyl once only 96/24?

Reply #24
Thats a bit rude isnt it?
Especially after all the really good advice and useful replies people have offered?
Re-read the mail I sent.
You can see its a thread I started and its a 'rhetorical question' in which I stated 'Another question maybe for another forum is' and was planning to do that!

Another question maybe for another forum is;
How important is Jitter in effecting things in audio? a lot of DAC/ADC make a big deal about jitter is it really that much of a problem and consequently charge the necessary money?

This has been discussed numerous times, which you will find when you search. Also, it is bad form to ask an unrelated question in someone else's thread, and you are much less likely to get the answer that you are looking for.