Topic: Analog to 24bit question

[Gutbucket]

This is true.  But the part that got my interest is this: " . . . but even WITHOUT any post transfer changes, you can still hear the difference between a 24-bit transfer and a 16-bit transfer --"

Yeah, that caught my eye too, and if so then that's the bottom line.

I would at least record in 24/44 for any processing that may need done, but thats just me and the purist in me

I don't get this argument, Bean. You can process the file in 24bit, 32bit, 64bit, or whatever your software can handle regardless of the bit depth of the recording.  In other words, the processing need not be tied to the resolution of the file you're working on.  I think the argument is capturing something you may miss with a 16 bit transfer, even if that isn't an extended dynamic range.

[Petrus]

Don't confuse bits with sound quality... to help determine what bit rate to use when converting an older analog cassette is to check the frequency response of the microphones used in the recording. But on that same note, frequency response does not guarantee a specific fidelity either.

Bit depth has nothing to do with frequency range. Bit depth governs the (theoretical) maximum dynamic range, simplified it is 6 dB for each bit. Sample frequency determines the maximum audio frequency, max frequency is sample rate/2 (Nyquist theorem). This is so basic that I find it funny to write about it here.

Not all converters are created equal, some are more accurate and have better analog stages. Also in the beginning times of digital audio the art of dithering was not as highly developed as it is now. 16 bit systems are just about perfect now.

We should also remember that even the best 24 bit systems are seldom if ever better than about 19bits in real dynamic range (115 db or so) thanks to less than perfect analog stages and power supplies.

[rich]

Cool...I have been taping for years, some good tapes some not so good tapes.  Either way, many thanks for the info Petrus.  I would be interested to hear your thoughts on DSD and 1 bit converters.  I used to run the Panasonic SV - 255 which I thought sounded amazing (1 bit PCM).  Thanks.

Rich

[Petrus]

It is not possible to compare one bit converters and "normal" PCM converters directly, the working principals are so different.

With PCM converters the signal is basically sent thru a gate, which flips when voltage is half of maximum. If it is, the most significant bit is set to 1, next to 1/4 gate, then 1/8 gate and so one. This happens once for each sample, typically 44100 times a second. We have relativelly few samples (but enough to describe the signal), but good measurement of the amplitude.

With one bit systems we have extremely high sample rates, but the lone sample just records the change, is the next sample bigger (higher voltage) than the previous. If yes we get 1, if smaller, 0. If we have enough samples this system also can record the waveform well enough. To describe a signal with accuracy of 16 bit PCM we need 2^16 as many samples per second or 44100*65536=2,889,037,600, almost 3 billion samples per sec. To compete with 24, even more. This gets unpractical and normal PCM is now prefererd. The high sample rate of one bit systems is nothing to write home about, it is not any advantage, just the way that system operates.

[Gutbucket]

... Good explanation by Nika Aldrich (who has a good book on digital audio) here:

http://www.users.qwest.net/%7Evolt42/cadenzarecording/DitherExplained.pdf
...

Excellent explanation.  One interesting thing he mentions that I was not aware of is that 'noise shaped' types of dither (which he refers to as 'colored' dither such a UV22, etc), should only be used at the very last wordlength reduction or could cause noise artifacts if dithered again.  Triangle Probability dither should instead be used if any additional processing might follow.

[dmccabe]

Bit depth has nothing to do with frequency range. Bit depth governs the (theoretical) maximum dynamic range, simplified it is 6 dB for each bit. Sample frequency determines the maximum audio frequency, max frequency is sample rate/2 (Nyquist theorem). This is so basic that I find it funny to write about it here.

You can't have sound without bit rate (dynamic range) AND frequency response. This is SO BASIC, I am still amazing you are continuing to post like you actually have experience in transferring analog cassettes to digital. Dynamic range IS DIFFERENT than frequency response, but they are directly related to each other. The fidelity of frequency response is also directly related to the quality of the bit converter. A sine curve at 24/96 has more detail than a sine curve at 16/44.1 -- even when the dynamic range is the same!

The human ear can't just listen to the dynamic range of sound... there needs to be frequency response at the same time.
Without going into great detail, the process in a pro A/D converter is not the same at different bit rates. At 16-bit, the A/D processes conversion DIFFERENTLY than at 24-bit... and you can hear that difference, even with the same source that has a limited dynamic range.

There are also playback D/A considerations that affect your perception of that sound. Bottom line, one should ALWAYS transfer in 24-bit to achieve the best fidelity when transferring analog recordings. To make a statement that is doesn't matter is misinformed. I can show you to many examples where an analog source was transferred to 24-bit/96 and 16/44.1... and the raw unaltered transfer will always sound better on the 24-bit version. I have never... ever seen an audio engineer recommend that an analog transfer be done in 16-bit... because there is "no benefit" to go up to 24-bit.

When you bring up the Nyquist theorem, it shows that you are trying to make your argument based on the mathematical limitations of dynamic range. Are you also saying that all A/D's running at the same bit rate and frequency sound the same?
Are you saying that any A/D sounds the same at 16-bit and 24-bit?

We use Mytek, Benchmark, Grace Design and Apogee A/D/A converters... they all sound different. Do you need to listen to some audio samples? Do you have the correct hardware to listen to a 24-bit/16-bit comparison? What is the monitoring path?

Old analog tapes propably have the noise floor at about -55 to -60 dB levels below peaks, even with 16 bits the lowest 30 dB at least will be hiss only. Using 24 bits will not give you any benefits, only almost 80 dB of hiss...

No benefits? To be blunt... are you sure you have any idea what you are talking about?

[Gutbucket]

...A sine curve at 24/96 has more detail than a sine curve at 16/44.1 -- even when the dynamic range is the same!

Isn't a sine curve a sine curve?  Doesn't the 16/44.1 version contain all the information to reconstruct the sine perfectly as long as it is within the frequency and dynamic limits of 16/44.1? Isn't the additional time domain / frequency data redundant in this simplified case? Of course this is closer to the mathematical theory side than a complex musical signal through real world converters.

...Without going into great detail, the process in a pro A/D converter is not the same at different bit rates. At 16-bit, the A/D processes conversion DIFFERENTLY than at 24-bit... and you can hear that difference, even with the same source that has a limited dynamic range.

There are also playback D/A considerations that affect your perception of that sound. Bottom line, one should ALWAYS transfer in 24-bit to achieve the best fidelity when transferring analog recordings. To make a statement that is doesn't matter is misinformed. I can show you to many examples where an analog source was transferred to 24-bit/96 and 16/44.1... and the raw unaltered transfer will always sound better on the 24-bit version...

^^^
I think this is the heart of the matter.   A 24bit tape transfer can sound better than a 16bit one.  At least in part because different circuits just sound different, but maybe for additional theoretical reasons too? 

Now here's the next question: After the initial Tape > A/D transfer at 24bits, assuming the musical portion of the signal is optimized in that available bit depth with plenty of range to spare, would a later wordlength reduction in the digital ream to 16 bits sound inferior?  Is it the A/D transfer stage that is the critical part here? or the depth of the eventual storage format?

Thanks to those of you with a deeper understanding and hands on-experience with this.

[Roving Sign]

Sorry to interrupt - just curious here....

Which would you prefer for 16 bit listening?

A) 24 bit master > dithered to 16 bits

B) Native 16 bit samples

I would tend to favor "B" - theoretically...

[boojum]

With the little knowledge that I have I would prefer 24 > 16 with dither.  I say this because the capture and any post processing would be done in 24 bit where there is less loss from this work, I believe.  I always record 24/48 and do post in 24/48 for this reason.  When I am done I mix down and dither to 16/44.1.

If it is just captured at 24 and then dithered down to 16 I do not see what benefit there can be.  It can never sound better than the original.  As has been posted earlier, it can sound different, but only with tweaking.  If I am wrong someone please tell me how the copy can sound better than the original.  Can a 24 bit copy sound better than a 16 bit copy?  I do not know.  Some think so.  But I would not put a lot of credence in that myself until it had been tested many times in double-blind tests.  As usual, YMMV.   

Happy Trails.

[Tim]

Sorry to interrupt - just curious here....

Which would you prefer for 16 bit listening?

A) 24 bit master > dithered to 16 bits

B) Native 16 bit samples

I would tend to favor "B" - theoretically...

depends on the dither but most of the time I would prefer option "A"

why did we all love those ad500 and ad1000's? It was their onboard dithering that gave them that famous Apogee sound

[Brian Skalinder]

Old analog tapes propably have the noise floor at about -55 to -60 dB levels below peaks, even with 16 bits the lowest 30 dB at least will be hiss only. Using 24 bits will not give you any benefits, only almost 80 dB of hiss...

Even with tape's limited SNR spec, it's still possible to hear signal below the noise, so it probably is still helpful to run reasonably close to peak on transfer to 16 bit.

These two statements get to the crux of the issue for me.  This topic perpetually gives me brain cramps, but I'll dive in anyway, if only with a couple questions.  Both questions assume the same sample rate, for simplicity's sake:

If an analog cassette provides 60 dB of dynamic range, and below -60 dB is only noise...

• am I correct that we should achieve the same results recording at 10-bit, much less at 16- or 24-bit?  (forgetting for the moment that none of us likely have a 10-bit ADC)
• does a 24-bit transfer capture lower level detail about the noise than 16-bit (i.e. noise below -96 dB) , and therefore provide greater accuracy in performing noise reduction (if one planned on doing so now, or in the future)?  (I suppose an implied question here:  if the noise on an analog cassette starts at -60 dB...how low does it go?  Below -96 dB?  If so, how far?)

If an analog cassette provides 60 dB of dynamic range, and below -60 dB is shared by both noise and signal...

• aren't there benefits to recording at 16- or 24-bit to capture the signal sharing those least significant bits with the noise?

Just found this in the rec.audio.pro FAQ suggesting that we hear signal below the noise floor:

5.14 - How can a 16-bit word length be enough to record all the detailin music? Doesn't that mean that the sound below -96 dB gets lost in thenoise? Since it is commonly understood that humans can perceive audiothat IS below the noise floor, aren't we losing something in digitalthat we don't lose in analog?

You're correct in saying that human hearing is capable of perceiving
audio that is well below the noise floor (we won't say what kind of
noise floor just yet). The reason it can do this is through a process
the ear and brain employ called averaging.

If we look at a single sample in a digital system or an instantaneous
shapshot in an analog system, the resulting value that we measure will
consist of some part signal and some part ambiguity. Regardless of the
real value of the signal, the presence of noise in the analog system
or quantization in the digital system sets a limit on the accuracy to
which we can unambiguously know what the original signal value was. So
on an individual sample or instantaneous snapshot, there is no way
that either ear or measurement instrument can detect signals that are
buried below either the noise or the quantization level (when properly
dithered).

However, if we look at (or listen to) much more than a single sample,
through the process of averaging, both instruments and the ear are
capable of detecting real signals below the noise floor. Let's look at
the simple case of a constant voltage that is 1/10th the value of the
noise floor. At the instantaneous or sample point, the noise value
overwhelms the signal completely. But, as we collect more consecutive
snapshots or samples, an interesting thing begins to happen. The noise
(or dither) is random and its long term average is, in fact, 0. But the
signal has a definite value, 1/10. Average the signal long enough, and the
average value due to the noise approaches 0, but the average value of
the signal remains constant at 1/10.

A somewhat analogous process happens with high frequency tones. In
this case the averaging effect is that of a narrow-band filter. The
spectrum of the noise (or simple dither) is broadband, but the
spectrum of the tone is very narrow band. Place a filter centered on
the tone and while we make the filter narrower and narrower, the
contribution of the noise gets less and less, but the contribution of
the signal remains the same.

Both the ear and measurement instruments are capable of averaging
and filtering, and together are capable of pulling real signals from
deep down within the noise, as long as the signals have one of two
properties: either a period that is long compared to the inherent
sampling period of the signal in a digital system or long compared to
the reciprocal of the bandwidth in an analog system, or a periodic
signal that remains periodic for a comparably long time.

Special measurement instrument were developed decades ago that were
capable of easily detecting real signals that were 60 dB below the
broadband noise floor. And these devices are equally capable of
detecting signals under similar conditions in properly dithered
digital systems as well.

How much the ear is capable of detecting is dependent upon many
conditions, such as the frequency and relative strength of the tone,
as well as individual factors such as aging, hearing damage and the
like.

But the same rules apply to both analog systems with noise and digital
systems with decorrelated quantization noise.

[Roving Sign]

Another interjection:

If I follow the dmccabe logic - shouldnt we all be remastering our old 16 bit DATs at 24 bits - just like the cassette process? If 24 bits makes cassettes sound better, why not 16 bit DATs???

[Tim]

Another interjection:

If I follow the dmccabe logic - shouldnt we all be remastering our old 16 bit DATs at 24 bits - just like the cassette process? If 24 bits makes cassettes sound better, why not 16 bit DATs???

I think many people are indeed doing just this

I know that's what I plan on doing

edit: Maybe I am just confused

[Brian Skalinder]

If I follow the dmccabe logic - shouldnt we all be remastering our old 16 bit DATs at 24 bits - just like the cassette process? If 24 bits makes cassettes sound better, why not 16 bit DATs???

If making edits to 16-bit masters after transferring to computer, then yes - you should do so with software that uses internal precision of 24-bit, or better yet 32bfp.  In this case, we're talking about the precision used in manipulation of the digital signal within the digital realm, not the transfer of an analog signal into the digital realm.  Different beast entirely.

There's no point in transferring a 16-bit DAT to PC at 24-bit, since it's all within the digital realm (DAT > bit-transparent S/PDIF > PC).  The least significant bits would simply be padded with zeroes.  I suppose one could transfer into the analog realm and then back (DAT > DAC > ADC > PC), but...not sure what the point would be, unless one wanted to take advantage of the particular sonic signature of the intermediate analog gear.

[dmccabe]

Another interjection:

If I follow the dmccabe logic - shouldnt we all be remastering our old 16 bit DATs at 24 bits - just like the cassette process? If 24 bits makes cassettes sound better, why not 16 bit DATs???

no, because the DAT 16-bit was already dithered... and adding another dither would be a destructive edit and hurt the overall quality. But we have found that if you run a DAT out through a high-quality DAC back to "the best analog you can get out of the DAT"... then go back in 24/96 to do any mastering... that sounds way better than upsampling a 16/44.1 to 24/96 for editing or remastering.

As noted in previous post... "noise"... especially analog tape noise is an integral part of the overall fidelity. Dismissing the noise as just wasted bits is not correct... it affects the overall sound. Forgetting about broadband noise reduction in 24-bit as a main beneift... as I stated: We have done many tests converting high quality analog reels through the same A/D in 16/44.1 - 24/44.1 and 24/96. Listening to those raw transfers, you can hear the difference between all three. Maybe alot of that has to do with playing the 24-bit back through a 24-bit DAC... but there you go... you can still HEAR the difference.

As for going to 16-bit from a 24-bit source...

Most top mastering engineers (I am not going to drop a name here, but think the entire Hendrix remastered catalog) take their 24/192 digital masters and send them back out to analog... then back into 16-bit/44.1 A/D of their choice for Cd master... rather than dither them in software. Anyone who states that once it is "digital it should stay digital"... is not doing what the top mastering engineers are doing in LA today.