Topic: Analog to 24bit question

[Brian Skalinder]

Are we talking about increasing the bit depth and leaving the sample rate unchanged or is the sample rate being increased as well?

Both bit-depth and sample rate are in play within this discussion.  Personally, I'm trying to understand the issues involving bit-depth first, before moving on to sample rate, hence my focus strictly on bit-depth in the last few posts.

[dmccabe]

I think I misspoke when stating all DACs "upsample" before the analog stage... some DACs do not. (I got slapped for that statement by someone who knows a lot more than me
If you look at different D/As, they use different methods, so there is no one way in the D/A stage.
It all depends on the actual chips in the particular hardware. Some DACs use 1-bit chips, so there is no dither at all in the D/A... but the point is not to get all hung up on the "engineering" of the specs.

The bottom line is "more bits" the better the signal to noise ratio.
The more frequency response, the better the detail on the transients.

They go together... you want both. That is why if you transfer a cassette at 24/96 rather than 16/44.1 it will sound more "open"... more analog.

And yes, "generally" the more dithers, the more chance for unwanted artifacts.

When you get into individual manufacturers of hardware you can then throw in a coloration equation.
Each piece of gear can make the sound different. So then it comes down to which sound you like best.

I don't design or make the hardware, I just use it

[dmccabe]

So I still don't understand how/why one is better than the other.

I am not stating one method is better than another... you need to compare specific hardware and/or software.

If you read that article I posted about the analog summing, the writer starts the entire comparison by stating he is not trying to say which one is "better".
For each type of music, there might be one method that sounds better. The key is to know all your options.

For some people, once they are in digital, they stay all digital.
Others like to use analog in their mixing or mastering stages.

But to get back to the very first original post, "does it make any difference to go to 24-bit when transferring a cassette: -- I'll answer YES.
Your mileage may vary.

[Brian Skalinder]

The bottom line is "more bits" the better the signal to noise ratio.
The more frequency response, the better the detail on the transients.

And...we're back to page 1.  Oh, well.

I don't think anyone disputes this, in general.  The question from the very beginning of the thread, batted about in various forms since, is:  if the SNR and frequency response are limited in the analog master (for example, 60 dB SNR and 16 kHz), how does transferring at higher bit-depth and sample rate improve <a> SNR, and <b> frequency response / transient detail?  Nothing I've seen here suggests that it does, but I do suspect that the closer approximation of the analog waveform - while not necessarily improving SNR or frequency response - provides better precision for future editing in the digital realm.

Time to poke around on Google to dig into the above, and subsequent, unanswered questions (why the "best" way to get from a 16-bit digital master to 24-bit is through DAC > ADC, how introducing an analog stage somehow helps with the negative effects of double dither, etc.).  Nothing personal, it just seems we're not going to get to the answers here.

[dmccabe]

On a related note:  does the software MOTB uses for edits use 24-bit or 32bfp internal precision?  And if the latter, wouldn't it make sense to convert 16-bit > DAC > ADC > 32bfp?

There are many different editors, each uses their own DAW gear... so it's not all the same for every release. If you look at the lineage on each show, it is well documented.
And not all plug-ins are the same quality as well. Everyone uses what they can afford. We do get donated gear... and are very thankful to our sponsers, but we do this all for free, we don't get paid or anything.  But I think I can say generally most of us are going for 24/96 for our own personal files. Those files are big enough. Going to 32bfp is sooo much bigger... there is a point of diminishing return.

And now we are doing DSD stuff... each raw transfer ends up around 10megs per show. People have to ftp files... etc. 24/96 seems high enough for now. As cpu and internet bandwidth get faster...
who knows what we will all be working with in a few years. Last we knew 16-bit was the "best"!

[dmccabe]

Time to poke around on Google to dig into the above, and subsequent, unanswered questions (why the "best" way to get from a 16-bit digital master to 24-bit is through DAC > ADC, how introducing an analog stage somehow helps with the negative effects of double dither, etc.).

You are probably not going to find any studios who get paid for transferring DAT masters -- what bands used 2 track DAT for studio work? -- maybe ADATs but..., thus very little online discussion about the best methods of editing ADATs. The best way is to test yourself. Do you have a high-end DAC? If not, send me a DAT, I can transfer it through a Benchmark DAC-1, then back to digital via Mytek at 24/96. Then ftp you the file, you can do your own tests of that file compared to a raw 16/48 transfer.

Then also edit both files and do another comparison.

Nothing personal, it just seems we're not going to get to the answers here.

Oh how I agree... there is no manual for sound engineers that says "this is how it is supposed to be done".
And if there was, no one would pay the pros to do it!

I just consider myself an addicted audiophile with an expensive hobby. Others in our group make a living mastering audio.

[Gutbucket]

If you come across anything you find enlightening Brian, please post here.

I still suspect that increasing bit depth only in the DAW is inconsequential as it just pads the word with zeros. In contrast to that, I suspect changing the sample rate either up or down (other than by a simple multiple, such as 44.1>88.2 or 48>96) is where the digital conversions cannot (yet?) do as nice a job as converting to analog and back to the target rate like MOTB & the mastering world are doing.

The reason I suspect that's the case is because SRC is a more complex mathematical operation (multiplication) not unlike digital summing (addition) that the guys going to analog summing that dmccabe posted the article about are avoiding.

Good discussion. dmccabe, thanks for sharing your on-hand experience with this stuff and especially for your MOTB work.

[boojum]

^^^^  Thank you!

[bensyverson]

Warning: huge post -- I'm just thinking out loud.

Here's my perspective as an image processing engineer... The same issues of bit depth and sampling rate (resolution) obviously come up in image processing. (I have also built a lot of audio processing circuits using pure data)

First, the difference between 44, 48 and 96k sampling rates. This is analogous to image resolution. In image processing, fine details are referred to as "high frequency," and this corresponds to high frequency audio. The better the sampling rate, the better you're able to reproduce high frequency detail. 44k has 0.92 times the resolution of 48k, and 96k has twice the resolution of 48k.

Here's 44k:


And 48k:


And 96k:


You can see how small the difference is between 44k and 48k. Both have enough resolution after Nyquist to cover the range of normal human hearing, which tops out at ~20k. There are some fine details you can see better in the 48k image, but you're squinting.

The 96k image is obviously larger, and there are more details, but... uh, unless you pitch shift that audio, it's just resolving details you can't hear. It would be like using a 20 megapixel camera when your lens could only resolve 10 MP of information. Just my opinion.

Bit depth is another issue. Bit depth is not "resolution." It simply defines how many steps of amplitude there are between 0 and 1. (1 usually being 0db). Here's the 24bit color image (8 bit per color channel, commonly but confusingly called "8 bit"):


Here's the same image resampled to 4 bit (16 amplitude values), so you can see the steps clearly:

Obviously this is why no one records at 4 bit.

However, here's the same image in 4 bit, but dithered using the algorithm generally accepted as "best" -- it's called error diffusion, or just "diffusion" for short:

Clearly, that is much, much better.

Here's a less impressive dithering technique ("pattern dither").

This demonstrates that the dithering style does make a difference -- pattern dither introduces distracting artifacts. IMO, for best results, use gear/software with error diffusion dithering.

Some audio gear actually just takes the bottom 16 bits of a 24bit signal (someone said Quicktime does this?). This is disastrous, since anything that reaches above 2/3 of the way to 0db will be unceremoniously clipped off. That would look like this:


24 bit vs 16bit
Okay, so how big of a difference is there between 16 and 24 bit? Well, in order to make the difference more clear, I'm going to filter out the high frequency information by blurring the image. Here's a 24bit image:


And here's the 16bit version (resampled to 24bit jpg for viewing on the web):


Wow, okay, the 24 bit version looks loads better! Well, that 16bit conversion didn't use any dithering. If we were converting from 24 > 16bit and used diffusion dithering, you would be hard pressed to spot the difference.

There's actually another factor here related to dithering. Error diffusion works by introducing minute errors (noise) to the signal. But your signal already has noise if it came from a mic, went through a pre, and passed through A/D. So even if you're not working with gear that dithers, odds are that your signal is doing the dithering for you.

Here's the same 24bit image with a small amount of noise added to emulate the amount of noise in a nice, clean recording:


And the signal with the same amount of noise at 16 bit:

It looks very close to the 24bit image. The 24bit is maybe a tiny bit nicer, but you have to be looking for it.

This is why Sound Devices says that at full signal level, 24 and 16 bit sound "largely identical." They should know.

Headroom
Of course, as SD points out, the world is not perfect, and you can't always be kissing 0db. So how does 24bit stack up against 16bit if you record with your levels down, and normalize (multiply) them later? Lets say you leave enough headroom that most of your audio peaks at -18db. You want to be sure that you won't clip if the dude next to you yells. That means you're only using 1/8th of the available levels -- in 16bit, 8192 levels; in 24bit, 2,097,152 levels. In other words, now your 16bit audio is really 13bit, and your 24bit is really 21bit.

The real issue, however, isn't the loss of amplitude fidelity. It's the fact that we're amplifying the ADC noise. Let's imagine that our ADC introduces an amount of noise that equals about 2 amplitude levels. So out of the 8192 levels used in our 16bit file, 2 of those are noise. And of the 2,097,152 levels used in our 24bit file, 2 of those are noise. See the issue? In the 16bit file, 2/8192 = 0.024% of the signal is noise, versus 0.0001% noise for the 24 bit file. Relative to the signal, the 16bit file has exactly 256 times more digital noise than the 24bit version.

That said, ADCs are all different, and will have different "characters," so if you do plan on running with a lot of overhead, it's probably best to try a few out and see what sounds good to you.

Dynamic Range
So, what about dynamic range? Dynamic range is one of the most misunderstood terms in both digital photography and digital audio. That's partially because DR figures are almost always given in logarithmic scale (stops in photography, dB in audio), and people always get tripped up with log numbers. It's also because people aren't sure what affects DR.

There are only two things that can affect DR: noise and bit depth. Noise is simple: your mic has a self-noise, your pre adds some noise, and your ADC adds some noise. Whatever you have left between the noise floor and 0db is your raw dynamic range. Bit depth affects DR, because audio is typically encoded linearly. So in 16bit, you have 65536 levels. Half of those levels (32768) cover 0db to -6db, half of the remaining levels (16384) cover -6 to -12, half of that (8192) cover -12 to -18, etc. By the time you get to the range between -66db and -72db, there are only 16 levels of amplitude to describe the waveform -- pretty gritty. By the time you get to the range between -84db and -90db, there are only two levels -- a square wave, either on or off. Of course, noise takes over long before we get to that point.

So the theoretical dynamic range of 16bit is 90db, but the last 30db or so are pretty rough. This is why some people think very soft sounds start to sound bad in 16bit -- for example, the oft-cited "end of the decay of a cymbal." There are plenty of microphones that have over 66db of dynamic range, so they can expose the limits of 16bit.

With 24bit, you start off with more levels. There are over 8 million levels to describe the amplitude between 0 and -6db!!! Obviously massive overkill. But the result is that between -66db and -72db, you have 4096 levels available vs 16 levels in 16bit. It takes 24bit a bit longer to get clipped to 16 levels -- you have to get down to the range between -114db and -120db. The theoretical limit to the DR in 24bit is 138db, because the range between -132 and -138db gives us only two levels.

Luckily, no microphone is capable of capturing that. Even microphone/pre/ADC combos capable of reaching 80db of DR will still have 1024 levels with which to describe their noise floor in 24bit.

[bensyverson]

LOL

[bensyverson]

I just wanted to add something... for those people concerned about dithering "adding noise" to their recordings. First, dithering only occurs when going from a higher bit depth to a lower one. So you want it when going 24 > 16, but not 16 > 24. When going from a lower bit depth to a higher one, all your hardware will do is "add zeros" to the end of the values. Actually, you shift the bits, but "adding zeros" is a good way to think about it. The point is, going UP the bit depth chain in the digital realm is absolutely lossless. You literally lose nothing from the original. You also gain nothing unless you're doing processing -- that is, there's no reason to bump 16bit DAT materials to 24bit unless you want to apply effects in 24bit space. You can't improve your 16bit recordings by resampling them to 24bit.

Okay, back to the noise added by dithering. The noise added is always less than one value in the destination bit depth. So if you're going from 24 > 16, the amount of noise added is less than one value out of 65,536. Nothing to worry about.

The noise is basically added to round numbers -- say you have an amplitude sample that's roughly 33.33% of the max. In 24bit, maybe the value is 5,592,550. If we divide this by the max value (16,777,215) and then multiply it by the max value of 16bit (65535), we get 21845.566. Without dithering, we'd just chop it off at the decimal place and call it 21845. Error diffusion will add a random value between -1 and 1 before the rounding happens. So for example, it might add 0.6 to 21845.566, to get 21846.166. Once we round the number, now we have 21846. Or we might get 21844. That way, we get a smoother gradation between values.

The point is: I wouldn't sweat the dithering "noise."

[dmccabe]

Here's my perspective as an image processing engineer... The same issues of bit depth and sampling rate (resolution) obviously come up in image processing.

You can see how small the difference is between 44k and 48k. Both have enough resolution after Nyquist to cover the range of normal human hearing, which tops out at ~20k. There are some fine details you can see better in the 48k image, but you're squinting.

The 96k image is obviously larger, and there are more details, but... uh, unless you pitch shift that audio, it's just resolving details you can't hear.

Very good analogy... but as indicated in red, "unless you" -- meaning you are going to make an edit... so, if you were to make a retouching edit to an image, the 96K image would certainly give the best quality image for the edit - not on the smaller one, even if you then had to res it down to a smaller size later... but still needed the bigger size -- for your DVD-A burn...

so if you need to do a...
- pitch shift
- eq and tonal adjustments
- cross fade from patches
- broadband noise reduction
... what bit rate and frequency would you use on a cassette master?

--- what bit rate and frequency would you use to make the same edits on a DAT master?

You can't improve your 16bit recordings by resampling them to 24bit.

I don't think anyone is claiming resampling is "improving" anything. We are talking about destructive editing. And the initial thread started with an analog master.
The topics seem to shift around with people pulling out their slide rulers to demonstrate how brand X sounds so much better than brand Y.

[Roving Sign]

Warning: huge post -- I'm just thinking out loud.

Here's my perspective as an image processing engineer...

Wow thanks for this post! I was just reading the Wikipedia entry for optical resolution - as I thought there had to be an analogy there. Thanks for spelling it all out.

[bensyverson]

Very good analogy... but as indicated in red, "unless you" -- meaning you are going to make an edit... so, if you were to make a retouching edit to an image, the 96K image would certainly give the best quality image for the edit - not on the smaller one, even if you then had to res it down to a smaller size later... but still needed the bigger size -- for your DVD-A burn...

so if you need to do a...
- pitch shift
- eq and tonal adjustments
- cross fade from patches
- broadband noise reduction
... what bit rate and frequency would you use on a cassette master?

--- what bit rate and frequency would you use to make the same edits on a DAT master?

Very good point, and there is something that I missed that gives 96k an edge right off the bat. Your ADC will introduce errors in the form of digital noise -- at 48k, that noise is happening at (duh) 48k. If you record in 96k and output to 48k, it means you're essentially oversampling that noise by 2X. That helps smooth out errors in the ADC.

To return to my analogy, as I said, 96k is like a 20 megapixel camera, even though your lens (in this analogy, your ears) can only resolve 10 MP of information. But downsampled, the 20MP camera will make a less noisy 10MP image than a 10MP camera with the same S/N ratio. So for every one pixel in the 10MP image, there are two pixels in the 20MP image, which helps average out the noise.

Oversampling is a common method to improve your effective S/N ratio, so I have to admit, 96k has some real appeal.

How much of a difference will this make, considering 48k is already oversampling (ie, above Nyquist) the limits of human hearing? Who knows, but if you know you're going to be manipulating a signal, it always helps to start with more of it! If you think of 48k as roughly 2X oversampling human hearing, then 96k would be a 4X oversample, which is standard in image processing to avoid aliasing. I would think that 96k would make even a slightly noisy A/D effectively transparent compared to the same A/D running at 48k.

I don't think anyone is claiming resampling is "improving" anything. We are talking about destructive editing. And the initial thread started with an analog master.
The topics seem to shift around with people pulling out their slide rulers to demonstrate how brand X sounds so much better than brand Y.

I think that could easily be true (brand X is better than brand Y). What you'll need to faithfully reproduce an analog master will depend on the format (2" RTR? Mini cassette?), but it comes down to figuring out the noise floor of the analog master, and then determining exactly how well you want to reproduce that noise. If you basically want a digital clone of a high-quality analog tape (ie, 60db dynamic range or above), 24bit is necessary, and you'd want to sample at either 48k or 96k, depending on the quality of your ADC.

[Brian Skalinder]

... what bit rate and frequency would you use on a cassette master?

--- what bit rate and frequency would you use to make the same edits on a DAT master?

Higher the better, as most have agreed.  But most of the (well, my, anyway) questions and discussion revolved around your suggestion that it's better to convert from a lower bit-depth (or sampling rate) to a higher via a DAC / ADC analog stage, and why (which was never really addressed).  Anyway...gonna let that one go now...

The topics seem to shift around with people pulling out their slide rulers to demonstrate how brand X sounds so much better than brand Y.

Not sure where that came from - you tossed out more brand X, brand Y, engineer X, and engineer Y comments than anyone else.  :shrugs:

  Thanks bensyverson for your lengthy posts - they made a lot of sense to me.  I still need to review another time or two to digest fully, though.