why bother with 24 bit

[crosstudio]

Here is where Farview is going:
If your sampling rate is not sufficiently high, you won't even be able to capture those higher frequencies because you won't be able to capture the changes in the wave

Here is where Danny.Guitar is going:
the more samples you take the more accurate your digital depiction of the analog sound will be.

and you are both absolutely right.

 

[Farview]

which is why voices sound more digital when recorded at grossly lower sampling rates. that has nothing to do with frequency and has everything to do with waves and time.

When voices are recorded at gossly low sample rates, that swishy sound is the audio interacting with the sample frequency (aliasing) Also, grossly low sample rates are also accompanied with low bit depth, adding to the horrible audio.

 

[SouthSIDE Glen]

Here is where Danny.Guitar is going:
the more samples you take the more accurate your digital depiction of the analog sound will be.

But once again (man, this is getting repetitive; it just goes to show how ingrained miscomprhensions can become), the "accuracy" is in increased frequency response only. Danny's wave pictures demonstrate in it well. Those little hashmarkds in the second picture represent smaller segnemts of time (higer sample rates.) Any deformation in the wav that is pictured that can only be measured by such an increased sample rate are deformations that happen at higher frequencies than the wav pictured.

Slice a wave into 8 pieces (for example), and you are effectively measuring at ~4x that wave's frequency. If that's a 20kHz wave, you are then sampling at 160kHz (20kHz x 8) and getting an effective "resolution" - or more correctly, frequency response - of something below 80kHz. Now try slicing it into 16 pieces. That's a sample rate of 320kHz with a frequency response of around 150kHz or so.

But the 20kHz sine will still be interpolated *exactly the same* whether you sample it at 8 locations or 16. It doesn't matter; the converter will fill in the "blanks" (so to speak) exactly the same way.

But - and this is the most important part that people have the most difficult part with - any detail that is picked up at 16 slices that "falls between the cracks" at 8 slices is not a detail that's happening at 20kHz, it's higher frequency information. The fact that it happens to be "riding upon" a 20kHz wave is as irrelevant to this discussion as the fact that the 20kHz wave is probably itself riding upon all sorts of mixed frequency waves of lower frequencies than it.

Picking up on such a detail is nothing more than picking up on higher frequency information, which is just a fancy way of saying that you are increasing your frequency response.

G.

 

[Farview]

Here is where Danny.Guitar is going:
the more samples you take the more accurate your digital depiction of the analog sound will be.

and you are both absolutely right.

Not entirely. If there is no signal above 22k, A higher sample rate will not get you a more accurate picture of anything. The only thing a higher sample rate can do for you is record higher frequencies. If there are no frequencies above the audible spectrum, there is no more 'detail' to record.

 

[NL5]

Not entirely. If there is no signal above 22k, A higher sample rate will not get you a more accurate picture of anything. The only thing a higher sample rate can do for you is record higher frequencies. If there are no frequencies above the audible spectrum, there is no more 'detail' to record.

Face it man, some people just DON'T get it. They are the kind of people that all the 192K hardware sellers are counting on. ....and there's alot of them.

Do any of you know who Dan Lavry is? Have you read his credentials? Did you read the fucking article he wrote?

If you discount the inherent shortcomings of the converters, 44.1 is COMPLETELY accurate in reproducing the AUDIO band. Once you have complete accuracy, there is nothing more. Sorry, but that is a fact. Now, if you read Dan Lavry's article, there are reasons for going above 44.1, but it has NOTHING to do with the ability to reproduce the wave with 44.1khz sampling - it is to make up for the converters. According to Mr. Lavry, everything after about 60khz is a waste, and 192 is a joke, and can actually be WORSE.

Read the fucking article - PLEASE.

edit - and go listen to Joshua Judges Ruth by Lyle Lovett - recorded on crappy 16 bit 44.1khz converters - no tape, no 96khz, no 192.

 

[danny.guitar]

Face it man, some people just DON'T get it. They are the kind of people that all the 192K hardware sellers are counting on. ....and there's alot of them.

I had understood it just fine not even half way into the thread. I was just posting what I always had thought about sample rates, and apparently what a lot of other people thought about sample rates.

 

[NL5]

I had understood it just fine not even half way into the thread. I was just posting what I always had thought about sample rates, and apparently what a lot of other people thought about sample rates.

My post wasn't directed at you at all. I read what you said, and understood. Three pages of people insisting that "more must be better" without really understanding the basic concept is just getting a bit tiring. Then to add insult to injury, they won't even read Dan Lavry's .pdf - but hey, what does he know, right? The people arguing for higher sample rates aren't even arguing the legitimate reasons for doing so.

Peace.

 

[danny.guitar]

My post wasn't directed at you at all. I read what you said, and understood. Three pages of people insisting that "more must be better" without really understanding the basic concept is just getting a bit tiring. Then to add insult to injury, they won't even read Dan Lavry's .pdf - but hey, what does he know, right? The people arguing for higher sample rates aren't even arguing the legitimate reasons for doing so.

Peace.

Oh ok my mistake.

I guess I can save my CPU power now and not record some of my songs at 96KHz/24-bit. 90% of the time I use 44.1/24-bit but for very few songs with maybe 1 or 2 tracks I used 96. Guess I can save my disk space now.

 

[Tim Gillett]

Ok but you don't mention the fact, which I believe is critical, that digital is a sampled analog signal (before it comes out as analog) and analog to analog is a linear process, with an infinite sampling rate, if one wishes to make a comparison to digital in that fashion. So no, I still say you can't compare apples and elephants. Two different technologies. Your earlier post made it sound as if the quality of sound recordings, for the Kind of Blue sessions ('59), is equivalent to below 16bits and that 16bits would be a dream for those sessions back then. One would then assume, especially someone who may not be familiar with analog all that much, that the technology back then enabled only tinny and hissy sounding recordings. That's misleading but that's what can be derived from your post. Your statement that "tracking even at 16 bits gives options that were only a dream back in '59" is ridiculous. In what way ? S/N ratio again ? Maybe but have you ever heard a good 50's recording straight from a master tape or good vinyl ? It's HUGE!



Yes, it's a measurement but doesn't tell the whole story, just as frequency response. Specs on paper mean little. If specs were all that then the Compact Disc, for example, should rightfully sound much better than good analog but it doesn't.



This is only important if you want dead silence. I personally would take a good analog source over better S/N ratio. Both may not be mutually exclusive however. I've heard some pretty quiet 50's recordings, shockingly good.

Daniel, the points you raised no one else has picked up on, probably because it's tangential to the main discussion. Still, I'll respond in this thread as you posted it here.
I mentioned the '59 Miles recording not because it was analog but because it has a s/n far smaller than 96db and yet it still sounds fine (I love listening to it) and is a classic recording. This was in response to people who were saying that you must track at 24bit (potentially 144db s/n). I was merely bringing in a point of comparison.

But back to analog since you mentioned it. The standard analog recording books of that era say that the recording process , while greatly improved on the old disc recording methods, was still the weak link in the chain. The best mics and preamps of that time had a considerably better s/n than the best tape machine could reproduce. Hence the later NR systems of Dolby and dbx were a great improvement, bringing the tape technology of the day much closer to the level of the mic and preamp technology of the day. This is just history.

Perhaps you need to learn from the old analog tape guys you seem to think you understand.

Regards Tim.

 

[Farview]

This was in response to people who were saying that you must track at 24bit (potentially 144db s/n). I was merely bringing in a point of comparison.

The other difference between the 50's and now: The S/N isn't as big of a deal when you are only dealing with a couple tracks. In the 50's no one was in danger of having 48 tracks all with a noise floor of -40dbVU getting mixed down to another deck with the same noise. All this stuff adds up.

 

[Tim Gillett]

Agreed, although this isnt the full picture. 48 tracks obviously raises the noise floor greatly over a few tracks but it normally also raises the program level too! The inportant thing is the overall signal to noise ratio, not simply the increased noise because of track addition.
We could say the same about multiple mics. Each extra mic/preamp adds to the noise but also adds to the overall output.
Track addition does not of itself weaken signal to noise, although in practice it probably does.


But as if partly to counteract that, we have the luxury today of easily muting all tracks in silent sections so that the noise of those tracks doesnt intrude on the final mix (so long as we can get away with it by not making the mix sound weird) Before automation that was tricky.
Because of the phenomenon of noise masking, the noise tends to be apparent when the track's program is low or silent.

Tim

 

[snow lizard]

OK, let me see if I've got this right.

Since PCM sampling relies on phase information to reconstruct audio from the digital samples, and since phase information is reliant on the accuracy of the clock to an extent, as well as other electronic processes required to make the alarm go off several thousand times per second (accurately), and since any inherent inaccuracy caused by jitter or such will effectively chump the audio and make it sound thin (phase cancellation), then having the same electronic problems occur several thousand more times per second just to hog up more processing resources in an effort to piss the dog off isn't necessarily going to help me to understand what time it is?

It seems to me that a converter designed for meticulously accurate operation that only offers a maximum sample rate of 44.1 KHz would sound better than a stock MP3 player converter at 384 KHz.

If I want to drive to work and red line my engine in second gear, will the extra engine cycles increase the accuracy of my commute?

And what about the math?


sl

 

[crosstudio]

Samples per second affects TIME. Inaccurate frequency reproduction is a side effect.

I press record. I wait N seconds and then I pluck my guitar string one time.

In a 88khz based recording, my string pluck started at:
sample 5 of 88,000 samples in the Nth second

In a 44.1khz based recording, I plucked that string at:
sample 2 (not 2.5, you lost time accuracy didn't you) in the Nth second

In a 22,050hz based recording, I plucked that string at:
sample 1 (not 1.25, you lost lots of accuracy didn't you) in the Nth second

you can see that samples per second does affect recording accuracy because it affects TIME.

The number of TIME SLICES (samples per second) affects frequency only because there are waves that happen so quickly that the start and end of the wave isn't captured without sufficient TIME SLICES.

a less than necessary detailed depiction of TIME is the root cause of missing frequencies.

edit:
BTW, if you downsample from 88khz to 44.1khz and then upsample back to 88khz, the guitar pluck that started at sample 5 of the Nth second is now located in the 4th sample of the Nth second. thus proving that it is TIME that is affected by reduced slices.

 

[Farview]

You scraping your pick across a string takes way more than 1 sample at 44.1k. Hell the attack of a kick drum is 12 samples at 44.1k.

Your theory assumes that there is actually something you can do that is only one sample long. Which is impossible because that would be above the nyquist frequency, therfore It wouldn't make it on the CD anyway.

Frequency IS time. Anything that happened to only one sample would be above the nyquist and not get recorded. The actual timing difference, if there was any, would be 1/44100 of a second. The delay you get from being a foot away from your speaker is 1/100 of a second, just to put it in perspective.

The timing thing is a non-issue.

 

[NL5]

Samples per second affects TIME. Inaccurate frequency reproduction is a side effect.

I press record. I wait N seconds and then I pluck my guitar string one time.

In a 88khz based recording, my string pluck started at:
sample 5 of 88,000 samples in the Nth second

In a 44.1khz based recording, I plucked that string at:
sample 2 (not 2.5, you lost time accuracy didn't you) in the Nth second

In a 22,050hz based recording, I plucked that string at:
sample 1 (not 1.25, you lost lots of accuracy didn't you) in the Nth second

you can see that samples per second does affect recording accuracy because it affects TIME.

The number of TIME SLICES (samples per second) affects frequency only because there are waves that happen so quickly that the start and end of the wave isn't captured without sufficient TIME SLICES.

a less than necessary detailed depiction of TIME is the root cause of missing frequencies.

edit:
BTW, if you downsample from 88khz to 44.1khz and then upsample back to 88khz, the guitar pluck that started at sample 5 of the Nth second is now located in the 4th sample of the Nth second. thus proving that it is TIME that is affected by reduced slices.

You do realize this is complete bullshit, right?

 

[SouthSIDE Glen]

The number of TIME SLICES (samples per second) affects frequency only because there are waves that happen so quickly that the start and end of the wave isn't captured without sufficient TIME SLICES.

That is exactly the point. Work with that.

The key phrase there is "sufficient time slices". The thing about starting on sample 2 or sample 2.5 or sample 5 is really misleading, because waht is important is actually not the digital picture, but the analog reconstruction of that digital picture. Jut because a smaller sample rate may not catch the upstroke of a wave until "after it started" does not mean the converter will reconstruct that wave starting at that point. The converter will examine the sample value (and the ones surrounding it) and determine that the wave did not start at that sample but instead started a calculated time before that. The delay is not induced as you described. There may be delays and timing issues caused by clock inaccuracies, jitter, etc. but not by the actual sample rate itself. And the Nyquest frequency (basically a sample rate twice that of the desired frequency, plus a ltittle extra for technical housekeeping) is a fast enough sample rate - providing enough time slices - to be able to properly calculate and reconstruct that full wave from the available data, including parts of the wave that fall before or after those samples.

That leaves any deviations from that wave that happen completly between samples. Such deviations cannot be inferred from the sample data itself, and therefore will not be reconstructed in the conversion back to analog. But, as you say, such deviations start and stop faster than the samples can catch, and therefore at a higher frequency than the samples can reproduce. It simply means that the lower sample rate is not able to reproduce the higher frequencies.

G.

 

[Whatmysay]

If time slices is not the issue then what about the claims of 1-bit recording with mega high sample rates

http://www.korg.com/service/downloadinfo.asp?DID=1253

Or is it to do with the reconstruction of the sample back into analogue – or are manufactures playing on our (at least some of us) naivety? I understand the significance of reducing bit rate has a lot to do with improved quality

But I can not understand how sample rates are not relevant to the quality even at higher bit rates – surely having 20 pictures of a wave form is better than 10 when it comes to reproducing the analogue sound. Basically with less pictures you are leaving bigger gaps for the software to interpret- even if the frequency if the sampling is higher than the frequency being sampled.

I’m less concerned about the number of pictures being taken and more interested in how the software deals with the gaps – particularly with any sort of distortion where a multitude of frequencies are present.

If you more knowledge people feel you have already dealt with this and I am one of those who just doesn’t get it pls ignore me – as I probably wouldn't hear the difference anyway?

Because while you’ve all been talking physics, biology has been diminishing my hearing everyday since 25 and without years of studio experience to train the hairs in my ears to have muscle memory – I’m a lost! It is not a disease I've got that is nearly everyone on the planet

I like irony, irony is healthy

PS Glen great tutorials thanks

 

[snow lizard]

If time slices is not the issue then what about the claims of 1-bit recording with mega high sample rates
<snip>

The 1 bit formats are a very different technology. Everything in this thread until now was talking about Pulse Code Modulation digital audio.

I'm not so sure about manufacturers playing onto naivety, but I haven't heard DSD yet either. Direct Stream Digital is the technology behind Super Audio Compact Disc. 1-bit, non PCM audio. It might be a great thing if such a format takes off later on because it's supposed to sound pretty good.

Every computer recording setup for the most part is PCM digital. I don't think anyone's come up with wave editing or plugin capability for DSD systems. You can record and play back, that's it. It's still a very new technology compared to PCM, but it looks promising so far.

Different can of worms than PCM.

Sample rates in PCM are relevant to what people like SSG have been talking about. The frequencies you're capable of recording. Once you're beyond the range of human hearing, what else is there? It takes a lot more processing power to accurately reproduce 80 KHz, but we can't hear it, so why use it?

As to how the software deals with the gaps, it isn't the software. It's the converter, and it's the PCM process. If you were to line up 2000 samples, each one of them individually would look like a small spike. It won't give you anywhere near enough information to reproduce a sound. So if you've got 2000 of them, essentially you've got 2000 very small spikes. But that's not how they come out. They interact with each other by phase relationships and bend. Kind of like what a magnet does to an electrical field, or the way a prism bends light. The samples don't come up with a connect-the-dots facsimile of the original wave. They reconstruct the original wave in all its detail. (At least they're supposed to, which brings up the overall quality of the conversion process, thus the converters. Good ones cost more for a reason, and it doesn't have much to do with sample rate anymore). The limit on what frequencies they can reproduce is the Nyquist limit. Whatever falls in "the gaps" and can't be reproduced accurately (higher frequencies) is filtered out.

Higher frequencis require more samples, so you have to run the process fast enough to cover human hearing, and the capabilities and limits of our mics and speakers.


sl

 

[Whatmysay]

Thanks got it

 

[NYMorningstar]

But once again (man, this is getting repetitive; it just goes to show how ingrained miscomprhensions can become), the "accuracy" is in increased frequency response only. Danny's wave pictures demonstrate in it well. Those little hashmarkds in the second picture represent smaller segnemts of time (higer sample rates.) Any deformation in the wav that is pictured that can only be measured by such an increased sample rate are deformations that happen at higher frequencies than the wav pictured.

Slice a wave into 8 pieces (for example), and you are effectively measuring at ~4x that wave's frequency. If that's a 20kHz wave, you are then sampling at 160kHz (20kHz x 8) and getting an effective "resolution" - or more correctly, frequency response - of something below 80kHz. Now try slicing it into 16 pieces. That's a sample rate of 320kHz with a frequency response of around 150kHz or so.

But the 20kHz sine will still be interpolated *exactly the same* whether you sample it at 8 locations or 16. It doesn't matter; the converter will fill in the "blanks" (so to speak) exactly the same way.

But - and this is the most important part that people have the most difficult part with - any detail that is picked up at 16 slices that "falls between the cracks" at 8 slices is not a detail that's happening at 20kHz, it's higher frequency information. The fact that it happens to be "riding upon" a 20kHz wave is as irrelevant to this discussion as the fact that the 20kHz wave is probably itself riding upon all sorts of mixed frequency waves of lower frequencies than it.

Picking up on such a detail is nothing more than picking up on higher frequency information, which is just a fancy way of saying that you are increasing your frequency response.

G.

This is one of the best explanations I've heard yet. It reminds me of quantum physics and higher dimensional relationships.