Music Player Network
Previous Thread
Next Thread
Print Thread
Page 2 of 11 1 2 3 4 10 11
Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
[/QB][/QUOTE]Nika-san,

With absolutely no experience with Pro Tools whatsoever, I can't make any statements about slamming a 2 bit - err 2 mix bus. However it would appear that there are some systems (analog and digital) that do a horrible job of summing 30, 40 whatever number of signals together in a coherent fashion.

Is it phase shift? Or just poor summing - this is outside the scope of discussing calculators and how numbers are added within a given digital system.[/QB][/QUOTE]

Drew,

Yes, digital (and definitely analog) mixers can be poorly designed. This does not mean that they inherently are, or that digital mixers inherently do bad math. The math is quite simple and errors done in the process would cause distortion, not phase shifts. Also, it really is as simple as discussion calculators, but that still does not guarantee that it's easy to design such mixers. I'm afraid that the math is really as simple as summing multiple channel's worth of words.

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by mixsit:
I recorded some simple acoustic music at -70-80 once. It came back noisy as hell, but not muted or raspy.
Imagine what it would have sounded like if what you put in was noisy already...

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by GZsound@hotmail.com:
So..does recording to digital media at low levels have an effect on the recording quality? And what is considered a low level? KISS...
It's all a matter of how large the dynamic range in what you're recording is. If it has a dynamic range (difference between peak and the quietest thing you can hear, potentially into the noise floor) or 50dB then you need to record it hot enough that you get 50dB worth of dynamic range about the noisefloor of the converters. If the converters have a dynamic range of 120dB then you need to record your material hotter than -70dB to maintain full fidelity.

Establishing how hot is "enough" is all a matter of knowing what it is you record. To answer your question, "low level" would be anything less than that, in my opinion.

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by Marshall Simmons:
btw--- the jaggies i was describing are called aliasing
Marshall
No. The jaggies are called "quantization error" and have nothing to do with aliasing.

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Marshall,

Your approach is the intutive approach but does not yield the correct answers.

Quote:
Originally posted by Marshall Simmons:
[QB]If you record at low levels at a fixed point bit rate, the lower the level, the less amount of points that can be used to describe the amplitude of a sine wave.
So far, so good.

Quote:
For instance, at full 16 bit resolution, you have 32768 points to describe the amplitude, and at 24 bit, its around 16 million. But if you record at a really low level, you USE a fraction of those points. IF you use 8 points to describe a sine wave, you get jaggies along that sine wave. When you normalize that file then, it'll still have those jaggies, but will take up the full amount of points your bit rate has to offer.
And that is also OK.

Quote:
Those jaggies will crate high frequency content because of the sharp edges (like square waves have infinite high frequency harmonics because of the sharp edges) THis high frequency content is called Quantenization noise (i think i spelled that wrong)
Sort of. The jaggies are not "high frequency." They are completely random content that does indeed yield quantization noise, but that noise is actually just white noise (RPDF white noise, to be specific, because it is completely random and has no specific probabilities that are determinable)

Quote:
This has nothing to do with the theoretical frequency range of a sound file. The theoretical frequency range is determined by the sampling rate that you recorded in... 44.100KHZ will give you a theoretical frquency range of 22.05KHZ. I say theoretical because all AD and DA converters have lowpass filters in them to stop higher frequencies then the Nyquest rate from passing. If there wasn't any lowpass filters and a frequency passed higher then the Nyquest rate (1/2 the sampling rate) you would get what is called Foldback. Foldback is basically when a frquency gets to high for a sampling rate and cannot be described. The points are then recognized as a lower frequency then it actually is and is rendered as that lower frequency (there is a formula for it, but i can't remember it off the top of my head)
Correct.

Quote:
BY the way, if you mix and record in floating point instead of fixed point, alot of these problems with BIT RATE goes away.
INCORRECT.

Your post did not give any account of what was "problematic" about bit depth to start with. There is problematic about bit depths. You just have to record such that any quantization noise is lower than the noise level in your recording material. Then there are no problems.

Recording in floating point does not solve any problems at all. It has the exact same quantization noise problems as fixed point does, though it can actually add distortion as well due to correlated noise artifacts at high levels.

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by Allan Speers:
"Think of it like this - you can record with peaks at -24 dbfs and yer still recording a 20 bit word..."

Brad, if Marshall is right about the "jaggies" (whatever you actually call them) then there would indeed be an advantage to recording fairly hot, even with a 24-bit system.
Allan,

First, Marshall is not completely right about the "jaggies."

But let's think about it... The jaggies represent completely random quantization error. That equals noise. So yes, you don't want that noise creeping into your signal. But what is the noise level of your track? If your track has a noise level that exceeds the level of "the jaggies" then the jaggies end up lower than the noise and are undectable. Think of how much "random error" (noise) there is in your tracks. Then we just need to make sure that any additional error caused by the conversion process (quantization noise) is lower in level than the random error you already have due to thermal noise, resistor noise, component error, atmospheric noise, transducer noise, room noise, etc.

Therefore, there is an advantage to recording higher, but only until the noise level of your track exceeds the quantization noise (there is a caveat we can go into later once we're all caught up with this concept).

Quote:
No matter what the peak level, the level of the upper harmonics, and of delicate room ambience is always significantly lower. That's the audio we need to be concerned with.
Correct. So what you need to do is make sure that the quietest thing you can hear ends up above the level of quantization noise. Once you've done that you're golden.

Quote:
Some have reported that recording a 24-bit signal at very low levels sounds fine. Why then does a track that is converted to 8 bit (even 12 bit) sound so edgy and, um, lacking in resolution (did I use the "R" word? Oops)
Because you're letting some of the audible signal end up below the noise floor. At THAT point you're too low in level, or using too few quantization steps.

Quote:
Why did the early 12-bit samplers sound the same? Is it possible that the accepted "1 bit lost for every 6db down" only refers to dynamic range, and not to ALL the deficiencies of lower bit-rates?

Surely there is SOMETHING gained from higher bit rates than just dynamic range?
Great question. There were two reasons that 12 bit samplers sounded like that. One of them is the one you've identified - not enough dynamic range, thus some of the signal is lost below the noise floor.

The other part, and the part you're really asking, is that older converters used what was called an "SAR" A/D converter that was dependent on a resistor ladder that was very difficult to calibrate. The result was that the lower you got the less linear the converter was, resulting in distortion. As you recorded hotter you got most of your signal out of the distortion range. This is why we've always been told to record hot with 16 bit converters - because 16 bit converters were not particularly linear. Now, with 1 bit delta sigma modulator and multibit modulator converters the linearity problems have gone away and the dynamic range is the only aspect of waveforms related to "bit depth."

Good question!

Quote:
The question then becomes, how many db's down ARE the harmonics and room ambience from the peaks in a typical track, and just how many bits down can you go before there is audible degradation?
That depends. What's the room noise in your room? Subtract 14dB from that number. Then subtract that number from the peak SPL of the performance. That's not an exact value but pretty durn close.

Room noise = 30dB SPL.
Subtract 14 = 16dB SPL
Singer sings at 80dB SPL
Dynamic range = 80-16 or 64dB
Need 11 bits.
A/D converter has 120dB of dynamic range:
Can turn signal down up to 56dB with no change in signal quality (120-64)

Quote:
My gut feeling is to agree with you that a -24dbfs peak @ 24-bit is completely acceptable, but I wish i had some hard numbers to look at.
There are your hard numbers.

As for recording -24dB FS peak, it all depends on the dynamic range. If you use converters with 120dB of dynamic range and you turn the signal down 24dB you get 96dB of dynamic range. If the signal is greater than that then you turned it down too far. If not then you're just fine.

Quote:
BTW: Nika, with a PT Mix system, you most definitely CAN overload the "mix bus." I used to do it constantly, and it sounds pretty bad. Digi has always maintained that you can simply pull-down your master (2-mix) fader to solve this. This is not true. When you do that, the distortion goes away, but "something" is not right with the sound. I don't care what Digi claims, I did very careful tests and am sure. You must lower the individual faders, much like with an analog console.
Allan, I, too, have done some very extensive tests on this while on the phone with some designers at Digi, including analysis in Spectrafoo, etc. The only caveat that I can see is A. Whether of not you used the dithered mixer, and B. Which implementation of the dithered mixer you used.

Cheers, Allan. Good post.

Nika.

Joined: Jul 2001
Posts: 413
Senior Member
Offline
Senior Member
Joined: Jul 2001
Posts: 413
I've not used Pro Tools so this may not be the case. But I once imported a file thinking it was recorded at 24 bit but was actually recorded in 16 bit. The files level was very low. Is it possible that the files you received are actually 16 bit??

Marc


Searching for a new sig
Joined: Apr 2002
Posts: 68
C
Senior Member
Offline
Senior Member
C
Joined: Apr 2002
Posts: 68
Nika,

I still see some misunderstanding in this thread from folks reading your answers.

Please answer this: Between two 24 bit recordings of a bass guitar(with the noise floor of each far above the noise floor of the converters) , if one is at recorded with peaks at -20dbfs and the other at -1dbfs does one have more quantization error than the other?

I believe not. But it looks implied that it is when you agreed with this statement:

Quote:
For instance, at full 16 bit resolution, you have 32768 points to describe the amplitude, and at 24 bit, its around 16 million. But if you record at a really low level, you USE a fraction of those points. IF you use 8 points to describe a sine wave, you get jaggies along that sine wave. When you normalize that file then, it'll still have those jaggies, but will take up the full amount of points your bit rate has to offer.


Kevin Perry
Nashville, TN
Joined: Sep 2000
Posts: 1,277
A
Platinum Member
Offline
Platinum Member
A
Joined: Sep 2000
Posts: 1,277
"The jaggies are called "quantization error"

-thanks, Nika! That was driving me nuts....
------------------------

One new question for you:

You wrote,

"older converters used what was called an "SAR" A/D converter that was dependent on a resistor ladder that was very difficult to calibrate. The result was that the lower you got the less linear the converter was, resulting in distortion. As you recorded hotter you got most of your signal out of the distortion range. This is why we've always been
told to record hot with 16 bit converters - because 16 bit converters were not particularly linear. Now, with 1 bit delta sigma modulator and multibit modulator converters the linearity problems have gone away and the dynamic range is the only aspect of waveforms related to "bit depth.""

Hmmm. that first part I understand, and that's quite interesting. however, I don't understand the last part. Who's using 1 bit delta sigma modulators and multibit modulator converters?
Am I? I thought the one-bit system was that Sony thang, and that all the current 24-bit converters are just that: 24 bit.

Either way, you seem to be saying that ALL modern ADC's are immune from the "SAR" ADCs' non-linearity problems. yes?

Joined: Jan 2001
Posts: 6,074
MP Hall of Fame Member
Offline
MP Hall of Fame Member
Joined: Jan 2001
Posts: 6,074
So..does recording to digital media at low levels have an effect on the recording quality? And what is considered a low level? KISS...[/qb][/QUOTE]What you are experiencing is most likely impedance mismatching, and or level mismatching.

Check your equipment (read the specs, and run tests with tones and a volt meter) to ensure that level "X" on your mixer corresponds to level "X" on your ADATs.

At the risk of sounding like a goddamn elitist gear snob, do you guys read the fucking manuals and have a basic understanding of electronics/levels etc?[/QB][/QUOTE]

Why risk sounding like a goddamn elitist gear snob when you can write a post like that and remove all doubt?

Impedance mismatching? Thanks. Read the manuals? Thanks, I ar stupid and didt do that. I think you have a fairly serious X and Y level balance problem.

Using the direct outs of my board means the only control over the input to the ADAT is with the input level control on the board. No faders, eq etc. Setting the levels to avoid mic clipping has meant the level to the ADAT is low..around mid scale on the meters. The question is regarding lack of fidelity with lower than optimum recorded levels. Do you have an answer?

What really gets my balls in an uproar are people like you that react instead of responding. It must be nice to have all the answers.


Mark G.
"A man may fail many times, but he isn't a failure until he begins to blame others" -- John Burroughs

"I consider ethics, as well as religion, as supplements to law in the government of man." -- Thomas Jefferson
Joined: Dec 2002
Posts: 80
L
Leifski Offline OP
Senior Member
OP Offline
Senior Member
L
Joined: Dec 2002
Posts: 80
Yo Hey,
Thanks for the answers... The more I think about these files I want to go in and re-record the whole mess. As far as over loading the mix buss, I have had the experience when I record say a guitar track hot, if I add an EQ and cut the bottom end it will immediately distort. I end up have to lower the gain on the file and then add the eq. Strange... It seems that less is more.

Joined: Feb 2001
Posts: 3,324
V
MP Hall of Fame Member
Offline
MP Hall of Fame Member
V
Joined: Feb 2001
Posts: 3,324
Quote:

Why risk sounding like a goddamn elitist gear snob when you can write a post like that and remove all doubt?
One of the things I strive for in my personal and professional life, is being succinct. \:D

Quote:

Using the direct outs of my board means the only control over the input to the ADAT is with the input level control on the board. No faders, eq etc. Setting the levels to avoid mic clipping has meant the level to the ADAT is low..around mid scale on the meters. The question is regarding lack of fidelity with lower than optimum recorded levels.
Quote:
Do you have an answer?
Yes.

You do not necessarily have to experience a "lack of fidelity" with lower than optimal levels. Nika provided a very detailed explaination on why this is: If the lowest levels of your material never get close to the noise floor of your system, you should be fine.

I have several feasible solutions to this scenario (levels mismatch from console to ADAT). If I were in your shoes, the 1st possible thing I might probably do is replace the operator (myself).

Quote:

What really gets my balls in an uproar are people like you that react instead of responding. It must be nice to have all the answers.
Mark G,

Don't take it personal, cuz it seems you have. My comments were not in any way directed at you, or your question. It is more an observation on the state of things. Seeing oneself quoted does that sometime huh?

My speculation still stands. Excluding yourself then, since you take affront to my question, how many of us that use and operate pro audio gear for any protracted period of time are diligent in reading and assimilating the information provided by the gear manufacturers, and, strive to have a comperhensive understanding of audio engineering principles and conventions?

And no, I don't have close to 1% of 1/1000 of the answers to anything.

Nika,

I understand that phase shift would not be the reason why summed digital audio sounds less than pristene. What then? And what is the design flaw that makes for poor analog summing?

Joined: Aug 2001
Posts: 707
Gold Member
Offline
Gold Member
Joined: Aug 2001
Posts: 707
Quote:
Originally posted by GZsound@hotmail.com:

What really gets my balls in an uproar are people like you that react instead of responding. It must be nice to have all the answers.
I don't think he meant to be so aggressive in that statement. Truth is, most of these basic concepts are often covered in pro-gear manuals. If they aren't, usually that company will have some sort of published material that covers this information in layman's terms. If you want to get really deep, buy a reference. There are all kinds of groovy books available that cover everything from basic acoustics, recording and miking techniques, all the way up to the latest digital signal processing.

But that doesn't mean you shouldn't ask...

On the other side of the coin - in the situation concerning having to keep the levels lower to avoid clipping - a good limiter on each channel will help immensely.

Now - my amateur take on the whole bit-rate thing:

I think it's still wiser to record at the top end of the bit ladder. This helps with things like reverb tails that, in theory, go into infinity. But, when you're limited to a final Least Significant Bit, it has to be either on or off and there is no in between. If you record a reverb tail at a low level, you're going to expose the quantization error more so than if you record at a higher level. Capich? So - just like tape, we're still fighting a kind of noise floor - even though you can make a sine wave that is half of the LSB still sound like a sine wave with a certain kind of dithering...

So - I've made my points, backed them up, argued them, and totally exposed my lack of knowledge in the process...


studioman adam
[][][][][][][][][][][][]
"uh........OK?"
Joined: Nov 2001
Posts: 168
F
Senior Member
Offline
Senior Member
F
Joined: Nov 2001
Posts: 168
Quote:
Fredo,
You say that Normalizing "doesn't help." Doesn't help with what, exactly?
In digital audio, any bit gives you 6 db. and any 6db DOUBLES the real possibilities of positions used by your audio signal.
In a 16 Bits format signal, you get 96db (16 bits x 6 db). A 16 bits signal has 65536 digital positions to store the audio.
If you are mixing into a 16 Bits signal, and you don't reach the 0db clipping point, letīs say you mix only to -6 db, you are using only 90db, and only 15 Bits (90db/6db).
15 bits gives you 32768 positions to store your digital data. (HALF than 16 Bits)
AS YOU CAN SEE, THE 6 "LAST" DB before clipping DOUBLES your data positions.
Then if you make your mix at -10db you are only using 20642 positions to store your digital signal (against 65536 that you would be using if you where using the complete dynamic range of a 16 bits digital signal).
Then, if you make a "Mastering" or "normalize" that -10db signal, then you still having 20642 positions, USED by your music, and the rest of positions, with "0" data. (20642 with MUSIC and 44849 with CEROS)
This same principle is aplyed to 24bits or 32bits signals.


Fredo Gevaert
Temple Of Tune
fredo@temple.be
http://www.temple.be
Joined: Dec 2002
Posts: 80
L
Leifski Offline OP
Senior Member
OP Offline
Senior Member
L
Joined: Dec 2002
Posts: 80
What was meant by "Normalizing won't help" is that the damage is done. The signal or information recorded will not get any better it will only get louder. The point is to record at a higher level in order to get the highest dynamic from the musician. That dynamic is lost and by normalizing the file will only increase a sixteen bit signal. It is the same as transferring data from a 16 bit adat machine. The signal might be on a 24 bit system but the tracks still sound like a 16 bit file becasue they were initially filtered and colored through a lesser system. I do believe I understand this right.....

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by ckevperry:

Please answer this: Between two 24 bit recordings of a bass guitar(with the noise floor of each far above the noise floor of the converters) , if one is at recorded with peaks at -20dbfs and the other at -1dbfs does one have more quantization error than the other?
Yes. It does. And therefore one has more quantization noise than the other.

BUT! If those noise levels are lower than the noisefloor of your signal path leading up to this process then neither are detectable.

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by Allan Speers:

One new question for you:

You wrote,

"older converters used what was called an "SAR" A/D converter that was dependent on a resistor ladder that was very difficult to calibrate. The result was that the lower you got the less linear the converter was, resulting in distortion. As you recorded hotter you got most of your signal out of the distortion range. This is why we've always been
told to record hot with 16 bit converters - because 16 bit converters were not particularly linear. Now, with 1 bit delta sigma modulator and multibit modulator converters the linearity problems have gone away and the dynamic range is the only aspect of waveforms related to "bit depth.""

Hmmm. that first part I understand, and that's quite interesting. however, I don't understand the last part. Who's using 1 bit delta sigma modulators and multibit modulator converters?
Am I? I thought the one-bit system was that Sony thang, and that all the current 24-bit converters are just that: 24 bit.


Yes, we all are. "Modulator" based converters are the staple of the industry and have been in use for about 12 years. I don't think a single company makes an SAR based converter anymore.

24 bit converters do not actually convert the audio using all 24 bits. They do it at an extremely high sampling rate using fewer bits and then filter the material in such a way that provides the additional quantization steps.

Quote:
Either way, you seem to be saying that ALL modern ADC's are immune from the "SAR" ADCs' non-linearity problems. yes?
That is correct. For the past ten years or so modulator based converters have really been the only type made (certain hi-fi pieces of equipment continued to use R-2R ladder DAC's for a while, but these, from my understanding, have also been eliminated from the market. I don't think a 24 bit SAR A/D converter exists. Non-linearities are thus no longer a problem with converters, and converters are linear throughout the entire range of operation.

For more you'll have to call me at the office. That is a big topic and explaining how converters work takes time - that I don't have for typing.

Cheers!
Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by NYC Drew:

Nika,

I understand that phase shift would not be the reason why summed digital audio sounds less than pristene. What then? And what is the design flaw that makes for poor analog summing?
Drew,

Ahh! Sorry. Analog? All bets are off.

Problems in analog mixers are noise and distortion. Phase would not be a problem of the summing itself (at least not that I can understand) but would be a problem at all other places throughout the desk.

Design flaws in poor analog summing? Sorry, I'm not an analog guy and don't really understand analog design well enough to answer. The biggest problem with analog mixing, however, is that it uses analog circuitry and analog components, each of which is inherently noisy and non-linear.

Sorry that didn't help much. Maybe Eveanna or someone else here with analog design chops (Frindle) could chime in and answer the question for us? I'm curious as well.

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by soldierman adam:
I think it's still wiser to record at the top end of the bit ladder. This helps with things like reverb tails that, in theory, go into infinity. But, when you're limited to a final Least Significant Bit, it has to be either on or off and there is no in between. If you record a reverb tail at a low level, you're going to expose the quantization error more so than if you record at a higher level. Capich? So - just like tape, we're still fighting a kind of noise floor - even though you can make a sine wave that is half of the LSB still sound like a sine wave with a certain kind of dithering...
Adam,

You are making good deductive analysis but without complete information.

What happens when the reverb tail goes into the noisefloor of your equipment. Do you need better quantization than the noisefloor of your equipment can support? The human ear is not capable of hearing very far below the noise floor (that's another thread in a biology and physiology forum). How far below the noise floor do you need to accurately capture if it's for the sake of the imperfect human ear?

Nika.

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by Nika:
Fredo,
You say that Normalizing "doesn't help." Doesn't help with what, exactly?
Fredo,

FYI, this was a rhetorical question.

Quote:
Originally posted by Fredo:
In digital audio, any bit gives you 6 db. and any 6db DOUBLES the real possibilities of positions used by your audio signal.
OK. Hold on, there. In digital audio any bit gives you 6dB of dynamic range. We have to keep straight on what we're talking about, here.

Doubling the number of bits does give you twice the number of quantization steps, but those steps only give you additional dynamic range. The more quantization steps the less RPDF white noise (quantization noise) is added to your signal due to quantization error, providing a lower noise floor and thus higher dynamic range, or the relationship between the loudest level you can put into the system and the level of your quantization noise.

If your signal only has 6dB of dynamic range (6dB between the peak level and the noise floor from mics, mic pres, room noise, atmospheric noise, etc.) then how many bits do you need to accurately represent this?

Quote:
In a 16 Bits format signal, you get 96db (16 bits x 6 db). A 16 bits signal has 65536 digital positions to store the audio.
Correct. One is defined by the other. If you only have 65536 quantization steps then your quantization noise will be at an amplitude of -96dB (half the amplitude of the LSB).

Quote:
If you are mixing into a 16 Bits signal, and you don't reach the 0db clipping point, letīs say you mix only to -6 db, you are using only 90db, and only 15 Bits (90db/6db).
15 bits gives you 32768 positions to store your digital data. (HALF than 16 Bits)
Correct. And if my signal has a noise floor of -48dB peak (48dB of dynamic range) then any added quantization noise from the reduction in quantization steps is still well below the noise floor of my material and the signal does not undergo any degredation. It is still completely accurately captured.

Quote:
AS YOU CAN SEE, THE 6 "LAST" DB before clipping DOUBLES your data positions.
And the effect is that it lowers the quantization noise floor by 6dB.

Quote:
Then if you make your mix at -10db you are only using 20642 positions to store your digital signal (against 65536 that you would be using if you where using the complete dynamic range of a 16 bits digital signal).
Correct. But if my signal only had 48dB of dynamic range then what changes when I record at a lower level?

Quote:
Then, if you make a "Mastering" or "normalize" that -10db signal, then you still having 20642 positions, USED by your music, and the rest of positions, with "0" data. (20642 with MUSIC and 44849 with CEROS)
Right. But again, if the quantization noise is lower than the amplitude of the noise that I recorded into the signal from other sources then, when I normalize the following happens:

The peak is raised 10dB
The room/mic/preamp noise is raised 10dB
The quantization noise is raised 10dB.

If the quantization noise is far enough below the signal noise floor that it is inaudible then normalizing doesn't suddenly make it audible because the entire signal raises the same amount.

Quote:
This same principle is aplyed to 24bits or 32bits signals.
Right, and this shows why 32 bits is a completely silly way to record. When will the noisefloor of your signal ever be lower than the quantization noise in 32 bit recordings?

Fredo,

You're approaching this with very logical and intuitive assesments, but with incomplete information. It won't make sense until you assess the following:

Why is it that each bit gives me 6dB of additional dynamic range?

Nika.

Joined: Apr 2002
Posts: 68
C
Senior Member
Offline
Senior Member
C
Joined: Apr 2002
Posts: 68
Quote:
Correct. And if my signal has a noise floor of -48dB peak (48dB of dynamic range) then any added quantization noise from the reduction in quantization steps is still well below the noise floor of my material and the signal does not undergo any degredation. It is still completely accurately captured.
Well said Nika. This is the key to this thread that is still misunderstood by many. Recording at lower levels does NOT imply "less accurately captured." (Here again, dependent on dynamic range of the signal.)


Kevin Perry
Nashville, TN
Joined: Jan 2001
Posts: 6,074
MP Hall of Fame Member
Offline
MP Hall of Fame Member
Joined: Jan 2001
Posts: 6,074
I don't think he meant to be so aggressive in that statement. Truth is, most of these basic concepts are often covered in pro-gear manuals. If they aren't, usually that company will have some sort of published material that covers this information in layman's terms. If you want to get really deep, buy a reference.
But that doesn't mean you shouldn't ask...

Thanks. I must have posted the question incorrectly. Rather than giving the background of what I am using the gear for at the time of the recording, I should have simply asked if recording to my ADAT at less than optimum level would give me a lower fidelity recording. I cannot find any analysis of the waveform after digital transfer into my DAW that shows me the file is less than what it should be.

To the best of my knowledge there is not a manual or reference that specifically explains using the channel inserts to get a signal to an ADAT and how to set optimum levels for a large condenser mic in a live concert environment while sending optimum levels to the ADAT via channel inserts.

My main concern has always been to get the best live sound possible and then worry about the recording. Sometimes there is not enough time between acts to get the levels to the ADAT perfect and I have made do with lower than normal levels.

I have read every manual I own repeatedly cover to cover. I have read every article I can find, every book I can find and talked to every professional sound guy I can find.

After 32 years in the studio business I can still get confused from time to time.

Re-read all the posts on this topic and it appears there is a lot of confusion among you experts also. 'scuse me.


Mark G.
"A man may fail many times, but he isn't a failure until he begins to blame others" -- John Burroughs

"I consider ethics, as well as religion, as supplements to law in the government of man." -- Thomas Jefferson
Joined: Nov 2001
Posts: 168
F
Senior Member
Offline
Senior Member
F
Joined: Nov 2001
Posts: 168
Quote:
OK. Hold on, there. In digital audio any bit gives you 6dB of dynamic range. We have to keep straight on what we're talking about, here.
Nika,
Although I do not agree with your dynamic range statement, I highly appreciate your input here.

This is how I see it, so correct me where I go wrong.
A digital signal is represented (and calculated) on two axes.
The vertical Axe represents the Bit Depth or Word length.
The horizontal Axe represents the sample rate.
Mathematically you have a point in saying that the vertical line only affects the dynamic range.

Can I ask my question in a very stupid way?
The whole system of placing bits in "squares" is to connect these points in order to recreate the (smooth) curve as it should have been in an analog environment.
So, if the vertical bit-resolution is “devided” in 32 “slices” instead of “16”, the curve that will be produced from the 32-bit calculation will be much more accurate than the 16 bit.
Please correct me where I go wrong.


Fredo Gevaert
Temple Of Tune
fredo@temple.be
http://www.temple.be
Joined: Feb 2001
Posts: 3,324
V
MP Hall of Fame Member
Offline
MP Hall of Fame Member
V
Joined: Feb 2001
Posts: 3,324
Quote:
Originally posted by GZsound@hotmail.com:
[QB]

To the best of my knowledge there is not a manual or reference that specifically explains using the channel inserts to get a signal to an ADAT and how to set optimum levels for a large condenser mic in a live concert environment while sending optimum levels to the ADAT via channel inserts...

...I have read every manual I own repeatedly cover to cover. I have read every article I can find, every book I can find and talked to every professional sound guy I can find.
Mark,

Again, the solution is quite straight forward.

1. Determine what the level is coming from your direct outs at, or close to the onset of clipping

2. Determine what levels are required on your ADAT's for 0dbFS.

From what you have posted, it appears that both units are not ideally matched. This is either an impedance mismatch, or a level mismatch.

In the manual for each of those pieces of gear you use, it (they) should specify what those levels are, what the expected impedances are, and if it's balanced or unbalanced from the console.

If I were to guess, I would wager that the direct outs are simply lower than the ADAT's require. Have you measured the levels?

Also, have you ever tried using a mic pre and split the send to your console and ADAT simultaneously? Are your cables unbalanced and your connections require balanced wiring?

Quote:

After 32 years in the studio business I can still get confused from time to time.
That happens to the best of us.

Quote:

Re-read all the posts on this topic and it appears there is a lot of confusion among you experts also. 'scuse me.
Yeah, but it's important to know who the REAL experts are. No confusion among those dudes...

Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by Fredo:
Quote:
OK. Hold on, there. In digital audio any bit gives you 6dB of dynamic range. We have to keep straight on what we're talking about, here.
Nika,
Although I do not agree with your dynamic range statement, I highly appreciate your input here.
Glad I can help. This is clearly an area of a lot of confusion.

Quote:
This is how I see it, so correct me where I go wrong.
A digital signal is represented (and calculated) on two axes.
The vertical Axe represents the Bit Depth or Word length.
The horizontal Axe represents the sample rate.
Mathematically you have a point in saying that the vertical line only affects the dynamic range.
Right. A change in sample rate affects the highest frequency that can be recorded. A change in bit depth changes the level of the quantization error. Do you agree with that? That if I have more quantization steps I have less error in my sampling process?

OK. Now what does error give us in our signal? Distortion? Phase shift? Frequency modulation? What characteristic of a waveform (amplitude, frequency, phase, dynamic range) is affected when we have more error? The answer takes a bit of math and a visual analysis to easily "see", but it is noise. When you add quantization error you add noise. More error equals more noise. This means that the more bits the greater the dynamic range of your signal, and nothing more.

Quote:
Can I ask my question in a very stupid way?
The whole system of placing bits in "squares" is to connect these points in order to recreate the (smooth) curve as it should have been in an analog environment.
So, if the vertical bit-resolution is “devided” in 32 “slices” instead of “16”, the curve that will be produced from the 32-bit calculation will be much more accurate than the 16 bit.
Please correct me where I go wrong.
You go wrong in that you assume that the ear can hear all of that. The ear can only discern audio between peak and the noise floor*.

Think of "the jaggies" as described above. Let's say that the random noise jaggies have an amplitude of 100 quantization steps. This means that the noise level in the audio is whatever the amplitude of 100 quantization steps is. Another way to look at it: take a sine wave, and for every sampling point randomize it within 100 quantization steps. That is exactly the same result: 100 LSB amplitude noise on top of a sine wave.

OK, now add some quantization error of 1/2 of a quantization step to that. Audible or no?

The part of your equation that doesn't work is that you assume that the sine wave is perfect to start with and that adding quantization noise makes it unperfect. You're forgetting that there is random variation in any waveform called "noise". We just want the quantization error to be at less amplitude than the inherent noise in the waveform. That noise comes from, again, mics, preamps, room, atmosphere, etc.

In a math textbook we get to look at perfect sine waves, and in those situations we do have to be concerned about quantization error. The real world has noise.

Nika.

Joined: Aug 2001
Posts: 707
Gold Member
Offline
Gold Member
Joined: Aug 2001
Posts: 707
OK.

Quote:
Originally posted by Nika:
You are making good deductive analysis but without complete information.

What happens when the reverb tail goes into the noisefloor of your equipment. Do you need better quantization than the noisefloor of your equipment can support? The human ear is not capable of hearing very far below the noise floor (that's another thread in a biology and physiology forum). How far below the noise floor do you need to accurately capture if it's for the sake of the imperfect human ear?

Nika.
You've got an interesting perpective (probably more practical than many of us) on the science of audio reproduction. I know I downloaded a paper you wrote that spelled out your personal ideas based on the limits of human hearing (right?).

I totally agree with you in that, through pushing the limits of digital sampling and storage, we're wasting time due to the fact that we are manipulating objects beyond our natural senses.

I respect that point of view for it's practicality. I'm not really a practical person, though. I think we should always push the limits of everything just to see what we can do. Something is likely to be useful on some level or another - probably something not even related to audio.

Now, trying to present that data to a discussion like this one probably points out way too much common sense. I've learned in the Army that, if it makes sense, it's counter-productive.

Quote:
Originally posted by GZsound:
My main concern has always been to get the best live sound possible and then worry about the recording. Sometimes there is not enough time between acts to get the levels to the ADAT perfect and I have made do with lower than normal levels.

I have read every manual I own repeatedly cover to cover. I have read every article I can find, every book I can find and talked to every professional sound guy I can find.
My bad, dude. You totally caught me off guard. I completely understand your frustration. I've been there, done the twelve-step and everything. All you have to do is first admit that the ADATs are a problem. I understand ADATaholism. It's rough. But in the end - it's just wasting all your money on a funny VCR. All that video head cleaner causes brain damage, man. You need to jump on the bandwagon... ;\)


studioman adam
[][][][][][][][][][][][]
"uh........OK?"
Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
Quote:
Originally posted by soldierman adam:
You've got an interesting perpective (probably more practical than many of us) on the science of audio reproduction. I know I downloaded a paper you wrote that spelled out your personal ideas based on the limits of human hearing (right?).
Not that I recall, but maybe I wrote something?

Quote:
I totally agree with you in that, through pushing the limits of digital sampling and storage, we're wasting time due to the fact that we are manipulating objects beyond our natural senses.

I respect that point of view for it's practicality. I'm not really a practical person, though. I think we should always push the limits of everything just to see what we can do. Something is likely to be useful on some level or another - probably something not even related to audio.

Now, trying to present that data to a discussion like this one probably points out way too much common sense. I've learned in the Army that, if it makes sense, it's counter-productive.
I was under the impression that the person posting this was discussing "audio" specifically, inferring sound intended for human hearing, so the limits of human hearing enter into the equation in a major way. We have been discussing (or I hope we have been discussing) sound quality and sampling needs to record perfect sound quality. With that in mind my statements have been accurate. If we want to talk about other applications than what is required for human hearing then the equations certainly do change.

Nika.

Joined: Apr 2002
Posts: 68
C
Senior Member
Offline
Senior Member
C
Joined: Apr 2002
Posts: 68
Quote:
Right. A change in sample rate affects the highest frequency that can be recorded. A change in bit depth changes the level of the quantization error. Do you agree with that? That if I have more quantization steps I have less error in my sampling process?
Everyone having trouble grasping this should go here to see this point illustrated nicely.


Kevin Perry
Nashville, TN
Joined: Sep 2001
Posts: 2,151
K
MP Hall of Fame Member
Offline
MP Hall of Fame Member
K
Joined: Sep 2001
Posts: 2,151
I beleive that, if you ignore the noise floor/dynamic rnage issue (meaning that you're recording above the floor and the dynamic range of the source is less than that available between the noise floor and the peak level recorded) then 24 bit recording is still better than 16 bit.

This is cause there is better resolution (meaning a smaller db step between each bit) even though your are not at the upper end of the recording level.

For typically compressed rock music I beleive the conditions described above often ( -no not always) apply

So 24 bits recorded at a low level is still better than the same signal recorded at 16 bits even at a higher level.

I beleive this is the point that was misunderstood at the beginning of the thread.

Nika, Have I got this wrong?

BTW - great thread.
Nika - thanks for your patience.


Check out some tunes here:
http://www.garageband.com/artist/KenFava
Joined: Mar 2001
Posts: 2,938
N
MP Hall of Fame Member
Offline
MP Hall of Fame Member
N
Joined: Mar 2001
Posts: 2,938
[QUOTE]Originally posted by Kendrix:
I beleive that, if you ignore the noise floor/dynamic rnage issue (meaning that you're recording above the floor and the dynamic range of the source is less than that available between the noise floor and the peak level recorded) then 24 bit recording is still better than 16 bit.

No no no!!! Your post here implies that there is some sort of benefit besides dynamic range given by higher bit depths. There is not. None. Nada. Zero. Zilch. Ixnay. There is no benefit other than dynamic range from having more quantization steps. Sorry.

This is cause there is better resolution (meaning a smaller db step between each bit) even though your are not at the upper end of the recording level.

Go back and reread the thread. There are only four characteristics of any waveform: amplitude, frequency, phase, and the amplitude of the random variation with in, called noise - thus, dynamic range.

Now, you use the word "resolution." What characteristic of a waveform (of those four) does that additional "resolution" give you? Take an "unresolved" signal and a "highly resolved" signal. What is the difference between them? Make sure you phrase your answer in terms of the four charactistics of waveforms identified above. Then stop using the term "resolution" as it is very indescriptive and does not speak to whatever type of "distortion" you claim the signal is undergoing. The word "resolution" is a red herring. If you use that word you'll be thrown all kinds of logically "intuitive" but incomplete answers. Describe what actually happens to the signal as opposed to what the cause of it is.

So 24 bits recorded at a low level is still better than the same signal recorded at 16 bits even at a higher level.

Nurp!

I beleive this is the point that was misunderstood at the beginning of the thread.

Still is. \:\)

Nika, Have I got this wrong?

Yurp!


Nika - thanks for your patience.


It is pretty evident that this particular issue is one of the greatest myths of the audio industry. I'm happy to help when I can.

Nika.

Page 2 of 11 1 2 3 4 10 11

Moderated by  gm 

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.5