24bit/96khz & 192khz sorry for audio noob question

[amrrahmy]

how much of a difference is there between one channel 24bit/96khz and 192khz

is studio quality 24bit/96khz considered a good quality for a film?

[Andrew M.]

Well, it depends what is your final product.
If it will end up on Blu-ray in 5.1 surround then go for 96 or even 192.
It mean, you can bring audio on 48 but for manipulation in post use 96 minimum.

Like with the video, Mysterium is 12 bit but color manipulation, we do in 32 bit if possible.

[amrrahmy]

96 minimum, meaning that there would be a clear difference in quality between 96 192.
is that difference also noticeable for on set recording(if on set was 96 and the rest of the sountrack and effects were higher quality)?

[Philippe Vandendriessche]

We mostly record 24 bits 48 kHz during shooting.
Good enough for dialog...

Music recording / sound effects: 24/96 is better.

I have never seen any project recorded in 192. I think the human ears cannot hear any difference between 192 and 96 recordings.
Between 16 & 24 bits there is a big difference.

[amrrahmy]

thanks dude.

[Andrew M.]

We mostly record 24 bits 48 kHz during shooting.
Good enough for dialog...

Music recording / sound effects: 24/96 is better.

I have never seen any project recorded in 192. I think the human ears cannot hear any difference between 192 and 96 recordings.
Between 16 & 24 bits there is a big difference.

Human ear maybe can't see the difference between 96 and 192 in terms of frequency response but in terms of phase of the signal coming to your ear we can detect 20 deg difference in the phase.

Just make simple experiment.
Get the software for surround sound editing and go to the sound placement editor.
Human can detect the direction of the sound based on the main 3 factors and number of secondary factors.
First is ..............no no loudness of the signal it is phase of the signal
Second, loudness
Third, frequency response of our ear lobe.

Now change the placement of the 2000Hz sound just by altering the phase by 20 deg.
Now switch it on and off, can you hear the difference?

We can hear 2000 Hz in the music very well. To reproduce the phase of this frequency you have to sample it say 3 times per each degree of the phase you want to detect.
Well 2000 x 3 x (360/20) = 108,000 (108kHz) Now what if you want to do it with 4000 Hz sound?

So even for the interview on the camera when I want to reproduce in just two channels the position of the speaker and then if I do zoom out on his face and zoom out to show the whole street where interview takes place I use phase shifting. Viewer expect sound to focus on the speaking person when you zoom in and to hear more of surround sounds when you zoom out.

[Kreisky]

48.000khz is broadcast standard !

I have a question to the sound guys here in forum:"
I think most tracks are mixed out at 44.000khz in a music studio,
is this for CD produktion ?
or do you mix them for music videos out on 48 ?"

thnx
bruno

[Kevin Halverson]

The value that higher sample rates (above 48 kS/s) provide are much less about the ultrasonic content than they are about improving the capabilities of the anti-aliasing filtration mechanism. Given that to satisfy the Nyquist requirement the amplitude of the inbound signal has to be at or below the noise floor by .5 F, this requires a fairly aggressive LP filter when sampling at single speed data rates (44k1 or 48K). Consider that the same order of LP filter can be used with higher sample rates (2x or 4x) and achieve much improved image rejection while still providing adequate pass band performance.

The "phase" argument is not accurate as it misses the very nature of a sampled system. If one modulates the amplitude of a monotonic stimulus "...switch it on and off...", the resultant output has higher order components. The slope of the amplitude modulation is just as bound to the Nyquist requirements as is the fundamental energy component itself. A very simple illustration of this is RF amplitude modulation (AM). Obviously the carrier is monotonic but as soon as modulation is introduced sidebands appear. These sidebands exhibit increased bandwidthrequirements; the exact same effect occurs in the "...switch it on and off..." situation. On and off imply square wave amplitude modulation and this modulation profile requires a much greater bandwidth system to capture the modulation. In fact, to completely capture a perfect square wave modulation (or fundamental) the sample rate would have to approach infinite.

"Phase" is a term that is often misused and in fact, there is no actual property called "phase" rather it is the relationship of a signal or signals to another. An example of its misuse is "invert the phase" when what should have been said was to invert the absolute polarity.

Directional localization in human hearing relies upon several different systems; including arrival time differences (this is why localization of impulse sound events is more pronounced) relative amplitude differences (but only at midband and higher frequencies) and by moving the receptors (changing head position which is often done without a conscious decision on the part of the listener) and listening for the delta in the arrival events. Human hearing is also sensitive to absolute polarity (at least some humans are) but the original signal must be both recognizable and have significant asymmetry for this to be much of a contributor.

The argument for greater bit depth is certainly easy to make as this is simple resolution improvement, the argument for higher sampling rates is more complex for the reasons that I have sighted and many others.

In my work, we often use extremely high sample rates (8x or 384 kS/s) for the most important recordings, but we rarely deliver this all the way through to the end product (often DVD-A recordings). Rather we use the very high sample rates to ensure that we have nearly perfect passband performance while still satisfying the stop band needs for the bit depth that we are utilizing.

So, is 48 kS/s enough? It really depends upon the event being recorded, but given the types of microphones used in sound for film field recording, it would be more difficult to justify higher sample rates than for situations where transducers that have more extended bandwidth are employeed. Greater bit depth (24b) is easily justified no mater the sample rate.

Kevin Halverson

[amrrahmy]

Human ear maybe can't see the difference between 96 and 192 in terms of frequency response but in terms of phase of the signal coming to your ear we can detect 20 deg difference in the phase.

Just make simple experiment.
Get the software for surround sound editing and go to the sound placement editor.
Human can detect the direction of the sound based on the main 3 factors and number of secondary factors.
First is ..............no no loudness of the signal it is phase of the signal
Second, loudness
Third, frequency response of our ear lobe.

Now change the placement of the 2000Hz sound just by altering the phase by 20 deg.
Now switch it on and off, can you hear the difference?

We can hear 2000 Hz in the music very well. To reproduce the phase of this frequency you have to sample it say 3 times per each degree of the phase you want to detect.
Well 2000 x 3 x (360/20) = 108,000 (108kHz) Now what if you want to do it with 4000 Hz sound?

So even for the interview on the camera when I want to reproduce in just two channels the position of the speaker and then if I do zoom out on his face and zoom out to show the whole street where interview takes place I use phase shifting. Viewer expect sound to focus on the speaking person when you zoom in and to hear more of surround sounds when you zoom out.

i think that goes to a point and stops.
u hear a diff sound because u changed the original frequency,(i think) the lower u go the more it gets pitched but when u reach the original freq the sound is the same, there is no change in quality.
what i was wondering was, if i recorded both 96 and 192(i wanted to know if i can record with a hardware that does 24bit/96khz). would i find a diff quality, or would it be the same?

not if i converted a 192 file to a 96khz, of course there would be loss from the original file and i would notice the diff, but that's due to the software and how the files were recorded and not the max ability of the 96khz(i think)

i wanted to record (24bit/96 if possible with 3 "duet" by apogee with a macbook pro)

[Kevin Halverson]

i think that goes to a point and stops.
u hear a diff sound because u changed the original frequency,(i think) the lower u go the more it gets pitched but when u reach the original freq the sound is the same, there is no change in quality.
what i was wondering was, if i recorded both 96 and 192(i wanted to know if i can record with a hardware that does 24bit/96khz). would i find a diff quality, or would it be the same?

not if i converted a 192 file to a 96khz, of course there would be loss from the original file and i would notice the diff, but that's due to the software and how the files were recorded and not the max ability of the 96khz(i think)

i wanted to record (24bit/96 if possible with 3 "duet" by apogee with a macbook pro)

Your example illustrates my previous point. If you record something at 192 kS/s and then later properly down sample it to 96 kS/s (or lower) you will retain much of the perceivable benefit of the higher recording sample rate. This is mostly due to the improvement of the performance of the anti aliasing filtration mechanism and not due to the increased ultrasonic content of the original higher sample rate (since it will not survive the down sample process).

By the way, the correct way of expressing sample rate is not to use frequency units (Hz) but to use samples per unit time (S/s or Samples per second). Instead of 96 kHz, the correct term is 96 kS/s. This is an industry wide misuse of a term that has persisted for years (with a few notable exceptions).

Kevin Halverson

[amrrahmy]

sorry i was writing when u posted, so i missed it. i only saw it after.

[amrrahmy]

384 kS/s that hardware is not something portable i guess(right?), u cant record outside the studio with something like that?, can u even use that on a closed set in a shoot?

[Andrew M.]

The "phase" argument is not accurate as it misses the very nature of a sampled system. If one modulates the amplitude of a monotonic stimulus "...switch it on and off...", the resultant output has higher order components.


Phase term I used as a time of arrival of sound to our ears.
It is the most directional information used by our brain, not the amplitude.
You do not need to switch it ON or OFF, just move the phase knob by 20 deg and you will see if 48kHz system will take it. Then try the same on 192.

[Kevin Halverson]

Phase term I used as a time of arrival of sound to our ears.
It is the most directional information used by our brain, not the amplitude.
You do not need to switch it ON or OFF, just move the phase knob by 20 deg and you will see if 48kHz system will take it. Then try the same on 192.

It is the relative arrival time, not PHASE, that human hearing uses as a localization clue. Amplitude is but one of the inputs used to localize a sound source, delta T is another. The dominance of the importance is dependent upon the spectral characteristics of the signal (in essence its frequency components) and is dominated by the shape of the human head.

Your use of the term "phase" is yet another commonly misused one in audio. When one rotates a "phase knob" what is actually occurring is the changing of the delta T parameter of an all pass filter (in essence delaying a signal by a given number of samples in a digital system or by an R/C time constant in an analog one).

"Phase" or angular displacement of a signal is only applicable when it is compared to another signal! When you are noticing the localization change when you rotate a "phase knob" this is because some other signal (perhaps the other channel) is unchanged, hence there is an arrival time difference between the two channels. Don't become confused by the misused terms in most DAWs (and having designed a lot of the software and hardware that is used by audio professionals, I know how easily these terms get misused).

Regardless of it being mislabeled "phase" in your DAW, it is the arrival time delta that provides the audible effect. You can easily prove this to yourself by routing a signal through a "phase control" and listening ONLY to this signal (no other path). Other than the latency, there is no other change to the character of the signal.

Kevin Halverson

[Kevin Halverson]

384 kS/s that hardware is not something portable i guess(right?), u cant record outside the studio with something like that?, can u even use that on a closed set in a shoot?

Currently, products based upon Si that is capable of 384 kS/s data rates are confined to non portable products (and there are but a very few of these currently available). There will be some portable ones introduced early next year (2009) assuming that the development continues on its current pace.

Again, the advantage of these types of sample rates is the improved image rejection made possible by the improvement to the stop band characteristics of the LP filters that proceed the ADC. Currently, there are no delivery vehicles for this type of data rate and in the standards committees that I work with, none are contemplated at this time.

Kevin Halverson

[amrrahmy]

dude i'll have some problems just recording 5.1ch audio with a macbook pro, and i'm not sure if i'll also need a san or raid system to go with the macbook pro. 384 is not something for film in the near future, it's just too much for todays hardware to handle if it would even make a difference in the quality which is kindof doubtful

[Kevin Halverson]

dude i'll have some problems just recording 5.1ch audio with a macbook pro, and i'm not sure if i'll also need a san or raid system to go with the macbook pro. 384 is not something for film in the near future, it's just too much for todays hardware to handle if it would even make a difference in the quality which is kindof doubtful

The great thing is that you don't have to record (store) high data rate material (96 kS/s, 192 kS/s, 384 kS/s) if you do a proper down sample first!

This is the wonderful thing about the concept, if the ADC runs at high data rates and the anti aliasing filter is designed to assure that the stop band performance meets the requirement of the bit depth you are recording, then all that is necessary is to pass the samples through to a highly accurate low pass filter (either hardware or software DSP), you can store (record) the data at much lower rates and still retail nearly all the benefit. I can assure you that many listeners can easily perceive the differences that this type of process makes (and I am not just talking about the "golden ear" types).

Don't worry about today's hardware not being capable of accomplishing this, it actually already is capable and this type of recording systems are coming from a number of manufacturers.

Kevin Halverson

[amrrahmy]

not something that u can attach to a camera.
and i think what u say would have some delay, not something that u can notice easily, however there would be some 0.0000etc ms delay(right?).

[amrrahmy]

sorry, "on" a camera

[Kevin Halverson]

not something that u can attach to a camera.
and i think what u say would have some delay, not something that u can notice easily, however there would be some 0.0000etc ms delay(right?).


Actually, some of these products are fairly small, the size of, or slightly smaller than many of today's field recorders. Though not intended to be mounted on a camera, they are certainly well within the size constraints of the intended application.

As for the latency of the process, its completely academic as it's in the recording signal chain. No matter the delay (up to a reasonable point) it's meaningless. Consider how long the delay is between the time something is recorded and when it's played back? This is not an issue at all.

Kevin Halverson

[amrrahmy]

but your talking built up delay, the more is recorded the more delay u get.
in a 10min scene u would get trouble by the end.

[Kevin Halverson]

but your talking built up delay, the more is recorded the more delay u get.
in a 10min scene u would get trouble by the end.

You are confusing delay with sample rate drift. These are not the same things at all. The delay is simply latency, it is constant and is therefore irrelevant (it does not accumulate as you have suggested).

What matters is maintaining the sample rate of any external audio recorder (analog or digital) to the sample rate of the video system (or film system in those cases) and the edge or time code meta data. This can be accomplished by either jam syncing the camera and field recorder (locks the meta data) or by a better approach of using a word clock signal (locks the two by sample frames). Word clock systems can be thought of as a similar system to the old pilot tone system used in the past.

Regardless of the method, as long as both sample clocks are accurate, there is nearly no drift (or effectively minimal drift with the occasional resynchronization) when jam synced. When locked via a word clock, the two recording systems (camera video and separate audio) remain perfectly synced no matter what the duration of the recording event. Which a word clock could be 10 minutes, 10 hours or 10,000 years, and the two systems they always remain synchronized.

The presence of the DSP in the signal chain of the recording system that I am describing has no effect on this relationship between the audio and video/film recording systems.

Kevin Halverson

[amrrahmy]

the constant process in the middle after recording the sound and before the final sound is recorded has a constant continuous delay, no matter how low it may be and trying to resync in sound pauses (no sound and low data in room sound recording) is not good enough and would mean that u have ur own cpu running to resync, that would be way too eeexxxxppppenssssssssive and not better than realtime 96k(i think)

[Kevin Halverson]

the constant process in the middle after recording the sound and before the final sound is recorded has a constant continuous delay, no matter how low it may be and trying to resync in sound pauses (no sound and low data in room sound recording) is not good enough and would mean that u have ur own cpu running to resync, that would be way too eeexxxxppppenssssssssive and not better than realtime 96k(i think)


Nope, completely off the mark. The time through the LP filter (number of clock cycles) is always constant regardless of the sample amplitude. It's simply a set of taps - multipliers and accumulators, it is always the same and is based upon the requirements of the filtration, not based upon the data in any way.

There is already a lot of processing "...in the middle after..." including DSP functions, such as level changes, channel assignment, etc. The additional processing burden of a hardware DSP on the CPU is zero. This LP filter is in the hardware chain after the ADC, not within a post processing CPU. I can assure you that the costs are trivial (in terms of hardware DSP) as these products are already in development (and in some cases, within the ADC Si itself).

Kevin Halverson

[amrrahmy]

There is already a lot of processingno there isn't.
level changes can be manual as well(and more accurate), so no processing is no biggy and u can do a 6ch through 3 duel on a macbook pro(or at least, it's my dream if enough money came from the sky and landed on my lap.(that would be nice))

[amrrahmy]

also if the epic came with the same quality of nikon d3, that would be sweet.

[Kevin Halverson]

Yes, there is. I have designed a number of ADCs over the years and other than single cycle flash (not appropriate for most audio applications), all have serious hardware processing engines (often in Si).

Nealy every ADC that is employed in digital audio applications have decimation filters (unless we are talking about single bit processes) so there is already a powerful DSP (even if it isn't called a DSP) in the signal path and it is already doing a lot of processing on every sample. I can show you the die area plots that are dedicated to various aspects of the Si (chip) and this will easily confirm the considerable processing that is occurring. Further, dedicated hardware filters are much better at doing this type of processing than some CPU (not unlike why we design visual processing engines with dedicated GPUs).

I think you are still confusing the host computer ("macbook pro" in your example) with the ADC hardware. These are different pieces in the digital audio chain.

As for "manual" level changes, are you referring to scaling the analog signal prior to the quantizing step? If so, this is certainly outside of the processing chain as it is purely and analog function.

Kevin Halverson

[amrrahmy]

"There is already a lot of processing", i'm not saying there is no processing, i'm saying the processing can be minimal enough that one laptop can handle 6ch of audio and be a monitor at the same time. also check the signals and the waves, see any imperfections, any distortions before and during the recording, see and limit room noise(i think).

[amrrahmy]

seeing the sound on a monitor is very important(to me at least) not the level the actual wave.

[Kevin Halverson]

I guess I still am not getting the point across.

If the ADC has an additional and accurate DSP to provide the down sample, the computer will never know that this occurred upstream of it. All the displayed information (waveforms) will be identical to an ADC which lacks this feature. I am fully aware of the necessity for a good UI (user interface), this is simply a way of getting better recordings, not changing the application level software.

The computer doesn't have to do any (not even one clock cycle) more with the data being thrown at it. This is an upstream process. Your DAW application will behave EXACTLY the same.

Kevin Halverson

[amrrahmy]

that's exactly what i'm talking about.
u have a full painful hard on equipment process of down sampling that has some fishy delays and some processing with clocks and time comparisons made to resync happening(that would kill any processor and also requires some kind of super rams) before the difficult process that needs a powerful pc and a fast hd, equipment for the down sampling has processing in it that requires at least a couple of ships if not a pc.

[amrrahmy]

that is expensive
and not 6ch 1050$ expensive

u r talking specially made ship expensive

[amrrahmy]

u 'll need like a ps3(or 2x) wired into process to make it happen realtime without delays.
with rewiring all the gpu power to processing

[Kevin Halverson]

Nope, this is a real time process and it isn't "fishy" at all. There is no "resync" as it is naturally always in sync, the memory burden is fully within the X & Y memory space of the ADC's DSP section (just a few k), so no external memory is required. The number of taps is not excessive and the number of clock cycles (all single cycle MAC operations, multiply/accumulate, not a reference to Apple at all) is remarkably few given the parallel nature of the processor.

Best of all it isn't out of the price range of commercial products.

Kevin Halverson

[amrrahmy]

what's the price range?and how many channels?

[amrrahmy]

and aprx dimensions?

[Andrew M.]

I started to use Audition 3 for surround editing and once I installed ASIO compatible sound card in my computer I can do a lot of work real time.

http://en.wikipedia.org/wiki/Audio_Stream_Input/Output

Can you use ASIO compatible device for high end work, or it is just for standard audio editing?

Surround editing in Audition.

http://www.reduser.net/forum/uploaded/805_1211760905.jpg

[amrrahmy]

software seems a bit week compared to logic
and editing would be pain

[Kevin Halverson]

Audition CS3 really is a wonderful program, I use it for a lot of standard projects (have been using it long before Adobe bought up the application from Syntrillium) and really like the integration between it and the other Adobe products in the suite.

For high end work we don't really use a DAW at all. It's all command line driven tools running under Irix on a SGI workstation. No GUI, just text, but it runs fast and we know exactly what is being done with the data.

When it comes to the SoundField work in 6.0 (3D surround including a height components) there are only a few tools out there, but I prefer to just process the data as individual fundamental signals (W,X,Y & Z) for editing purposes (treating it as a 4 channel project) and then post process the fundamentals into the delivery channel count.

For more conventional footprints (like 5.x) the commercial tools (like Audition) are more than adequate for the task.

Kevin Halverson

[Kevin Halverson]

what's the price range?and how many channels?

The die area is not information that I can disclose, but I can say that a 2 channel ADC fits in an 64 pin QFP. Price for the Si will be well under $80 per channel pair in manufacturing quantities.

Kevin Halverson

[amrrahmy]

but u r not seeing the harmony nor your playing very well with moving sounds. ur missing the surround point, it's not in the channels, it's in the connection between them, the space between them.it's what happens in conjunction between the channels in the space away from the channels.(i think)

[Kevin Halverson]

amrrahmy,

I have no idea what point you are trying to convey here, so perhaps you might want to restate it.

As for SoundField recordings, they uniquely capture all the information that arrives at a single point in space from all three dimensions along with the pressure (W) information. They, in my experience, convey the best representation of the natural acoustic space of any recording style (when played back with a system that can recreate 3D sound space).

Kevin Halverson

[amrrahmy]

that's the reason why u don't need more than 4 channels.(any professional would only use 4 ch to make 5.1). it's ridiculous to put an odd channel between four channels. what were they thinking?. if u want to add more channels u need to add more channels all around the space.it needs to be 8.1 now

[amrrahmy]

sound editing include importing external audio tracks that might not have been recorded on set, sound effects and other stuff like music or whatever needs to be made surround, u don't have to limit yourself with only the recorded on set to be surround, that's when a gui would come in handy

[amrrahmy]

music, i dont mean exactly soundtrack, well more like in movie music that isnt on set

[amrrahmy]

or that trans into soundtrack

[amrrahmy]

that's why it's called surround editing, other wise it would be normal sound editing.

[amrrahmy]

and 3d is diff than surround

[Andrew M.]

Is there any good GUI based 3D editing tools for PC that you would recommend?
Audition has one major problem, it doesn't let me process phase(delay) for each Track when I do monitor the final Master or at least the bus A or B. Or I didn't figure it out how to do it?

I have to decide up front the track content, mix it and compose it in to the single track.
Once I create one single track from such mix I can edit such track again.
No multi track edit in Audition, only multichannel, one, two, or 5, that is all.
Even Premiere allows me to do multi-track edit (limited functions though) but not Audition.

What I need is to be able to adjust each track position in the space when listening to the final (master or bus) output.

[Kevin Halverson]

Hello Andrew,

I don't have a suggestion for another PC DAW that I really like and certainly none that is going to integrate as well into an Adobe workflow. Now that I understand what you are trying to accomplish, let me see if I can come up with a workaround that will accomplish the task.

Might be a day or two, but I will let you know what I learn.

Kevin Halverson

[Andrew M.]

Thanks! Kevin highly appreciated.

Also what are you using for the height vertical placement/adjustment of sound.

[Kevin Halverson]

Hello Andrew,

It's really nothing more than a bit of calculus to obtain the coefficients for the primary SoundField signals to the delivery channel configuration. We do A to B format conversion in hardware and delivery channel configuration formating in software. Since we have a specified location for the playback channels, there is no guessing necessary, just apply the math and the results are nearly perfect every time.

Kevin Halverson

[Andrew M.]

Thanks!

What do you think about Wavelab-Lab-6.1
it supports 24-bit 32 at up to 384 kHz


http://www.steinberg.net/128_1.html

http://making-music.blogspot.com/2007/11/steinberg-wavelab-wavelab-studio-610.html