Is there room for 128 bit 2 channel sound?

Greg Voevodsky

Greg Voevodsky

Active member
(LET ME CORRECT MYSELF - I meant 32bit / 128hz on 2 channels as kindly pointed out below.)

I was wondering if there is any chance that RED could offer recording in 128 bit - 2 channel. I am planning on using Holophones new H4 Supermini surround mic that records 5.1 channels to 2 channels. - http://holophone.com/products_h4.html

Is there any chance or room left to record at 128 bit on 2 channels rather than just 16 bit? This would be great for my ambient nature dvds. Otherwise, Im looking into a seperate recording system.
 
Last edited:
Greg, I think maybe you have your units mixed up? If not, you are asking for something that is WAY overspecced for sound! :)

For high end sound I would definitely go for a separate solution (sound is my main field besides directing so I am familiar with it). Could I suggest something like the Sounddevices 722 (2 channel) or 744T (4 channel)?

http://www.sounddevices.com

Both feature up to 24bit 192kHz uncompressed recording with some of the best DSPs I've seen at that price range. Team it up with a Rode NTG-2 directional mic (the one I use and prefer) and you have a formidable sound recording solution for not a lot of money.

The other thing is that most surround sound begins life as multiple mono tracks, mixed correctly to give a sense of space. If you want 5.1 wilderness sounds for example, record five lots of sound and mix them into each channel with appropriate fades etc.

The mic you link to is interesting but for really pro sound I would go the separate route.

HTH
 
Last edited:
Greg Voevodsky said:
Is there any chance or room left to record at 128 bit on 2 channels rather than just 16 bit? This would be great for my ambient nature dvds. Otherwise, Im looking into a seperate recording system.
Click to expand...

There is no such thing. The highest quality of digital audio recording to date is 32bit/192kHz.
most professional digital audio recording/editing/mastering software solutions offer a 32bit floating point for processing, but only up to 32bit/kHz (and in most cases only up to 24bit/kHz).

128bit audio recording simply does not exist, at least not in the professional recording world. I guess you got confused...
 
Yes, I am a bit confused - that's why Im asking these questions. thank you - So Any chance for 32bit/192khz? Also, for the sound pros out there, can you really hear the difference in 'ambience' like at the beach which is mainly white noise anyway or is the camera standard enough?
 
[SIZE=+0]Quantizing audio (or any real world quantity) at 32 bits is impossible. 2^32 has a dynamic range that can't be achieved. Quantizing at 24 bits is slightly beyond the limits of any real world sampling device when done at sample rates typically used in digital audio (48 kS/s and above). There are quantizers that are 24 bit accurate, but these operate at very low sampling rates (typically under a few hundred samples per second) and are used in instrumentation not for digital audio.[/SIZE]

[SIZE=+0]Now, if you are going to process a 24 bit sample, using a longer word length is an appropriate choice, but not at the quantizer step. Audio sampled with a 24 bit quantizer is essentially "perfect" for any practical application.[/SIZE]
 
Greg Voevodsky said:
Yes, I am a bit confused - that's why Im asking these questions. thank you - So Any chance for 32bit/192khz? Also, for the sound pros out there, can you really hear the difference in 'ambience' like at the beach which is mainly white noise anyway or is the camera standard enough?
Click to expand...
Short answer in your scenario - Yes, camera standard is enough.

Long answer - As per khmuse's comment, 24bit already exceeds the range of human hearing. To explain a bit better, 24bit audio recording gives you a much higher signal to noise ratio, allowing for better distinction between the highs and lows (to put it in layman's terms). In more technical terms, the dynamic range of 16 bit audio is around 96dB, but 24bit audio is around 144dB.

To give an example, if absolute silence is 0dB, and a mouse squeaked at say 6dB while at the same time a jet roared overhead generating 120dB (above pain threshold!), a 16bit recording leveled at 0dB would clip the jet noise quite substantially but still capture the squeak, OR you could level the recorder to capture the 120dB jet without clipping but because of the 96dB range the recorder would treat the mouse squeak as being below the threshold of zero and not record it at all (extreme and unrealistic example, but anyway!). Now if you recorded same scenario using 24bit, the dynamic range of 144dB means that not only would the jet not clip at all, but the recorder would also capture the squeak of the mouse in a way that could be measured or extracted (I say this because human hearing doesn't have the dynamic range of 24bit to discern both at the same time). If the recorded frequency patterns of the jet and mouse were distinct (most likely) then you could effectively extract just the mouse squeak from that recording.

Hopefully that makes it a bit clearer and helps you understand just what the bitrate means in audio recording :)

One nice advantage of recording in 24bit is that if you stay in that realm right through post production and only down-convert right at the end to say 16bit/48kHz (DVD audio), ALL the introduced noise due to imperfections in filters, algorithms etc will still be lower than the lowest threshold in a 16bit file, so you effectively introduce NO unwanted artifacts into your final mix.

Keep in mind that the newer formats (HD-DVD, BluRay etc) all have the capacity to use 24bit/96kHz audio files. Personally 24/96 is my recording format of choice (uncompressed WAV) as it will give you audio that basically exceeds the threshold of human hearing. 192kHz is excessive to my mind, though you never know when audio might for some reason transition to that!
 
Last edited:
To apply your mouse/jet example to the real world though 144db is only the theoritical dynamic range of 24bits. By the time you you go through all the circuitry and noise the s/n of the actual equipment will come in -30db or more then it's theoritical dynamic range (ie the 744t's dynamic range is actual 110db and most 16bit prosumer cameras have a dynamic range around 60-70db). And even that would be with optimum recording, processing, and broadcast after the fact.

For sample rate anything over 96kHz is going to be a waste of space and processing power. 192kHz only really has a potential advantage in sound f/x recordings.

Cheers
 
wabbit said:
To apply your mouse/jet example to the real world though 144db is only the theoritical dynamic range of bits. By the time you you go through all the circuitry and noise the s/n of the actual equipment will come in -30db or more then it's theoritical dynamic range (ie the 744t's dynamic range is actual DB and most bit presumer cameras have a dynamic range around 60-DB). And even that would be with optimum recording, processing, and broadcast after the fact.

For sample rate anything over kHz is going to be a waste of space and processing power. kHz only really has a potential advantage in sound f/x recordings.

Cheers
Click to expand...

The signal to noise ratio is effected by both the self noise (noise floor) of the system and the amplitude of the signal. There is a concept used to simplify the understanding, specifically it is to express the product as the equivalent input noise (consider a known amplitude noise source summed at the input of a theoretically perfect gain block to get an idea of the concept).

Using audio spectra as a criteria, extremely high performance A>D converters can exhibit noise floors (power spectrum) of about -120 dB, these devices are able to resolve a monotonic signal presented to their input which is about 2^23 below their FS (full scale) amplitude.

Higher sample rates have the obvious advantage of increased bandwidth (slight less than 'sample rate'/2) and also show better time domain characteristics at even lower frequency. Testing has shown that a qualitative improvement can be repeatable observed for higher sample rates (though with quickly diminishing returns) for sampled systems even when the delivery format is reduced from the acquisition sample rate.

I would agree that 96 kS/s is likely sufficient for nearly any application other than perhaps extremely high performance music reproduction. Frankly, if data rate is an issue, I would trade off sample rate for increased word lengths, but only up to 24 bits for the reasons previously stated.
 
Surround Audio

Surround Audio

We are always open to innovative ideas, so if anyone can summarize what the audio data requested to be recorded actually is, and how it could be physically presented to the camera, we'll take it under consideration -

- but no promises ... O.K?
 
My summary for audio recording would be this (assuming 2 XLR inputs which looks realistic given your rendered images of the camera):

Each mono XLR input:

- 48V phantom power
- selectable 16bit/48kHz, 24bit/96kHz
- uncompressed WAV file (same filename as video take but .WAV extension, same or separate folder structure)
- adjustable input level.

The other half of this would be having a good DSP in place to get maximum audio quality from the source signal when coverting analogue to digital. I wouldn't expect audiophile levels of quality as that is not the primary function of the camera, but something able to accurately record the 24/96 option should be pretty effective.

HTH
 
Last edited:
Bit rate of audio

Bit rate of audio

If you have a well adjusted input gain, recording at 16 bit / 48 KHz sounds to all but the most "Golden Ears" the same as recording at 24 bit / 48 KHz.

But an ideal input gain is never the case, especially recording outdoors. So you record at a lower level (because you dont want to risk overdriving the AD´s) - thereby giving away bit resolution.

Or you record high gain and trust the automatic input limiter (mostly awfull when they are actually doing something).

The actual loss of bit resolution when recording at 16 bit can be quite dramatic for low signals. Down to a point where you have to deal with a 12 bit audio that you then have to boost at the dubbing stage (not nice).

A 24 bit input signal can be still quite ok even when the location sound guy is asleep at the boom. (Not regarding other influences like poor mic preamps, ambient noise, etc.)

So, 24bit/48KHz will be quite good for most purposes with a possible 4 channel recording being a great spec. (With phantom powered inputs - but of course you know that)

Having a *possibility* of 2-channel 24bit/96KHz for concerts etc. will be quite sensational. Although those events usually have multitrack recording setups budgeted in that a camera cannot compete with.

32bit floating point is of course the ideal, but it is used only inside some mixing desks or software (Euphonix 5, Logic Pro). It would be quite an overkill to do that. I dont know if there even are AD converters with that spec. And the floating point is good for processing the audio - there´s very little audio-processing going on inside the cxamera.

What really would be nice would be a usable input-limiter / compressor for the audio with decent user control. At an interview I want a different audiolevel compressor than when I shoot racecars (where I basically just need a brickwall limiter - and earplugs).

Just an example: Today I can only set the automatic on or off. So with no racecar present the auto sets a high gain that gets dialed down when a car roars up to you (mashing the crescendo in the process). After it passed the gain setting stays down for a bit, then goes up again - mashing the decrescendo.

A decent automatic control would allow me to set a treshhold where I want the auto to only work at the absolute loudest peak of the sound.

At an interview the person might speak very low but there may be strong passages. I now want the automatic to boost the low parts AND very subtle compress the loud parts.......

Ok, that ´ll cut into the expertise of the sound recordist - but I think we´ll see a lot of small / single crew setups with the RED.

Jochen
--------------------------
www.jochenschmidt.de
 
Last edited:
If high sound quality is what you want, then I think it would make the most sense to record sound seperately. With a laptop + audio interface, you can get very high quality sound for low cost.

For filmmaking use, you likely either want:
-Record sound seperately, in which case audio capability doesn't matter.
-Record a lot of digital inputs (ADAT carries 8?)... which has the slight advantage of saving you time. However, you'd have to be tethered to the camera, which won't work in all situations. You might as well do double system anyways.

For ENG, low-budget work you likely don't need extremely high quality audio. Well-implemented 16bit/48khz audio should suffice- good dynamic range, low aliasing, a compressor/limiter with soft knee (not auto-gain control). In typical recording environments, there'd be lots of ambient+background noise so high S/N ratio may not be necessary.

If you want better sound, it shouldn't be that expensive to go double system.

2- Limiter: A soft knee on the limiter would make it better sounding than the typical limiters where the transition point is hard (i.e. in some computer filters). This is probably nice in run and gun situations (i.e. ENG, documentary) where there may be unexpected variations in audio. You can get 10+dB more dynamic range with minimal distortion (to my ears anyways; 20dB is more like it).
As a point of reference, the (very expensive) Sonosax mixer has a soft knee limiter and it sounds very nice + transparent at low amounts.
There are many low-cost compressors available which also sound nice (i.e. RNC).
 
I guess the main advantage to recording on board the camera is sync if both begin recording at the same time.

Being computer file based, it is extremely easy to have the audio file keep the same name as the video file with only extension differing, then in post you simply line up the start of both on a timeline and off you go, ready synced footage with no effort at all :)
 
You must log in or register to reply here.
Back
Top