16 bit or 24 bit?

HerzogisGod

HerzogisGod

Well-known member
obviously 24 bit. But for some reason my editing software (new tek, speed edit) will not recognize 24 bit recordings so I am forced (for now) to record 16 bit.

I use a Fostex fr2le to record sounds. This recorder can capture at very high bit rates.

Is 16 bit acceptable for indie film sound work; FX, voice overs, ADR?

I use an HVX200, which I believe records 16 bit audio. This will at least make all my additional field recording audio on an equal playing field as my camera's audio.

Am I okay recording 16 bit, or should I wait and figure out a way to import 24 bit into Speed Edit?
 
16 bit works just fine; music CDs are 16 bit. Yes 24 bit would be better but in practical terms most people can't tell the difference between the two.
 
With digital audio people have lost their touch with the reality a bit. 16 bit audio provides much better dynamic range than any analog systems used for cinema from the fifties to nineties (remember those Nagras which cost 10 k$ per channel, they equalled about 12 bit resolutions at best). To ask if 16 bits is good enough for indie productions? 16 bits is as good as the real thing, people can not tell live signal apart from 16/48 signal, this has been tested many times. Also practically nobody has playback systems which could fully use even 16 bits (96 dB dynamic range with like 30 dB SPL ambient noise means 126 dB SLP peaks...).

Another matter is the added safety margin provided by 24 bits. Levels setting is safer and still you can get "perfect" 16 bit transfers for delivery. There is one caveat, though: cheap 24 bit recorders are nowhere near 24 bit in real life, many of them are not any better than 16 bits actually, as the mic pres and ADC are so noisy that 24 bit files only have 8 bits worth of more noise than 16 bit files. I am afraid Fostex belong to that group. Even the highest end recorders like digital Nagras, Devas and SD7xx have 19 bits of real resolution (dynamic ranges arond 114-117 dB) and the very best dedicated converters like Lavry end Weiss can manage 21 bits out of 24. Real 24 bit audio does not exist.

Theory is theory, reality is reality.
 
Petrus you get so pumped up. Is that like one in the hand is worth two in the bush or what? But your right again many people can't hear the difference and even when its converted way down for streaming some people cant seem to notice the difference. Its like having a Farrari that does 260 mph that you can't use when a corvette for one fourth the price is almost as fast and seems pretty nice when driving. 24 bit or as high as you can get helps a little when you are going to be processing the signal alot or streching it.
 
Soundmen of the past decades made great soundtracks for the movies with the ancient analog equipment, now we worry if we do not have the best theoretically possible systems.

I can quarentee that any indie production sound is not going to trip over the fact that only 16 bit recorders were used, it is the microphones, placement, mixing and post which matters, and did I mention mic placement?

Even if 16 bit recorders are used, all respectable DAWs use 32 bit processing which hides the rounding errors.

I myself record in 24/96 as I have equipment for that which actually works almost as adverticed, at about 19 bits in my case (SD722). But people should be aware that cheaper 24 bit recorders are not that and should not rely on getting more headroom with using 24 bit sample depth, as they are not.
 
there is no any problem recording dialog in 16 bit , i do it for the last 15 years with hhb dat and even now with 788 or deva .
 
i record in 16/48, and have yet to meet someone else here who doesn't. never noticed a difference worth mentioning
 
For video I record at 24/48 to save my ass if I judge the levels grossly wrong. Has not one iota of quality difference in the end product (exept maybe if I blew it, but that was a mistake, not that 24 sounds better).

Just once it has been usefull: had to take sound from the soundboard and could not trust the levels given and no possibility to correct after the show started. Recorded safe at 24 bits; file peaked at -32 dBFS. it was possible to raise levels 28 dB without hiss.

In real life 16 bits is plenty good and possible to record hot with good limiters. I would rather get a SD302 mixer than 24 recorder any day.
 
Petrus and I disagree a bit (I can easily hear the difference between 16 & 24) but 16 is more than fine for production sound. There is a whole era of films made (with digital recorders) when pretty much nothing recorded over 16 bit. Then the raging question was is 44.1 better than 48... There is always some debate going on in sound. I would argue that it is always better (if you have the option) to record at 24. Even with your low priced recorder that may only be getting a useable 16-18 bits. 16 bit recordings don't actually get 16 bits either, you need some head room unless your a psychic. So at 24 your at least getting more than at 16 even if it's not the full 24 (and it never will be).
 
phd said:
Use camera format 16/48. Every change of format gives distortion
Click to expand...

That is a very misleading and not totally correct statement. On a pure science level you are probably technically correct, since any change would be considered a "distortion" from the original. On a real life practical level your not. That "distortion" will not be audible to anyone unless your using really crappy software to do the conversion and that is a software issue not something inherent in format conversion. As has been pointed out above your DAW or NLE is going to be "format conversions" every-time you do anything to the sound, like change a level or play more than one sound at the same time. Post on films is typically done at 48/24 even if the tracks were shot at 16 because of advantages in processing.

Anyway right answer wrong reason.
 
Noiz2 said:
Petrus and I disagree a bit (I can easily hear the difference between 16 & 24) but 16 is more than fine for production sound..
Click to expand...

The question I'd have is not so much at what quantization an audio signal is digitized at... but what is the internal representation of the digital value, 16, 24, or some form of floating point representation, to prevent rounding/truncation errors in 'mixing' or processing the signal. With a signal with 'high' noise, having extra bits of noise may not be all that benefical.
 
If it's on DV its 16. Protools Mix bus is something like 48 bit fixed in HD and 32 float in Le. One can get very technical about what is in theory better but the bottom line is well recorded 16 sound great, badly recorded whatever sound , well, badly recorded.

We happened to do a bunch of double blind tests with a Nagra V at a studio that I work at sometimes. Everyone picked 24 over 16. 96 over 48 was slightly detectable depending on material. Some didn't hear it some did. I'm a guy and over fifty so my hearing has some built in roll off. Folks with better hearing might be able to hear more of a difference with 96??? But some of our listeners were young so???
Now the Nagra has really good electronics but we also could hear the difference with stuff recorded with a Marantz, and that has less "really good" electronics. So I was sold.

Post beats the hell out of audio so the finer points of theory go out the window. You production could be at anything, it's going to be mixed with a wide variety of sound, some of which are spotty at best and some have been processed to hell and gone. Then the whole mess gets mixed down to linear PCM if your DV or a file that gets converted to AC3 if your going to DVD. Finer points of dither got lost the third or forth time the file got destructively processed.
 
There is no important difference between 16bit and 24bit if you are using the same recorder and recording really hot or with a limiter.

24bit has a big advantage because it can record equal digital quality at much lower signal level. An accidental -36dB level on 16bit recording is only 10bits of digital resolution so it can have audible noise/alias when you boost the level in post. The -36dB signal on 24bit theoretically has very good audio quality. This is only advantageus if the 24bit ADs are very good though, field recorders are not exactly audiophile because the typical microphone signal has high noise level and ambience which hides the design flaws. Some 24bit recorders are of very low quality.

A CD sounds excellent at 16 bit but it is optimised and mastered for 16bit delivery making excellent use of the dynamic range using compression and limiting. It is usualy recorded at much higher precision in the first place. A 16bit field recording that leaves 12dBs or 18dBs for headroom is very compromised compared to the typical audio cd.

Changes inside the editor are usually non destructive. If the editor uses floating point in its calculations, even multiple processing passes have excellent audio quality. If the recorder processes and stores 16bit, the quality could be very low after a few passes. The high processing resolution is good even for low resolution input.

It's better if the recording, processing and delivery is done at the same sampling rate. 48khz for most video uses. I doubt anyone will notice the loss of resampling in a video, but it's better if we avoid it.

The only advantage of 96khz over 48khz is in the limits of audile range, 18khz and up, because 96khz is pushing the nyquist limit and the AD artifacts one octave higher. Depending on 48Khz implementation, this 18-20khz area could be compromised. We don't hear much above 20khz. The point of >40khz sampling is to leave some space above 20khz for AD implementation needs.
 
24-Bit

The important think about *recording* at 24-bits is that you have adequate and proper headroom.

Once edited it can be down-converted to 16-bits.

Recording at 16-bits you have to be a *lot* more careful with levels, have less headroom and run the risk of distortion with unexpected peaks.

That many people can't tell the difference is unimportant - it's the *recording safety* that is the important thing about recording 24-bits.
 
AMEN!

That is, if you have a 24 bit recorder which actually has better than 16 bit S/N ratio. Cheap ones do not, at least not with mic-ins.

Use line in with good mixer.
 
Otis Grapsas said:
Changes inside the editor are usually non destructive. If the editor uses floating point in its calculations, even multiple processing passes have excellent audio quality.
Click to expand...

Sort of depends on how you define it. True most mixing is "non destructive" in that you don't rewrite the source file. But if you do a volume pass at the clip level and a volume change in a sub and another on the master the stream has gone through three processing passes and the stream was destructively altered on each one. I agree that on any DAW worth the name your not going to hear any degradation, and that was my point (responding to the poster who claimed any change "distorted" the sound).

However it is also true that on a film work flow a LOT of the recordings are destructively processed, meaning the file is replaced with the processed file. Noise reduction, time syncing alt takes or ADR, EQ matching between clips (some times) and probably most FX sound are destructively processed.

A side note is that FX are more and more being recorded at 96/24 for higher budgeted films and commercials. I don't think production is done much at 96 though.
 
Noiz2 said:
Sort of depends on how you define it. True most mixing is "non destructive" in that you don't rewrite the source file. But if you do a volume pass at the clip level and a volume change in a sub and another on the master the stream has gone through three processing passes and the stream was destructively altered on each one.
Click to expand...

Volume changes are single multiplications with a float parameter. You can probably have hundreds of these in a row without an audible change in quality. The software should be doing them in real time because they cost nothing (cpu wise). They are usually implemented at the last stage of processing in real time and are always active regardless of volume setting. They are destructively processed only on mixdown.

Multipass destructive processing is an issue. Material that needs to pass through a lot of different systems and editors, should be stored in a high precision format. This is not as critical as it sounds though because even at a 16bit intermediate, a single channel will not be going down destructively 40dB in one stage and up 40dB in the next stage. Generally, the changes in volume will be small and alias will not be a problem. As long as original levels and intermediate levels are hot, it will be ok. Final levels can be low of course, this reduces the alias further.

This is a similar situation to 8bit vs 10bit video. You can make the 8bit file looks grossly inferior if you reduce levels to 1/16th (reduce brightness), store destructively in 8bit, then increase the level by 16x. But then you are comparing 4bit video to 6bit video (not 8bit to 10bit). This is not typical of everyday editing because there is no point in reducing exposure destructively just to increase it again and nobody records video with 4stops of underexposure (1/16x level). It does serve the 10bit marketing purposes well though because it magnifies the difference of the two formats using an unrealistic use scenario.

We do know 8bit video and 16bit audio are good enough for professional results if they are used properly. It's in 99% of the camcorders out there:) In most of them it's not even a full 8bit and 16bit because compression is used, sometimes extreme mpeg2/4 compression with mediocre PSNR, which performs far below the capability of the specified bit depth.
 
Last edited:
Petrus said:
16 bits is as good as the real thing, people can not tell live signal apart from 16/48 signal, this has been tested many times.

...and then....

Real 24 bit audio does not exist.

Theory is theory, reality is reality.
Click to expand...

Petrus,

Um, I'm going to disagree with the part about 16-bit being OK. 16-bit isn't reall 16-bit either, It's usually more like 12-14 bit.

And it's a fact that most of the problems people had with digital audio went away when the industry switched from 16-bit to 24-bit. I don't think we really need 96 kHz sampling, 44.1 or 48 will do.

Regards,

Ty Ford
 
You must log in or register to reply here.
Back
Top