bit rate/kHz for ambient nature recording

As for the 16-bit vs 24-bit audio debate goes, I suggest people take a look at this thread on the matter...

24bit vs 16bit, the myth exploded!

...I know that in theory 24-bit audio should make a difference ( other than dynamic range ), but I have yet to find any 24-bit audio samples that clearly demonstrate the superiority of 24-bit audio, other than the much better dynamic range.

I would love for someone to post a link to a great piece of 24-bit audio that I can download, and then I can dither and downsample this audio to 16-bit, so I can try and hear the difference with my own ears. I work with a couple of recording engineers, so I will even get them to give it a try, as I honestly doubt that the 24-bit audio format is going to make a perceptable difference other than a much better dynamic range. Anyone got a link ?
 
Admittedly I did not read through all 50+ pages of the thread you posted, but what I don't see addressed is dynamics processing of recorded samples of 16-bit versus 24-bit. This is where I think the 24-bit makes the biggest difference.

To make a very easy example, if you record video with say any of the 8-bit cameras and you adjust the curves too much, you start to see polarization in the video. Since you are already at your target bit depth, and now you've expanded it, you can see this clear as day, as I've done it myself on many occasions.

Back to audio, if you use a lot of dynamics processing where you've got very quiet parts followed by loud parts where in post you are going to squash those together, the 24-bit is going to make a difference. The more you expand the lower resolution portion into a higher resolution, the more you are going to create steps in the audio. It's the same situation.

Just like 8-bit video camera output, what matters is how far you are going to push the image, in this case audio.
 
Rick Burnett said:
what I don't see addressed is dynamics processing of recorded samples of 16-bit versus 24-bit
Click to expand...

Much like with 8-bit vs 16-bit images, judicious use of dithering can help to maintain the integrity of lower bit-rate media. I think a better equivalent would 16-bit graphics vs 24-bit graphics.

24-bit images can be better, but often the difference is very subtle.
 
Dingo, that article is just wrong. I'm sure its been pointed out somewhere in the 53 pages of discussion. It's simple: values in dBFS range from 0 to 1 (-infinity dBFS to 0dBFS). 1 (0dBFS) is always the maximum value. -48dB is not the same as 48dB; in fact, -48dB is about .004, while 48dB is about 8. The extra resolution is there throughout the range. The extra .004 of dynamic range is at the bottom of the scale.

If you want to Play with this, record something at 4- bit, 8-bit, 16-bit. Those differences are easier to hear. Then extrapolate what 24-bit will add.
 
Or just use something like Digidesign (er, Avid)'s D-Fi plugin that lets you reduce the sample size and hear the differences.
 
ullanta said:
Or just use something like Digidesign (er, Avid)'s D-Fi plugin that lets you reduce the sample size and hear the differences.
Click to expand...

Haha, I use bit reduction for effect in Ableton, it's fun! But then again, I like to sound design as well.

I can say this, I have an Apogee Rosetta 200 audio interface and an MOTU 896HD. The apogee cost me $1800 whereas the 896HD $695. I took at $2k microphone (Blue Kiwi) and fed it into a Focusrite ISA220 which is also $2k when I bought it. I say this because I want you to know I used REALLY good equipment, but certainly not THE best (and that is subjective because I think more expensive preamps color the sound a lot, especially the Manley stuff to me).

I split the input with an isolation transformer with two XLR outputs and fed them both into each ADC. I then used a test tone of pink noise and set each of them to the same levels, using the digital input on the computer at high resolution to make sure they were at the same level.

Both devices were clocked with a Lucid Genx6 which is VERY stable.

I then recorded acoustic guitar that I played into the mic.

I asked a handful of other people to listen to the recordings and tell me

1) are they different and explain the difference.
2) which do they prefer.

Answer 1 was a mix of yes and no, but they couldn't explain why

Answer 2 was they liked both. There was nothing that stood out

I myself listend to the samples over and over and over. The differences between the two (as I inverted one and summed it to the other after aligning the waveforms) was in the transient response. The Motu 896 converters didn't detect the wavefront as fast. But the difference was VERY VERY VERY small.

Given the noise floor between both was below the room noise, I couldn't explore at the time if I had a quieter space or used an anechoic chamber, but it was a good lesson for me.

1) Expensive equipment might exceed the environment you work in OR the conditions you will use the material.

2) While one may technically be better, most people can't tell the difference. Know which statistics count.

3) Maybe I don't have the most amazing ears and am not qualified to answer my own question.

I kept the Apogee and use it for anything I run a mic through, but I know it is overkill for me.

:)
 
TheDingo said:
As for the 16-bit vs 24-bit audio debate goes, I suggest people take a look at this thread on the matter...

24bit vs 16bit, the myth exploded!

...I know that in theory 24-bit audio should make a difference ( other than dynamic range ), but I have yet to find any 24-bit audio samples that clearly demonstrate the superiority of 24-bit audio, other than the much better dynamic range.

I would love for someone to post a link to a great piece of 24-bit audio that I can download, and then I can dither and downsample this audio to 16-bit, so I can try and hear the difference with my own ears. I work with a couple of recording engineers, so I will even get them to give it a try, as I honestly doubt that the 24-bit audio format is going to make a perceptable difference other than a much better dynamic range. Anyone got a link ?
Click to expand...

In the final product the noise level in a perfect 16 bit file is about -90 dBFS, which means that you would have to play it at about 120 dB SPL peaks to raise the noise floor above the listening room ambient noise. Thus the dynamic range difference between 16 and 24 bit files is next to impossible to hear. The great thing in recording at 24 bits is the extra safety it provides in level setting. In real life there are really nothing that you can record with more than 90 dB dynamic range, and it is next to impossible to reproduce more than 90 dB dynamic range, which 16 bits can handle.

I have done a simple test, which I failed: I had an original 24/96 file of classical orchestral music (from Linn). I cut the file into 30 second snippets and converted them to 16/44 and back to 24/96. Then I, by flipping a coin to guarantee randomness, glued those down converted snippets with the original untouched snippets so that consecutive 30 second snippets were either 16/44 or 24/96. No matter how hard I tried I could not hear any difference in the file (where the resolution changed), it just played smoothly. All high-rez fans say "you can notice the difference immediately"! Well I could not, and neither could any of those 150 other people who have listened to it so far.
 
I agree, Petrus, I don't think the difference is so significant in the final product for most things. The main benefit is the added flexibility 24-bit allows when recording and in post. I think a better test to try is to record something with peaks at -20dB (or just drop the levels in post) in both 16 and 24 bit, then normalize them to proper levels, and see if there is any noticeable difference there. In fact, if someone is bored, it would be interesting to try this at a few peak levels (-30, -20, -12, -6...?) and see if we can determine where the difference really starts kicking in.
 
I need to figure out if I can get my devices to record in 16-bit (MOTU 896HD) so I can do some tests.

The first thing I did was drop a sine way into a track and set the peak at -54dB. I then exported this as:

24-bit
16-bit no dither (truncates bottom 8 bits)
16-bit pow-r 3 dithering.

I brought all 3 samples back in and increased them to a listenable level.

The 24-bit was clean.

The 16-bit no dither had TONS of artifacts in the sound. It was horrible.

The 16-bit pow-r 3 dithering was super noisy.

Now granted, this is extreme, but if you are really pushing the dynamics on a sound, or, something got recorded really low, it's amazing how bad these were. The source had ZERO noise as it was digitally created and not recorded through something. I just found it fascinating.

What I learned was:

NEVER reduce a 24-bit audio sample to 16-bit unless it is the absolute final mastered version. If you have to make any changes, the dithered noise can possibly be heard if you gain something up too much.

What I'll try and do is play the sine through my Adam A7s and record with the Kiwi. The goal will be

1) decently loud at both 16-bit and 24-bit. Do not expect any change here.

2) middle of the range on both, use a compressor to bring it up, expect a little difference.

3) very low sound, mainly to see how much it degrades when certain tones are brought really far up.

Hope to do this tomorrow.
 
Cool, Rick! Remember, a sine wave is actually the best case... While you're playing, try some other more complex waveforms like dialog or music (or nature sounds to stay on topic).
 
The vital thing about 24-bit is not as a LISTENING thing, but as a MASTERING thing.

24-bits gives the headroom and safety margin for MASTERING and doing the original recording.

When all is finished, you can then down-convert to 16-bit.

It's nothing to do with hearing the difference between 16 and 24 bit - but the extra safety you get when recording and that you can manipulate the recording in post production without degrading the file.
 
John Willett said:
The vital thing about 24-bit is not as a LISTENING thing, but as a MASTERING thing.

24-bits gives the headroom and safety margin for MASTERING and doing the original recording.
Click to expand...

...But there is a problem with many recorders in the prosumer and low-end price range, where self-noise of the audio circuitry defeats the extra headroom of using the 24-bit format.

For example: I own an Olympus LS-11 audio recorder which has very nice mics and fairly good pre-amps, and has a very nice over all build quality to it. Performing the noise-floor test I described earlier, the self noise of the unit is about -75 dB, which is fine for much of the work I shoot. Previously I owned an Olympus LS-10 recorder, which showed self-noise of about -62 dB, which was a problem sometimes when recording very quiet sounds.

A couple of years ago I tested several Tascam recorders and found that they were no better than the Olympus recorders. ( self noise was in the range of -48 dB to -73 dB )

Sony has recently discontinued the PCM-D50 recorder, which I am hoping that this means a newer replacement is coming out, as the PCM-D50 is quite a good recorder. ( I owned one a few years ago, but never tested it for noise )

Later this Spring I want to test some of the newer low cost ( under $600 ) recorders that came out in the past year, to see if the self-noise situation has improved at all.
 
An interesting twist, doing SRC after recording at a high sample rate. As others alluded to this may not have much bearing on whether RECORDING at 24 bit has advantages over 16 bit. We did some blind tests recording the same things at 16 and 24 and at 44.1, 48, and 96. Our conclusions were that 24 bit was worth it and 96 maybe on high bright and impacty stuff, especially if you were planning on slowing it down.

Of course what this could mean is that the recorder worked better at 24 bit? Maybe it is always at 24 and is dithering on the fly down to 16? Either way we could clearly hear the difference. We had assumed it was purely a bit depth thing but?


The other thing to remember in all this as it relates to film is that in most cases you are recording the raw material that will get hammered and shaped into the final product. What would never fly as a one shot high end music recording is often not even in the sights. If your recording FX you are mostly in the wild in uncontrolled environments. When you get back those sounds are going to be manipulated and tweeked and layered up with other recordings maybe some library stuff from the BBC (mono analog converted to 44.1 CD and then digitized and converted to 48). If your lucky to have access to a good library you may be pulling sounds that were recorded on all kinds of weird formats that have gone through god knows how many conversions and or generations.

In other words it's always best to start with the best you can get but it doesn't pay to get too obsessive about it.

Dialog tends to be a little more "raw" but it is still going to be heavily processed by the time it's finished.

The irony is that the closer you work to the no budget/ no post aesthetic the bigger the problem with noisy recordings.

When I say "noisy" I'm talking the kind of noise we are talking about on this thread not the "dialog swamped with a ground loop interrupted by an intermittent connection" noisy that no-post folks are forever trying to find a plugin to "fix".
 
So I did some tests with my Sony PCM-D50 because my audio gear in the studio appears to only do 24-bit and I could not find any selection to change that. I created a test tone on the computer and recorded it (440Hz) at -40dB at both 24-bit and 16-bit and compared that to something recorded at -6dB at 24-bit. I used a compressor set to always be on to just use it as a gain up to the same -6dB.

The room I am in is not very quiet, there is a bit of hum, and because of this, the bottom floor of what I am recording is not limited by the recorder, it's limited by the room tone. While not always as bad as this, this has ALWAYS been the case when I am on set recording (well except for the H4n which I could hear the opamps). This is true with the audio in my FS100 as well, and the reason it is SO important to use a microphone if you can to focus ON the sound you want. In other words, it really doesn't matter how expensive the recorder is if you can't get a good separation of the signal you want and the ambient noise you don't want as much of.

With the OPs question about ambient sounds, this is very different in some sense, but then again, it isn't in terms of what is limiting the system.

-40dB was pretty quiet. What was interesting was the only difference I really perceived was more crispness in the noise in the 24-bit version. Now, who knows what the real difference is between 16-bit and 24-bit in the Sony PCM-D50, I certainly don't. What I was surprised at was the sine wave didn't sound different to my ears in all three recordings after the -40dB was compensated for.

So, what does this mean?

Well, it means that the limiting factor in most situations that we deal with is the actual noise in the scene, and not the converters/op amps for dialog situations. For ambient sounds, I think switching between 16 and 24 bit is going to do nothing if your analog path is noisy (again, h4n) because it's the relative levels that matter and the bit depth doesn't change this. When you record at 16 or 24 on the pcm-d50, the input is still the same on the scale.

As stated, your dynamic range which is the same since the noise coming through the microphone alone is eating ALL of the negative end of the spectrum so if your ambient noise is at -50dB, everything below -50dB really doesn't matter does it?

Now a sine wave isn't an example of everything, so I am not testing high frequency bit depth concerns here, but in this simple test, I was expecting A LOT more difference.

Now, had the room been quieter and I could record at even lower than -40dB maybe I could find the point when there will be a difference, and I believe there will be, that said, it's not realistic. -40dB is SO LOW that in the field, I'd never record a signal that low because the noise would get brought so much up, it would not sound right.

What I will say is this, EVERY piece of recording software that is not integer based is going to convert your audio to 32 bit. The beauty of working with 24-bit is that it fits nicely into the 24-bit of the 32-bit reserved for the mantissa (where as the remaining 8 bits are used for exponent). (there are 64-bit formats as well but I don't use them). If you use 16-bit, there is a conversion process or it just fills the top 8 bits with zeros (I have seen both). What I am not clear on is:

a) what does the 16->32 bit conversion do?
b) how does processing affect that original 16-bit material.

Thinking of this from an application specific standpoint, filters can really be hard on audio, mostly low pass. But, I don't use ANY lowpass filters on film audio typically, if anything, it's a high pass filter to remove something from the bottom end. So really, how much does it matter? I mean, this stuff is pretty easy to test in your situations. (And i mean stuff that I record, not stuff I might use from an audio library)

Given that I can hear the difference between 8-bit and 16-bit, but not 16-bit to 24-bit, and that the software I use likes to use 24-bits of precision, I am going to stick with 24-bit. To me it seems the best practice especially if I really start to alter the clips with multiple levels of tools. I think it still comes down to how much you are going to manipulate your sounds in the end, and, how good the converters are. That's my two octaves anyways :)
 
...I was DOP on an Indie short this weekend and our sound guy had an Zoom H4n recorder, which I borrowed for half an hour to do some quick noise floor tests.
( years ago I owned the absolutely horrible Zoom H4 recorder, which I got rid of as soon as I could, but I have always wondered how improved the newer H4n was )

Using the method I described in an earlier post ( using my MixPre mixer ), here are the results:

48 kHz 16-bit : noise-floor = -66 dB
48 kHz 24-bit : noise-floor = -66 dB
96 kHz 24-bit : noise-floor = -63 dB

...I don't know why but 96 kHz 24-bit on the Zoom H4n recorder shows 3 dB of more noise than 48 kHz does, which seems a little odd.

For comparison my Olympus LS-11 recorder shows a noise-floor of -75 dB at the same settings.
 
Rick Burnett said:
For ambient sounds, I think switching between 16 and 24 bit is going to do nothing if your analog path is noisy (again, h4n) because it's the relative levels that matter and the bit depth doesn't change this. When you record at 16 or 24 on the pcm-d50, the input is still the same on the scale.
Click to expand...

Rick, thanks for providing all the great testing information.

I think the statement above is at least theoretically not quite appropriate... since you're on a testing bent, maybe you can test! But I would say that the relative levels ARE affected, especially near the noise floor, because the resolution becomes increasingly important down there. 24-bit will have 256 times more precision with which to disambiguate the noise from the very quiet signal.... that is, some distinctions that are representable in 24-bit will not be representable in 16-bit due to precision, and many would be blurred.

Maybe that's what makes the noise crispier! (I've been trying to throw the phrase "crispy noise" into every conversation I have since you wrote it.)
 
I do lots of ambient nature recordings with a Marantz PMD661. I always use 24 bit, and mostly at 44.1 kHz, sometimes at 96kHz. You won't hear a difference between 44.1 and 48 k, but you may hear a slight improvement at 96k. However, for video it probably makes sense to use 48k. I use 44.1 because of the ease of dropping files into other audio projects done at 44.1. At 96k your storage requirements become a heck of alot larger. You need to decide if that makes it worth it or not for you.
The preamps in the Marantz are ok. There is however, a drop in sound quality from these compared with my expensive studio preamps.
 
You must log in or register to reply here.
Back
Top