bit rate/kHz for ambient nature recording

With the final product 16 versus 24 bits will not matter much, but using 24 bits for recording gives some extra safety, as you can set the levels a bit (yes, maybe 6 dB*...) lower; accidental clipping will not happen so easily and often.

*) the difference between 16 and 24 bit recording is not 8 bits in reality, as even the best 24 bit recorders give only 21 bit resolution, many cheap "24 bit" recorders no more than 16 bits. This is because the analog stages are never perfect, and cheap converters even less perfect.
 
Petrus said:
With the final product 16 versus 24 bits will not matter much, but using 24 bits for recording gives some extra safety, as you can set the levels a bit (yes, maybe 6 dB*...) lower; accidental clipping will not happen so easily and often.

*) the difference between 16 and 24 bit recording is not 8 bits in reality, as even the best 24 bit recorders give only 21 bit resolution, many cheap "24 bit" recorders no more than 16 bits. This is because the analog stages are never perfect, and cheap converters even less perfect.
Click to expand...

Don't forget though that many people dither their 24-bit into 16-bit which gives I believe a perceived range of about 21-bit (I use powr-3 myself). Also, if you do any sort of dynamics processing in post, that extra detail can be used if needed.
 
Petrus said:
With the final product 16 versus 24 bits will not matter much, but using 24 bits for recording gives some extra safety, as you can set the levels a bit (yes, maybe 6 dB*...) lower; accidental clipping will not happen so easily and often.

*) the difference between 16 and 24 bit recording is not 8 bits in reality, as even the best 24 bit recorders give only 21 bit resolution, many cheap "24 bit" recorders no more than 16 bits. This is because the analog stages are never perfect, and cheap converters even less perfect.
Click to expand...

Yes, but then 16-bit really isn't 16-bit, so the difference is........WTF knows, but appreciably better.

Regards,

Ty Ford
 
My vote is for 24/48, the best compromise between quality and storage size. It doesn't seem like you plan on effects work, just ambience. Larger sample size has a much more noticeable benefit in most cases than higher sample rate.

For the record, there is discussion earlier in the thread about theoretical noise floors of -96 and -144 d. For 16 an 24-bit, respectively. This is not correct. This is the dynamic range that can be represented by the respective formats. Noise floor is a result of analog circuitry and is the same in 16 or 24 bits… that is, if a particular recorder has a noise floor of -89 db, then it will have a noise floor of -89 db whether you switch it to 16 or 24-bit recording.

Further, more bits does NOT provide more headroom. More bits provide more footroom, and higher resolution, allowing you to record at a lower level (and thus indirectly give you the effect of more headroom). But if the analog signal you're trying to record is -3db at 16 bits, switching your recorder to 24 bit mode will not change that...the level will still be -3db. If clipping at 16, it will clip at 24.

Sorry, just some extra precision in terminology here. You can adjust the analog signal path so as to get more effective headroom when using 24 bit recording, but 24 vs. 16 bit recording without analog signal level changes provide exactly the same headroom.
 
ullanta said:
For the record, there is discussion earlier in the thread about theoretical noise floors of -96 and -144 d. For 16 an 24-bit, respectively. This is not correct. This is the dynamic range that can be represented by the respective formats. Noise floor is a result of analog circuitry and is the same in 16 or 24 bits… that is, if a particular recorder has a noise floor of -89 db, then it will have a noise floor of -89 db whether you switch it to 16 or 24-bit recording.
Click to expand...

Wikipedia: Audio bit depth

By increasing the sampling bit depth, quantization noise is reduced so that the S/N is improved. The 'rule-of-thumb' relationship between bit depth and S/N is, for each 1-bit increase in bit depth, the S/N will increase by 6 dB.[1][2] 24-bit digital audio has a theoretical maximum S/N of 144 dB, compared to 96 dB for 16-bit; however, as of 2007 digital audio converter technology is limited to a S/N of about 124 dB (21-bit)[3] because of real world limitations in integrated circuit design.
Click to expand...

My understanding is that once the noise floor has crossed the threshold of what the 16-bit digital audio format can reproduce ( -96 dB in theory ), then a higher bit-rate format is required to accurately reproduce this audio.
 
I probably should have clarified that changing the analog gain is what I meant by "trying to retain more detail" when referring to the 16-bit. When I used to record in 16-bit days for 16-bit release, I'd push those levels up, and sometimes, even use a compressor just to tame the highs down. These days, I give myself that additional headroom on the digital side by reducing the gain to allow for more peaks. Granted, by the definition of headroom from an analog standpoint, it is not really more headroom from the analog perspective, but most people I speak with from a digital standpoint understand what is being explained. It also makes it more complex because while you are speaking of acquisition in 24-bit, you are really planning on 16-bit, and thats part of the equation.

I will say that I've also gone so far as to record two channels, with the second channel from -6 to -12dB down so that if I got an event I didn't expect, I could use both recordings to fix the over in post. This works well.

With the Sony PCM-D50, it actually has a mode of doing this automatically. It actually works fairly well although I really never want overs anyways.

All that said, the part about the noise is very important. If you are using a noisy recorder and you lower the analog gain, then the signal to noise ratio has been reduced meaning that the incoming analog levels are now closer to the noise. So while you've given yourself more dynamic range to work with, you've also brought up your noise since the difference between signal and noise has shrunk in the analog domain. This is because the noise is the same in both 16 and 24 bit on a given recorder.

When I started recording ambience, you REALLY notice converter noise. Would love to pick up a SoundDevices one day.
 
TheDingo: this can get increasingly technical (which I like!)... but "noise floor" is not generally referring to quantization noise. It is used generally to refer to analog circuit noise present in any device. Further, quantization noise is not a fixed quiet "noise signal" like we think of the "self-noise" of the noise floor... it is a result of the mathematics of A/D conversion. 24-bit has quantization noise that's 1/256 of the quantization noise of 16-bit conversion. It is this "quantization error" - or "quantization precision" - which is the primary difference between 24 bit and 16-bit recording. The extra resolution of 24-bit lets us record at lower levels without loss of signal quality due to the conversion process itself. The noise floor, as Rick points out, remains the same... so in lowering levels to gain extra headroom, we lose a little signal/noise ratio. On good equipment these days, it's a good tradeoff. But something to be aware of and plan for.

And, yes, if the noise floor is actually below -96dBFS, then quantization error in a 16-bit recording will start to dominate the noise floor. But remember, we've got an analog noise floor on the mic, the preamp, AND the recorder, and while any one of these might, given great equipment, be below -96dBFS, it's unlikely in any real-world situation that the combination will. And even then, that the actual noise level of any real-world environment will not be low enough for that to matter.

There's a lotta kindsa noise!
 
Last edited:
ullanta said:
but "noise floor" is not generally referring to quantization noise. It is used generally to refer to analog circuit noise present in any device.
Click to expand...

I completely understand this, but with higher end equipment the combined noise of the entire audio chain may still exceed what the what the 16-bit format can faithfully record, and thus 24-bit recording becomes necessary.

With lower end gear, self noise of the combined audio path defeats any reason to use 24-bit recording, because the gear itself can't come anywhere near the -96 dB format threshold, thus 16-bit records exactly the same information as recording in 24-bit. There's no "extra resolution" in the 24-bit format, just un-used dynamic range if the audio chain is too noisy.
 
TheDingo said:
I completely understand this, but with higher end equipment the combined noise of the entire audio chain may still exceed what the what the 16-bit format can faithfully record, and thus 24-bit recording becomes necessary.

With lower end gear, self noise of the combined audio path defeats any reason to use 24-bit recording, because the gear itself can't come anywhere near the -96 dB format threshold, thus 16-bit records exactly the same information as recording in 24-bit. There's no "extra resolution" in the 24-bit format, just un-used dynamic range if the audio chain is too noisy.
Click to expand...

What gear?

http://www.avisoft.com/recordertests.htm

NONE of these has a noise floor worse than -96dB. I think you might be thinking of MUCH older gear.
 
TheDingo said:
With lower end gear, self noise of the combined audio path defeats any reason to use 24-bit recording, because the gear itself can't come anywhere near the -96 dB format threshold, thus 16-bit records exactly the same information as recording in 24-bit. There's no "extra resolution" in the 24-bit format, just un-used dynamic range if the audio chain is too noisy.
Click to expand...

This is not at all true. 24-bit recording has exactly 256 times the resolution of 16-bit. That is, in a 24 bit recording, there are 256 times as many possible representable signal levels between, say, 0dBFS and -1dBFS as there are in a 16-bit recording. This is consistent throughout the range.

This is why, to open a new can of worms, the effective dynamic range of a 24-bit recording is not much greater than a 16-bit recording. The top of the scale is pinned at 0dBFS; the extra dynamic range is at the bottom of the scale. The difference between -96dB and -144dB is exactly 1/256 of the difference between 0dB and -96dB... so the dynamic range of 24-bit is 1.004 times the dynamic range of 16-bit, and the difference is in what the quietest sound you can record is...
 
I think people say dynamic range a lot, but what they really mean is resolution. If you record with higher resolution than your target resolution, you have the freedom to stretch sound volumes up and not introduce as many issues. For instance, if you are recording a vocal in a song and you record it clean, no processing. Then, in post you want to use a compressor so you can raise the volume of the piece and bring down the high parts so they don't clip. The problem is, if you recorded that in 16-bit, let's say your average audio peaked at 8-bit. Now you've just brought it up to 12-bit and compressed some peaks hitting at 15-bits down. What you are doing now is causing 8-bits to be stretched into 12, which means you are now stretching out the quantization now and that is certainly something I don't desire. Not that software can't interpolate, but at the same time, the further you stretch, you are still missing detail that would have been there in the 24-bit resolution.
 
Rick Burnett said:
What gear?

http://www.avisoft.com/recordertests.htm

NONE of these has a noise floor worse than -96dB. I think you might be thinking of MUCH older gear.
Click to expand...

Hey Rick, that's a tricky chart, and this is a tricky issue. We're kind of dancing between dBu and dBFS. But in that chart, the interesting number is on the right - under their definition of dynamic range... and invert it to get the kind of noise floor we're talking about in dBFS. So the best of those is around a -87 dBFS noise floor.

(Note the clipping level in the second column, and imagine that the clipping level is 0 dBFS).
 
ullanta said:
Hey Rick, that's a tricky chart, and this is a tricky issue. We're kind of dancing between dBu and dBFS. But in that chart, the interesting number is on the right - under their definition of dynamic range... and invert it to get the kind of noise floor we're talking about in dBFS. So the best of those is around a -87 dBFS noise floor.

(Note the clipping level in the second column, and imagine that the clipping level is 0 dBFS).
Click to expand...

Ahhhh, hadn't noticed that before. Thanks for pointing that out. That is a tricky graph. But what you said earlier still holds true. 24-bit will have more resolution than 16-bit for the same noise floor. That's why it's good to use even if you still recorded at the same input analog level.
 
Rick Burnett said:
I think people say dynamic range a lot, but what they really mean is resolution.
Click to expand...

This is indeed one of the most frequently misunderstood aspects of sample size... that and the logarithmic nature of the dB scale itself...

The only reason that there is ANY difference in dynamic range between 16 and 24-bit recordings (or 8-bit or 2-bit) is because of the resolution. That is, the smallest step above 0 that a 16-bit system can represent is -96dBFS (or 1/65536); the smallest step that a 24-bit system can represent is -144dBFS (1/16777216). The high end... the largest signal that can be represented... is exactly the same. So in non-logarithmic numbers, the actual difference in dynamic range is the size of the smallest step... so 24-bit has a dynamic range that is 1.004 times as great as 16-bit, and the extra is at the quiet end.
 
Rick Burnett said:
24-bit will have more resolution than 16-bit for the same noise floor. That's why it's good to use even if you still recorded at the same input analog level.
Click to expand...

Yes! People often think I'm arguing against 24-bit recording when I say the dynamic range is not that much different. I'm not! I would always recommend 24-bit recording, for the resolution benefits.
 
What a great chart! I think that REALLY illustrates the difference. You make a really good point as well, people are so focused on the 0dB range of the meters, and how big the difference is between 0dB and -6dB, they might not realize how small a difference there is between -96dB and -144dB.

This is why when I compare recorders, the first thing I look at is SNR. This is a good indication of your limiting factor on gaining up ANYTHING you record. Like I mentioned earlier, the H4n I bought was so noisy for ambient recordings, I just couldn't use it.

48/24, that's what I'd use for film work.
 
Rick Burnett said:
What gear?

http://www.avisoft.com/recordertests.htm

NONE of these has a noise floor worse than -96dB. I think you might be thinking of MUCH older gear.
Click to expand...

Yes Rick, I've read the Avisoft tests, but I've never been able to reproduce anything even close to the noise floor levels that Avisoft states.

...The simple hands on audio noise floor test that I do when I buy a new camera or audio recorder is as follows. This test does not take into account any mic characteristics ( no mics are used at all ), so it's an optimal conditions test that will likely never exist in the field. ( all locations are too noisy for threshold noise-floor tests, and many mics will have self-noise that is higher than the recorder's noise floor )

1- Using a professional field mixer ( I use a Sound Devices MixPre ) that has a 100+ dB dynamic range or better ( MixPre is rated as 110 dB ), connect your mixer to your camera or recorder using a balanced XLR cable. ( either Mic or Line level )
2- Turn on the mixer's 1 kHz test tone, and set the recording level of your camera or recorder to your normal recording level. ( most people use -12 dB for prosumer gear )
3- Record 10 seconds of test tone, then switch off the test tone and continue to record for another 10 seconds.
4- Play the audio that you just recorded in to your audio editor and measure the audio level of the test tone ( it should be -12 dB or very close to it )
5- Next see where your audio meters drop to when you reach the part where you shut off the mixer's test tone. You may have audio tools to more accurately measure the level of this "silent" noise, but even the audio level meters of your editor will give you a good idea of what the audio level is.

For all practical purposes, the "silent" noise level you observe with this test is the noise-floor of your audio recording device. I always use Line level audio for these tests, as that is usually what I record with my cameras or audio recorders.

It is surprising how well some audio recorders do with this simple test, and the results are not always related to the price of the gear.
 
That's definitely true about the price of gear. We recorded to the FS100 two days ago with a Rode NTG-2 and I gotta say, it sounded fantastic. I used that same mic with an AF100 and the AF100 converters/preamps were pretty noisy. Test test test, that's the best thing you can do. Find the limit of the equipment you want to use.

I still recommend the Zoom recorders for people on a budget that just need dialog, they work great for that. But I tell them, the noise floor is pretty high in comparison to other gear not that much more. It might not have all the bells and whistles, but when we recorded to the FS100 and the Zoom, the difference was VERY clear.
 
You must log in or register to reply here.
Back
Top