Topic: Question about 24-bit recording with an R-09

[dorrcoq]

I don't see the point of doing 24 bit on this recorder.  The specs of the chip inside say the max SNR is 92dB.  That means, the noise floor can be no lower than 16 bits (-96dB).

I verified this.  I put a 1k resistor load on each channel of a miniplug, plugged into line in, and recorded in 24 bit.  If this was a perfect recorder, there should be zero bits of noise.  In fact, the lower 9 bits were noise.  So, go ahead and record in 24 bit, but you'll just be recording 8 bits of extra noise.

AFAIK top notch ADCs will get maybe 18 or 20 bits maximum.  So, 15-16 bits is not bad.  It just doesn't warrant recording in 24bit.

  Richard

OK, if you record at 24 bit and only "actually" get 15 or 16, what do you get if you record at 16 bit?  Surely not 16, then.  So isn't 24 bit still higher quality even if it isn't truly 24 bit?

[illconditioned]

I don't see the point of doing 24 bit on this recorder.  The specs of the chip inside say the max SNR is 92dB.  That means, the noise floor can be no lower than 16 bits (-96dB).

I verified this.  I put a 1k resistor load on each channel of a miniplug, plugged into line in, and recorded in 24 bit.  If this was a perfect recorder, there should be zero bits of noise.  In fact, the lower 9 bits were noise.  So, go ahead and record in 24 bit, but you'll just be recording 8 bits of extra noise.

AFAIK top notch ADCs will get maybe 18 or 20 bits maximum.  So, 15-16 bits is not bad.  It just doesn't warrant recording in 24bit.

  Richard

OK, if you record at 24 bit and only "actually" get 15 or 16, what do you get if you record at 16 bit?  Surely not 16, then.  So isn't 24 bit still higher quality even if it isn't truly 24 bit?

If you record in 16 bit, you get 15 (or 16) bits.

My point is you get the same, regardless of what you record in, so no benefit from keeping 24 bits...

The best analogy is that of reading a meter, like a volt meter or a digital scale.  The measuring instrument has a limited accuracy, no matter how many "digits" or "decimal places" you care to read out.  That is what SNR (signal to noise ratio) says.  How accurate you can measure the signal (music) relative to the noise.

  Richard

[illconditioned]

Is this really true though, at least in theory?

Here your "true" signal would be variations in air pressure right at your microphone capsule.  The microphone will add some noise onto this signal.  The preamp will add some noise onto that.  The analog stage of the ADC, more noise.  And then quantization to 16 or 24 bit more noise onto that.  So (while the difference may well be trivially/inaudibly small compared to all the noise already added), quantizing to 24 bits will still result in less cumulative noise than to 16 bits.

I read a "rule of thumb" somewhere, I believe in regards to preamp selection, that you want to have 10 db better signal to noise ratio in say preamp vs. microphone for the later element in the chain to not add significantly to the noise floor.  If you accept that rule of thumb, that would seem to argue that there can be some benefit to using more than 16 bits to record an incoming signal with a 90 dB SNR (although whether most of us get signals nearly this clean coming out of the mics is another matter).

OK, there *may* be something in your argument.  You're saying that the signal quantization noise should be 10dB less than the ADC noise (which is -92dB).  In that case, we should record with more than 16 bits.  Hmm.  I'll have to think about this...

  Richard

[Will_S]

Is this really true though, at least in theory?

Here your "true" signal would be variations in air pressure right at your microphone capsule.  The microphone will add some noise onto this signal.  The preamp will add some noise onto that.  The analog stage of the ADC, more noise.  And then quantization to 16 or 24 bit more noise onto that.  So (while the difference may well be trivially/inaudibly small compared to all the noise already added), quantizing to 24 bits will still result in less cumulative noise than to 16 bits.

I read a "rule of thumb" somewhere, I believe in regards to preamp selection, that you want to have 10 db better signal to noise ratio in say preamp vs. microphone for the later element in the chain to not add significantly to the noise floor.  If you accept that rule of thumb, that would seem to argue that there can be some benefit to using more than 16 bits to record an incoming signal with a 90 dB SNR (although whether most of us get signals nearly this clean coming out of the mics is another matter).

OK, there *may* be something in your argument.  You're saying that the signal quantization noise should be 10dB less than the ADC noise (which is -92dB).  In that case, we should record with more than 16 bits.  Hmm.  I'll have to think about this...

  Richard

Heh, I actually thought about this a bit more and deleted my original post because I convinced myself it didn't make sense in light of your results actually looking at the bits when you recorded no input.  That maybe the same rule of thumb doesn't/can't apply to the actual A/D conversion, where it comes down to (or not, depending on dither?) a simple yes/no for the last bit.

But at any rate, to me it makes sense to record in 24 bit as long as I'm not worried about filling up my SD card before the show is over, because it leads to fewer worries if I decide to do some post-processing later, as I'm not worried about the impacts of multiple generations of "rounding error" when applying multiple edits to a 24 bit file.  But maybe it would be just as well to record at 16 bit and then convert to 24 bit if I wanted to edit it...

[Gutbucket]

I don't see the point of doing 24 bit on this recorder.  The specs of the chip inside say the max SNR is 92dB.  That means, the noise floor can be no lower than 16 bits (-96dB).

I verified this.  I put a 1k resistor load on each channel of a miniplug, plugged into line in, and recorded in 24 bit.  If this was a perfect recorder, there should be zero bits of noise.  In fact, the lower 9 bits were noise.  So, go ahead and record in 24 bit, but you'll just be recording 8 bits of extra noise.

...

Considering that dithering a signal works by adding a form of low level noise before truncating the bottom 8 bits (essentially allowing us to hear details beneath the noise floor that dissolve into the dither noise instead of into quantization artifacts), would the noise you measured in the bottom 9 bits of the R-09's 24bit recorded signal act as a form of dither allowing better than 16 bit performance?

[Brian Skalinder]

But at any rate, to me it makes sense to record in 24 bit as long as I'm not worried about filling up my SD card before the show is over, because it leads to fewer worries if I decide to do some post-processing later, as I'm not worried about the impacts of multiple generations of "rounding error" when applying multiple edits to a 24 bit file.  But maybe it would be just as well to record at 16 bit and then convert to 24 bit if I wanted to edit it...

FWIW, most audio editors perform edits at 24-bit or 32bfp, or are configurable to make edits at 16-bit, 24-bit, or 32bfp.  Odds are your software already makes edits at 24-bit or 32bfp.  Probably worth checking how your s/w is configured, though, as I know some editors - Audition/CEP, for example - allow configuration to always internally dither back to 16-bit after every edit.  Obviously, once we make our first edit, we want to remain in the 24-bit or 32bfp realm until we're done with edits, and only then dither down to our target format.  So just check your editing s/w to make sure it's configured to edit at 24-bit or 32bfp.  No point in "converting" 16-bit to 24-bit, as the s/w should do that for you when making edits.

[Will_S]

But at any rate, to me it makes sense to record in 24 bit as long as I'm not worried about filling up my SD card before the show is over, because it leads to fewer worries if I decide to do some post-processing later, as I'm not worried about the impacts of multiple generations of "rounding error" when applying multiple edits to a 24 bit file.  But maybe it would be just as well to record at 16 bit and then convert to 24 bit if I wanted to edit it...

FWIW, most audio editors perform edits at 24-bit or 32bfp, or are configurable to make edits at 16-bit, 24-bit, or 32bfp.  Odds are your software already makes edits at 24-bit or 32bfp.  Probably worth checking how your s/w is configured, though, as I know some editors - Audition/CEP, for example - allow configuration to always internally dither back to 16-bit after every edit.  Obviously, once we make our first edit, we want to remain in the 24-bit or 32bfp realm until we're done with edits, and only then dither down to our target format.  So just check your editing s/w to make sure it's configured to edit at 24-bit or 32bfp.  No point in "converting" 16-bit to 24-bit, as the s/w should do that for you when making edits.

Yeah, I use Sound Studio which does internal processing at 32 bits.  But if I want to make a few edits, save, and return to the file to edit some more later, it will dither back down to 16 bits for the save unless I tell it not to.  But none of that prevents starting with a 16 bit file and being sure to save it at 24 bits+ if I plan multiple editing sessions.

[Brian Skalinder]

Yeah, I use Sound Studio which does internal processing at 32 bits.  But if I want to make a few edits, save, and return to the file to edit some more later, it will dither back down to 16 bits for the save unless I tell it not to.  But none of that prevents starting with a 16 bit file and being sure to save it at 24 bits+ if I plan multiple editing sessions.

So...just tell it to save in 32bfp.    And saving a 16-bit file at 24-bit won't matter, since the editings at 32bfp.  You'd want to save the 16-bit file as 32bfp.  (Whoops, just saw your + sign...so...you're already there, but I'll leave my comment for newbies in case it helps.)

[BayTaynt3d]

Doesn't 24-bit bit allow for more headroom during recording?

[Will_S]

Doesn't 24-bit bit allow for more headroom during recording?

Not if the last 8-9 bits are noise regardless of input level, which Richard's results suggest may be the case.  (Now, whether the last couple bits might be jointly determined by signal and noise....)

[calvinroots]

sorry for comming in late and maybe sounding dumb...I am to some of this stuff, but trying to learn.  So correct me if I am wrong but the best possible recording with the R-09 is to run 16/48 ...does that make sense if the 24 bit is basically worthless.  And oh  yeah...why do they bother having the "24"function if it isn't even 24 ...just to advertise 24??? that ain't cool!

THANKS

[Will_S]

Actually, I've been thinking about this a bit more.

The reasoning that 16 bit = 96 dB dynamic range and so using more than 16 bits for anything with a SNR less than 96 bits is maybe a bit misleading or oversimplifying.

One, the noise floor of an actual recording is a complex function of frequency, and different shapes of noise (that may still integrate to something around-96 dB) will be audible to different degrees.  Digital quantization noise may be be more objectionable than analog noise that is summarized by the same single number.

Two, I think it's pretty well established that dithered audio sounds better than undithered.  Dithered 16 bit audio has a signal to noise ratio of about 91 dB (depending on the exact form of dither).  So maybe the noise of the R09's A>D is just a bit (hah!) below the 16 bit noise floor after all.

Also, going back to my earlier post, I do think all noise is additive, be it analog or quantization noise.  So it seems like 24 bit could be quieter than 16 bit even with an input that is technically noisier than the theoretical limit for 16 bit.

[Gutbucket]

Most 24bit gear is limited by their analog stages or the gear proceeding them in the chain and cannot fully utilise all 24 bits.  In this case there may be only a few bits beyond 16 that are usefull, but even a few bits are alot and when recording and unsure of where the levels will peak their main benefit is headroom.  Not long ago 18 and 20 bit converters were the bees knees. 

I believe there are benefits of running the R-09 in 24bit mode, even if there are only just a couple usefull bits past 16.  What you must weigh that benefit against is the increased storage size for the larger files and the increased workload of dithering the 24 bit down to 16 bit for CD playback.  Potential quality vs size & convenience.

I haven't done compairson listening tests or a bit analysis of the files, so I'm relying on the reports of others that there may be a few usefull bits past 16.

For me the storage space isn't a big deal and I playback the 24bit files through the computer so the decision is an easy one.  I have a very critical ear but I know that I will impove the resolution and ability of my playback system in the future.  There may be things in my recordings that I don't take notice of now that I will be able to with better playback.  I look as it as an insurace policy with out much cost.

Will_S posted while I was typing & I agree with his thoughts.  I'm certainly no authority on this.

[calvinroots]

thanks for more info...I am all about QUALITY --- Space, time, anything...doesn't matter.  You only get one chane to make a recording (live at least!) so I say do all you can to make it the best you can with what you got! 

>>>my thoughts - calvinroots