Topic: Any drawbacks to recording in 24/96

[aaronji]

Even though there may well never be an advantage (re the future proofing example) where there to be an advantage or strong reason in the future, better to have the extra bits (imho) than to not.

It's really not possible for there to be some advantage in the future.  As long as you sampled at more than twice the highest frequency of your source, you can perfectly re-create the original signal...You can't improve on perfect.  Although if bionic ears become widespread, that might change the equation a bit!

To my previous point, that the bandwidth of the signal is already considerably limited before it even gets to the recorder, see the attached figure.  It's the frequency analysis of a 24/96 recording I made a few months ago (moe., Jam in the Dam).  The red line is 20 kHz.  There is nothing above 18 kHz or so.  This is pretty typical and is probably due to the PA.  Spot checking some other files, I haven't found any where there's signal past 23 kHz.

I recently read the Bob Katz book on mastering. One interesting thing Bob points out is that every ADC uses anti-aliasing and anti-imaging filters on the source material before conversion. These filters usually degrade the sound quality by adding (usually inaudible) distortion. With a 96kHz sampling rate, these errors are spread across a bigger bandwidth, with less errors ending up in the audible band. This might not matter much as these errors are usually inaudible. However, when doing post-processing, these errors can accumulate, making them audible eventually. This is highly theoretical, though. Other than that, Katz does not seem to see a reason for using sample rates higher than 48kHz.

It's my understanding that this problem has been greatly minimized by sigma-delta ADCs of the kind usually found in today's crop of recorders (even the cheaper ones, like the Zoom H2 have an oversampling sigma-delta)...

[George2]

[/quote]
It's my understanding that this problem has been greatly minimized by sigma-delta ADCs of the kind usually found in today's crop of recorders (even the cheaper ones, like the Zoom H2 have an oversampling sigma-delta)...
[/quote]

Hey, could you point me to a link about the Zoom H2 using Sigma-Delta?

[aaronji]

Hey, could you point me to a link about the Zoom H2 using Sigma-Delta?

I think I first saw that on the Zoom forum (http://www.2090.org/zoom/bbs/viewtopic.php?f=15&t=9745&sid=eb0abb9770ef451047defec8550562cb&start=17, the post from trevmar 7th from the top).  The manual for that chip states the sigma-delta'ness (http://www.ti.com/lit/ds/symlink/tlv320aic32.pdf)...I have also read that the M10 has a sigma-delta; I thought that was on taperssection, but I can't find it at the moment (although this post says the DAC is, which might imply that the ADC is too:  http://taperssection.com/index.php?topic=139638.msg1836579#msg1836579).  I was also told that the PMD620 and 661 have sigma-delta ADCs by a well known mod'er of those machines...

[aaronji]

^^^ Thanks, Jon!  Maybe I can put a few questions to you?  If the initial analog signal is already bandwidth limited to somewhere below the Nyquist frequency for a given sampling frequency, is aliasing even possible?  My understanding is that aliasing occurs when the signal contains frequencies between Nyquist and the sampling frequency.  If those frequencies don't exist in the sample, it would seem that aliasing is a non-issue.  If that is the case, would the quality of the filters matter anymore? 

The concerns you listed seem to relate to lesser quality chips.  Is the quality of these chips in typical taper's recorders, such as the Sony M10 or the R-05, poor enough that these are pertinent?  I assume the higher-end stuff, like the SD boxes, have no problems?  Also, is there a rule of thumb for interpreting the passband attenuation and stopband attenuation measures (cut-off values or a difference or something)?

Really, I am most interested in having good quality audio.  I am by no means wedded to 48 kHz sampling; from what I have read, the advantages of higher sampling rates are negligible in terms of the sound.  The reduced computing and storage demands tipped the balance for me thus far.  I would switch to 96 kHz in a second, though, if there was a real, audible benefit to doing so...

Interesting tip about upsampling for some of the processing, by the way...And thanks for all of your excellent posts on this topic!  I have learned a lot from you...

[TimSmith]

I dont use 24/96, only 24/48. Please correct me if I'm wrong (I'm a noob):
1) most people cant hear much higher than 20kgz
2) cheaper mics (like CA14) and cheaper playback systems have flat freq. response up to only 20kgz (maybe little higher, but not 48 anyway)

So I dont see any need for myself in taping in 24/96. Now I even want to start using 24/44. Less steps in post processing (usually make two filesets - 16 and 24 bit) and for video in mkv container I dont have to use 48, 44 is ok.

PS. Funny when someone uploads 24/96 MASTER recorded with $100 unit like Zoom H1. 8-) Pity that it cant record 24/192...

[aaronji]

Technically, yes, but an analog signal is not going to be naturally bandwidth-limited to the extent that it would need to be to avoid all aliasing (>-100dB).  By that, I mean all alias products must be below the noise level of the ADC.  That's a pretty strict criterion that is probably not relevant in concert recording, but that's what an engineer would aim for.

Pretty close, though.  I looked at the 24/96 files I have (an admittedly small sample of six, but fairly representative of the range of stuff I record).  Three were below -100 dB at 24 kHz, one was essentially -100 (-99.998), and the other two were close (-98 and -99.45).  Even at 22.05 kHz, one was below -100 and the other five were all below -92.  I interpret that to mean that, at least for my shows/venues/gear, 96 kHz won't result in any improvement, even if the ADC filters are poor.  Really, at 44.1, it seems like I wouldn't have any audible artifacts, given the levels of ambient noise are way above -92 dB (I assume that's what you meant about concert recording?)...

Thanks for your help, Jon!

[TimeBandit]

in most venues the PA doesn't go further then 20 khz so theres no sense with 96k.

for full acoustic, nature , atmo, 96k makes sense

[stevetoney]

I don't record in 24/96 because I record live music.  If I were recording in a studio when I could control everything, I'd go 24/96 or 24/192.

For live music, even with my Schoeps rig, I think it's overkill to record music with ultra high bit rates, when most of the time the sound is kinda crappy anyway and especially when you can't hear a difference between 24/48 and 24/96 in that situation anyway.  All this technical mumbo jumbo is great when you put on your gear head hat, but where the rubber meets the road for live recordings, IMHO it just doesn't matter. 

Another reason I don't use 24/96 is it takes twice as long for my software to load and process the files.  Not a big deal, unless you go to lots of festivals and have 30 or 40 shows to master and you don't want to spend twice the amount of time waiting for your software to process the files. 

Finally, if you ever want to share your music over the web, your files are double the size so it takes twice as long to upload/download them. 

[yates7592]

I don't see any of the "drawbacks" noted previously as real issues - SDHC memory, battery life, time - most of us have all of these in abundance.

My philosophy has always been to try and maximise the sound quality, and that to me, means once I have bought the best gear I can afford and positioned myself in the best spot (subjective I know), I then record at 24/96 (or sometimes 24/88.2).

I am puzzled about the screenshot in the earlier post showing nothing in a 24/96 recording above a fequency of 20kHz. Here's my last recording using B3 mics, a Church preamp and R-09HR. There is much data above a frequency of 22kHz and 24kHz, whether I can hear it is another question! This was from the front, so not much PA sound, mainly drums, amps and monitors.

[aaronji]

If you are at a large venue it's likely the audio has passed through a digital conversion somewhere which would explain the lack of ultrasonic content.

They weren't particularly large venues (the biggest has a maximum capacity of 1500; the others range from 300 to 1000), but I am positive that several had digital soundboards...

My philosophy has always been to try and maximise the sound quality, and that to me, means once I have bought the best gear I can afford and positioned myself in the best spot (subjective I know), I then record at 24/96 (or sometimes 24/88.2).

So what you are saying is that the ADC in the R09HR introduces audible artifacts at/below 48 kHz sampling during concert recordings?  Otherwise, what do you mean by "maximize the sound quality"? 

I am puzzled about the screenshot in the earlier post showing nothing in a 24/96 recording above a fequency of 20kHz. Here's my last recording using B3 mics, a Church preamp and R-09HR. There is much data above a frequency of 22kHz and 24kHz, whether I can hear it is another question! This was from the front, so not much PA sound, mainly drums, amps and monitors.

As Jon pointed out, it is likely that there was a digital conversion somewhere in the venue's sound system.  Up front, most of that ultrasonic energy is probably coming from the direct capture of cymbals and some other percussion, maybe some horns, and not from the amps/monitors...

[stevetoney]

I don't see any of the "drawbacks" noted previously as real issues - SDHC memory, battery life, time - most of us have all of these in abundance.

My philosophy has always been to try and maximise the sound quality, and that to me, means once

If you have time in abundance, perhaps you can come to my place and help me with my weed patch that doubles as a lawn. 

I don't disagree with your statement, assumging 24/96 provides increased sound quality over 24/48.  I do recall a long time ago a blind sound test was conducted and people couldn't reliably choose 24/96 over 24/48 for live music recordings, which is again why I think most people use 24/48. 

The point is, for live recording, there seems to be a point of diminishing returns.  To drive this home, I'm sure some people would use 24/384 or 24/768 (or whatever it would be) if it was available, but what difference does it make if you can't hear any difference?  You certainly wouldn't convince me that the 24/768 file is a higher quality sound than the 24/48 if I don't hear any difference on any playback systems.

[yates7592]

No, the high frequency patches you can see on the spectrogram (?) come mainly from the lead guitar, you can see how the high frequency (30-35kHz) concentration drops off in patches, that is where the guitarist stepped back and the horn player stepped in....so, I would guess that all that high frequency business is actually part of most, if not all, shows.

[DSatz]

yates7592, a good P.A. system might have a 12 kHz bandwidth. During a live performance there will almost certainly be some sound energy above that limit as the result of non-linear distortion. (Spurious sound energy is also produced at frequencies in the main part of the audible band, of course; it doesn't just "go up" in frequency.) Whether that distortion is beneficial or harmful is entirely a matter of personal opinion. It's part of the live sonic experience, and I could agree with anyone who feels that it should be captured in a recording. Also, I don't know what anyone could do about it if they considered it to be harmful; you can't "record around it."

The same issue came up in the 1980s when CDs were introduced to the public. Defenders of vinyl showed that the playback of vinyl LPs often includes sonic information above 20 kHz which the CD couldn't reproduce. However, when you consider the limits of studio microphones, tape recorders, disc lathes and the whole vinyl pressing and playback process, you realize that very little of that "information" was put there by the record company! Rather, it was mostly tracing distortion--a fact which the vinyl advocates didn't generally choose to mention.

Most home loudspeakers are doing well if they can reproduce 18-20 kHz cleanly. If you record signals beyond 20 kHz, playing them back through an amplifier and loudspeakers can cause audible artifacts. Again this may be experienced as subjectively pleasant, unpleasant or neutral--but since this can also happen with signals that are below the Nyquist limit for 44.1 kHz sampling, as far as this is concerned it might not even make much difference what sampling rate is chosen for the recording.

Some 96 kHz converters are technically and, potentially at least, audibly superior to some 44.1 kHz converters, while some 44.1 kHz converters are technically and, potentially at least, audibly superior to some other 96 kHz converters. There's no such thing as 44.1 kHz sampling having one certain level of quality and 96 kHz sampling having a certain other level of quality. There are only specific implementations that use those two sample rates. In many cases the bandpass filtering rather than the conversion (as such) has the greater effect on sound quality.

--best regards

[yates7592]

I understand that the benefit of recording at 96kHz / 192kHz and upwards is not proven, but my simple spectogram (based on a stage-side recording of "raw" music, not processed through a crappy PA) shows that some frequency content >24kHz is there (and it is instrument-specific).

As I previously stated, I'm not sure I can hear that sound, but for archiving purposes at least, if you have that capability to record, save to disc with acres to spare and playback, without any inconvenience or cost above recording at 24/48, then why not? I see at a festival you might be wise to step down, but for a one-off show, what do you have to lose?

To answer the original question, no, there are not drawbacks to recording in 24/96, only (possible/probable) benefits.

[vanark]

To answer the original question, no, there are not drawbacks to recording in 24/96, only (possible/probable) benefits.

This is your opinion.  Several of us have pointed out the drawbacks related to increased processing time, at the very least.  This may not be important to you, but to some of us, it is a large enough drawback to not record at 24/96.