Sample rate for music preservation

[David Marker]

I'm going to be heading off to Italy later this month to make field recordings of Southern Italian folk musicians who play rare types of musical instruments and musical styles in an effort to preserve this music as much as possible, as these traditions are threatened by the onslaught of modernization and homogenization. Currently I don't know exactly what I'm going to be using the recordings for - most likely they will end up on CD, but there is always the possibility that I could use them in film as I am a filmmaker. But most important is that they are of high quality and are "futureproof", as I view this type of recording as a sort of intangible cultural artifact.

I will be making the field recordings on a Sound Devices 702T and am using a stereo pair of Audio Technica 4051a condenser mics. I recognize that the mics are the weaker link of the set up but they are the best I have now and I am familiar with them. They have a frequency range of 20-20,000.

I'm trying to figure out what the best sample rate to record should be (at 24 bit of course). I was originally thinking I would go with 88.2 or 96, but today I was told by several professionals that there is absolutely no need to use anything above 48 because my mics wouldn't even be good enough to get frequencies above 20k. Also it would be harder to work with the larger audio files in post.

However, I just want to make sure that given my hardware that I am using the best digital acquisition rate to preserve this rare and difficult to acquire music.

Any expertise would be nice. I'm not looking to start a sample rate debate or get some long mathematical theoretical response on frequency rates etc., but just some real world advice on what I should use. These recordings are for posterity and preservation, but also usability. Thanks in advance.

[Perrone Ford]

Originally Posted by David Marker

I'm trying to figure out what the best sample rate to record should be (at 24 bit of course). I was originally thinking I would go with 88.2 or 96, but today I was told by several professionals that there is absolutely no need to use anything above 48 because my mics wouldn't even be good enough to get frequencies above 20k.

but just some real world advice on what I should use. These recordings are for posterity and preservation, but also usability. Thanks in advance.

Hollywood records feature films at 48k. They have budgets in the tens of millions of dollars. If they don't really see the need to go above 48k, do you really think you need to?

[Bob Hart]

I think your important task is to also make a visual recording. 48K fits more closely that choice if you make it.

If the craft of playing these instruments is rare, facing extinction, then the techniques associated with a particular sound, flourish or effect are important for anyone in future who seeks to recreate and master these instruments.

Otherwise, future performers who rely on emulating a particular effect may never be able to actually get to it without the visual cues.

It would be nicer to have the even crisper fidelilty but a more valuable record is to preserve the visual "how to" information as well so that the genuine sound to the human ear can be reproduced "live" in future.

[Jim Andrada]

Well, if there's no particular reason NOT to record at 96k, I'd do it - in fact, since all you're talking about is disk space why not go for broke and use the 192k the recorder is capable of:)?

Lately the audiophile crowd seems to be wanting gear that will exceed 20kHz. I know that Schoeps is offering an enhanced mic body amp that they rate to 40kHz+. There seems to be some feeling lately that even though it's generally accepted that people can't hear above about 20kHz in the normal sense of hearing, they can perceive "something" at higher frequencies. True or false I know not.

48k is probably fine for all practical purposes but at the price of CF cards, I'd go for 96k and feel more righteous about it.

Enjoy Italy! Wish I were going.

[Rick Reineke]

24 bit, Absolutely. As for sample rate, it has been said, that going from a 88.2kHz sample rate, down-sampled to 44.1 for CD release, yields better quality. If video is involved go with 96KHz. (half 48K)

[Jimmy Tuffrey]

I'd say the music you archive is ultimately more important than the sampling rate.
Workflow becomes the next question.
Whatever enables you to work efficiently and quick enough to enjoy the experience and thus repeat it is most important.
If you want to go at a high rate to start with, 88.2khz, and you find it gives you too much work in file management then reduce it to 44.1khz.

You want an easy way to review your recordings and edit them as soon as possible after recoding. Or during perhaps. Keeping a trimmed down catalogue of field recordings will help immensely.

[Rick Reineke]

As Jimmy said..... and I failed to mention... It's the music that matters', your re'pore with the players and make them feel comfortable around you to get a great performance. There are no recording schools that can teach that.

[Jeff Kellam]

Originally Posted by Perrone Ford

Hollywood records feature films at 48k. They have budgets in the tens of millions of dollars. If they don't really see the need to go above 48k, do you really think you need to?

Lots of movies have soundtracks at 24B 192KHz nowdays, although most are 24B 48KHz, with everything in between.



Unofficial Blu-ray Audio and Video Specifications Thread - AVS Forum

[David Marker]

Thanks for all the responses.

To respond to Bob - I agree 100% with what you are saying about capturing visual. In 2008 I recorded a feature length documentary about Southern Italian folk music that has been in post production for over a year. It's essentially done now and I'm submitting to festivals. One of the last things I'm dealing with in post right now is mastering the audio to remove some of the high end clipping that I picked up on some of the tracks. Because I shot the film myself - running and gunning like a mad man - I recorded the audio directly to my XHA1 and thus was limited to 16bit 48htz and the Cannon's crappy pre amps and internal processing and lack of limiter.

As much as I'd love to bring the camera back this year to film more, it just creates way too much work for me to then feel obligated to do something meaningful with the video. I've put countless hours of work into this documentary.

Anyway, now I'm just focused on going back and doing really good audio recordings, still photos, and detailed field notes about my experience.

My entire recording rig (including a hack-sawed mic stand) breaks down and fits into a Kata backpack. I will not be bringing a laptop so all of my editing will take place back in the US after I return. Thus the bit rate I record at in Italy won't effect my work other than taking up more space on my CF cards.

I'm leaning towards 88 or 96 and I don't really know how to decide between those two as well.

I have a question. Does a high bit rate only extend the dynamic range of what is captured - ie. beyond the normal 20-20k, or does it also just generally add more texture to the normal sounds found within the normal hearing range? If increased bit rate does the latter than I think I'll go with a high bit rate. But I don't really understand exactly what the real world implication of a high bit rate is over a lower one...

[Jim Andrada]

I think the question is whether people hear (or in some other way) perceive anything usefull above about 20kHz. Some insist no, others (see below link) say yes.

http://www.earthworksaudio.com/tech/...yond_20khz.pdf

A lot of people (particularly audiophile type people and the high end gear makers who pander to them) these days feel that CD sampling rates are insufficient and produce a somewhat "cold" sound.

The other question is how effectively the recording system in total limits the peak frequency.

For a sampling frequency of 44.1kHz, any frequency above about 20kHz that sneaks through, whether from the original sound or from high frequency noise in the system will be reflected as noise (aliasing) at a frequency lower than 20kHz (Ref "Nyquist limit")

Anyhow, as a huge over simplification, the higher the sampling rate, the lower the likeleyhood of adding "shmutz" to the desired material.

Of course, if you think that people DO perceive something over 20kHZ and you spend a bunch of money for a system that will record/reconstruct such signals, then again, higher sampling rates are essential.

Another factoid is that so called Audio DVD can support up to a 192kHz rate for 2,0 stereo and up to 96 kHz for 2,1 stereo.

So what does all this tell us? Basically that higher sampling rates are better than lower sampling rates, and some (small percentage of) recorded material is sampled at up to even 192 kHz.

If it were me I'd sample at 96kHz and feel good about it, although not much difference between 96k and 88.2k.

[Jimmy Tuffrey]

I generally stick to the rule that if it's for music then 44.1 or 88.2. I f it's for video then 48 khz. What is the point of recording music at 96khz an then having the to convert to 44.1 for music release or cd audio ? The quality will suffer compared to halving the sample rate from 88.2 to 44.1. It makes no sense to me to record music at 96 khz.

Also if you are making stereo or mono recordings, then what is the need to be in the high end domain? If it was a multi track project with various levels of mixing post production then that would make more sense.
If your recordings are good in the field as they should be then they should need only editing which can be done in the final format.
So I would say make that final format choice and record in it. Saves a lot of work.

My experience is that saving work is paramount as the post process always takes a long time. When you have just recorded something you have immersed your self in it and can make decisions. Back at home you have to spend all that listening time again and then you will not be so immersed. I've learnt to try very hard to limit the amount I record and focus on quality of performance and not too many choices of take. It's surprising how much less fun it is a few months later. One has moved on. Also it might in retrospect not be music that you are wanting to listen to again.

[Jeff Kellam]

David:

A lot of these guys are smarter and more experienced than me on audio, but I say that anything worth doing, is worth doing right. For me, that means at least capturing within the upper limits of the current standards. I can always manage to mess it up later in post.

For bit depth, it's an easy choice. For PCM recording where the frequency range extends to the Nyquist limit (when it dosent, i.e. oversampling, DR is even greater), the dynamic range in decibels is equal to 1.76 + 6.02 X bits, or about 6 dB per bit. So, 96 dB for 16-bit CD audio and 144dB for 24-bit.

Looking at the resolution of 16 and 24-bit, the resolution of 16-bit is 65,536 and 24 bit has a resolution of 16,777,216. 24-bit is just massively superior to 16-bit.

As far as sampling rate, thinking about the rate being based on the nyquist limit (half the upper band limit) completely overlooks the whole area of oversampling and digital signal processing. Oversampling (not 44.1 upsampling, that is a different subject) is mandatory for any high end audio as it allows analog anti-aliasing filters to achieve high resolution A/D conversion which would be otherwise very difficult to implement with the sharp cutoff necessary to maximize use of the available bandwidth without exceeding the Nyquist limit. The point of 96 & 192K sample rates is not totally highfrequency band capture, but oversampling which ultimately results in a far better DAC in playback.

As a bonus with 96K and up sampling, we do get the HF captured as far as the mics allow it. With HF, like we all know, most of us cant hear much above 18 kHz. However, most acoustic instruments produce usable output up to around the 30 kHz mark and the ultrasonic content of multiple instruments blend together to produce audible beat frequencies which contribute to the overall timbre of the sound. If you record a string section at a distance with a stereo pair the interactions will have taken place on the "sound stage" of the venue before your microphones capture the sound. You can record this audio with 48K sampling and not lose the HF interaction as it has already happened as it's captured.

However, if you recorded the same string section with a mic on each instrument to a separate channel into your mixer, the close-mic technique does not pick up any interactions. The HF interactions that produce lower beat frequencies will have to happen in the mixer or post. But that could never happen with a 48K sample rate capture as the HF content is not there. So, high sampling rates allow the flexibility of using different mic techniques with better results.


Also, some research suggests that the human brain can discern a difference in a sound's arrival time between the two ears of less than 15 microseconds which corresponds with the time between samples at 96 kHz sampling. So higher sample rates certainly increase stereo and surround imaging accuracy.

I would reccomend that the oversampling rate be an even ratio to the lower preferred delivery format for post capture editing downmixes to low-fi delivery methods. That is always mathematically more accurate and keeps the digital filtering of the dithering errors to a minimum.

I think any hi-fi capture, particularly music, which is inherently complex sound, would use 24-bit 88.2 or 96K at a mimimum.

[David Marker]

Thanks for the responses. I'm probably going to go with 88.2. It might be overkill for some but I want to cover my bases.

I'm still not sure if this question has been answered yet for me:
Does a higher sample rate do anything other than expand the frequency range that is obtainable by the recorder? Other than giving you more bandwidth for analog anti-aliasing filters (which is way over my head...), does a higher bit rate also give you a "better" or more "textured" quality to your sound - and I'm talking about mics here that only have a frequency range of 20-20k????? - If a higher sample rate only enables you to capture sound at a higher frequency rate, than with my AT mics that top out at 20k, wouldn't it be pointless? But if however a higher sample rate also gives you "better" sound, even within the confines of a frequency rate that my mics produce than that would be a completely diff. story - RIGHT?

Could somebody explain this to me as if I were a 5th grader? I feel like a moron, but this Nyquist limit stuff is a little too technical for me to understand.

[Steve Oakley]

whats good for dialog recording and whats good for archival music recording are two different things. 48K is good enough for dialog on most voices....

as for music, the higher sample rate simply records more detail in everything. its not so much just to capture the HF you can't hear. even at 10khz, you'd rather have 9 or so samples at 96khz vs 5ish at 48k for a waveform. just live with the fuzzy math, you get the idea. FWIW, many major audio studios are recoding at 192khz because "it sounds just like tape" its got that level of detail.

[Jon Fairhurst]

Unless you are doing audiophile work, don't lose sleep over the sampling rate. Sure, if you will deliver in 48kHz, and can record and process everything in 48 kHz, do that, simply to avoid an unnecessary conversion.

In the real world, film composers use samplers. Even when they can hire and orchestra, they ofter layer it with samples. And all of the major sample platforms work at 44.1kHz. They convert to 48kHz, and nobody has ever detected the conversion with their ears.

Some perfectionists use a high end 44.1kHz D/A and connect it to a high end 48kHz A/D, rather than do digital conversion. Some claim that it sounds better, but if you ask me, it's really on the margins. If you have access to that equipment and have the time, sure, why not? Just don't expect the audience to notice.

I have two workflows... As a composer, I do the music at 44.1kHz, since that's what my sampler and libraries use. When the mix is done, I upconvert the music to 48kHz, if the dialog is in 48kHz. Otherwise, I keep everything at 44.1 kHz and upconvert as the very last step.

It's not a matter of choosing one rate over another. It's simply a matter of minimizing the number of conversions.

On the topic of higher sample rates, I avoid them, unless I know that I will slow the audio for a special effect. Regarding bit depth, I work at 24-bits. I record at 24-bits if the equipment chain actually provides a 24-bit result. I dither down to 16-bits as the final step - immediately after any 48kHz conversion as needed.