Jump to content

Sampling rates


Tony Johnson

Recommended Posts

Higher sample rates help you to better capture high frequencies both audible and inaudible. the more true representation of transients, harmonics, etc you can capture, the more interesting things you can do.

Higher sample rates help you capture higher frequencies, but do not increase accuracy of frequencies below half of the sample rate.

Link to comment
Share on other sites

  • Replies 106
  • Created
  • Last Reply

Top Posters In This Topic

The ear is very sensitive to directional information conveyed in very short wavelengths; consider the tiny effects that folds in your outer ear have on an approaching sound... we learn a lot about directionality from those microsecond differences.

 

HOWEVER, we're talking about film tracks, folks. Dialog is one mic (or a number of mics blended into a single signal), played on a center channel in a theater. Directionality isn't an issue.

 

Music and sfx are, at most, four other channels*. The directionality in those sources is usually created in post, by operations that take place in the audible range.

 

So what difference could the directional information in tiny phase differences have to do with the files we're recording or sending to post? Until we're recording binaural dialog with artificial ears -- and playing it to the audience through wideband headphones -- it's not relevant to the process.

 

--

* I'm not counting the LFE track in a discussion about ultra-high sample rates.

 

subsonic and supersonic frequencies interact with the audible range of frequencies

 

 

 

Okay, got a problem with this one. Yes, supersonic frequencies will interact with audible ones. But any interaction is in the supersonic range. Again... it's not relevant to our work, or to any other media sound. 

 

Even the extreme case of two separate supersonic sources being close enough together that they create an audible beat, the beat would be reproduced by normal s/r. Otherwise it wouldn't be an audible beat, would it?

 

And subsonic isn't even part of this discussion. Unless you're asserting that you need 96 kHz s/r to carry a 10 Hz signal.

Link to comment
Share on other sites

You don't record room tones? Or try to capture sounds other than dialog on set?

It might not benefit YOUR work, but it sure benefits others. Sound for picture is a lot more than just what you do on set. Overall, 96 or higher is beneficial for the ENTIRE sound for picture workflow.

It doesn't have anything to do with binural sound. It has everything to do with sound quality and creating or recreating realistic seamless sound.

Link to comment
Share on other sites

 

48khz is a standard, it is also a great balance in between DSP power requirements, equipment design and hard drive space/transfer speed.

 

My two cents.

 

I agree here. In particular because it's difficult to have a complete set of gear that runs at a higher sample rate/doesn't roll off the higher frequency response, and you're only as good as your weakest link.

My two cents.

Link to comment
Share on other sites

Also the interaction is NOT only in the inaudible range and has many benefits in all kinds of media.

Your argument is very narrow minded, and I'm not sure you understand these concepts.

Maybe take a class on psychoacoustics.

 

Er... Ah... Jay Rose has not only studied the science of sound extensively, but is considered by many to be somewhat of an authority on it.  He has not only been widely published but has also conducted highly regarded classes and seminars.

 

His post production credits include an extensive variety of nationally broadcast television programs and sound design for feature films.

 

He is considered by many of his high profile clients to be somewhat of an audio guru. 

Link to comment
Share on other sites

Additionally, you may also wish to peruse his books: http://dplay.com/book/

 

Mr. Rose knows what he's talking about.

 

Now, as a previous poster mentioned, if you can provide a link to the source of your argument about the supersonic interactions, I'd certainly be interested in checking it out. If the science is there, then I'd be willing to change my mind on the subject.

Link to comment
Share on other sites

Any two or more frequencies can and will interact with each other and possibly create a new frequency, the "beat frequency" as Jay pointed out above. This may happen across the entire spectrum, but it is only relevant to us, if the resulting "beat frequency" falls in the audible range. If and when it does 48kHz can easily sample th result of this effect. Once the signal has been digitized there will be no more interaction of frequencies. Only once this is played back could there be new beat frequencies, created by the room - if any.

So for this phenomenon, as Jay says, we don't need more than 48kHz.

Actually, since we don't want random interaction of frequencies and beat frequencies, this is an argument for lower sample rates

Link to comment
Share on other sites

Here are 2 articles I could quickly find. One is a scholarly peer reviewed article, one is wikipedia (but still good)

http://scitation.aip.org/content/asa/journal/jasa/59/3/10.1121/1.380913

http://en.m.wikipedia.org/wiki/Sound_localization

Its also important to remember that sound in the real world is made of complex waves with a lot more character than true sine waves. The lower the sample rate the less of this character you capture.

I don't know what Jay's qualifications are but I have degrees in acoustics, computer science, and telecommunications. All with concentrations in audio. I have been trained by, and continue to be trained by the best engineers regionally, notionally, and internationally. My favorite being the man that wrote this book I think everyone should read.

http://www.amazon.com/Technical-fundamentals-audio-Ted-Uzzle/dp/0872887014

Physics and psychoacoustics are very complicated things and these discussions don't translate to internet forums and large groups of people well.

Talking about the interaction of sound waves both audible and inaudible is a physics discission. One I'm not prepared to take part of from my phone.

Link to comment
Share on other sites

Music and sfx are, at most, four other channels*. The directionality in those sources is usually created in post, by operations that take place in the audible range.

Not when it comes to stereo, 5.1, 7.1 sources recorded in the field. Both during production and post.

Is there anything on a dialog track we need that 48 doesn't cover? No probably not, but a sound track is not only made up of dialog. It's also made up of music, effects, Foley, etc... All things that can benefit from higher sampling rates. More definition = cleaner/truer sound that is not only easier to manipulate, but also has more manipulation possibilities.

You need to think of the big picture. The job of the location sound engineer is to gather as many resources as possible for the post guys to do their job. And their job is better done at 96 or 192 or etc...

There are reasons why companies like Neve, focusrite, ssl, API spent time and money researching and building audio tools that had higher bandwidth than that of human hearing. Those frequencies make a very clear and audible difference in frequencies we can hear.

Link to comment
Share on other sites

Here are 2 articles I could quickly find. One is a scholarly peer reviewed article, one is wikipedia (but still good)

http://scitation.aip.org/content/asa/journal/jasa/59/3/10.1121/1.380913

http://en.m.wikipedia.org/wiki/Sound_localization

Its also important to remember that sound in the real world is made of complex waves with a lot more character than true sine waves. The lower the sample rate the less of this character you capture.

I don't know what Jay's qualifications are but I have degrees in acoustics, computer science, and telecommunications. All with concentrations in audio. I have been trained by, and continue to be trained by the best engineers regionally, notionally, and internationally. My favorite being the man that wrote this book I think everyone should read.

http://www.amazon.com/Technical-fundamentals-audio-Ted-Uzzle/dp/0872887014

Physics and psychoacoustics are very complicated things and these discussions don't translate to internet forums and large groups of people well.

Talking about the interaction of sound waves both audible and inaudible is a physics discission. One I'm not prepared to take part of from my phone.

 

First one seems to be behind a paywall, is it available elsewhere? In sympathy with your argument I would point out that there isn't a great deal of new data it seems, the paper you posted was published in 76. 

 

"You need to think of the big picture. The job of the location sound engineer is to gather as many resources as possible for the post guys to do their job. And their job is better done at 96 or 192 or etc..."

 

This I don't agree with, I have never been asked for 96 or 192 and you would assume they would ask no? Are there loads of post houses working with full 96K chains? This is an honest question. 

 

Cheers

 

Nick

Link to comment
Share on other sites

You don't record room tones? Or try to capture sounds other than dialog on set?

 

Honestly, I don't. I haven't recorded anything on set since my back went bad, and I moved exclusively to post.

 

But I do expect the production folks to record tone, and I look for it when I'm editing dialog. (Many times it's not usable because the environment has changed during the course of a single setup, so I have to harvest from the bought takes... or synthesize using a convolution engine... but that's a subject for a different thread.)

 

Also: when I'm brought in as designer or supervisor before production -- which happens with some of the better producers around here -- I'll make special requests for wild effects and prop noises that the production team can record on set. Those help a lot.

 

But think about room tone: what do we use it for? (that was rhetorical, here's the answer:)

We cut it in with the dialog! It has to match the dialog exactly, and it gets played through the dialog channel. Back to a single speaker, and primary interest in the dialog range.  Usually ditto for pfx, which should sound like part of the characters' world.

 

96 or higher is beneficial for the ENTIRE sound for picture workflow. 

 

 

 

Only if it's going to be released in 96k, or if you're doing a DSP effect that'll attempt to stretch waves and will modify that ~40 kHz audio into something in the audible range.

 

AND it's really easy to upsample, if you need compatibility with wider bandwidth stuff in the final mix. 

 

AND we're in the digital domain. Subsequent generations don't have the kind of HF loss (or noise buildup) we had to worry about 20 years ago.

 

And this discussion was about production sound. 

 

And (and this is an esthetic decision, so feel free to disagree)... when I'm designing a track, I leave certain ranges basically clear so the voice formants and consonant noises can show through. Then I expect sfx and music to fill in other parts of the band.

 

But really, I don't care much about over 20 kHz sound... because very few people can hear it, and there's very little up there, and I know almost everything I work on is going to be radically lossy compressed -- Dolby Digital is notorious -- and I don't want to mess up the algorithm wasting bits on things that are less important. 

 

There are reasons why companies like Neve, focusrite, ssl, API spent time and money researching and building audio tools that had higher bandwidth than that of human hearing. 

 

I can think of a few:

 

  • Neve, focusrite, ssl, etc are in the studio hardware business. Their biggest market is music production, not film. I don't think anybody is using any of their gear on a set (except maybe Focusrite, who makes some nice USB<>audio boxes.) It's easier to make an argument for extreme HF, when some playback customers will actually be able to access it.
  • They are honestly committed to better sound in the abstract.
  • It gives them bragging rights, which translates to ads and eventual profits.

 

Have any listening tests been published that have demonstrated that the difference is audible? (Honest question)

 

 

 

 

I'd honestly be interested in that also. Back when we had a more scholarly discussion of sample rates in all media, on usenet, the best we could find was that there's a difference that couldn't be defined or measured, but could be reproduced... but only when testing two complete systems. When we subtracted the effects of the anti-aliasing and -imaging filters, using the same converters and limiting s/r with oversampled math in the digital domain, the difference disappeared.

 

 Physics and psychoacoustics are very complicated things and these discussions don't translate to internet forums and large groups of people well. 

 

That's worthy of a new thread, on the philosophy of the scientific method. I believe large forums of well educated professionals are a good way to explore complex subjects and arrive at something near the truth.

 

You may disagree.

Link to comment
Share on other sites

 Are there loads of post houses working with full 96K chains? This is an honest question. 

 

 

...worthy of a careful answer.

 

As far as I know, every competent post house can handle 96k s/r, and many can do 192. Even my modest setup.

 

But do we use that function? 

 

Usually not. It was supplied with the gear, because today's computers are remarkably fast and it's no big deal to double or quadruple the number of samples.

 

And the clients don't care.

 

And some of us designer-engineers don't care also, because we know how the track will be used. But see previous answer for a discussion of that.

 

---

FWIW, we do go the extra distance on bit depth. That makes an audible difference, particularly in processes with a lot of math. Even if the ultimate release is 16 bits, I'll work at 24 (and run processors and mix engines a lot deeper than that)... and then dither after I've mixed.

Link to comment
Share on other sites

Thanks Jay, and it's interesting to note the bit depth point. How do facilities handle monitoring? A quick look and Dynaudio and Genelec sites don't really show any interest in supersonic tweeters but it could just be that the specs are incomplete.

Link to comment
Share on other sites

This is a very interesting thread.

 

After reading some of the back and forth, I agree that the reason we don't use 96k or 192k in production is (and would only use it is if), as Mr. Rose brilliantly described:

 

Only if it's going to be released in 96k, or if you're doing a DSP effect that'll attempt to stretch waves and will modify that ~40 kHz audio into something in the audible range.

 

Even any of those sounds that you could record in higher sampling rates (and probably should), such as music and sound effects, will eventually end up at 48kHz sampling rate for playback, and so, as location sound recordists/sound mixers, is it really worth recording higher than that, when for the most part all we are recording is dialog, and when we are not, we are attempting as best as possible to match dialog (as in the case of roomtone and maybe even wild sound effects)?

 

I think at this point we start going into subjective arguments as opposed to objective ones. Hearing is a much subjective experience, as not everyone can hear all the same frequencies and at the same capacity.

 

This AES paper, posted by ncg, is interesting indeed:

Not that this proves the opposite but it's worth reading.

 

http://www.aes.org/e-lib/browse.cfm?elib=14195

 

This is not to dispute the argument that inaudible frequencies may affect those within audible range, but rather to make the argument that in production sound, the audible difference, IMHO, is so minuscule (perhaps even non-existent), that we need not concern ourselves with recording higher sampling rates for dialog, unless posts requests it for manipulation.

 

I also agree that a new thread on the physics and psychoacoustics would also be worthwhile.

Link to comment
Share on other sites

I should clarify that I'm talking big picture - from production to distro being at 96, otherwise, you are correct. It would be useless.

All the hardware I mentioned are used in post production, and would benefit from the higher sample rates. If post is at 96 or 192, production should be at 96 or 192.

Again I'm talking big picture for film and tv.

There are a lot of "extras" and transients, and harmonics, and etc... That make sound waves very complex.

It's that complexity that were used to hearing in the real world.

Lower sample rates fundamentally reduce complexity and definition.

Again ill mention I'm traveling and away from a computer. I have not carefully read the above posts and it's difficult for me to cite sources and find references using this method.

Link to comment
Share on other sites

CCalandro, thank you for the links. I'll peruse them further later. 

 

My point of interest in this thread was mainly if the post house's request of the OP to use higher sampling rates could not have been satisfied by merely importing 48k sampled files and oversampling in post.  This is why I asked the OP if he used a wireless or hardwired setup. The only way there would be any information above @ 22k is if he were using a hardwired boom, and then that would also depend upon the mic's response as well. If the OP was using a totally wireless setup, then everything would most likely be capped by the TX's sampling rate (assuming wireless systems which utilize digital signals from Zaxcom, Lectro, etc.) 

Link to comment
Share on other sites

cc: " a physics discission. "

Vin is a degree physicist...

and there is also higher math (Calculus) involved...

 

while there are mathematical and other technical arguments, what is ultimately relevant:(and in answer to NewEndian's ?)  listening tests of all sorts have not conclusively shown any "improvement" in the sound as perceived by human listeners.

Also, just as all digital audio is not created equal, not all 96kHz (or 192kHz) audio is created equally... like the HSR audio of a Zoom..?

 

sync: " The only way there would be any information above @ 22k is if he were using a hardwired boom,

it depends, but certainly not much useful information... and isn't that what this discussion is about..?

and then that would also depend upon the mic's response as well. "

that, too...

Link to comment
Share on other sites

Even if the advantages of higher sample rates may not be immediately obvious, what would be the downside? We could just record @96kHz couldn't we?

 

True. Though not necessarily a "downside" (depending on how you want to look at it), but 96k will definitely produce larger files. That is not to say that I don't disagree with recording at 96k, but currently, I still don't really see a use for it while the finished medium still plays at 48k.

 

 

Also I think most IMAX films are done in 96

I might be wrong about that though

 

 

Not really sure if this is the case. Would love to read specs about this.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...