Jump to content

Sampling rates


Tony Johnson

Recommended Posts

- Higher sample rates yield better results during manipulation. Elastic audio, noise reduction and other tasks performed by post every day yield fewer artifacts when recorded at 192/24. Post will often upsample before manipulation.

This above quote taken from Brian Liston post on "What specs really matter" an interesting read and makes the point re artifacts heard when sound recorded at 48/24 is manipulated in post which nowadays is a common occurrence.

On the hobbit we recorded at 96/24 for dial that was going to be manipulated after, ie Smaug/ necromancer and a few others.

Does anyone else record at 96khz or higher on a regular basis on films?? Will it become the norm soon.

Tony

Link to comment
Share on other sites

  • Replies 106
  • Created
  • Last Reply

Top Posters In This Topic

Yes it did especially as it went through manipulation, pitch change etc. What I am hearing from post is that when noise reduction notch filtering or RX is applied to dial, which in the case of The hobbit was often as there was a lot of unwanted Bg noise on set, then they can hear the artifacts applied when the dial is recorded at 48khz. The more processes the dial has to go through then the less 48 holds up.

Tony

Link to comment
Share on other sites

"On the hobbit we recorded at 96/24 for dial that was going to be manipulated after, ie Smaug/ necromancer and a few others.

Does anyone else record at 96khz or higher on a regular basis on films?? Will it become the norm soon.
Tony"
 
I think there are very few that record dialog at 96K except in those instances, as you mention, where there is going to be extensive manipulation. I would not consider use of eq, compression, noise reduction schemes typically employed on dialog tracks to be extensive manipulation. Though differences in "artifacts" may be discernible when comparing the same processes applied to 48K vs. 96K, I don't believe these differences are compelling enough to make 96K a new standard for production sound recording. I think all the current production recorders can do 96K, I just don't think there will be much call for it.
Link to comment
Share on other sites

Are the artifacts lower because the dialog was recorded at 96k? Or because the processing engine is running at a higher sample rate, meaning there's less interpolation between samples?

 

In other words, could you get the same improvement by recording 48k and up-sampling before you process?

 

Or if the issue is that up-sampling would just cause 2 or 4 adjacent samples with the same value, and so interpolation would be the same… could you add a little low-level HF noise when up-sampling to 96k or 192k, to simulate just about anything that the higher s/r would have been picking up on the set? 

Link to comment
Share on other sites

I don't see how up scaling could be as good as higher sampling for the original. A lot of post sound people record all of the effects at 96 these days I wonder if the music is being recorded at 96 too. Of course we have capabilities to record production sound at 96 but mirroring takes a long time plus in some recorders it reduces the track count available.

Tony

Link to comment
Share on other sites

I don't see how up scaling could be as good as higher sampling for the original.

 

 

First the theory: If there's nothing on an original track > 20 kHz besides electronic noise, then 48k s/r in a good oversampled recorder would be as good as higher sampling rates. 

 

So the question is, is there anything on a film set worth recording up there?

 

Remember also, once you're in the digital domain, there isn't any HF falloff from multiple generations. So if your final release medium is 48 k s/r...

Link to comment
Share on other sites

Of course we have capabilities to record production sound at 96 but mirroring takes a long time plus in some recorders it reduces the track count available.

 

We have a good number of people's from manufacturers here. Will be "challenge accepted".

 

:)

 

Best

V

Link to comment
Share on other sites

Famous Author: " First the theory: If there's nothing on an original track > 20 kHz besides electronic noise, then 48k s/r in a good oversampled recorder would be as good as higher sampling rates."

well, I'm not quite onboard with that theory, and somewhat agree with CC: " Higher sample rates create a more analogous digital capture. " though I would phrase it differently.

When I studied Calculus (circa 196n), the professor kept saying that the more we integrate a function, the more accurately we can determine the area under the curve...

Digital Audio is calculus, and the more we sample (integrate) the wave, the more accurate our reproduction of that wave can be.

Link to comment
Share on other sites

Digital Audio is calculus, and the more we sample (integrate) the wave, the more accurate our reproduction of that wave can be.

Yes of course, that much is pretty much a given. The question is how much more accurate than 48k do we need the wave to be?

Jay is right in that we don't need more than 48k as far as frequency response is concerned, however, and this was discussed before here and in other threads, when it comes to audio manipulation in post, especially time/pitch shifting and others, higher sample rates are clearly preferable. (Sorry for the long sentence)

However, if post can up-sample the audio and if that is really working just as well (is it?), then maybe that is a good way.

Although, it would probably still be better if we recorded everything at higher rates to begin (and yes, I did mean rates as in money and samples)

Link to comment
Share on other sites

When I studied Calculus (circa 196n), the professor kept saying that the more we integrate a function, the more accurately we can determine the area under the curve.

 

That's probably still true. But the area under the curve is defined by the mics, preamp, and ultimately those harmonics in the human voice* that are louder than both the ambient noise and any noise contributed by the electronics. 

 

*When I studied the human vocal tract (circa 196n), the vocal folds, throat, and head resonators were mostly soft tissue - even the bony parts had a tissue covering. So very high frequencies are going to get absorbed...

Link to comment
Share on other sites

I thing  like Jay, , a 1khz signal recorded at 8khz sf. will sound the same that a 1khz signal recorded al 48khz sf or 96khz sf. if it is reproduced with a properly designed converter.  I thing that there is a general missunderstanding of converters, and the details of the real limitations that every process have. 

 

I think the diferences in the processing of signals recorded at diferent rates is related to the algorithm used, If the process is bad wen using  48khz , and is good when upsampling the same signal to 96khz...   "that" is software related.

Link to comment
Share on other sites

  • 3 weeks later...

Ultimatly, it is about giving to post what they want and need not about what we think is the best or what makes sense or not.  Of course it is also our job to suggest ( and also sometimes strongly suggest ) certain things.

 

BUT, if post request 96khz because down the road of the audio chain, it is required, then it is what we need to give them.

 

If it was possible to someone who knows the definite answer to this, to answer if in order not to have artifacts on manipulation of the files, an oversampled ( 48khz up sampled to 96khz ) is the same or it actually needs to be recorded at 96khz to avoid them.

 

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.

Link to comment
Share on other sites

The pros and cons of 96kHz and 192kHz sampling frequencies were discussed some years back at this link:

 

 

Me personally, I think it's silly given the limitations of most locations and sound stages, but I can see where it would make a difference if the sound was going to be drastically altered in post by the SFX editors. I don't think it makes any sense for dialogue, but I do think 24-bits is a standard and is expected most of the time for production sound on serious projects with realistic budgets.

Link to comment
Share on other sites

I just wrote a whole big thing and then safari crashed and it was gone...  It was well thought out, written, and better before.  Thanks Safari...

 

In short - 

 

subsonic and supersonic frequencies interact with the audible range of frequencies.  being able to capture and manipulate these frequencies creates a natural more realistic product. 

 

The majority of perceived spatial relationship is created within high and supersonic frequency ranges.

Link to comment
Share on other sites

The majority of perceived spatial relationship is created within high and supersonic frequency ranges.

The majority of sound localization is derived from time, volume, and spectral differences perceived by your ears.

For various reasons it is difficult to localize low frequency sounds, but we are talking about very low frequency, i.e. below 75hz or so. Above that, localization is possible and not just limited to the high frequencies.

Link to comment
Share on other sites

What's your source for this supersonic information?

 

I wouldn't mind seeing this too. The last time I dug through all this I came across this from the Apogee marketing department.

 

[On the face of it this is quite absurd. Do we need to capture “audio” signals at up to 96 kHz? Obviously not – such signals don ’t exist. However, some recent research suggests that the human brain can discern a difference in a sound's arrival time between the two ears of better than 15 microseconds – around the time between samples at 96 kHz sampling – and some people can even discern a 5µS difference! So while super-high sample rates are probably unnecessary for frequency response, they may be justified for stereo and surround imaging accuracy. However, it should be noted that many authorities dispute this conclusion.]

 

http://www.digitalprosound.com/Htm/SoapBox/soap2_Apogee.htm

 

I couldn't find the papers they referenced but did find and interesting response from Dan Lavry, here.

 

http://www.lavryengineering.com/lavry_forum/viewtopic.php?f=1&t=479

 

I actually agree with CCalandro in that, if there is no technical reason not to do it and It doesn't screw workflow, why not. I just haven't found any evidence to support the claims of increased HF giving better audible localisation and imaging. Not losing sleep over this, just interested if there's any new data?

 

Cheers

Link to comment
Share on other sites

Not only do they exist but they interface and interact with the "regular" 20 - 20 range.

 

The majority of left vs right localization happens through time.

 

The majority of hight localization happens through frequency and the shape of your ears. In digital audio these high frequencies are sampled less, and sometimes creates aliasing. (depending on a lot of factors) 

 

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.

 

I can get pretty deep into this stuff, but it will quickly turn into a discussion about the ear, physics, and other smartie pants kind of stuff I don't feel the internet would benefit from.  

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...