Sample rate 44100 vs 4800?

[Joe Riggs]

Hello,

Calling all audio experts.

I'm doing a bit of quality control for a project and I just got the first batch of dailies.

The audio has been recorded at a sample rate of 44,100/32 bit.

Now I thought the minimum one should record audio is at 48,000 khz. Isn't this industry standard?

By it being recorded at 44,100, is that bad?

How big of a quality difference are we looking at vs 48,000?

Does the fact that it was recorded at 32bit, make the 44,100 superior to a 48,000 recorded at 16 or 24 bits?

Thank you

 

[Dustin Dooley]

44.1 has been the standard for CDs ever since they came out. If you find that CD quality should be good enough you will be fine. DVDs have had a 48k standard since they came out which is why for video 48 is the standard. In an everyday listening situation you would be hard pressed to hear the difference. If you had an audiophile system in a tuned room you might be able to tell. Folks that have been in audio for years and years might also be able to tell a difference if they have golden ears ;D

 

[Dave Blackham]

What Dustin says.

But 44.1 will have a slightly lower HF extinction frequency, bit depth may be more important to you. 48k 24bit would be a good standard to capture Audio and has been the features standard for quite some time. These no real cost to it and standardising your workflow will save costs in the long term.

 

[Hrvoje Simic]

Does the fact that it was recorded at 32bit, make the 44,100 superior to a 48,000 recorded at 16 or 24 bits?

It is highly unlikely that audio was recorded at 32 bits because there are no 32bit audio recorders or audio interfaces.
Someone may have confused 32bit float internal processing or audio file software export with recording bit depth. If you have 32bit exported files, they don't have any increase in quality and origin of the files is either in 16 or 24 bits.

24 bits are the high-end standard and enough to capture the perceivable DR.
44,1 vs 48...48 is a broadcast standard but if you received a 44,1 file there is nothing significant to be bothered about, the difference is too small.

24 bit 44,1 kHz would be superior to 16 bit 48 kHz if one would have to choose.
If you visualise a sound wave, kHz are representing its digital sampling in time on X axis and bits in dynamics on Y axis.
44,100 vs. 48,000 values per second compared to 65,500 vs. 16,7M of dynamics values.

 

[Serge Tsipis]

I'd consider 44.1/32 inappropriate for video.
There is a risk of sync problems. Get those files loaded into NLE and double check if it is in sync with the picture. It could deviate drastically in 3-5 minutes down the timeline. And there is no cure without degrading the sound quality, unfortunately...

The difference comes from re-sampling or re-quantizing the recording. If you don't want to dig into sound technology much, then avoiding any sort of conversions is always better. Generally, video comes with 48000Hz, 16 or 24bit sound. And my advice is to stick to that. You can multiply 48kHz by any number your equipment is capable, like 96kHz or 192kHz, if you need to record concert multitrack out of a FOH desk with a pristine quality. But any sort of TV/lav/RF/boom work is bullet-proof in 48kHz/24bit and mostly done in 48kHz/16bit to save HDD and CPU.

You can simplify understanding of Sampling Rate as how high you can go with a pitch, and Bit Rate as a dynamic range - how quiet you can go and still hear it. Bigger is always better. But there is no point going crazy about that numbers.

Bit rate:
24bit is more than enough for most live recording applications. It is possible but there is no point recording 32bit. It's primarily used for internal DAW computations like mixing, dynamics, effects, etc.

Sampling rate:
This one can trigger disaster! Imagine sampling like sawing a stick into equal chunks. If you have 96 and need 48 chunks, you can glue them in pairs and it's done. But if you have 88 or 44... It's gonna be a challenge! Well, that's exactly what happens to the sound.

When to consider going higher than 48/24 standard? When you need to get "RAW sound" for the mixing engineer to work with - edit, mix, master. Otherwise, it's likely to be a waste of storage + huge risk of post sync issues. It's like shooting REDRAW, but never grading.

I suspect any NLE software nowadays performs an internal conversion "quietly". That is ok unless one loads up live event captured with different sample rates and attempts to sync files. In 3-5 minutes of a playback time it starts to hit the fan! I hope it's not your case, or at least you don't have such a long takes!

 

[PatrickFaith]

48khz/24bit is the standard of what people expect for film. Anything else will just confuse things for low budget. Foley is often 96 khz, but that is recorded separately and normally you will get that as a stem at 48khz/24. Only guys I see recording everything at 96khz/24bit is the people going for atmos.

Sometimes the stem output from protools is 48khz/32 bit, this is mainly because some people mix the temp with compression (for the exec's) and with 32 bit re-recording & mastering can remove that and do it better.

 

[Paul Russell]

Just use Soundforge or Audition to resample it to 48/24. No-one will notice the difference.

 

[Marc Wielage]

Just use Soundforge or Audition to resample it to 48/24. No-one will notice the difference.

That is somewhat true with real-world material, but it's hopelessly inept for anybody to record 44.1kHz in the field for production. 32-bit is equally insane.

As Patrick says above, 48kHz/24-bit has been the standard for at least 10-12 years that I know of. It's always a good idea to establish the entire workflow for picture and sound prior to the start of the project, have a conversation with the DP, DIT, sound mixer and editor, and make sure everybody is on the same page up front. This is particularly important for mixed camera formats, timecode issues, metadata, file-naming protocol, and so on.

I hope Paul discovered this problem early on so he can nip this problem now, before it starts snowballing out of control.

 

[Paul Russell]

I hope Paul discovered this problem early on so he can nip this problem now, before it starts snowballing out of control.

You mean Joe?

Just a couple of audio clarifications.

As has been rightly stated, 48KHZ is the standard video sampling rate, and as such offers an audible frequency range of 0Hz to 24KHz. Very few human beings can hear above 20KHZ. 44.1KHZ offers a range of 0-22.05KHZ. Still adequate for ordinary audio of conversations etc. It's important when doing the conversion up to 48KHZ that resampling occurs, and not just 'interpreting' the file at 48KHZ, as that will cause sync slippage/

Bit depth is all about dynamics: the higher depth, the greater dynamic range. It's a logarithmic scale. The standard for audio production these days is 24bit, and that's what most digital interfaces record at as it offers a dynamic range of 144dB. That's from silence to launch of a Space Shuttle, and this is perfectly adequate for most, if not all, audio acquisition strategies. Most DAWS work in 32bit or 64bit ranges for post processing in order to maintain a higher degree of accuracy when summing, compressing, EQing or otherwise messing around with the audio. Final output to CD is 16bit. To DVD/Blu Ray material it's 24bit.

Occasionally some projects (like live gigs destined for the audiophile market) can be recorded at 96KHZ, 24bit (96/24) but this really is only for audiophiles and bats. 96KHZ contains audio up to 48KHZ, well beyond the range of the human ear. So unless the finished product is for the audiophile market, 96/24 is overkill, imho.

So in summary, Joe, you can do the conversion once in this case, but it will be a PITA to keep doing it. As you are receiving dailies, best to advise the sound engineer (who really should know all of this already) to make any further acquisition in 48KHZ/24bit.

HTH, HAND etc

 

[Dustin Dooley]

In regards to the syncing issue with 44.1, you are correct that it can cause some real problems with long takes whether that is a live event or other. Resampling usually does the trick as Paul mentioned, but in at least one case with a Zoom H2 I had to do some pretty mind numbing math to determine that it was off by .001% and brought it into Nuendo and time warped it. Luckily it was such a tiny variation that you couldn't hear a difference.Just as a side not I also got some H4n files in 44.1 and resampled and didn't have an issue at all.

 

[Elsie N]

I'd consider 44.1/32 inappropriate for video.
There is a risk of sync problems. Get those files loaded into NLE and double check if it is in sync with the picture. It could deviate drastically in 3-5 minutes down the timeline. And there is no cure without degrading the sound quality, unfortunately...

The difference comes from re-sampling or re-quantizing the recording. If you don't want to dig into sound technology much, then avoiding any sort of conversions is always better. Generally, video comes with 48000Hz, 16 or 24bit sound. And my advice is to stick to that. You can multiply 48kHz by any number your equipment is capable, like 96kHz or 192kHz, if you need to record concert multitrack out of a FOH desk with a pristine quality. But any sort of TV/lav/RF/boom work is bullet-proof in 48kHz/24bit and mostly done in 48kHz/16bit to save HDD and CPU.

You can simplify understanding of Sampling Rate as how high you can go with a pitch, and Bit Rate as a dynamic range - how quiet you can go and still hear it. Bigger is always better. But there is no point going crazy about that numbers.

Bit rate:
24bit is more than enough for most live recording applications. It is possible but there is no point recording 32bit. It's primarily used for internal DAW computations like mixing, dynamics, effects, etc.

Sampling rate:
This one can trigger disaster! Imagine sampling like sawing a stick into equal chunks. If you have 96 and need 48 chunks, you can glue them in pairs and it's done. But if you have 88 or 44... It's gonna be a challenge! Well, that's exactly what happens to the sound.

When to consider going higher than 48/24 standard? When you need to get "RAW sound" for the mixing engineer to work with - edit, mix, master. Otherwise, it's likely to be a waste of storage + huge risk of post sync issues. It's like shooting REDRAW, but never grading.

I suspect any NLE software nowadays performs an internal conversion "quietly". That is ok unless one loads up live event captured with different sample rates and attempts to sync files. In 3-5 minutes of a playback time it starts to hit the fan! I hope it's not your case, or at least you don't have such a long takes!

Copied and pasted into my notes for future reference. Thanks

 

[Dustin Dooley]

One more thing to remember if your files are really 32bits, is that you need to apply a dither at the end of the signal chain when converting down to 24bit. This adds a small amount of "noise" in order to make sure that there are no quantitization artifacts. Basically it is white noise you can't really hear to combat the very audible artifacts that might be introduced when converting from 32-24bits. You only need this if you are reducing the bit rate. If you are going up you don't need it. It is also not applicable for resampling diferent khz. All recording software should have dithering that you can add. Just make sure you add it after any effects that you might have applied as an EQ or compressor might raise the dither noise to audible levels.

I usually run into this problem of 44.1khz and usually 16bit files if the audio tech is used to recording for CD's and not video. I always make sure to check the settings if I hire someone I know might be more familiar with recording bands rather than audio for moving images.