Record at 44.1k -or- 48k?

[K8ch]

Hi,


I like to write pop tunes, so I generally record at 44.1
However, I would like my music to be considered for TV and film, as well.

I understand that the Film/TV industries use music that's recorded at 48k.
Furthermore, I heard that it's easy to convert 48k to 44.1, whereas converting 44.1 into 48k (for film use) is not easily done without artifacts.


Do any of you have experience to comment on this matter?




Peace,

[serr]

Actually its converting between the lower sample rates that gets a little compromised (48k to 44.1k and vice verse). If you're wanting to be critical, go HD in your recordings. 24 bit 96k is a modern HD release format now anyway and converts to the lower sample rates just fine for those intended targets. (24/88.2 is also just as accepted as a release format and possibly converts to 44.1k a little cleaner but this is less of an issue with modern sample rate conversion algorithms.)

[The Telenator]

movie / film soundtracks must be finished and submitted in 48k.

the standard for CDs and music in general for listening, retail purchase, etc., is 44.1k.

These are the industry standards. That's all there is to this discussion.

[urednik]

Originally Posted by The Telenator

movie / film soundtracks must be finished and submitted in 48k.

the standard for CDs and music in general for listening, retail purchase, etc., is 44.1k.

These are the industry standards. That's all there is to this discussion.

Ijust wonder: did you read his post?

[The Telenator]

Originally Posted by urednik

Ijust wonder: did you read his post?

YES ... but the remarks have no basis in fact.

Get Voxengo's r8t brain converter (free or commercial version) or just use, maybe, Audacity. Both are top-notch converters, no artifacts, no BS.

I really don't understand these 'artifacts', etc., he's talking about.

Oh, and one more thing -- I'm not going to get into another of these silly sample-rate, bit-depth crazy 20-page fools' threads, but the fidelity and quality of audio is just FINE in 44.1 or 48. Let's not go into more of this myth nonsense (again).

[urednik]

I am sorry, but you do not understand his question. He is just wondering what we suggest him to use in order to resample. And he has to resample anyway.

Serr did already answer him in a professional way, what makes you think, we want to hear your opinion on a matter you were not asked at all?

K8ch, I would suggest you keep it at 44.1kHz. I always prefer audio sample rate. On TV or video you anyway are not really able to notice any artefacts made by resampling.
This way you keep small files etc.

[Fran Guidry]

Originally Posted by K8ch

Hi,


I like to write pop tunes, so I generally record at 44.1
However, I would like my music to be considered for TV and film, as well.

I understand that the Film/TV industries use music that's recorded at 48k.
Furthermore, I heard that it's easy to convert 48k to 44.1, whereas converting 44.1 into 48k (for film use) is not easily done without artifacts.


Do any of you have experience to comment on this matter?




Peace,

I can only speak definitely of the NLE (non-linear editor, video equivalent of DAW) but Edius is perfectly happy to ingest 44.1 and resample to produce the final output.

I'd be surprised if any current generation NLE lacks this ability.

Fran

[karbomusic]

This is why 48 KHz over 44.1 KHz. NOISE FLOOR.

[The Telenator]

Originally Posted by urednik

I am sorry, but you do not understand his question. He is just wondering what we suggest him to use in order to resample. And he has to resample anyway.

Serr did already answer him in a professional way, what makes you think, we want to hear your opinion on a matter you were not asked at all?

K8ch, I would suggest you keep it at 44.1kHz. I always prefer audio sample rate. On TV or video you anyway are not really able to notice any artefacts made by resampling.
This way you keep small files etc.

C'mon, I understand where he's going with this. Noise floor stuff now? This is digital -- noise floor should be down near -120dB! What the hell are you chaps talking about?

If this is headed into a 'I need 96k or it won't sound good', you need to go read some technical papers, then return to the thread. Look, 44.1 at even 16 bits is good enough for all pop and rock.

You know all those plugins that upsample (4x, 8x, etc.) automatically or when set? THEY are doing conversions constantly -- up, then back to track level. No artifacts. (Or none you'll ever hear, anyway.)

[The Telenator]

uREDneck: 'Serr did already answer him in a professional way, what makes you think, we want to hear your opinion on a matter you were not asked at all?'


I don't give a flying fuck what you want to 'hear'. If I see something I feel like commenting on, then ...

You got much bigger problems than not liking my replies, that's for certain.

[whiteaxxxe]

Originally Posted by bikerider

This is why 48 KHz over 44.1 KHz. NOISE FLOOR. 48 KHz has a lower noise floor compared to 44.1.

bullshit.

sorry, but I stated only a fact. no pun intended. (indeed, pun is intended, but I try to be nice ... )

SR has nothing to do with digital inherent noise floor. and with not digital inherent noise floor it has also nothing to do. if there is outer digital noise you will sample that noise regardless what SR you choose.

so please stop talking nonsense about things that you obviously didnt get. please. and dont talk after someones blahblahing you have read on the internet. understand in the first place.

as said, outer digital noise you cant do nothing about with SR or bit depth. inherent digital noise floor is put down with a higher bit rate. 24bit is the way to go AND: bit depth has nothing to do with samplerate.

got it?

[David Else]

Originally Posted by K8ch

Hi, I heard that it's easy to convert 48k to 44.1, whereas converting 44.1 into 48k (for film use) is not easily done without artifacts.

This is not true if you use a modern SRC like SOX. If you use Foobar, this is your best solution:

http://www.hydrogenaud.io/forums/ind...howtopic=67373

If you have a good CPU then I recommend you record at 88.2 and then down-sample the final mixdown to whatever you like. You will get a little better audio quality, especially if your plug-ins benefit from the higher sample-rate (some do, some don't).

The ultimate purist solution (and the easiest) is just record in the sample-rate that the final medium requires. I think CD is dead and 48k is playable just about everywhere, so nothing to loose using that over 44.1!

[Fergler]

whiteaxxxe is right, bit depth controls noise floor not sample rate or bit rate.

fwiw you should try to record at double the sample rate if you're gonna record to a higher sample rate than to be exported. It makes sense for the math involved, slightly better preservance of small peaks and valleys.

I record vocals and horns at 88200 whereever possible in case there is any time editing needed, it makes the edits smoother (more room to stretch before noticeable artifacts).

It's more important to have a good quality converter to get the analog signal faithfully produced than it is to worry about sample rates, but if you have the option you should worry a little bit. There are some noticeable effects for sure.

[urednik]

K8ch

in order to get your answers you have to ignore replies made by some guys here. I guess you can understand that already, but just in case, cause some people love to spread fraud.

[The Telenator]

Originally Posted by whiteaxxxe

bullshit.

sorry, but I stated only a fact. no pun intended. (indeed, pun is intended, but I try to be nice ... )

SR has nothing to do with digital inherent noise floor. and with not digital inherent noise floor it has also nothing to do. if there is outer digital noise you will sample that noise regardless what SR you choose.

so please stop talking nonsense about things that you obviously didnt get. please. and dont talk after someones blahblahing you have read on the internet. understand in the first place.

as said, outer digital noise you cant do nothing about with SR or bit depth. inherent digital noise floor is put down with a higher bit rate. 24bit is the way to go AND: bit depth has nothing to do with samplerate.

got it?

Uh, right. You chaps go study up instead of spending all your energy being offended so easily. I was actually doing you a favour by trying to save you needless hours of worry over what is a load of unfactual crap.

[urednik]

Originally Posted by The Telenator

I was actually doing you a favour by trying to save you needless hours of worry over what is a load of unfactual crap.

And what is this suppose to be:

Originally Posted by The Telenator

Get Voxengo's r8t brain converter (free or commercial version) or just use, maybe, Audacity. Both are top-notch converters, no artifacts, no BS.

I am gone here. Enough for me.

[The Telenator]

well ... don't go away all mad....

[SaulT]

It basically states that the higher the sampling rate, the farther apart the Frequency impulses are. This gives a wider area of decay of the sample's wave form. Thus there is less artifacting / interference / noise / bleed through from sample to sample. 44.1 gives a noise floor of around -3 to -12db. 48 gives a noise floor of around -60db.

Every sentence in this paragraph is incorrect, e.g. the higher the sampling rate the closer the dirac impulses would be considered to be, because you're sampling more times per second, but even then we're not talking about dirac combs or impulse trains. The noise floor numbers are ridiculous... anyone else remember when the sample rates were below 20k? Gee whiz, I don't remember them being howling static boxes.

Any thing under -60dB is considered silent. My Fast Track Pro with nothing connected to the inputs puts out -56dB. That's pretty freaking quiet in recording terms.

I hate to break it to you, but no, that's not. Even with nothing connected that's still a little high. What happens when you connect your mic and add some gain? Don't get me wrong - I've been there, my last audio interface had a noise floor around there, and I had to record pretty hot to make up for it, but even then I still had some noise floor issues once I had added some gain and compression. You need to go lower than that if you want to record anything with a lot of dynamics, for instance. Consider - recording at an average of -15 dB gives you around 12 dB to peak. That's not all that much...

I can appreciate that you're well-meaning, but you don't understand what you're talking about.

[timlloyd]

Originally Posted by whiteaxxxe

bullshit.

It's actually not, it's just irrelevant. Sample rate can affect noise floor (though the rest of BR's post is rather odd), but the following should have no impact on the OP's decision between recording 44.1 and 48 kHz (why? read this).

All other factors being the same, when analogue noise is dominant (which is pretty much always the case), doubling the sample rate will increase the noise by approx 3 dBs (though noise level in the audible range will remain the same). If the dominant noise is quantization or dither noise, then increasing the sample rate distributes the noise over a greater frequency range, lowering it in the audible range ... also, perceptual noise shaping can be done more effectively at higher sample rates, which can further lower the noise level in the audible range (though not lower than the limit set by the noise floor of the AD conversion!).

As for the original question -- assuming that you are using a decent sample rate convertor (Reaper's is fine), there should be no noticeable difference between starting at 44.1 or 48k. Just pick one and stop worrying about it 48 kHz has a slight edge imo, since 44.1 is only really necessary if your primary release target is CD. If it's not, then just do everything at 48.

It is a waste of time to worry about this stuff unless you're an anorak

[Tod]

Originally Posted by K8ch

I understand that the Film/TV industries use music that's recorded at 48k.
Furthermore, I heard that it's easy to convert 48k to 44.1, whereas converting 44.1 into 48k (for film use) is not easily done without artifacts.

Hi Keith, I've made a few videos to YouTubes and I've submitted them in both 44.1k and 48k sample rates, however if music is included I did make sure they were at 48K. I honestly couldn't tell the difference.

I have used the Voxengo's r8t brain converter for this but there is one guy who has replied who I really trust with this stuff, Fran Guidry. I'm not familiar with what he's suggesting here but I will be checking it out myself.

Originally Posted by Fran Guidry

I can only speak definitely of the NLE (non-linear editor, video equivalent of DAW) but Edius is perfectly happy to ingest 44.1 and resample to produce the final output.

I'd be surprised if any current generation NLE lacks this ability.

Fran

EDIT: Aah, timlloyd, popped in while I was typing, another person I've learned to trust.

[SaulT]

movie / film soundtracks must be finished and submitted in 48k.

If the end target has to be this, then record at least at this - 24-bit at 48 kHz shouldn't put a strain on your system. Should be just fine.

[Tod]

Originally Posted by SaulT

If the end target has to be this, then record at least at this - 24-bit at 48 kHz shouldn't put a strain on your system. Should be just fine.

I totally agree, however, if your using sampled instruments, there's a good chance they're already going to have audio at 44.1K. Somewhere along the line a conversion is going to have to take place. Heh heh, unless you've bought libraries that are only 48K/24bit.

[K8ch]

Yikes!


I was hoping for a nice, simple "yes" or "no".

I really appreciate everyone taking time to discuss this.
Everyone.

I watched the video and it was very interesting, but quite a bit over my head.



Still not sure what to do, though.
Maybe record at 96...and buy a larger hard-drive to hold the bigger files.


I wonder if 96 convert to 44.1 as well as 88.2 would convert to 44.1?



Peace,


Keith

[peter5992]

Keith, here is a very practical approach.

When you say you would like your music to be considered for film and TV, what does that mean? Oftentimes music for film and TV is placed through libraries or intermediaries, and a lot of those guys have specific requirements to deliver (wav / aiff, 44 kHz / 48 kHz), and how it is delivered.

Yes it is true that the standard in film and TV is 24 bit / 48 kHz, and I don't into the technical fidelity differences, I would start at the end: what are the specs you are being required to deliver? And then work your way back from where you currently are.

[chas51]

I know little to nothing about the details discussed here.

but I want to tell you guys one thing.

I have made alot of videos.

when I recorded at 48k, synching the audio with the video, was easy.

when I recorded at 44.1k, it was impossible.

so, I thought that the OP's question was really a synching audio with video question.

to elaborate briefly (I'm sure everybody already knows this, but just in case),
48k sliced the audio into small enough pieces, that in my video program (Sony Vegas), synching up the better audio that I made using Reaper, with the video, and then muting the audio track that was on the video, was pretty simple.
44.1k,forget it. synching..........uh, no.......

[Time Waster]

Originally Posted by Fergler

fwiw you should try to record at double the sample rate if you're gonna record to a higher sample rate than to be exported. It makes sense for the math involved, slightly better preservance of small peaks and valleys.

From my reading, this is not necessary. Intuitively it makes sense but in practical terms it make no difference. The algorithms are good enough to estimate the waveform pretty closely. You'll get artifacts near the top end regardless.

[Fergler]

Originally Posted by Time Waster

From my reading, this is not necessary. Intuitively it makes sense but in practical terms it make no difference. The algorithms are good enough to estimate the waveform pretty closely. You'll get artifacts near the top end regardless.

Very subtle differences, for sure. Mostly noticeable in low end. I suggest taking a fat kick drum sample of 96k and downsampling to various target sample rates and comparing it. Can get a little (a LITTLE) washy down there which you can reveal nicely with null tests.

I don't care much about mixing sample rates, tbh. The only reason I like higher sample rates than final form is because of editing, it really has made the biggest difference for me than any one tool in my arsenal.

[whiteaxxxe]

Originally Posted by The Telenator

Uh, right. You chaps go study up instead of spending all your energy being offended so easily. I was actually doing you a favour by trying to save you needless hours of worry over what is a load of unfactual crap.

?? you think you are smart? well ... thats your problem.

and: you spend your time and energy talking BS? well ... thats too your problem. especially when you are not able to admit being wrong.

[planetnine]

Come on guys, life's too short for this.



>

[Giano]

Originally Posted by planetnine

Come on guys, life's too short for this.



>

Life is short ? Average life expectancy is 75 years !!!, so enough time for most of you to find out whether 44.1 or 48 k is the better choice !

[willowhaus]

Here's a simple answer: less conversion is ALWAYS better. So if your aim is to send your music to film & TV, them record and export in the expected format of 48k/24 bit.

In reality, any decent-quality re-sampling will likely be unnoticed anyway if not overdone, and is probably inevitable at some point (note the point about 44k/16 VSTis, or using different sample libraries). However, any conversion process will cost something from the original file, be it padding or removing data.

In practical terms, if your music doesn't have to be re-sampled on the back end then it's one less thing your customer (or you) will have to worry about.

[poplomop]

Higher recording sample rate reduces recording latency. Try it! (You can always convert to lower sample rate later to save on resources etc...)

[Fergler]

Besides, at some point the audio has to get turned into vibrations in a gaseous volume, that is the WORST part of the conversion process. It totally destroys your sound! :P

[Irvin]

Originally Posted by poplomop

Higher recording sample rate reduces recording latency. Try it! (You can always convert to lower sample rate later to save on resources etc...)

I record at 24/96 for that particular reason, along with the fact that some plugins will not need to do any oversampling on their own. Those two reasons are good enough for me.

[Fran Guidry]

Originally Posted by chas51

I know little to nothing about the details discussed here.

but I want to tell you guys one thing.

I have made alot of videos.

when I recorded at 48k, synching the audio with the video, was easy.

when I recorded at 44.1k, it was impossible.

so, I thought that the OP's question was really a synching audio with video question.

to elaborate briefly (I'm sure everybody already knows this, but just in case),
48k sliced the audio into small enough pieces, that in my video program (Sony Vegas), synching up the better audio that I made using Reaper, with the video, and then muting the audio track that was on the video, was pretty simple.
44.1k,forget it. synching..........uh, no.......

I'm not sure what you consider "alot of videos" but my two YouTube channels are pretty full. This was an issue in the last century but I haven't had any problem syncing different sample rates in a _long_ time.

I've never fooled with Vegas, so perhaps there's an issue there. What version are you using?

Fran

[chas51]

hundreds for me.
my Sony Vegas is a few years old now.
I'd have to check what version when I get to my desktop.
maybe it is a Sony Vegas issue.
never would have thought of that.
thanks Fran.

[Serenitynow]

Originally Posted by poplomop

Higher recording sample rate reduces recording latency. Try it! (You can always convert to lower sample rate later to save on resources etc...)

Can someone explain this to me? Why does it reduce latency?

[Lazarus]

Originally Posted by Serenitynow

Can someone explain this to me? Why does it reduce latency?

The way it works is that the DAW gets fixed size blocks of samples to process at a time, this is your buffer/block size setting in Preferences and/or your audio interface driver window (sorry if that's too basic). So a sample rate of 44.1kHz with a block size of (for easy arithmetic sake we won't use 1024, 512, 256 etc) 1000 samples it gets 44.1 sets of 1000 samples to process every second. If you double your sample rate to 88.2kHz then it gets 88.2 blocks per second. So two blocks are sent for every one at 44.1kHz. This halves latency because the system is now set up so it needs to process a block in half the time.

Obviously this means that the CPU needs to work harder in the same way it would if you just simply halved your buffer size, only the higher sample rate would mean that plugins are processing twice as much information and so there is a lot more overall work done. It's a solution if you are running at your audio interface's lowest buffer size and want to reduce latency, but apart from that...

[HugoRibeiroDotCom]

I record pretty much everything at 48KHz 24bits which is what most of my clients want.

[The Telenator]

Originally Posted by HugoRibeiroDotCom

I record pretty much everything at 48KHz 24bits which is what most of my clients want.

And that would mean you are doing pretty much only videos? Why otherwise use 48k? the difference between 44.1 and 48 is so minimal, especially now in the digital age, that it makes this entire thread ridiculous from the get go -- unless someone is asking what the standard accepted format is for whatever you need to produce.

The argument that Fergler put forward earlier: "... fwiw you should try to record at double the sample rate if you're gonna record to a higher sample rate than to be exported. It makes sense for the math involved, slightly better preservance of small peaks and valleys." The maths part makes perfect common sense to the human way of thinking: easier to divide 4 by 2 than 5 by 2, just for example -- except computers are not humans and the difference in the maths' difficulty is nonsense to a PC's ability to 'do the maths'.

As to the 'preservance' of those 'small peaks and valleys', that sort of sounds like common sense, too, and theoretically it is provable ... except there is one small problem with this thinking: NOBODY can hear these theoretical benefits. And 'no hear'ums' just don't win arguments in music (or physics, either, for that matter).

If anyone really likes recording or (perhaps slightly more defensible) mixing in very high sample rates, well, by all means, don't let me or anyone stop you! Just be aware that it is more of a personal choice than anything else. Any gains are going to be likely immeasurable to completely nonexistent.