Discussion in 'Recording/Live Sound' started by JoeB63, Jul 13, 2005.
What do you guys recommend for PC-based recording?
24-bit and 48K? or what?
Yep, 24 bit and 48k. I think the 24 bits are way more
important than 96k. If you've got a ton of storage space and
you want to do 24/96, go for it, but you'll be more than fine
When you go to burn a CD, there'll be a conversion to 44.1k.
Some folks feel there is some loss in this conversion (which is
true) if not done with a good sample rate converter algorithm
(poorly implemented filters), but most of the decent programs
out there do a good job.
I do 24/44.1k because I usually mix down to a masterlink and don't want to rely on it's SRC. Also because my digital mixer seems to clock better @ 44.1k. If you burn it to CD but still want to use a higher res, try 88.2k (the math is much easier). If you mix to dat, try 48k or 96k.
The choice really hinges on your software and hardware.
One minute of digital audio recorded at 24/44.1 requires 7.6 mb of disk space. Multiply that times 3:30 in song length and multiply that result times 30 or so tracks and the required space is not only demanding of the drive, but of the processor and the data transfer process. The worst aspect is the inability to backup song data to hard media without a dvd burner:
3.5 minutes x 7.6 mb x 30 tracks = 798 mb, a size that exceeds cd storage capacity.
Recording at 24 bits requires software that can effectively dither down from 24 bits to 16 bits. Cakewalk just started offering their first decent offering for dithering effectively, but at a cost of an extra $100 for the producer edition of Sonar4.
Without an effective dithering process built in to the software, an extra expediture for mastering software is usually necessary to get the best results when dithering down.
There is no advantage whatsoever to recording at 48k rather than 44.1 if you are mixing in the box down to 44.1 for burning to a CD. It's a waste of time and disk space. All your computer does when you convert is discard samples.
For that same reason, recording at 96k is also a waste if you are converting inside the box. It only makes sense if you are mixing to tape using very high grade D/A converters, and even then it's the subject of some debate.
If you're eventually going to end up on CD, GO WITH 44.1!!! PLEASE!!!! You'll thank me for it later. Unless you're using a VERY high quality sample rate convertor to convert from 48-->44.1, which you're probably not due to their high cost, it's better to record at 44.1 and not have to ever convert the sample rate. You will lose more in the conversion than you'll gain in the higher recording rate.
24 bit is a given, but be sure that you use something to "dither". If you don't you're just deleting the extra bits when you go down to 16bit (which is what CD is) and the loss is audible. I fear the people who think this is OK. Truncating bits is audible and there are no professional engineers who will say otherwise. Use dither. Dither is an intelligent way of getting rid of the extra bits while preserving some of what was recorded inside of them. Use it!
Depending on what software you're using to record, it will probably be listed as either an option when you go to make the final .wav or .aif, or need to be added as a plugin on the master fader (a la ProTools).
I do have to say however that 48k is by far the most popular sample rate in the pro studios, especially where there are great sample rate convertors available. For film and television, 48k is the standard. In fact, if your CD ever makes it to television, it's going to be sent up to 48k, or 96k in some cases.
ProTools has "DigiDesign Dither" as an option if you don't have/use a dithering plug.
I'm not familiar with the DigiDither other than as a plugin. I know it's not an option when you bounce to disk. Is it a preference? In any case, the Pow-R dither is amazing, but must be used as a plugin, and that was going to be my suggestion if dither came up again. I haven't used Digi's dither since Pow-R became an option.
Anyone whos saying there's no point recording at 48k when you're going to redbook audio is mistaken. When you tweak a fader or insert DSP or make any changes to the audio, the math always looks better at 48k... Yes you'll eventually dither down, but when you keep at a higher sample rate all along it makes for less signal degradation as you process. All of your DSP (including mixing) is done at the higher sample rate and resolution.
For most projects I do 16/48k (what has been the standard for film sets using digital) for good projects I do 24/48k. i rarely go higher then that for sample rates unless the project is going to SACD or DVD or some other such format. The other exception is when i have to import a lot fo audio at a different rate, in which case I'll stick to whatever the imported audio is, to aovid the conversions.
Not really. Dithering is a term that is more applicable to bit depth than sample rate. It means, roughly, that noise is added to the audio which is then re-sampled at the lower bit depth. Recording at a higher bit depth and dithering with a good algorithm has proveable benefits in terms of headroom and overall audio quality.
On the other hand, sample rate conversion at its best is nothing more than dropping samples - just throwing them out the window - with as few undesireable artifacts as possible.
The advantage to higher sample rates (if any) is in analog playback. Not "dithering."
But if you think you're getting better results, go for it. Who am I to argue with another man's ears?
My appologies.. I honestly didn't intend to use the term 'dithering'... I meant sample rate conversion.
The way it was alwasy explained to me is that when you do your math (processing) at higher rates (sample rate AND bit depth) and THEN convert down, you'll get better general quality, then if you do all the processing at the lower rate. I find the difference to be audible on a high quality, properly set up playback system, albiet only a touch. I'm more then willing to be convinced otherwise...
Gravity - you're not wrong, but there are a few factors to consider.
In general, it IS better to capture and process the audio image at a higher rate, but there are times when the anticipation of later procession, like sample rate converting, will make the advantage of using a higher source rate more damaging than it's worth.
For instance, if you're forced to use an inferior sample rate convertor BECAUSE the source material is at 48k, but you need to end up at 44.1, you'll end up doing more harm than good. It would have been better to simply stay at 44.1 the entire time. But if you have the best sample rate convertor money can buy, it would make perfect sense to record and process at the higher sample rates because, yes, there will be more information kept intact as the audio gets processed and mangled and manipulated. Audio processing as simple as turning up the volume slightly could create poorly-rounded mathematical strings that audibly alter the original sound. This is why internal processing is done at the highest sample rate as possible, sometimes at 4x the sample rate. It keeps things intact.
THEN, at the end, use a solid sample rate convertor to get down to your destination rate and you'll have the best final product you can.
Thats basically exactly how it was explained to me years ago when i was learning digital audio theory.
I was sort of assuming good equipment/conversion software when i made my initial statement. My bad!
Thanks for explaining it more clearly then i could put it.. it's been a while since i've had to put my theory into words ;-)
What constitutes a 'good' sample rate converter in your opinion?
I've always worked at 44.1. Wondering now about doing some 48 stuff and SRC-ing it to compare.
It's not really even a matter of opinion -- it's the convertor that does the least damage to the final audio as it converts from A to B. Ideally, there would be NO sonic difference between the two.
I've auditioned machines that were so good that I couldn't tell which was the original and which was the converted audio, but they were priced into the stratosphere. There is a very popular piece of software called BarbaBatch that is very good, and is somewhat reasonably priced ($400 list).
Logic users rejoice: New in Logic Pro 7.x is true world class SRC, which compares with very high end hardware SRC (and BarbaBatch, of course). (SRC in Logic Pro 6 and below was worthless).
DSP Quattro 2.x also has rather nice SRC, but not as good as BarbaBatch or Logic Pro 7.x. (SRC in DSP Quattro 1 was worthless).
Anyone know how good the SRC in Nuendo 2 and Wavelab 4 is?