Looking for some terms and definitions for standard digital audio settings...

Hello, all.

So I've finished up with my meatworld obligations and I can put my head back into focusing on my new rig and X1 so I thought I'd take a minute to pick your brains on a couple basic concepts that I kind of know about but am still a little muddled on. It might help me know a bit more about the correct way to phrase questions and not sound like such a dum dum all the time too so reading my threads and comments won't be as painful for you power users.


So I've recorded lots of stuff in the past but always just set it and forget it based on parameters my recording buddies recommended for my old system and what I was trying to do. Now I would like to really understand what exactly these particular settings mean, how they effect my recordings and what applications they would be used for.

If you guys have detailed links, articles, etc that I can bookmark for further study I am a HOUND for that type of stuff but just straight answers in thread are obviously more than appreciated as well.

I'll just break it down by the setting/concept type. Please excuse any mistaken terms, numbers, ideas or outright brainfarts. These are the things I'm actually trying to clear up in my mind.

There seem to be three main settings I should be looking at.

Buffer: This would be the thingy I keep being told to adjust to avoid dropouts and stuff. Currently at 512 (I think) increased from 200something. What is this and why does it do what it does?

Bitrate: I know 16 bit is what you want to burn a CD and 24 is what I was told to set my projects to on my old system for recording, mixing, etc.. but I had the option of 32 bit float and now I think there is a 64 bit option. What is this and what does it mean? Also what are some of the reasons and applications I would want to use these different settings for?

This last one I THINK is the samplerate... the one that sets a project or file to 44,000 or 96,000khz, etc... I used to try to stick to 96,000 for recording in the past (I probably screwed up the exact numbers but you know what I mean).

So yeah... just some basic definitions and ideas of what these do and how one would use them effectively and some external resources to peruse would clear up a whole pile of mind gack on my end.

If there is anything else I should know about obviously I'm all ears.

Thanks in advance. You guys rock.

No doubt you've opened a can o'worms.

Buffer is the amount of time your computer takes to communicate between the interface and itself.  What kind of interface takes such a big buffer?  If you are recording there will be a noticable delay between hitting a soft synth key or a guitar string or a vocal chord and when you hear it back.  SMaller buffers mean less time but also require a faster computer.  That depends also on how well the DRIVERS are written - drivers being the bits of software that the hardware uses to communicate w/ the computer.  No doubt others will chime in, but that is all about as simple as I can make it. 

Bit depth and sample rate are how we define how much info is devoted to the sound.  16 bit, CD quality has a maximum, theoritical 96 dBs between abosolute silence and representing the loudest possible signal.  That is more than enough, much more than say cassette or even reel to reel.  However, that is theortical, no practical.  There is always a noise floor, and no electronics is going to deliver the full 96 dbs.  24 bit, naturally, has more info, being bigger.  However, again, no converter ever reaches the theoritical maximum - they top out somewheres -120 db.  If a typical noise floor is 20-30 dB recording at 16 bits only leaves you 70 "steps" or so of loudness while 24 gives you closer to 100.  More is not always better, but in this case is.  In practical terms, you can set your levels coming in and worry a lot less about getting digital overs, or clipping, which ain't nice.

Since your interface can't make use of higher bit rates it doesn't make much sense to use them recording and there are no 32 bit converters.  It can help when performing math on a 24 bit sound, tho.  I always use higher depths when editing in sound Forge and operations like that.  But note that doesn't have much to do w/ inputting or outputting sound - your interface will just use the bits it can.

As far as sample rate - that is how many samples per second the computer uses to define your sound(s).  Due to alaising (see Nyquist),  to capture/produce sound at 22 kHz to 24 kHz, the range of human hearing for the high end, you need to record and output a sound at 44 to 48 kHz.  That is twice what you can hear but digital needs to filter the sound, and that takes twice the bandwidth or you get artifact.  You can read up on it or take it as gospel.  You don't need the forumla for gravity to know that falling hurts.

Many, if not most people use 24 bits at 44.1 kHz, the CD standard.  Some swear they can hear a difference using 48 kHz, the digital audio standard.  See for yourself which works better.  Some claim to hear a good difference using  96 or 88.2 kHz.  Many consider that overkill or more suitable for dogs and bats to listen to, but again when performing math functions it is possible to get more precise w/ bigger numbers.  A lot of the 'good" effects 10 years ago "upsampled" as part of their routine.  Now, most effects do this.  And all the digital filters, esp. older ones, don't perform as well and some interfaces may actually sound better using 96 kHz.  But again, listen.  If you can pick out the 96 in a blind test consistently, well, that answers your question.

Hope this helps a little and just doesn't confuse you more, but you got to start somewhere.  Don't let anyone tell you - use your own ears.

@

And I guess I'm kind of wondering what the numbers mean. Like what is the significance of 512 in regards to buffer? Stuff like that. I know you touched somewhat on it and it is certainly a big topic. Just kind of looking for the basic premise and words I can punch into google so I can research stuff further. Sorry... I'm a pain in the butt, I know.

Bit depth is about the dynamic range. How many steps from the lowest to the highest loudness. With digital the good thing is the loudest will stay that loud but we loose the very lowest when we go from a high bit depth to lower ones. The good news is we wont miss those low volumes because they are to low to hear.

Sample rate is the frequency of sampling. One would think that that would mean the higher the sample rate the more precise the sample. This is not so. What actually happens is the range of frequencies becomes greater. The bandwidth increases not the precision. All a higher sample rate does is allow frequencies beyond our range of hearing to be recorded. In most cases there are none. And even if they were there we can't hear them.

One thing to bare in mind is no audio interface can process above 24 bits. 32 bit audio can't be heard for example anywhere. It has to be truncated to 24 bits to output.

Thanks, John. So why do we have the option of 32 or more bit depths in our DAWs?

I think I grasp the "Dithering" concept (actually forgot to ask about that one so thanks for bringing it up) but the bit rate thing is still kind of bending my brain. No biggie... I think I've got a good start anyway to hunt around for articles. So I guess in my search terms I should use the terms "Buffer", "Sample Rate", "Bit Depth" and "Dithering" and that should point me to the right info. Anything else I should add to that list? Thanks again guys. Super helpful. I'm actually working on another thread topic about capturing the best input signals for different purposes using the gear I have on hand but that's gonna be pretty epic and will have to wait for another day. Not feeling so good right now. Cheers.

Beepster are you OK? Your last line is that you are not feeling too good.

I'm just really tired after the last week. I'm not a healthy dude so just getting this place cleaned up for the home care therapist kicked my ass. I should feel better tomorrow. Thanks for asking.

Heh... yeah. I really wanted to do my system optimization tonight but if I can't shake off the cobwebs in the next hour or so it may have to wait until tomorrow. I'm just extremely anxious to take another stab at the tutorials even if I have to switch to my internal soundcard for now. I think I'm gonna try to work through the help topic tuts 3 times over just so it all settles into my brain properly. Then I'll start screwing around with some of the Craig Anderton stuff. Oooh that's what I can do tonight. Watch the second Anderton vid. I don't understand half the stuff but it's neat seeing what's possible.

Beepster


Sooo... those bit depths would be used for what? Mixing? Adding effects? Mastering? Soft synths? Stuff like that? Then once you have everything good to go you would mix down to the lower bit depth? If so, and you record at 24 bits, does that mean you convert the project to the higher bit depth after your take is done? I know the answers are out there for me to find but as long as we're here I figured I'd ask. I remember seeing some vids with some mastering gurus talking about how generally you'd send a mastering house a 24 bit project but that in recent years some people choose the higher bit depths and that it does have it's advantages for the final mastering engineer.

I record at 24 bits, process internally at 32 bits and use the 64 bit engine for export.

As I understand it there's no point in increasing the bit depth of a 24 bit file 'cos all that happens is a load of zeros get added. Processing is a different ball game though. The higher the rate the more accurate the calculation thus less truncating (potential for distortion)

@FBB... okay, I think I'm starting to understand. Could you maybe be a little more specific as to what you mean by "processing"? Thanks again, man. You rule.

Oh and back in the day when I first started I was recording at 32 bit float (I figured the higher the number the better) but when I mixed down I got lots of pops and clicks. My engineer buddiy said I should keep the project at the same bit depth (24 bit specifically) to avoid that and only drop it down to 16 bit for my final master. So I started recording at 24 bit, mixing and mastering at 24 bit, do a premaster mixdown in Nuendo, toss it into Wavelab for some final EQing and from Wavelab export to 16 bit for my final product. I'm assuming things have changed somewhat since those days.

Beepster


@FBB... okay, I think I'm starting to understand. Could you maybe be a little more specific as to what you mean by "processing"? Thanks again, man. You rule.

By processing I mean the 'space' for the math to happen, or how accurate the calculation is. Remember it's binary so the more bits you can operate on the better.

Here's the best way I can explain it, I'm sure someone else will be able to do it better than me. If the numbers get too big the 'extra' bits get tossed out and there's your distortion. A calculation at 16 bits is more accurate than at 8 bits. 32 more accurate than 16bits and 64 more accurate than 32. Hence the higher the bit rate the more accurate the calculations the less chance of distortion.

A file at a set bit rate is just that though. If you increase bit depth you give the file more 'room' but there was no bits there in the first place so increasing a file to a larger bit depth is a bit pointless. Reduction is a different matter though because then you are potentially reducing existing bits and therefore able to distort that file, hence the need to dither.

Hope that's not too goobledegook.

Okay... I'll probably need to just keep researching this as it is obviously a very in depth and technical topic and brain is starting to go sideways but this has been extremely informative. For now I think I'll just do things the way I used to. Set my Sample Rate at 96,000 (if my interface can handle it... Focusrite says it can) and my Bit Depth at 24 Bit and just keep everything set that way until I do my final mixdown. You guys see any problems arising from that? My computer should be more than fast enough if that is a concern.

Beep,

your 96/24 sounds (pun intended) sounds good.  Be sure and click on the 64 bit engine in sonar.  What FBB was explaining about the maths uses this for greater calculations.  What happens is when dividing numbers (esp.) is you get some numbers w/ fractions.  These get rounded up or down.  This doesn't matter much isolated, but your computer is performing thousands of these a second and individual numbers get rounded and mulitplied or divided by other rounded numbers again and again - think each process in mixing.  The gain goes down - numbers are rounded.  Then that number is shunted through several stages of EQ, compressor, bused to a reverb, rounded again and again and errors start to creep in.  If you use a thousand point number instead of a hundred point number the errors are much smaller.

So use the 64 bit engine.

Get yourself well.

@