whats best?

[multitrak_ent]

ok i tried asking this in the 24 vs 16 44 vs 96 thread but i guess i have to start a new one to get an answer.
Now say i am going to record what would be better in the end, going 16 bit 96khz or 24 bit 44.1 khz?
is it better to go up in bit rate or sample rate, which is more important to getting the best sound?
Im asking because im looking at how am i going to get the most power out of my system at the same time the best sound. iv been experimenting as much as i can with this, and im leaning in one way towards pushing sample rate before bit rate, when i first thought about it ( without realy thinking about it ) i went with bit rate being more important, a friend said a few things now im leaning towards sample rate. but what i really want to know is scientificly what is better on paper what is more important with an end result. ( im hopeing this wont go into quantum phisics and im stuck with everything is related to much to make a difference)

[shevsound]

Hi,

Depends what you mean by best. I use Pro Tools Digi002rack and a Mackie SDR 24/96. Both of these units can record at the higher rates. Really you have to view whats best regarding to your situation.

I usually record at 44.1kHz 16-bit (mainly due to the fact that I know this works fine and I have paying clients who expect results).

However I am beginning to get into 24-bit recording. The main thing about 24-bit is that youve got the extra "bits". Whats so good about this? Well if you mix down a track to stereo in your Audio Sequencing program, then export it to something like Soundforge or Wavelab and process it (EQ, Compression, Limiting, Normalization) each time you process, the resolution is decreased.
CD quality is 44.1kHz 16-bit. So if you are processing a 16-bit file, the resolution actually drops below 16-bit and you end up loosing some of the quality.
Working on 24-bit files means that even though your resolution is dropping each time you process the file, it is still higher than 16-bit. Then you can convert the file to 44.1kHz 16-bit before burning it to CD and have the maxiumum resolution possible.

Also recording in 24 bit gives you more headroom (ie you can get a good "hot" level on your recorder and still have room above for any sudden peaks) which gives you less risk of clipping.

But, like I say, it depends on your setup and 24-bit WAV files are much larger than 16-bit and thus take up more memory and hard drive space etc, so could lead to problems with playback etc, and at the end of the day, are you more interested in getting your music recorded well with the minimum of fuss - or are you aiming for audio perfection?

Hope that makes things clearer.

[apl]

By the math, there is no advantage to recording at a higher sampling rate than you're going to publish in. So, if you're going to CD, you may as well record at 44.1kHz.

Converting to 48kHz or 96kHz to 44.1kHz involves some nasty math because the ratio is 1.088435374. See this explanation. (Scroll down to sample rate conversion).

Bit depth conversion is much simpler. Using 24 bits of dynamic range when recording makes it easier to get good levels above the noise floor without overloads. The conversion to 16 bits does not introduce artifacts, so 24 bit is the way to go.

[lpdeluxe]

This has been covered recently in another thread, but I have this to add to what apl and Shevsound have posted:

Higher bit depth, all other things being equal, will give more natural sounding recordings than higher sample rate. For example, if you record at 24 bit, and add reverb, your reverb fades are going to sound more natural because they won't be so obviously dropping from each level to the next. Converting down to 16 bit for CD burning is an easy, troublefree task for the software.

Whether you increase sampling rate or bit depth, you're going to increase the size of your files, so you might want to go with the method that yields the greater bang, i.e., more bit depth.

I personally record at 44.1K/32 bit for the reasons above and that Shevsound listed.

[Blue Bear Sound]

Originally Posted by multitrak_ent

ok i tried asking this in the 24 vs 16 44 vs 96 thread but i guess i have to start a new one to get an answer.
Now say i am going to record what would be better in the end, going 16 bit 96khz or 24 bit 44.1 khz?
is it better to go up in bit rate or sample rate, which is more important to getting the best sound?
Im asking because im looking at how am i going to get the most power out of my system at the same time the best sound. iv been experimenting as much as i can with this, and im leaning in one way towards pushing sample rate before bit rate, when i first thought about it ( without realy thinking about it ) i went with bit rate being more important, a friend said a few things now im leaning towards sample rate. but what i really want to know is scientificly what is better on paper what is more important with an end result. ( im hopeing this wont go into quantum phisics and im stuck with everything is related to much to make a difference)

-------------------------------------------------
DEFINITION - BIT RATE

The ratio of the number of bits that are transferred between devices in a specified amount of time, typically one second.

Bit rate is the same as data rate, data transfer rate and bit time.
-------------------------------------------------

Note that "bit rate" has NOTHING to do with word size or bit depth of digital audio (ie - 16-bit or 24-bit)

Carry on..................

[lpdeluxe]

OHMIGOD I DIDN'T SAY BIT RATE DID I?

pound pound pound pound pound

[shevsound]

I dont think anyone's gonna pound you for that! Its all meant to help.

[Massive Master]

24-Bit - For sure.

High sample rate - Eh, I can take it or leave it. If a project starts and stays in the box until it's on a CD you can keep it at 44.1kHz.

If it's REALLY GOOD and very dynamic - A chamber orchestra for example - I'd be at 88.2/24-bit. There's information up there that you don't want to lose. It's going to make a difference in the finished product.

If it's pretty standard rock 'n roll, it's not nearly as important IMO.

[shevsound]

I recently recorded a "Harp Ensemble" - 21 junior harpists on multi track at 44.1 16bit (to be safe). Do you think that recording in 24bit would have made a noticable difference to the sound of the recording? ie would the detail really stand out?
I transfer my WAVS from Mackie SDR to Pro Tools LE using USB cable - already a 24 track 5 hour session takes forever to transfer - then i'm concerned that I'll be putting too much strain on my PC processor.

[lpdeluxe]

You will run into exactly the same problem with graphic programs: I bought a relatively expensive scanner for the graphics work I do and tested it by scanning in some prints at high rates (1200ppi, for example) with the result that a) I couldn't distinguish any difference between that and the the default sampling rate and b) my computer gagged when I tried to edit this huge monster of a picture file.

I'd call that analogous, but actually I guess it's digitalous.

As to whether it sounds better at the higher rate, the question is really, can I hear a difference?

[TexRoadkill]

To say that you should only record in 44.1khz because of CD is pretty short sided. DVDA and other high res formats will be standard in the near future. If what you are recording is important then it may make sense to have some higher res tracks that you could remix or remaster in the future. If you're just having fun or cutting demos then it's probably not a big deal.

[lpdeluxe]

Good point, but then I'll have to buy another sound card/mixer/etc. One thing at a time!

[Cloneboy Studio]

Personally I hear a major difference in articulation and detail between 16 and 24 bit. I cannot stand 16 bit... something always seemed to be missing. And the high end always seemed funny to me.

[TexRoadkill]

Originally Posted by Cloneboy Studio

And the high end always seemed funny to me.

Bit depth has nothing to do with the high end, only the dynamic range. You are probably noticing the difference in convertor quality.

[masteringhouse]

Here is a great read on the subject.

http://www.apogeedigital.com/pdf/apogeeguide.pdf

Personally I feel that bit depth is more important than sampling frequencies above 44.1 and 48K. However things like filter design come into play at higher frequencies. Think of it this way, if you are sampling at 44.1 and trying to capture 20Hz, you will only be able to sample that frequency at 2 distinct spots (since you need to capture a cycle and a cycle requires that you sample the wave at it's high and low points). When you sample at twice the frequency you will have more "point" to re-draw the wave form, so instead of the square wave you would get from sampling at 2 spots (on or off) you have a few more to choose from in order to recreate a sine wave (or complex wave form) more accurately.

Basically it comes down to this (IMHO), given a particular sample rate, it will become less accurate as you go up in frequency. Since most people can't hear above 18K, and a lot of cheaper systems don't produce these frequencies, most people can't hear the difference. The decision you will need to make is if you are content with sweeping this distortion "under the rug".

[lpdeluxe]

Very nice link, Tom.

John

[masteringhouse]

Originally Posted by lpdeluxe

OHMIGOD I DIDN'T SAY BIT RATE DID I?

pound pound pound pound pound

Don't beat yourself up too much. I'm sure that you were thinking ahead to SACD/DSD

[multitrak_ent]

thanx for the replies, interesting that everyone says bit depth is where to go up in not your sample rate. a mate of mine went the other way, for the fact of when you move up in sample rate you allow more harmonics into your feild .and chances are that no fundamental note would be under the frequancy of 20hz then its all the higher harmonics that get effected.
also when dealing with the missing fundamental which is our lower harmonics been created out of higher frequancy notes. now when i put the harmonic theorys inside music into this problem it seems that the sample rate becomes very important.
the reason why im interested is because when im working with clients so far i have kept both up 24 bit and 96khz but even though i have 1 gig ram 3200 amd chip i still want flexabilty with power on plugins, im using the digi 002 so all my plugins get processed by the amd chip ( wish i had a hd system for this reason).
i am doing tests my self to see whats best i really just wanted the scientific reason why bit depth is more important than sampleing rate for quality.

Has anyone looked at how harmonics are effected by sampling rate? and how much sound we destroy by not getting more harmonics, which we cant hear but do make the sound different, the anolgly i use for this is. imagine you just put some nice reverb on a vocal track and its there but now your used to it and it is not noticed, then take it off to be a dry vocal and you can tell a big difference. similer to an mp3 cuting the high end harmonics when it compresses it down and the first thing you notice is the highs sound stuffed up, anyone investigated this?

[masteringhouse]

Originally Posted by multitrak_ent

thanx for the replies, interesting that everyone says bit depth is where to go up in not your sample rate. a mate of mine went the other way, for the fact of when you move up in sample rate you allow more harmonics into your feild .and chances are that no fundamental note would be under the frequancy of 20hz then its all the higher harmonics that get effected.

I will only speak for myself here, but I never said NOT to go up in sample rate, only if choosing between 16 bit high sample rate vs 24 bit 44.1K sample rate I would choose the later. The reason for this is that quantization error is going to be more noticeable when doing a significant of amount of audio processing, than the distortion of the upper harmonics will be.

Originally Posted by multitrak_ent

Has anyone looked at how harmonics are effected by sampling rate? and how much sound we destroy by not getting more harmonics, which we cant hear but do make the sound different, the anolgly i use for this is. imagine you just put some nice reverb on a vocal track and its there but now your used to it and it is not noticed, then take it off to be a dry vocal and you can tell a big difference. similer to an mp3 cuting the high end harmonics when it compresses it down and the first thing you notice is the highs sound stuffed up, anyone investigated this?

A sample rate of 44.1K will capture audio in the 20K range, it will just not capture it as accurately as a higher sample rate will. So it's not really like what you would hear with an MP3 file (unless you want to talk about filters). It's an issue with accuracy and distortion as you move up the range of frequencies. Since our hearing starts to rolls off the frequencies in the upper range (along with most consumer speaker systems), there is a tendency for most folks not to be able to hear the difference.

[Bulls Hit]

Originally Posted by masteringhouse

Here is a great read on the subject.

http://www.apogeedigital.com/pdf/apogeeguide.pdf

That's one of the best summaries I've read on the subject.

One thing it seemed to gloss over though was when it talked about the anti aliasing/anti imaging filters and using oversampling as a technique rather than higher sampling rates to stop phase errors creeping down into the audible range.

Is oversampling used because it's better than using higher sampling rates?
If you are using 88.2 or 96 sampling rates, does that make oversampling unnecessary?

[multitrak_ent]

I will only speak for myself here, but I never said NOT to go up in sample rate, only if choosing between 16 bit high sample rate vs 24 bit 44.1K sample rate I would choose the later. The reason for this is that quantization error is going to be more noticeable when doing a significant of amount of audio processing, than the distortion of the upper harmonics will be.

sorry should had been more clear. i was meaning more the fact that you would choose to go up in bit before sampling rate.

A sample rate of 44.1K will capture audio in the 20K range, it will just not capture it as accurately as a higher sample rate will

so of course 48 88 and 96 capture everything at 20k better . but do they capture more, i was under the impression that 48 will get 24hz one way and 24khz the other way, likewise with 88 and 96, is this true?
meaning the higher sample rate the larger the spectrum, sort of like a bigger canvas for a painter.
( yes i do understand that our hearing is limited said to be 20 hz to 20 khz but gets worse as we get older)
but understanding and doing research into music theory and harmonics leavs me to believe that it makes a difference cutting out harmonics.

also if we are going to cut out the frequanceys anyway does it matter that there in there in the first place, why do we have the option of 96khz if we go down to 44.1?
Is moving up in sample rate giving you some buffer room?

[ds21]

I may as well jump-in, I will also say given one or the other, and both converters of high quality, I'd go with the BitDepth, I believe that the brain will adjust to Frequency (or lack of) more easily then dynamics. But for me this many comes form listening not theory.

[masteringhouse]

Originally Posted by multitrak_ent

also if we are going to cut out the frequanceys anyway does it matter that there in there in the first place, why do we have the option of 96khz if we go down to 44.1?
Is moving up in sample rate giving you some buffer room?

It's giving a designer "buffer room" in regards to the type of filter that he uses to prevent alias frequencies. So yes this is part of it. The fact that higher sample rate captures frequencies higher than 20K doesn't matter per se since these will be filtered out (hopefully) before going down to 44.1.

My main point was that higher sample rates capture higher frequencies more accurately. Let's take a 20K sine wave for an example. At 20K a 44.1K sample rate will sample this at 2 spots, hopefully at a crest and trough point of the wave. The the point at which the crest/trough is sampled will most likely not be at the maximum point of the wave so already we have "distortion" in terms of the amplitude of the wave. Furthmore we only have 2 points of reference for the wave (on and off) so we essentially have a digital representation of a square wave at this point. If we increase the sample rate we will have more points of reference and the wave will be represented more accurately in both amplitude and it's original sine wave form. Also note that when a sine wave is represented as a square wave there will actually be more harmonics generated on the D/A side. These need to be filtered out appropriately or you may get aliasing causing a harshness in the audio. This is where the filters come into play. They will need to filter out all frequencies above nyquist (frequency at half the sample rate) to ensure that these will not be generated. Creating a filter that does this isn't easy, so it's good to have some of the "breathing room" that a higher sample rate provides.

Hope this long-winded reply answers your question a bit better. If you really want the details of all of this (and if you have aspirations of being a pro audio engineer you should). Check out Ken Pohlman's book "Principles of Digital Audio".

[multitrak_ent]

thanx mastering house, now I understand how it works as a buffer, good explanation there , makes alot of sense now. thanx again for your time and help.

[multitrak_ent]

thanx mastering house, now I understand how it works as a buffer, good explanation there , makes alot of sense now. thanx again for your time and help.