What's the 'correct' sample rate to record at?

[San]

One for all you digital guys out there..

I have a feeling that the answer to this question is "as high as you can get" but I thought I'd see what you folks do!

Up until very recently I had been doing all my recording at 16-bit/48k which was OK but never inspiring. Last night I managed to coax my system to record at 24-bit/192k and the difference was night and day but my PC pretty much ground to a halt after a while! I'll need to upgrade to continue at this resolution.

Is 24-bit/192k overkill or is it best to keep everything as high quality as possible until mixing down to CD quality? What rate do you use?

Thanks,

San.

[Quinny]

Hi San.

I have the capability to go up to 24/96, my friends are able to go to 24/192. We all record at 24/48, apart from one who records at 24/96. The arguments for and against each are looooooonnnnng stories, and a lot of it is down to personal preference. On one hand, it's fairly easy to see from the theory and the numbers that higher resolution *should* sound better.... however, in practice I find it makes next to no difference (sonically anyway, the ramifications for storage and machine spec are pretty obvious).
Another argument for really high res recording is that you'll be able to capture as much detail as possible to give to the mastering guys, who can then do their thing.

I personally have recorded the same samples at all kinds of resolutions. The jump from 16 bit to 20 bit wasn't that obvious, however going up to 24bit made a big difference. I'd highly recommend that if you can stay at 24 bit to do so.
As regards sample speed, I sampled at 96 and 192 khz and couldn't really note a difference with my ears. Even the jump from 48 to 96 is negligible to my ears, so I stick to 48. 24/96 is fast becoming an industry standard though.

16 Bit will do the job well (especially if you have good a/d converters), for instance my old 16 Bit Pro Tools system sounds better than some 24 bit systems with lower quality converters. However, stick to 24 Bit and 48/96khz and you'll be fine. :)

Niall.

[San]

Thanks Quinny, sounds like good advice to me.

San.

[Quinny]

Might also be worthwhile doing the tests for yourself..... record a piece at different resolutions and note the differences in machine performance (storage, CPU usage, RAM etc) and what it actually sounds like. It's an interesting thing to do, aside from anything else.

Or maybe I just got too much time on my hands.... :(

Q.

[LSchefman]

Just my opinion:

96 K means you can record without aliasing up to 48,000 Hz. I don't know of any recording microphones that have measurable frequency response above about 24,000 Hz, and most are down 30 db or so at 20,000 Hz.

Although I can record at 96 K, I don't, because I fail to understand what anyone thinks they can hear, record, or even sense, above 24,000 Hz.

24 bit is actually 20 bits in the real world, due to user bits, parity, etc. I record at that rate, because why not? In theory, the resolution is finer.

I will also state that I liked the sound of my 16 bit Tascam DA-88s better than I like the sound of most 24/96 systems, and kind of wish I hadn't gotten rid of them when I became seduced by hard disk recording in the mid 90s.

Then again, I like analog best.

Now - practically speaking -

Tell me if you can honestly tell whether you are listening to a CD that has been tracked and mixed 24 bit, and one that hasn't.

Of course you can't, and most CDs don't give you that information. In fact, most CDs from major artists are still recorded to analog tape, perhaps transferred to pro tools or similar for editing, and then mastered to analog half inch tape.

It's easy to tell this: see if they are coded AAD (analog tracking, analog mastering, digital encoding), or ADD, or DDD, etc.


The point is, there is an awful lot of hype out there, and we musicians are often suckers for gear. But the really big differences in sound come from things like the summing amplifiers in $400,000 consoles, and high end analog processors, etc.

And in the end, it's the quality of the music that is most important, and not whether it's 16 bit or 24 bit or any bit. :D

[lchender]

Hey Les! I still use my Tascam DA38s. They rock. I'll let the studio spend the money on all the latest hard disk stuff. I'll track at home on my DA38s and then take a couple teeny little tapes with me to the studio in my briefcase. :D

[GaryNattrass]

I agree with Les on this one but with the following footnote.

Even though you record at 20 bits when a 20 bit signal is transfered to digital CD it will only take the first 16 bits of information. What is then resolved by the CD machine on playback is down to the quality of the machine, the early CD player's could only resolve 14 bits of information and that is why older CD machines and D to A converters sound dreadful.

It will not be compressed down from 20 bit to 16 like analogue it will simply only resolve the first 16 bits of the data stream.

It is best to record at the highest resolution so that you can hear more detail and make more subjective decisions regarding overall quality. I would think that 20 bit 48k will be quite enough for most uses as sample rate is not as benificial as bit rate.

I personally also record analogue at 44.1k and at 16 bits onto a Fostex D160.
The AudioFiles and Logic desks I use at work are running at 16 bit but at 48K because that is what Digital Betacam uses as its sample rate.

[GaryNattrass]

Interesting remarks there John and I suspect it waas the analogue bit that made the difference in the first place.

When reording the analogue tape the frequency response of a machine will be reasonable flat up to a certain point, in the instance of a good multi track it will be good all the way up to 20khz with decent noise reduction. There will then be a steady roll off of the frequencies but the potential is there to go alll the way up to 30k which as you can imagine covers a lot of harmonic content.
Most analogue desks will also go way up to 30khz and in the case of some Neve consoles they are passing frequencies up to 100k.
This also goes for mastering and the reason a lot of pro's master to half inch running at 30ips is that it is virtually flat up to 30khz.

Digital is different and the bit rate does affect the reslution of the final product. Therefore 24 (20 bits in reality) will be better than 16 bits and also give a better dynamic range. (The dynamic range of a Neve console is around 128db's by the way a lot more than 16 or 20 bit)
The frequency response is directly related to the sample rate and is usually half of what the sampe rate is. i.e. 44.1k will give you 22khz and 48k will give you about 24khz.
Digital is only good up to those frequencies and will not resolve or include anything above them.

So therefore an analogue recorder will always sound nicer than a digital one as it is including second and third harmonics way up to any beyond 30khz.
A digital recorder will cut off anything above 24k.

I know you cannot hear that high but in terms of sound the harmonic content of a signal is made up of all of the phase interaction way up in theory to 100k and way down to 20hz.

BTW the HiFi amp is fantastic John and works really well with the old Ormond with Jensen 10 inch speaker. thanks again for doing it for me.

[jas]

I was under the impression that the files (songs) on an audio CD were ALWAYS 44.1, 16-bit. That the CD players wouldn't play 'em otherwise.

Is that incorrect?

-John

[redmax61]

I have to agree with Les and John here. Most of my recordings are done at 24/48K with great results.

[Lyle Long]

The jump from 16 bit to 24 bit makes the largest improvement in sound quality for the least amount of computer stress. Each bit added doubles the dynamic resolution of the digital word. 16 bit is 2 to the 16 power. That equals 65,536 steps of resolution. That is a theoretical 96 db ( 90 db actual) dynamic range. A 24 bit word has 16,777,216 steps of resolution (2 to the 24th power). That is a theoretical 144 db range ( 120-126 db actual depending on the system converters) Each bit is worth 6 db's dynamic range. It is much more capable of finer harmonic detail because of this. The processing impact on the computer is greater by 1-1/2 times.

Upping the sample rate puts far greater stress on your machine. That is because the sample rate is the clock rate that the processing must happen at. The move from 44.1Khz to 48Khz is slight but moving to 96Khz puts a lot of strain on the machine. A single mono track at 16/44.1 uses about 5.2 MB of information per minute. A 24/44.1 track uses 7.5 MB per minute. A 24/96 track uses over 17 MB per minute. A single 24/192 track uses over 34MB of information per minute. A 24 track 24/196 recording must move information at 816 MB per minute or 13.6 MB per second. That is huge! That makes a single 4 minute pop song a 3.3 Gig file at 24 tracks for your system to deal with!!

There is an real argument to be made for recording harmonic information out of hearing range having an impact sonically in our hearing range due to harmonic phase interaction, but that interaction cannot be heard unless the system has enough dynamic range first! The bit depth is what gets you that dynamic range in a PCM digital system. The sample rate has to do with frequency response. The main goal is to keep as much resolution until the final step of mastering before duplication. It does matter sonically!

I hope that explains some of the issues for you. Whew!

[BIG GINGER GIT]

I know nowt about this stuff but I record at 24 and bounce to 16 for mixing down to CD ? :confused: :(

[Lyle Long]

Actually, you mix at 24 bit, creating a 24 bit-2 track mix. This is done by either internally mixing (summing-bounce to disk) or by routing the audio out through a mixer and rerecording it from the mixer as a separate two track interleaved stereo file. The second method works really good if you have a multi channel output interface and a good sounding mixer. Plus, you can use any hardware outboard gear you may have at that time as well. I would try both with your system. Summing methods are a hot topic with pro engineers, with plenty of positions on both sides that have validity. There will be an audible difference between the mix methods. Pick the best sounding one obviously.

Then during mastering, after all level optimization has been done, at the very last step, the bit rate of the stereo file is reduced to 16 bit with the introduction of Dither (noise). Basically the bottom 8 bits are removed and noise (Dither) is used to smooth out the truncated file. This leaves most of the signal's detail intact. Far more sonic detail than mixing at 16 bit and then mastering a 16 bit mix. Cubase uses Apogee UV-22 as it's dither system. Waves has their dither system which is called IDR. It is available in their L1+ and L2 Limiter plugins. Both work great. Hope that helps! Peace!

[Brian D]

Yep, dither is the key.