The Great 96 kHz Debate....all invited!

[Bob's Mods]

Well, I recorded my first tune using the 96 kHz sample rate and I must admit, I was pleasantly surprised. I easily heard improved track to track definition and improved plug-in detail. At least with my rack, the improvement in playback was noticeble without any ear strain.
I've used 44.1 since the beginning of time and experimented with 48 kHz, which, sounded the same as 44.1 kHz. I've never been able record in 96 kHz previous as my gear topped out at 48 kHz. My sound card, a WaveTerminal 2496 circa 1990, is supposed to do 96 but it had problems at that rate. I think the clock has too much jitter at 96 kHz. Since I picked up a dbx Quantum, it does work at 96 k without a problem, so I was finally able to record at the higher rate.
I've read various posts of others attempts at, and use of, 96 kHz. Some guys love it for the reasons I've described. Others, didn't find much improvment and felt it wasn't worth the hard drive space and cpu time. I feel each person's experience may well vary as gear quality and user expertise is all over the map. For me anyway, it was one of the bigger jumps in record quality I have experienced.
The opinion of many is "Why record at 96 when human hearing tops out 20 kHz"? This is the common argument. My feeling is that we all have been viewing this in the wrong light, its not so much the wider bandwidth that brings out this improvement, its more bits in the recorded range! More bits mean greater math precision, more detail. The unfortunate downside is the wasted portion of the bandwith. When recording at 44.1 you are "grabbing" one audio sample every 22.7 microseconds. When recording at 96 you more than doubling your sample rate grabbing a sample every 10.4 microseconds. Now, when you record for one second at 44.1 kHz this means 44,100 samples are recorded in one second which translates into one sample every 22.7 uS. At 96 kHz, you are grabbing 96,000 samples every second. This translates into one sample every 10.4 uS. As most of us are using 24 bit converters these days, I will use that as my baseline. In one second, for one mono track, at 24 bits, you have a total possible bit count of 24 bits X 44,100 = 1,058,400 bits. A one second mono audio track at 96 k would be 24 bits X 96,000 = 2,304,000 total possible bit count. The bit count has more than doubled over that in the 44dot1 world. Hence, there more bits to perform more detailed math calculations. This is why cpu cycle time goes way up, more bits to work with, more calculations to do! Your processor is getting warm! (I am not sure if your plugins do math on that part of the bandwidth where there is no audio data, hopefully it only calculates the audio data it sees).
The focus has been on the "wasted hard disk space", well I got a 120 MB Seagate, so that ain't a problem here. And my plugin use? I don't use a lot and the ones I use are cpu misers. If you like to use cpu hog plugs in realtime you may have a problem at 96 kHz unless you have one of the newer more powerful machines. The focus also has been on why record a bandwidth of 96 k when we hear 20 k max? These are valid points against using 96 k. The reason for using 96 kHz, more samples per second "in band" which means finer resolution for mixing tracks and plugin calculations. You've got to take the good with the bad if your gear can cleanly handle recording at 96 kHz.
The pain factor is a little higher when recording at this rate too. Its not just a HD, CPU intensive problem. For me to record at this rate I have to turn off all VST plugins. Something I don't have to do at 44.1 or 48k. My system easily handles plugins being on in realtime when recording at 44.1 and 48 but it generates "spikes" in 96 if the plugins are running when recording so they all have to be off. There is also the down sample thing too. Cool Edit Pro handles that without a problem though. And yes, the improved clarity does translate when downsampled to 44.1. Some guys where saying this wasn't the case for them but in my case, the sonic "hit" caused by downsampling to the CD standard was miniscule at best.
If your gear supports it, and you don't mind the extra fuss (and waste) related to working at this lofty rate - and your gear produces the sought after benefits (mine did), you may find the results to your liking.
My conclusion is track to track definition is not simply just a mic/preamp issue, it is also related to sample rate. If you find your tracks are not standing out in the mix with the detail you would like, 96 k should help. The detail in your plugs and esp reverb is really cool too. I suspect, in ages yet to come, we will migrate to a 96 kHz world. I think DVD sound is there now?

Bob

 

[Cloneboy Studio]

I've been recording a lot of 96khz stuff in PT. There's a difference, but not enough of one IMHO to freak out over. 48khz still sounds good to my ears.

 

[Bob's Mods]

I was intregued by the improvement, but it did not reach the "wet my pants" threshold. Results should be expected to vary widely from person to person. I believe there is some promise here.

Bob

 

[Solafide]

In my opinion a lot of the difference is in the converters you are using. They seem to have a "sweet spot." My motu 2408mk3 sounds significantly smoother at 96 khz (which for some things is good and others not). There is a little difference in the other rates, but nothing of major, and then it seems to just come alive at 96. I had an echo mia before that, and 48 seemed to be a nice rate on that one (if I remember right). I have nothing to back this up, it just an opinion bassed on limmitted observation.

 

[Bob's Mods]

I'm thinking lower end audio cards and gear that advertise that they can do 96 kHz recording probably don't do it well, if they can manage to do it all. Your "?better?" gear may be better at achieving suitable results in the higher rhelm.

Bob

 

[Teacher]

I'm thinking lower end audio cards and gear that advertise that they can do 96 kHz recording probably don't do it well, if they can manage to do it all. Your "?better?" gear may be better at achieving suitable results in the higher rhelm.

Bob

Actually its supposedly the other way around. Cheap converters will yield a better result @ 96khz then high end converters. Apogee animately advocates that if you make a great converter design they should sound identical @ 44.1/48 to 88.2/96khz. I personally don't think all that but I do believe high end converters differences would be more subtle. Its like comparing the old high end 20 bit converters to the prosumer converters today. Those old 20 bit converters are going to kick the prosumer converters ass despite being 20 bit and 44.1/48

I do agree that the math in programs sounds better @ 96 then in 44.1 but thats a difference you wouldn't notice unless a/bing 96 to 44.1. I personally didn't think it was worth the processing power and space lost due to increase size.

 

[HangDawg]

What about the people who don't mix in the box?

 

[mshilarious]

How did you perform your test? Is this 96kHz on your new converters vs. a lower rate with different converters, or the same converters?

I think to prove that 96kHz is better, I am aware of three arguments: the Nyquist theory isn't true; human hearing (and the capabilities of microphones and speakers) extends past 20kHz; or that implementation on 96kHz allows use of sonically superior anti-aliasing filters.

Let's leave aside arguments #1 and #2 (however it does seem that you want to make argument #1, which is far beyond my mathematical capability).

Here's my problem with #3: If a converter can operate at 44.1 or 96, wouldn't it use the same anti-aliasing filter? It wouldn't have to, but is a manufacturer going to use two different filters in the same converter?

Or maybe there is an issue with the implementation of a converter's clock: does it have less jitter at 96 than 44.1? Is that a design issue--meaning that the opposite could be true for a different converter?

I think it would be very difficult (expensive) to design a converter than would properly control for those factors in order to conclusively prove that 96 was better from a theoretical perspective.

 

[mrface2112]

for me, though, the defining thing has been "what's your final medium?". b/c if it's cd, then you're going to have to downsample from 96k to 44.1. unfortunately, this is NOT very clean math (and IS very complicated math), and when you combine it with the dithering that you should do (and like most of us, the dither we use is prolly not a very good one) you run the risk of compromising your signal when you go from 96 to 44.1.

conversely, by dropping to 88.2, you dramatically simplify that math since it's simple division.

personally, i still record at 44.1. i could record at 96 if i'd like, but i don't since everything i record is going to end up on cd anyway, i really see no point to eat the extra HD (and archiving) space for something that's just going to end up on cd. seems like wasted space, processing time and *possibly* fidelity.

if i had a 192khz interface, i'd prolly go ahead and give that a go, since i'd like to burn my own dvda's.....but i don't.

that said, i DO record at 24bit. the headroom gained from the 8 extra bits is worth the extra space and even the losses gained from dithering down to 16bit.

everyone here should go pick up Bob Katz' book "Mastering Audio, The Art and the Science" and you'll learn more about bitrates, sampling rates, dither, and general 1's and 0's than you ever wanted to know. it's a complicated (and slow) read, but it's certainly worth it.

not meant to be a counter-point or disagreement or anything (or to imply that Mr Katz espouses or agrees with what i was saying above).....it's just another take on this.


cheers,
wade

 

[chessrock]

personally, i still record at 44.1. i could record at 96 if i'd like, but i don't since everything i record is going to end up on cd anyway, i really see no point to eat the extra HD (and archiving) space for something that's just going to end up on cd.

Cue the following retort:

Well, back when I was very young, 1/8" cassette tape was the most popular medium for consumer playback. You never saw anyone in those days saying: "Well, if the finished product is going to be on cassette, I don't see the benefit of tracking to 2-inch."

Same thing with MP-3. Why not just track everything to MP-3 if that's what a significant amount of people are listening back on? The idea is for each track to be of the highest possible fidelity in order to minimize the damage that will inevitably incur later on down the line after everything is mixed down. Then you only undergo one generation of quality degredation, rather than a culmination of losses over every single track.

 

[mrface2112]

chessrock, in theory, you're right. the good rule of thumb is to always record at the highest resolution you can. if you see benefit from recording at 96khz, then by all means, do so.

longer word-lengths means that any distortion that's introduced by digital processing will be spread across a wider spectrum (and thus less audible), and obviously you should always process before downsampling. at minimum these are good things. there are plenty other "benefits" to this as well.

since i don't have a method of taking things to dvd-a, and since i mostly record myself and a small, selected clientele (all of whom want cds) 96k is still a little redundant for me at this point. i'm not going to archive the client's stuff for potential dvd-a's, so recording them at 96k is beyond my intentions (or pay level ). i could see some sense in recording my personal stuff at 96k, since i *would* like some forward flexibility. maybe i'll take it into consideration.

also be interesting if i could record the same tracks simultaneously into the DAW at 44.1 and 96k, and then judge the differences between the native 44.1 and the downsampled 96->44.1......but i don't think i can do that with my current situation. be an interesting science project.


cheers,
wade

 

[chessrock]

also be interesting if i could record the same tracks simultaneously into the DAW at 44.1 and 96k, and then judge the differences between the native 44.1 and the downsampled 96->44.1......but i don't think i can do that with my current situation. be an interesting science project.

All you'd have to do is track at 88, make a copy and downsample that to 44 with no filters, etc.

Do a mixdown for a session ... then go back and replace each track with it's corresponding downsampled counterpart. Convert the finished products to 16 bit - 44.1 and compare the results.

 

[mrface2112]

yeah, that's about as "close" to the test as i could do. still, a downsampled 88.2 to 44.1 isn't *exactly* the same as natively recording at 44.1.....but it IS as close as you're gonna get to doing 2 discrete/distinct recordings without actually doing it.

and like i said, recording at 88.2 would be the most optimal for eventual conversion to 44.1--both numerically and on the processing front *if* you're going to use a higher sample rate......however, since dvda is 96k, 88.2 doesn't do me a lot of good on *that* front. so i guess you need to (again) leverage your options. record at 96 for dvda, or record at 88.2 for cd?

still, i suppose you're right--i should at least be recording things destined for cd at 88.2, if not *just* from the advantages you gain on the processing front. i'm just lazy, i guess.


cheers,
wade

(edited to make me sound less of a dumbass)

 

[chessrock]

Oh not at all, Wade.

Most of us are just as dumb or dumber about this stuff.

I'm just trying to present the other side of the argument, and look at possible ways of testing this stuff out. I'm still tracking at 44.1, as are most of the people I know doing this stuff. And for the same reasons as yourself and everyone else, I might ad.

I just think it would be interesting to test this stuff out to see what kind of relevence there really is, and it's got me thinking. I suppose it might be fun, next time I get to track a project, to try and set something up. That way, we might, as a community, have some actual music samples with which to test our theories against, rather than all of these blind theories we keep throwing around in this endless debate.

It really does us no good to have someone come in to a thread and say: "I tracked a project last week at 44.1, and an entirely different project this week at 88.2, and I could hear the difference right away." Well, duh.

It would be so easy to set up a head-to-head test and have something tangible to base this on, I wonder why no one does it. Of course, I suppose I'm just as guilty and could easily do it myself, right?

 

[mrface2112]

LOL

glad to see i'm not the only one in this boat


i suppose the easiest test would be to get 2 DAWs running the same flavor input interface (identical firmware revs, etc) and run one at 88.2 and the other at 44.1, and just split the signal coming in off the mixing board/pres. sure you've got degradation b/c of the splitter, but it's the same degradation going to both recorders. and that's still not quite "identical", since it's 2 separate DAWs and interfaces. and the clocking will be different, so you're gonna have issues with jitter.....ugh, can nothing interesting be easy?

seriously.....a lot of it was initially over my head and took some re-reading to understand, but Katz's book is fantastic. it (and Mixerman) provided a good something to immerse myself in during the holidays at the inlaws'......


cheers,
wade

 

[mshilarious]

LOL

glad to see i'm not the only one in this boat


i suppose the easiest test would be to get 2 DAWs running the same flavor input interface (identical firmware revs, etc) and run one at 88.2 and the other at 44.1, and just split the signal coming in off the mixing board/pres. sure you've got degradation b/c of the splitter

The output of a pre can feed two cards with no troubles. And off a mixing board, well just use left for one and right for the other. Swap and repeat the test if you're fussy.

As for jitter, well I think Lavry wrote that an ideal clock would have separate crystals for each sample rate. A 'lesser' clock might very well have more or less jitter at different sample rates, so your only conclusion from your test is that particular converter is better at one sample rate vs. another.

You could control for that by testing a wide range of different converters

 

[regebro]

Well, I recorded my first tune using the 96 kHz sample [...snippetisnip...] definition is not simply just a mic/preamp issue, it is also related to sample rate. If you find your tracks are not standing out in the mix with the detail you would like, 96 k should help. The detail in your plugs and esp reverb is really cool too. I suspect, in ages yet to come, we will migrate to a 96 kHz world. I think DVD sound is there now?

Bob

Yup.

But, an interesting question is: Since it is not a 96k world, does it matter now. I.e., that 96 sounds better than 44.1 is an admitted fact, but will a recording and mixing done in 96 sound better even after it's sampled down to 44.1?

Theory sais: Probably not. What about practice? What do you think?

 

[mrface2112]

<<The output of a pre can feed two cards with no troubles.>>

absolutely. but the purists would whine about the fact that you're splitting the signal and thus minorly degrading the signal. *rolls eyes* they're right, but since you've got essential the same degraded signal going to each card, it's a non-issue as far as this test and i'm concerned.


<<And off a mixing board, well just use left for one and right for the other>>

well, i'd like to bypass as much within the board as possible--summing amps, channel strips, etc. i was thinking that tapping the insert (usually post gain and pre everything else) would be the best way to go about it. i've seen a number of boards where "left" runs a little hotter than "right", etc....so i'd like to bypass as much as possible. hence my thought of splitting a tapped insert.


you're right about Lavry's argument (Katz was on his side in that discussion, IIRC).....although of course, using a clock with separate crystals for each rate would invalidate the test, since the crystals would be different for 88.2 and 44.1. and how many of us are using Lavry-level gear anyway? i'm thinking about a Delta 1010, motu, or something else similarly "prosumer". decent for prosumer, but certainly nowhere near the level of Lavry's converters, etc.


my brain hurts


cheers,
wade

 

[Reggie]

Surely we can get somebody (if nobody has done this already) to do a test of recording a project with a full range of instruments (orchestra might be nice) both at 44.1 and 88.2 at the same time with as-identical-as-possible setups, and then convert the high-res track to 44.1. If there is no discernible difference between the two, then I am not going to worry anymore about recording high-res until I start doing DVD stuff.

Can anybody do this for the good of the audio community? Or maybe for a candy bar? Please?

 

[mrface2112]

<<If there is no discernible difference between the two, then I am not going to worry anymore about recording high-res >>

well, the recorded track is only about 1/2 of the ballgame. the other half (and the bigger half, IMO) is that you get massive benefits from 88.2 when applying digital processing (reverbs, eq, compression, etc) to the tracks. since the wordlength is longer, the effect of the effect is applied in a more "smooth" manner. virtually all process creates some distortion--and simply put, you've got twice as much resolution to apply the reverb to....and if you can blend that created distortion in over twice the resolution, it's going to be "hidden" better than if you apply the same processing to a 44.1 track.

so if you mix in the box, there's definitely some *immediate* benefit to mixing tracks sourced at 88.2 rather than 44.1. the question here, though, is how big of a difference this makes. my guess is that it's going to be noticible enough to make it worth the disk space. (yes, notice my turn of opinion over this thread) but my guess is that i'm still going to record at 24/44.1 b/c i'm used to it and i'm lazy.


cheers,
wade