|
|
Post by tonycamphd on Oct 13, 2014 11:39:44 GMT -6
I'm no expert, it's clear that Pop and small know a LOT about this stuff, and of course I respect science, I was more eluding to believing "my lying ears". As far as 44.1 vs 96, I did those experiments a while ago with my own tracking(scientific? Probably not), to me hearing is believing, it's pretty much that simple.
As far as the HD tracks are concerned, I've lived with them long enough now to say confidently that they sound better to me than the standard Cd's. But, I have been re mastering the re masters with a linear multiband, to give a little bump on the bottom, and bring up the overall volume to better match a listening system(far from crushed). My brother and I are guys who appreciate the thump in the chest low end, as this pleases personal taste. The audiophile tracks seem mastered anti hype, low level, and very flat with much less compression, the drums really stand out as sounding better IMO on the tracks I've heard, much more realism, body/tone.
but as always, to each his own 8)
maybe ill post a 44.1/16 cd track, and the same 96k/24 track before and aft my gorilla mastering to see what y'all think... Science be damned 8)
|
|
|
|
Post by popmann on Oct 13, 2014 12:18:23 GMT -6
It's certainly hard to quantify "twice as good"....or "half as much like what you fed it"....there's certainly a diminishing return. I like to say 48 is much more improved over 44 than 88 is over 48.
|
|
ericn
Temp
Balance Engineer
Posts: 16,103 
Member is Online
|
Post by ericn on Oct 13, 2014 13:55:34 GMT -6
Every time somebody goes back to Nyquist theory I remind them, it's a theory not a law!
|
|
|
|
Post by svart on Oct 13, 2014 21:11:32 GMT -6
Every time somebody goes back to Nyquist theory I remind them, it's a theory not a law! LOL. I'll go say this to the guys at work. They'll get a kick out of it! |
|
|
|
Post by svart on Oct 13, 2014 21:17:17 GMT -6
One thing I failed to mention.
When clocking ADCs and DACs, increasing the sample rate requires higher clock speeds. Higher speeds mean that jitter and wander matters more and more. In some devices you'll have deterministic jitter from other sources intermodulating into the clock, and sometimes it's random.
So some devices might sound fine at lower sample rates but once you raise them, they might not have the actual resolution you expect.
|
|
|
|
Post by formatcyes on Oct 13, 2014 22:22:14 GMT -6
Every time somebody goes back to Nyquist theory I remind them, it's a theory not a law! There is no such thing as a science law. Everything in science is a theory (everything) and subject to change as new (per reviewed) evidence presents its self. Nyquist theory is the current one. A lot of testing has been done in the audio world and it is still the current theory. We are free to have our own options but not our own facts. (nyquist is the current fact). Our audio gear isent run on pixie fart's and spell's. Internally the convertor chips are not running at the frequency you have set. If you set your rate at 48khz your chip (most modern chips) over sample so the true sample rate is much higher. Burr brown 4 to 8x over sampling as an example. The reason 50hz was raised as a sample rate in the past was to give room for the low pass filters not so important now. If you are noticing a difference between sample rate's it is the analoge section and/or the sample rate conversion that is causing the issue. Not the sample rate... I was going to type more stuff but svart's post say's it best of all. |
|
|
|
Post by Martin John Butler on Oct 14, 2014 8:39:05 GMT -6
I do respect carefully done scientific studies, and keep an open mind. That said, it's obvious, science is not finished. I used to have endless debates with audiophiles regarding things now commonly accepted. like jitter being an important factor. My ears have caught things major league producers have missed on too many occasions for it to be luck, so I've learned to trust my ears. When I said "feel might be a better word than hear" earlier, I really meant it. Sometimes a feeling comes on slowly, and you have to notice signs that aren't obvious. Like when I notice I listen longer with certain plug-ins and less when using others, but it's subtle.
I think we perceive far more than we're aware of now, and in time, it will be clearer why certain things like higher sample rates make a difference, despite those who can "prove" it's not so. I've been around the block a few times, and been there and done that. If I hear something, or feel something, it's there, if someone else doesn't, it doesn't bother me anymore, despite their protests to the contrary.
|
|
|
|
Post by Bob Olhsson on Oct 14, 2014 11:07:56 GMT -6
I made comparisons while recording and mixing a 30 song compilation in the box entirely from microphones.
My results in the final 44.1x16 CD were that anything above 44.1 was clearly better. Comparing 48k to 88.2 depended on the program material and 96k was always better. All bets were off on any signal processing that wasn't dithered to 24 bits which turned me into a total dither nut.
Considering that 48k has been the AES/SMPTE/EBU pro standard all along, it's safe to assume that most plug-ins aimed at pro users would be optimized for that sample rate.
Finally, the reason for higher than Nyquist sample rates is to minimize grunge below 20kHz. Modern A to D converters all sample at high rates and then an on the fly conversion is performed so it's really a question of when and how the down-sampling is being done because nothing is really being recorded at 44.1, 48, 88.2 or 96k.
|
|
|
|
Post by Bob Olhsson on Oct 14, 2014 11:22:08 GMT -6
I think we perceive far more than we're aware of now, and in time, it will be clearer why certain things like higher sample rates make a difference, despite those who can "prove" it's not so. Half-assed ABX tests are not real science and prove nothing. They are every bit as full of B.S. as audiophile mysticism. In fact I think they are worse because they often give audiophile mysticism credibility.
Real science from Bell Labs tells us that our brain can't handle all of the information that our ears present. We must always focus our hearing using what amounts to a very narrow depth of field lens. Anybody who claims to have "ears of God" is sadly mistaken. Interestingly people having more hearing damage are actually able to hear more unwanted artifacts than people having "perfect" hearing because they are less able to mask information.
|
|
|
|
Post by Martin John Butler on Oct 14, 2014 11:39:04 GMT -6
Just to clarify Bob, when mixing that project did you mean you found 96k to be the best sounding?
Also, I'm so unfamiliar with dithering, can you give me the skinny on what you find to be the best settings when bouncing a song in Logic to stereo for mastering?
Quoting Bob, "Half-assed ABX tests are not real science and prove nothing", thank you sir!
Me, perhaps I'm a reformed "audiophile mystic", but after more "big shots caught with their pants down" experiences I've had in studios than I have time to relate here, I simply trust my ears and choose what sounds best to me. I still appreciate the helpful suggestions and insights here though, as I might not have tried certain things without them.
|
|
|
|
Post by tonycamphd on Oct 14, 2014 12:28:07 GMT -6
Martin John Butler there are threads hear pages long on dithering, and bobs all over them, search it up bromee as far as the rest, I'm following the lead of the grand freaking poobah!, case closed 8)
|
|
|
|
Post by Martin John Butler on Oct 14, 2014 13:06:48 GMT -6
Me too, just waiting for clarification.
Am I right, Bob liked 96k and 24 bit dithering? Or did I misunderstand.
|
|
|
|
Post by tonycamphd on Oct 14, 2014 13:15:17 GMT -6
Me too, just waiting for clarification. Am I right, Bob liked 96k and 24 bit dithering? Or did I misunderstand. Mr Ohlsson quote "My results in the final 44.1x16 CD were that anything above 44.1 was clearly better. Comparing 48k to 88.2 depended on the program material and 96k was always better."this is beyond good enough for the girls i go with....8) |
|
|
|
Post by Martin John Butler on Oct 14, 2014 14:14:40 GMT -6
Thanks Tony,. When I read that, I didn't quite follow. Does Bob mean 96k was always better than 44.1 or better than all of the others?
|
|
|
|
Post by formatcyes on Oct 14, 2014 14:21:34 GMT -6
Early in this thread Mr Ohlsson posted "At the very least I record at 48k"
|
|
|
|
Post by popmann on Oct 14, 2014 14:25:36 GMT -6
He said ONCE IT WAS ON CD, 96khz mixes sounded consistently best. This has been my experience, too. It's not (solely) about what it sounds like at the working rate in the studio or for the .0001% of the world who will buy the "hi Rez" version. It's that some amount of "better" follows all the way to the end delivery.
IME, the logic of "if it's bound for CD, stay at 44" is utterly incorrect. That was a THEORY when digital mixing began to be viable. It's not true. Disproven countless ways and times. The sad part is that now that some engineers are being called upon to mix digitally for the first time (due to budgets) they're reading the same theory, which anyone who has BEEN mixing digitally for any amount of time will tell you is clearly not the case.
|
|
|
|
Post by tonycamphd on Oct 14, 2014 14:33:37 GMT -6
He said ONCE IT WAS ON CD, 96khz mixes sounded consistently best. This has been my experience, too. It's not (solely) about what it sounds like at the working rate in the studio or for the .0001% of the world who will buy the "hi Rez" version. It's that some amount of "better" follows all the way to the end delivery. IME, the logic of "if it's bound for CD, stay at 44" is utterly incorrect. That was a THEORY when digital mixing began to be viable. It's not true. Disproven countless ways and times. The sad part is that now that some engineers are being called upon to mix digitally for the first time (due to budgets) they're reading the same theory, which anyone who has BEEN mixing digitally for any amount of time will tell you is clearly not the case. Exactly! brought down from the mountain on stone tablets!! |
|
|
|
Post by Martin John Butler on Oct 14, 2014 14:42:11 GMT -6
Thanks guys, that helped
Now, how about for other mediums, say...iITunes,
Any consensus on what's best ?
|
|
|
|
Post by tonycamphd on Oct 14, 2014 14:54:37 GMT -6
If u can, just track and mix at 88.2 or 96k 24 bit, dither and bounce to whatever u need is MO
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 14, 2014 15:09:03 GMT -6
Every time somebody goes back to Nyquist theory I remind them, it's a theory not a law! So is gravity. Sorry, couldn't resist. Take it with a smile. And please don't let me explain the differences between theory, theorem, axiom, lemma, korollar etc.pp. Believe me, i hated this stuff, too.... |
|
|
|
Post by LesC on Oct 14, 2014 15:23:20 GMT -6
You guys have convinced me to try 96k, instead of 44.1k, so thank you for complicating my life! In order to help uncomplicate my life, can you please tell me the best options for sample rate conversion back down to 44.1k on PC? I know there have been threads about this, but no real consensus. Fear of that last SRC has always held me back from trying other sampling rates. If it helps, for recording I'm using Cubase 7.5, and I have an old Wavelab license that I could probably upgrade fairly inexpensively.
Popmann, I think you're a PC guy, what do you use? Or anybody else?
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 14, 2014 15:34:49 GMT -6
This discussion really is very interesting to me and i learn every day right here about new views on this seemingly simple question, Bob even brought aspects of the nature of human perception of audio into the discussion that are completely new to me and exciting, and if i come to a personal preliminary conclusion, then it may be that the topic includes physics, math, many detailed technical aspects, cybernetics, neurobiology and whatever, and real world observations and experience do not contradict science anyhow.
I take it as a wide field of things that can be learned along the road, and i want to thank every single participant of the discussion for the overall polite and patient athmosphere...hard to find such condensed experience and information elsewhere.
Had to say this...
|
|
|
|
Post by formatcyes on Oct 14, 2014 16:50:17 GMT -6
Its a complicated question because its not the sample rate that cause differences we hear but the architecture around the conversion. This is why I wish audio would stick to 48khz and build everything around that instead of having to go looking for the sweet spot on your particular interface, because of poor/lazy design or just compromises trying to allow for 44.1khz thru to infinity...
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 14, 2014 17:05:48 GMT -6
You guys have convinced me to try 96k, instead of 44.1k, so thank you for complicating my life! In order to help uncomplicate my life, can you please tell me the best options for sample rate conversion back down to 44.1k on PC? I know there have been threads about this, but no real consensus. Fear of that last SRC has always held me back from trying other sampling rates. If it helps, for recording I'm using Cubase 7.5, and I have an old Wavelab license that I could probably upgrade fairly inexpensively. Popmann, I think you're a PC guy, what do you use? Or anybody else? FinalCD 0.17. Free. Try the /f2 filter option. Killer SRC. As in...uhm...no namedropping. It is just damn good SRC, period. Measure, hear, believe, whatever. Also has a sharp filter option if there is trouble going on with this setting, depending on the program material. www.sonicillusions.co.uk/downloads.htm |
|
|
|
Post by Bob Olhsson on Oct 14, 2014 17:22:00 GMT -6
Bell Labs determined that dither is always necessary after DSP calculations in the 1950s. My experience has been that it always makes a noticeable improvement provided the signal has never been truncated.
I had to take this on faith until I started recording pop music again here in Nashville after a decade of post production and mastering. I literally went directly from 24 track analog rock recording to 24 bit Pro Tools acoustic/Americana. I quickly learned that lack of 24 bit dither prior to the monitor D to A makes the sound of the room, i.e. the sense of depth and dimension go right away. If there is very much truncation such as a Pro Tools TDM rig that isn't running the dithered mixer, the sound also becomes fragile and crunchy sounding from any added DSP. Not making that mixer and dithering in general the default has not helped the reputation DAWs and plug-ins at all! It's every bit as stupid as making people need to remember to turn the bias on when using an analog recorder.
Unfortunately every kid who took a little Unix in college thinks they are some kind of a DSP expert. I only know barely enough to be dangerous but I see so much utter BS spread around that it's insane.
|
|
