|
|
Post by Johnkenn on Jan 28, 2014 19:23:27 GMT -6
This occurred to me...(so we can defuse the nerd fight above)...I've heard people justify using 96 over 88 because it's easier math for the computer...like the computer gives a shit? It's a computer, not the fat kid at the Math Fair.
|
|
|
|
Post by Bob Olhsson on Jan 28, 2014 19:25:24 GMT -6
All I can say is that I've never been pleased to find higher sample rates sounding better because it's a huge hassle.
I think a lot of the story is what sample rate plug-ins were designed using although I've had very mixed results from up sampling a previous recording. The big jump comes from recording off the floor at 48 or 96 exactly like the big jump from using analog comes when it's used to record from the floor.
|
|
|
|
Post by LesC on Jan 28, 2014 19:42:35 GMT -6
The easier math thing is a red herring, I believe. It used to be true in the bad old days when something called synchronous sample rate conversion was done, then it really was "better" to convert 88.2 down to 44.1, 96 to 44.1 caused nasty artifacts. Today we have asynchronous sample rate conversion, and converting any sample rate to any other sample rate is similar mathematically. In this case, an exact multiple makes no difference. That's my understanding, please let me know if I'm wrong.
I used to record at 48 when I was using ADAT XT-20's, then when I switched to PC's I started recording at 44.1. I've always worried about the sample rate conversion, and have read some people's opinions that converting from 96 down to 44.1 causes worse artifacts than the problems associated with recording 44.1 in the first place.
I've always toyed with the idea of trying 96 or 88.2, but I know that Cubase is notoriously bad at sample rate conversions. What do you guys who do record at the higher rates use for the conversion? Do you just accept the SRC within your recording software? Or use specialized sofware for the conversion? Or run two separate PC's (or Macs), one out at the higher sample rate and one in at the lower sample rate? Or use a hardware device, such as the RME ADI-192 DD? Which one of these techniques would be best, any idea?
|
|
|
|
Post by LesC on Jan 28, 2014 19:46:35 GMT -6
I hope that somebody here can really do a definitive test. I think it would be invaluable. I, for one, would definitely switch to higher sample rates if there was a trustable blind test where I could truly hear the difference.
I also wonder if I'm wasting my Burl B2 ADC, running it at 44.1. I would just love to know if higher sample rates, once converted down to 44.1 for the mix or master, are actually worth it.
|
|
|
|
Post by tonycamphd on Jan 28, 2014 20:53:01 GMT -6
I would ask those who swear by 44.1, have you ever actually recorded at 48khz? 88? 96? More than one track once, didn't hear a difference and moved on? I'm always interested in the road that brings people to where they are as much as where they are. I do 48Khz all the time, any compositions I work on with film in mind start life at 48. I've heard 88 and 192. Sounded the same. I've done up ABX tests for MP3 before and had those who decried compress formats shirk away into the darkness. I can do the same for SR rates where I'd downsample a wav and then bring it back up to 88 (loosing all the information that make the 2 different) and see what people say. But I'd need a couple of people to be willing to participate. These kind of internet tests mean nothing, it's about your OWN personal findings with ur own equipment IMO |
|
|
|
Post by tonycamphd on Jan 28, 2014 20:57:43 GMT -6
This occurred to me...(so we can defuse the nerd fight above)...I've heard people justify using 96 over 88 because it's easier math for the computer...like the computer gives a shit? It's a computer, not the fat kid at the Math Fair. It's about truncation errors and correction which today's computers DO make, 88.2 is less prone, or some shit lol! Even though less problematic as les stated |
|
|
|
Post by matt on Jan 28, 2014 21:37:57 GMT -6
The easier math thing is a red herring, I believe. It used to be true in the bad old days I think this is correct. Here is a discussion from a recent Sound On Sound article: Why is 88.2kHz the best sample rate for recording?Here is the salient excerpt: Modern 'asynchronous' sample-rate conversion is far more sophisticated and works by analysing the source and destination sample rates and working out only the required sample values with huge precision. This achieves a technical performance that is significantly in excess of any real-world converter and very close to the 24-bit theoretical level — and that's achieved with any ratio of input-to-output sample rate. There is no measurable difference in performance between using simple integer ratios or complex ones.
In fact, it's interesting to note that some of the best performing D-A and A-D converters from the likes of Benchmark, Crookwood, Cranesong, Drawmer, and others, all use non-integer sample-rate conversion as an inherent part of their jitter-isolation process. For example, D-A converters using this approach typically up-sample the incoming digital audio to something like 210kHz, or the rate at which the physical D-A converter chip achieves its best performance figures: no simple-ratio conversions going on there, yet class-leading performance specifications!
So integer conversion used to be relevant, but not anymore, it seems. |
|
|
|
Post by LesC on Jan 28, 2014 21:44:03 GMT -6
Thank you Matt.
|
|
|
|
Post by matt on Jan 28, 2014 22:03:46 GMT -6
Sure, no prob. The truth is out there (or "in there"?). It's just a Google search away, sifted by a little common sense analysis to vet the "facts" for apparent veracity. The Web is most certainly the cradle of the biggest single collection of hyperbolic BS ever in the history of humanity. If it were a color, it would be Purple, I think. However, the greatest falsehoods always possess the ring of truthfulness. But I digress. |
|
|
|
Post by Ward on Jan 28, 2014 22:41:35 GMT -6
Here's a general question that I think I'll get some responses to:
Say you start a project at 48k, and you've done your preproduction, which (if you're anything like me) includes fake drumz, rough bass and guitars and keys and guide vocal all to a click track. Then you next go to line production and start bringing in the band and/or session musicians...and one by one replace all your roughs and guides with the real thing.
Do you then decide to convert the session to 88.2k or 96K?
And I wish to ask about mixing down, if you're doing it in the box, as to have that will affect things too...considering the math, but let's approach that next.
|
|
|
|
Post by jazznoise on Jan 29, 2014 6:30:57 GMT -6
Actually, Tony, since it would be your own equipment it would be your own findings. They mean nothing if they're giving you answers you don't want to hear, I suppose, but the idea that it's meritless is wrong. There is no basis for giving that sort of thinking any credit, it's just a defensive reaction not exactly dissimilar to the whole Science Vs. Art dichotomy. If there's a difference, there is. If there's not, there's not.
There are advantages to Higher SR's that have been mentioned - plugins using internal SRC can sometimes cheap out and introduce artifacts, like Bob mentioned, or you can reduce hardware latency (Since 128 samples at 88.2Khz is the same speed as 64 samples at 44.1 Khz)without using tiny buffer sizes. But the sonics of a raw .wav that is sampled above 20Khz and has effective anti-alias filtering should be identical because, asides from your hearing not hearing ultra-sonics, your ADC stage consists almost entirely of the same circuitry regardless. The signal is feed into a Sigma Delta running the Mhz and then truncated down to your desired sample rate.
The math thing is definitely bogus, though I don't have the programming muscles to shoot it down off hand, I've read papers that demonstrated as much.
Files will be up later. Hold on to your jimmies.
|
|
|
|
Post by popmann on Jan 29, 2014 15:37:10 GMT -6
For the record, I never called it meritless-nor did I see anyone else say that unless I missed something. It's simply not proving what you think it's proving. You're extrapolating results of applying a decimation filter equivalent of 48khz's to an end result mixed/mastered file making no difference to perception, into also applying equally to sample rate at the time of AD conversion of analog signal, ADA loops to analog gear and back, and all digital DSP algorithms that might be applied during production.
|
|
|
|
Post by tonycamphd on Jan 29, 2014 16:22:06 GMT -6
Actually, Tony, since it would be your own equipment it would be your own findings. They mean nothing if they're giving you answers you don't want to hear, I suppose, but the idea that it's meritless is wrong. There is no basis for giving that sort of thinking any credit, it's just a defensive reaction not exactly dissimilar to the whole Science Vs. Art dichotomy. If there's a difference, there is. If there's not, there's not. There are advantages to Higher SR's that have been mentioned - plugins using internal SRC can sometimes cheap out and introduce artifacts, like Bob mentioned, or you can reduce hardware latency (Since 128 samples at 88.2Khz is the same speed as 64 samples at 44.1 Khz)without using tiny buffer sizes. But the sonics of a raw .wav that is sampled above 20Khz and has effective anti-alias filtering should be identical because, asides from your hearing not hearing ultra-sonics, your ADC stage consists almost entirely of the same circuitry regardless. The signal is feed into a Sigma Delta running the Mhz and then truncated down to your desired sample rate. The math thing is definitely bogus, though I don't have the programming muscles to shoot it down off hand, I've read papers that demonstrated as much. Files will be up later. Hold on to your jimmies. That's just it, it's not my equipment?? Your recordings mean nothing to me, meaning I'm not intimately familiar with anything in your process, the act of trying to decipher any differences, is more than probably futile, there are too many factors to consider, it has nothing to do with anything more than that. Perceived differences matter only in the context of my OWN, controlled and tangible recording and monitoring environment. The assertion that I'm afraid I might hear something I don't want to hear is pretty funny though, thanx for that 8) |
|
|
|
Post by mobeach on Jan 29, 2014 17:14:24 GMT -6
44.1/24
|
|
|
|
Post by Bob Olhsson on Feb 5, 2014 11:33:18 GMT -6
The problem with digital is that we are always at the mercy of specific implementations so whatever is theoretically best is meaningless when it comes to real world use.
Dan's problem with going above 60 is the speed and accuracy of the over sampling required. Most modern converters operate at a fixed extremely high frequency and then do an on the fly conversion down to the rate we store. It's really a question of where the reductions happen in our workflow and their quality as opposed to the rate the A to D is operating at. I knew George Massenberg prefers 192 so it's interesting that Al hears the same thing for recordings from microphones. I assume this is with Avid HD interfaces.
On a related note, I've found 24 bit dither to be very important recording from microphones which is a reason the Slate and UA mike preamps with DSP leave me worrying about quality. I'd definitely record anything from such at 32 bit float in order to minimize the truncation distortion but I don't really trust most software developers to get it right, especially the ones who argue nobody can hear the effect of not dithering.
|
|
