|
|
Post by Martin John Butler on Dec 13, 2013 0:29:44 GMT -6
I just read this article about someone switching from 44.1 to using 96k sample rate to record, and how a UAD plug sounds better because of it: Tim Dolbear's Sound Advice: 96k verse 44.1k sample rates, my real life test. timdolbear.blogspot.com/2013/12/96k-verse-441k-sample-rates-my-real.htmlI prefer 48k to 44.1, I wondered what sample rate you guys are using.. I doubt I have the firepower to use 96k, but I'm still curious.. |
|
|
|
Post by lolo on Dec 13, 2013 0:36:52 GMT -6
48k/24bit
|
|
|
|
Post by unit7 on Dec 13, 2013 8:36:15 GMT -6
yay
|
|
|
|
Post by Martin John Butler on Dec 13, 2013 8:44:45 GMT -6
I realize that identifying the same mix run done at 44.1 and 48k played back to back would be a guessing game, but it's the feeling I get from the accumulation of thousands of little things, plugs, eq's, air, that has led me to 48 k, even though I'm aware of needing to dither to make it a CD.
What I'm especially curious about, is how in the article, the producer noticed a big improvement in the sound of a particular plug-in when using 96k. I'd have to try 96 in a complete production and see if I notice anything. Unfortunately, I'm at the edge of my processing power already, with my long in the tooth iMac. This means I may HAVE to create a master mix, and then add my mastering plugs in a new session, instead of mixing to the mastering plugs. I find that once I'm close to a satisfying mix without mastering plugs, I make subtle changes to the mix when the plugs are engaged. You know, a little more bass ata certain frequency, a pinch more LA2 on a vocal, etc..
|
|
|
|
Post by svart on Dec 13, 2013 9:40:56 GMT -6
24/44.1 for me. If for some reason I needed to go to higher sample rates I'd use 88.2 so the math is easier for the dithering/reduction at the end. Dividing by an even integer is a lot less "lossy" than dividing by a fraction. A lot less truncation is required, theoretically.
|
|
|
|
Post by jcoutu1 on Dec 13, 2013 10:00:39 GMT -6
24/44.1 for me. If for some reason I needed to go to higher sample rates I'd use 88.2 so the math is easier for the dithering/reduction at the end. Dividing by an even integer is a lot less "lossy" than dividing by a fraction. A lot less truncation is required, theoretically. I'm in the same boat here. 24/44.1. |
|
|
|
Post by Johnkenn on Dec 13, 2013 10:10:05 GMT -6
24/44.1 for me. If for some reason I needed to go to higher sample rates I'd use 88.2 so the math is easier for the dithering/reduction at the end. Dividing by an even integer is a lot less "lossy" than dividing by a fraction. A lot less truncation is required, theoretically. I see the theory behind that, but computers don't really have a hard time with some math over other math...like, they're not better at addition than subtraction... |
|
|
|
Post by Johnkenn on Dec 13, 2013 10:11:29 GMT -6
96 definitely sounds better...when you're at 96. I just wonder if it's all for naught once you dither down to 16/44.1...
|
|
|
|
Post by Ward on Dec 13, 2013 10:13:26 GMT -6
24bit/48khz here, same as Lolo...AGAIN! Waves plugins, SSL duende plugz and Pro Tools native plugz seem to work better at 48 as opposed to 88.2 or 96, for some weird reason. Plus, I really can't tell the difference between 48, 88.2, 96 and 192 anyhow. Like most people past 18, my hearing doesn't really hear well past 16.4k
|
|
|
|
Post by svart on Dec 13, 2013 10:29:12 GMT -6
24/44.1 for me. If for some reason I needed to go to higher sample rates I'd use 88.2 so the math is easier for the dithering/reduction at the end. Dividing by an even integer is a lot less "lossy" than dividing by a fraction. A lot less truncation is required, theoretically. I see the theory behind that, but computers don't really have a hard time with some math over other math...like, they're not better at addition than subtraction... It's not the type of math, it's the amount. A simple divide, is well.. simple. A fractional divide requires more bits to represent the number being created. I'm totally trying to explain this in layman's terms, so bear with me.. So 2x2=4. A single integer. 1/4x1/4= 0.0625.. Taking more bits from your bandwidth for other things. If you keep doing this over and over, your decimal numbers will become so big that they won't fit and thus will have to be reduced in any number of ways, either by more math or by straight truncation. Usually this truncation is done for speed. This truncation will CHOP off any extra numbers, thus REDUCING your fidelity at the bit level.. A savvy coder will avoid this, a coder who relies on DSP or CPU brute force might not. In either case, latency is the enemy and there will probably be a bit of both going on to balance fidelity and speed. So while it may "sound better", is the audio itself actually what it should be after the process, or is it some approximation thereof? Chances are it doesn't actually matter as long as you like the sound coming out, but make no mistake, what you like in what you hear is mostly likely NOT that the plug-in doesn't affect the sound fidelity, but simply the sonic imprint it makes on the audio signal. Similar to why folks love old Class A tube stuff even though it's chock full of distortions. Technically it measures like lead, but it sounds like gold. |
|
|
|
Post by tonycamphd on Dec 13, 2013 10:58:22 GMT -6
IMO Yay all day for 96k, especially for wide open acoustic stuff. I hate to convolute a simple Q, but i think a lot depends on the quality of your clocking/conversion at any sample rate.
|
|
|
|
Post by popmann on Dec 13, 2013 11:07:20 GMT -6
88.2 here. If a demo starts inside my Kronos its locked at 48...but, it's also a DEMO...
I think you guys have a serious misunderstanding of how resampling works. But, I won't pretend to be an expert--I think it treats it as an audio stream...not a file of samples to figure out which to chuck...regardless, the difference does survive being made redbook. Obviously, it gets more subtle...but, it's there. More importantly, like EVERY piece of gear, it's less about some result that can't be attained without it--it's about how easy (or not) it is to get there. I love that 88 sounds like the analog cue I'm listening to...there's nothing to "fix". I've been known to put an Oversmapling EQ on the two of 44 projects I'm sent with a HUGE 16-22khz boost and low midrange cut. Which binge me to the other part, when I switched and auditioned plugs...I chose the ones I did because they sounded the same at all sample rates-there were others that sounded much more useable at 88/96...but, like analog gear, I need my tools to be sample rate independent, as I get different rates sent for mixing. I don't need to mentally keep track of what I can use at one rate but not another...
|
|
|
|
Post by popmann on Dec 13, 2013 17:08:27 GMT -6
I also want to throw in that part of why this is so contested is there are different factors involved...capture AD quality...plug in DSP "quality"...how much you value input=output versus 44 sounding "good enough"--and how many actually ever HEAR the analog they're sending for conversion?
Also:correction...I don't know why I capitalized "huge" in the last post...the lift is rarely HUGE...I'm saying I consistently have to lift the tippy top at single rate and only occasionally do at double rate. But, it makes a cumulative effect of most of my single rate productions are actually "brighter" than double because of those little evening out of the uneven response up there. Capture at double rate, I only lift the highs when there's a specific tonal reason to do so because they were captured linearly well beyond where I'm sure I can hear.
|
|
|
|
Post by Martin John Butler on Dec 13, 2013 22:13:02 GMT -6
Guess I'll have to try it asap and see what happens. Great posts guys, thanks.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Dec 14, 2013 0:34:26 GMT -6
24/44.1 for me.
96 is a matter, because the converters are marketed agressively with this...why? Because converters are made to sell. Most of them (not all) sound better/are optimized for ONE sample rate.
The difference between 44.1/48 and 96 can even be percieved with the Mytek AD 8x96 i had.
With the 96k sounding better.
They should sound better because they are 96k, so they are made to sound better at 96k. That is all.
One other valid reason for 96k at tracking can be: The need for lower latency.
|
|
|
|
Post by popmann on Dec 14, 2013 11:03:07 GMT -6
Actually, the people making high end converters are the only ones to financially gain from the use of SINGLE sample rates, because that's where they sound markedly better. At double rates, the playing field is more even, IME.
So, if one wants to believe its marketing driven conspiracy, the fact is makers like Apogee and Mytec etc are the guys who benefit from people continuing to work at single rates. If you told me I had to work at single rate, I'd have a Symphony set up here stat. They've done something to make their converters more bright/open at single rates.
Larvy's rec is solid on mostly any level you want to look at it--88 if you can, 48 if you can't. It covers every hardware scenario and gets the best sounding sampling for legacy system and ones from the last decade. He wrote the white paper to explain why Larvy converters would not be quad rate capable. So, if anyone deserves science over marketing points...well, this is a done thing for me--I simply follow his rec...not just because he knows more than I do about the real world implementation tech...but, because it also explains what I've heard and said for years. Double better than single...quad is really no better than double. Though, I think we disagree on DSD...but, honestly, no one is making that available south of $50k+ Genex systems, so...it's functionally irrelevant for my work or most clients I mix for.
|
|
|
|
Post by tonycamphd on Dec 14, 2013 12:05:27 GMT -6
It's possible if the clocking/conversion isn't top notch, higher sample rates have been reputed to hollow out, don't ask me why, but there are plenty of guys who have complained about it, i'm guessing it's a combination of poor clocking, and the price point/design reality smallbutfine talked about, but of course i cant guarantee it.
|
|
|
|
Post by Martin John Butler on Dec 14, 2013 18:04:55 GMT -6
interesting, thanks guys.
|
|
|
|
Post by Bob Olhsson on Dec 14, 2013 20:45:16 GMT -6
Tim's experience pretty matches mine including MP3s sounding better. It's a lot like the difference between recording to analog multi-track vs. just mixing to analog. More information is preserved from the mikes.
The AAC spec also supports 96k. Too bad Apple doesn't get with it and support it. At the very least I record at 48k.
|
|
|
|
Post by tonycamphd on Dec 14, 2013 22:36:16 GMT -6
Tim's experience pretty matches mine including MP3s sounding better. It's a lot like the difference between recording to analog multi-track vs. just mixing to analog. More information is preserved from the mikes. The AAC spec also supports 96k. Too bad Apple doesn't get with it and support it. At the very least I record at 48k. hi Bob, did i miss something? who's Tim? |
|
|
|
Post by Martin John Butler on Dec 14, 2013 23:23:39 GMT -6
First post Tony, take a minute, it's worth the read.
Tim Dolbear's Sound Advice: 96k verse 44.1k sample rates, my real life test.
timdolbear.blogspot.com/2013/12/96k-verse-441k-sample-rates-my-real.html
|
|
|
|
Post by tonycamphd on Dec 15, 2013 0:33:31 GMT -6
First post Tony, take a minute, it's worth the read. Tim Dolbear's Sound Advice: 96k verse 44.1k sample rates, my real life test. timdolbear.blogspot.com/2013/12/96k-verse-441k-sample-rates-my-real.html Der, doy..yeah...uhhh huh huh....i should read the first post your sayin...?...huh Insert facepalm here _____ Alrighty then, just read it, it's nice to see others coming to similar conclusion as i've found. |
|
|
|
Post by swurveman on Oct 11, 2014 9:51:02 GMT -6
24/44.1 for me. 96 is a matter, because the converters are marketed agressively with this...why? Because converters are made to sell. Most of them (not all) sound better/are optimized for ONE sample rate. The difference between 44.1/48 and 96 can even be percieved with the Mytek AD 8x96 i had. With the 96k sounding better. They should sound better because they are 96k, so they are made to sound better at 96k. That is all. One other valid reason for 96k at tracking can be: The need for lower latency. Interesting information smallbutfine. How can a consumer identify which sample rate their converter is optimized for? |
|
|
|
Post by swurveman on Oct 11, 2014 10:05:36 GMT -6
First post Tony, take a minute, it's worth the read. Tim Dolbear's Sound Advice: 96k verse 44.1k sample rates, my real life test. timdolbear.blogspot.com/2013/12/96k-verse-441k-sample-rates-my-real.html Wouldn't the quality of the SRC play into this? I have Wavelab's Crystal Resampler, but have been doing everything at 44.1 anyway. Thanks for this link. I'll listen for the difference. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Oct 11, 2014 11:45:18 GMT -6
24/44.1 for me. 96 is a matter, because the converters are marketed agressively with this...why? Because converters are made to sell. Most of them (not all) sound better/are optimized for ONE sample rate. The difference between 44.1/48 and 96 can even be percieved with the Mytek AD 8x96 i had. With the 96k sounding better. They should sound better because they are 96k, so they are made to sound better at 96k. That is all. One other valid reason for 96k at tracking can be: The need for lower latency. Interesting information smallbutfine. How can a consumer identify which sample rate their converter is optimized for? Well, the consumer is the one who needs to be attracted to a new converter with arguments like highest sample rate as beeing better per se. Then, after a while, the higher sample rate becomes a standard, and the next marketing argument is beeing "future proof" (see actual marketing for 192kHz...and uplifted broadcast standards e.g. for archiving following). Since everyone profits from this (higher sample rate needs faster processors, more memory on media etc.pp., and also software producers can justify new versions due to new formats), there is not much anyone can do about this. So, if i would make an educated guess: More is better. So my guess is manufacturers have to optimize for higher sample rate to meet the buyers expectations. This may or may not be true for converter chips and/or their analog circuitry. As said, there are some strong arguments for using e.g. double speed sample rates nowadays. It is not marketing alone. But from my perspective, higher bit depth would be more desireable to get better conversion. 32bit or even 48bit would be nice, but there is only one or 2 converter chips working with 32bit depth available from the shelf right now, and no devices readily available, at least to my knowledge... As for SRCs ... they have vastly improved these days. Even free or open source SRCs can do very high quality sample rate conversion nowadays, like e.g. SoX, shebash and a couple of others. So most commercial DAW companies were forced to keep up in quality, which is a good thing. No real need anymore for expensive hardware SRCs like the Weiss Saracon, IMHO. BR, Martin |
|
