|
Post by EmRR on Nov 6, 2018 10:14:18 GMT -6
Can I ask your reasons for this sample rate? While it's completely true that a sine wave can be reproduced (assuming perfect filters) up to nyquist, there's a second issue that I haven't really seen discussed here. Let's say you're at 44.1K and you're recording a 22K sine wave. The sine wave--even a perfect one--might be recorded (with those pesky perfect filters). But the phase of that sine wave is another thing entirely. Yes! I've brought this up here and many other places, always to blank stares. I've posted the phase plots....more blank stares..... It may not matter for a heavy guitar record, but for classical work it is important. Think about it, if you do a couple of passes through converters while mixing a rock record at 48K, the top end may become 180º out of phase with the bottom end, with that long variable phase slope across the spectrum. Rock/pop records with liberal use of HPF's....same thing from the other end. This is why I'm told the British way back when didn't call them 'equalizers', but instead 'phase distortion units'; whether it's true or not they used that term I don't really know. Of course EQ is a whole other kettle of fish......
|
|
|
Post by Blackdawg on Nov 6, 2018 11:46:36 GMT -6
There is plenty of reason to be at 88.2 or 96 these days. Having the Nyquist filter that high up is where the benefit is. It is drastically better that 44.1k IMO at least. Which is the same sort of reasoning for having it up at 384k. Is it truly better? I'm not certain either. I've only been working in DXD for almost 2 years. Still learning it fully myself. I do believe that 24/96k should be the standard though at least now days. Computer are fast enough and storage is cheap enough. I see absolutely no reason to work in 44.1k anymore really, barely even 48k too. Either way, getting to recording in 384 and swap to 24/96 even when down sampling its not the same. Then it gets turned into AAC or mp4 for video or 24/48 and it just isn't ever as awesome sounding. Rather disheartening. That is my experience anyways. It is cool to be able to A/B the two different setups. Everything runs to Grace m802 mic pre amps. I use the Grace AD converters to run digitally to PT and that stays digital to the AX32 monitor controller and all the way to the Genelec 8351a's and 8041a's. FOr Pyramix it goes from the Grace out the built in splitter analog to 2 Hapi's then digitally up to Pyramix and then analog out of a Horus to the AX32. So really even my DXD stuff i'm technically monitoring in 24/96k. But it sounds much better still. We use PT as our backup recorder...even though Pyramix in Safety Record mode is bullet proof compared to PT. Then I do mixing in Pyramix for audio releases and PT for film releases. Even for the film I use the DXD audio and sample convert to 96k and it still sounds better than the PT audio. It's of course all pretty subjective. I realize that doing stuff at 384k is not normal for a lot of reasons. I'm just lucky I get to work at a place as I do and get to use it. Otherwise I'd probably do everything in 96k if I was freelance. I imagine it is disheartening.I think the same was happening back in the day after tape transfers and vinyl mastering and the like. Even Allen Sides mentioned that what he heard through the board during tracking wasn't the same immediately on tape playback. 10-15% lost in his estimation.
If everyone would get on the high-res bandwagon, digital could make that a thing of the past. Right from tracking to mastering and to the consumer without losing quality. That would be a first. Then there was the mp3. Somehow opportunity knocked and it all went backerds.
It would be fun to do testing here. I am certainly astounded with the quality I hear on my Lynx (n). Maybe I'll just run it at 192 from now on.
Never tried DXD though.
It's funny, as soon as 88.2k gets in on more and more computers, the less it seems necessary now. As I understand it, the use of that rate is simple math for the computer to halve the rate. Seems like computers now don't have any issues doing that with speed from other rates.
96k should probably be the standard, yea. 441 is so close to the audible range considering the filtering. There was a lot more room on tape, right? Bias was at what, 150khz or something on some machines and higher? Yes some tape machines had an amazing frequency response that was very high. It also all depending on the tape you were using. As well as the Bias of the machine which was a pain because you had to constantly check it and realign it. Even some vinyl had very high freq response. There of course was always some big limitations to the all analog formats that held it bad in various ways but over time was over come a lot with technology. There was 70 years of engineering into tape recording by the time digital took off. Just think where we will be in another 30 years with digital recording. Yeah the argument was 88.2 could down sample easier for computers to do the math. But now days, that is a mute point, computers are insanely powerful and always getting better. Can I ask your reasons for this sample rate? While it's completely true that a sine wave can be reproduced (assuming perfect filters) up to nyquist, there's a second issue that I haven't really seen discussed here. Let's say you're at 44.1K and you're recording a 22K sine wave. The sine wave--even a perfect one--might be recorded (with those pesky perfect filters). But the phase of that sine wave is another thing entirely. There are only two points of that sine wave actually recorded and they occur only at the clock ticks. The phase may have only a tenuous relationship to the actual phase of the input. I don't know what you'd call this, but it's something akin to jitter: even if your clocking is perfect, that sine wave has been shoehorned into the two available bins--one positive and one negative. Now consider that we're recording a piano (and let's make the false assumption that it doesn't generate any energy above 22K). A note consists of a stack of partials, going from the fundamental all the way up to 22K. The higher the frequency of the partial, the more iffy its phase. And when you consider that partials also change pitch in complex ways, then things get worse. There's sort of a 'grinding' as those partials are increasingly phase-shifted relative to the fundamental. It's a small thing, to be certain. But small things are what we focus on. Higher sample rates don't make this go away, but they do lower the amount of error. I also think they produce better imaging, but this doesn't appear to be much-studied. Great points.
|
|
|
Post by Mister Chase on Nov 6, 2018 14:07:38 GMT -6
Can I ask your reasons for this sample rate? While it's completely true that a sine wave can be reproduced (assuming perfect filters) up to nyquist, there's a second issue that I haven't really seen discussed here. Let's say you're at 44.1K and you're recording a 22K sine wave. The sine wave--even a perfect one--might be recorded (with those pesky perfect filters). But the phase of that sine wave is another thing entirely. There are only two points of that sine wave actually recorded and they occur only at the clock ticks. The phase may have only a tenuous relationship to the actual phase of the input. I don't know what you'd call this, but it's something akin to jitter: even if your clocking is perfect, that sine wave has been shoehorned into the two available bins--one positive and one negative. Now consider that we're recording a piano (and let's make the false assumption that it doesn't generate any energy above 22K). A note consists of a stack of partials, going from the fundamental all the way up to 22K. The higher the frequency of the partial, the more iffy its phase. And when you consider that partials also change pitch in complex ways, then things get worse. There's sort of a 'grinding' as those partials are increasingly phase-shifted relative to the fundamental. It's a small thing, to be certain. But small things are what we focus on. Higher sample rates don't make this go away, but they do lower the amount of error. I also think they produce better imaging, but this doesn't appear to be much-studied. Ok so let me try to follow along here.
So for instance, because a wave needs two points to be recorded digitally, a peak and trough, we need a sample rate twice as high as the highest frequency we want to record. That much I understand from Nyquist. So what you are saying, is because there are only two points on that example frequency, there isn't enough information to align phase, or at least guarantee it?
So does this mean that when a converter samples say, a 100hz waveform at 44.1khz, it's taking many samples of that wave, whereas only two on the max frequency? So the phase would likely be less and less coherent as the frequencies go up?
Very interesting. And, while I don't disagree with something like 384k, I can't totally understand why do it either. However, the phase thing makes sense. I just wonder if there truly are detractions to going higher and higher, besides noise. Is there really no downside other than disk space and high frequency noise? I suppose in some ways that noise up there doesn't matter since most speakers people will be listening on just won't produce those frequencies. Most mics won't pick them up either. Hmmmm.
|
|
|
Post by Mister Chase on Nov 6, 2018 15:47:43 GMT -6
It's interesting. I just recorded the same strummed chords on an acoustic at 44.1 96 and 192. Trying to do it the same each time. Miktek C5.
I found that switching back and forth there is sort of a crunch to the highs at 44.1. Its subtle. 96k it disappears and the sound gets more open and relaxed. 192 is quite close to 96 in terms of sound for my ears. Maybe just more of the same increases 96k saw but more subtle. It seems like something else is going on subtly but I can't put my finger on it. Pretty interesting. I feel like 96k is my favorite but I can't tell you why. Hm.
Crap I just realized I devolved this thread where it wasn't meant to go. Sorry. Sigh.
|
|
|
Post by Blackdawg on Nov 6, 2018 15:54:33 GMT -6
It's interesting. I just recorded the same strummed chords on an acoustic at 44.1 96 and 192. Trying to do it the same each time. Miktek C5. I found that switching back and forth there is sort of a crunch to the highs at 44.1. Its subtle. 96k it disappears and the sound gets more open and relaxed. 192 is quite close to 96 in terms of sound for my ears. Maybe just more of the same increases 96k saw but more subtle. It seems like something else is going on subtly but I can't put my finger on it. Pretty interesting. I feel like 96k is my favorite but I can't tell you why. Hm. Crap I just realized I devolved this thread where it wasn't meant to go. Sorry. Sigh. Thats what Im getting at. It feels better, more open and natural. Not really better sounding as a whole because you're getting more samples per say. Just feels better. Same goes for 384k for me.
|
|
|
Post by Guitar on Nov 6, 2018 16:07:41 GMT -6
It's interesting. I just recorded the same strummed chords on an acoustic at 44.1 96 and 192. Trying to do it the same each time. Miktek C5. I found that switching back and forth there is sort of a crunch to the highs at 44.1. Its subtle. 96k it disappears and the sound gets more open and relaxed. 192 is quite close to 96 in terms of sound for my ears. Maybe just more of the same increases 96k saw but more subtle. It seems like something else is going on subtly but I can't put my finger on it. Pretty interesting. I feel like 96k is my favorite but I can't tell you why. Hm. Crap I just realized I devolved this thread where it wasn't meant to go. Sorry. Sigh. I seem to prefer 96 over 192 on most of my converters. 192 almost seems too airy or something. 96 is a little darker but without the crunch of 44.1.
|
|
|
Post by Mister Chase on Nov 6, 2018 19:53:10 GMT -6
I recall having heard Tony Maserati describe 441 as having a crunch to it, and that he actually likes it for what he does. I feel like I can hear that.
But I also haven't lost sight of where that lies in the hierarchy of priorities in sound.
I didn't really get the sense that 192 was more airy, but it almost seemed a little too smooth or something. I can't describe it, however it was not a truly scientific A/B so that goes out the window to a degree. Repeatedly though, I can hear the differences in the top.
|
|
|
Post by Ward on Nov 7, 2018 8:12:48 GMT -6
The inevitable next question is,
Are all the advantages of higher frequency sampling rates for tracking and mixing lost when mastering down to a final product at 44.1Khz?
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Nov 7, 2018 9:24:40 GMT -6
I've found with my chain, working at higher rates, but ending up downsampling to 44, to sound better than working at 44 all the way through, and that's after years of extensive testing. But every plugin and converter handles different sample rates with different results, so I don't think there can ever be a final answer to that question other than "it depends, do whatever sounds best on your system after testing" kinda thing.
|
|
|
Post by Blackdawg on Nov 7, 2018 9:56:41 GMT -6
Same here.
We render down to 96k for our finals from 384k and it sounds better than starting at 96k.
|
|
|
Post by jeremygillespie on Nov 7, 2018 10:07:23 GMT -6
For the mastering folks - do you convert down to 44.1 via software, or are you playing the higher sample rate music through your system and outboard and then printing through new converters to the lower sample rate on another system? Is there a difference?
|
|
|
Post by Ward on Nov 7, 2018 10:17:54 GMT -6
For the mastering folks - do you convert down to 44.1 via software, or are you playing the higher sample rate music through your system and outboard and then printing through new converters to the lower sample rate on another system? Is there a difference? Excellent question! Answers gleefully looked forward to.
|
|
|
Post by Blackdawg on Nov 7, 2018 10:21:57 GMT -6
For the mastering folks - do you convert down to 44.1 via software, or are you playing the higher sample rate music through your system and outboard and then printing through new converters to the lower sample rate on another system? Is there a difference? In my limited experience with this, they both sound different and can be used different. Some have different converters just for different sounds to hit with after the outboard gear. In which case having two systems can be handy as you don't have to do dither to get to a lower resolution file. That said, I prefer the sound of capturing the gear through the higher resolution then dithering down to 44.1k. More of the stuff gets captured that way from the gear I just ran stuff through and sounds more open. That's just me though. I'm just as curious as to what the real pro mastering guys say!
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Nov 7, 2018 12:02:53 GMT -6
For the mastering folks - do you convert down to 44.1 via software, or are you playing the higher sample rate music through your system and outboard and then printing through new converters to the lower sample rate on another system? Is there a difference? I always use offline SRC at the end of the process as it sounds better than realtime SRC, and only requires one computer system/interface card. I've heard of folks doing the transfer and capture at different rates with different converters (and/or computers) but have never tried it myself. I would expect you could achieve excellent results either way. Also be careful with terms like "dithering down to 44.1". We don't dither down to another sample rate, we sample rate convert, and then add dither if needed. Sorry to be pedantic, but we should be clear on our terminology.
|
|
|
Post by christopher on Nov 7, 2018 15:59:48 GMT -6
I use 96 now. Something related I've noticed finally.. when using Reaper I've always felt like my offline mixdowns are never as good as they were in the old ProTools realtime render. I've told myself it's my imagination over and over for years. Especially annoying because even when I do max settings and realtime render in Reaper, I feel like something gets changed! But finally I did some tests, dither vs no dither, all the diffferent render settings, etc. And what my ears told me: If I render to floating point wavs, it sounds exactly like what I hear. 32bitFP or 64bitFP, both seem to keep whatever it sounded like, offline or not. This is a huge relief, because it makes sense.. I'm hearing floating point when mixing, not fixed point. Still don't know why PT seemed to get it right, oh well.. more tests later in life, for now I'm happy.
|
|
|
Post by Bob Olhsson on Nov 7, 2018 16:47:07 GMT -6
Something people forget is that we aren't simply recording and playing back audio. A lot of common digital signal processing involves Nyquist filtering so we are often talking about multiple filters in series.
|
|
|
Post by jeremygillespie on Nov 7, 2018 17:13:46 GMT -6
For the mastering folks - do you convert down to 44.1 via software, or are you playing the higher sample rate music through your system and outboard and then printing through new converters to the lower sample rate on another system? Is there a difference? In my limited experience with this, they both sound different and can be used different. Some have different converters just for different sounds to hit with after the outboard gear. In which case having two systems can be handy as you don't have to do dither to get to a lower resolution file. That said, I prefer the sound of capturing the gear through the higher resolution then dithering down to 44.1k. More of the stuff gets captured that way from the gear I just ran stuff through and sounds more open. That's just me though. I'm just as curious as to what the real pro mastering guys say! I know people that do both but never really put any thought into what was better. I’ve also had mastering guys tell me that even if I’m working at 44.1, that I should print my mixes off the console onto another system running at min 96k. I guess it makes some sense, but I never worked in a room setup like that and it would just take too much time and make recalls a total bitch to deal with.
|
|
|
Post by popmann on Nov 7, 2018 21:14:21 GMT -6
You've never worked in a room with a two track of any kind? Man, I feel old...that was EVERY room EVER prior to ProTools. It doesn't really change much of anything about mix recall procedure...because it's capturing the mix--not DOING it...however you do recalls NOW...would be exactly the same with a two track running 192.
|
|
|
Post by jeremygillespie on Nov 7, 2018 21:30:25 GMT -6
You've never worked in a room with a two track of any kind? Man, I feel old...that was EVERY room EVER prior to ProTools. It doesn't really change much of anything about mix recall procedure...because it's capturing the mix--not DOING it...however you do recalls NOW...would be exactly the same with a two track running 192. I’ve spent 15 years in a room with an ATR102, and use it all the time. I was saying that it’s easier to print a mix into my original session as I can take that print track and put it in and out of input while listening to A/B and make sure nothing funky is going on after the recall. Always nice to be able to do that.
|
|
|
Post by Brian Campbell on Nov 7, 2018 21:42:39 GMT -6
I work mainly in post so 24/48k. The same for non post work in house and mostly everything that comes in the door for overdubs/mixing.
|
|
|
Post by cowboycoalminer on Nov 8, 2018 13:10:08 GMT -6
I'm late to the party but 44.1 here. I can't justify the taxing on the computer for diminishing gains IMO. Eq and mix choices compensate very well I think.
That said, higher sample rates do sound better to me. I just don't have the patience to mess with the back and forth dance between CPU and gain from higher sample rates.
|
|
|
Post by mulmany on Nov 8, 2018 15:52:50 GMT -6
48k,88.2k, or 96k depending on client and project. Primarily 88.2 or 96 for music and 48 for post.
|
|
|
Post by Mister Chase on Nov 8, 2018 16:07:36 GMT -6
I recall having heard Tony Maserati describe 441 as having a crunch to it, and that he actually likes it for what he does. I feel like I can hear that. But I also haven't lost sight of where that lies in the hierarchy of priorities in sound. I didn't really get the sense that 192 was more airy, but it almost seemed a little too smooth or something. I can't describe it, however it was not a truly scientific A/B so that goes out the window to a degree. Repeatedly though, I can hear the differences in the top. The other interesting thing is I tried boosting a high shelf on acoustic guitar at each sample rate and the higher rates were definitely easier on the ears. Pushing up at 44 was increasing the junk there. I may be a 96 and 192k convert now, at least for a lot of things.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Nov 10, 2018 14:22:48 GMT -6
To the question of how to SRC to target sample rate (or even reducing bit depth to lower target), i do not see the point in doing this by a digital algorithm, IF you use an analog DAC - ADC loop for mastering processing with analog gear anyway, at least if you have the possibilities to run them in different resolution/on 2 machines. But of course it depends on the sonic quality you can actually achieve with the converters. I guess this was more of a serious decision, when digital SRCs were not as good as they are nowadays. High quality digital SRC is now available for everyone without investing thousands of bucks, like in the times when the Weiss hardware was kind of one of few options ... i guess that makes it easier to not screw things up, no matter which way you choose...
|
|
|
Post by stormymondays on Nov 11, 2018 10:14:29 GMT -6
I’ve just realized that all I need is a couple more cables and I can run my I/O at 96k (Focusrite Clarett Octopre into RME Fireface 802). I might just have to switch now!
I wonder how taxing it would be on the computer though. And I will need a disk space calculator to figure out the extra cost in disk space.
Now, I’m in the middle of a project where drums and bass have already been recorded to 48k in Logic. Would you upsample and continue?
|
|