|
Post by Guitar on Jun 15, 2022 22:15:45 GMT -6
Wow, smashlord did jmoose call you? Eerie coincidence if not.
Rhetorical question: if one was "requested" or "verbally coerced" into delivering 96 KHz files... Why not just upsample your own shit with RX, foobar2000, r8brain, dB Power Amp or whatever and send them the exact thing they're asking for?
Would anyone even notice? Even blink an ear? Would anyone "conference call you with their friends" to be "concerned" about sound quality?
Something to think about.
|
|
|
Post by christophert on Jun 16, 2022 2:46:06 GMT -6
|
|
|
Post by indiehouse on Jun 16, 2022 5:20:50 GMT -6
So, the consensus is that it's a non-issue? It's starting to drift into 48 vs 96 territory, but I guess I'm more interested in any technical issues with 48 upsampled to 96? I plan on running these drum tracks through outboard, so there's going to be more conversion trips through my Aurora (n). Then again through the drum buss and master buss.
|
|
|
Post by Guitar on Jun 16, 2022 5:28:05 GMT -6
So, the consensus is that it's a non-issue? It's starting to drift into 48 vs 96 territory, but I guess I'm more interested in any technical issues with 48 upsampled to 96? I plan on running these drum tracks through outboard, so there's going to be more conversion trips through my Aurora (n). Then again through the drum buss and master buss. If you want it to be "absolutely perfect" in techincal terms, some SRC algorithms are cleaner than others. Here's a handy reference: src.infinitewave.ca/Altough, I wonder, if downsampling is "dirtier" (due to filtering,) than upsampling, which in my mind just adds a bunch of "nothing bits" above the previous cutoff frequency. Maybe @tomegatherion would know.
|
|
|
Post by svart on Jun 16, 2022 6:28:57 GMT -6
I worked with an artist who's album was primarily tracked by another producer where I was brought in to cut some overdubs because I was local to the artist. When things were tracked and it came time to mix, he liked my roughs better than his producer's mixes, so I ended up eventually getting the mix gig. I did a couple of the mixes that the artist loved and then got a conference call from the producer saying he was "concerned that we are compromising on quality" because I was sending back mixes @ 48khz instead of 96kHz. Meanwhile, while he was concerned with what our furry friends can hear, down in the audible spectrum, the sessions he was sending me were littered with bad edits and poorly intonnated guitars. So yeah.... seems like people get too hung up on the wrong things sometimes and miss the bigger picture. So many stories like this. I had a guitarist (seems they are generally more anal about small things than anyone else, second only to vocalists..) who would nitpick various small things in his tracks like single string plucks.. But not be concerned that string bends would land on a sour note or that his guitar intonation was off at higher notes. Or a vocalist that was super concerned about a hard T sound in one word on one track and we did maybe 30 takes over a day on this one phrase and I could barely tell a difference between the takes.. In the end the artist was so upset it wasn't working she was about to ditch the whole tune out of anger. I ended up ducking the T a little in automation with a little de-esser and it worked well enough she was OK with it.
|
|
|
Post by svart on Jun 16, 2022 7:14:38 GMT -6
So, the consensus is that it's a non-issue? It's starting to drift into 48 vs 96 territory, but I guess I'm more interested in any technical issues with 48 upsampled to 96? I plan on running these drum tracks through outboard, so there's going to be more conversion trips through my Aurora (n). Then again through the drum buss and master buss. If you want it to be "absolutely perfect" in techincal terms, some SRC algorithms are cleaner than others. Here's a handy reference: src.infinitewave.ca/Altough, I wonder, if downsampling is "dirtier" (due to filtering,) than upsampling, which in my mind just adds a bunch of "nothing bits" above the previous cutoff frequency. Maybe @tomegatherion would know. So here's how it works. Upsampling, or adding more sample points to an existing waveform, would normally analyze the existing points and between each point would estimate a curve of some type. The algorithm would then estimate the value of the new sample based on this interpolation curve. The drawback is that the quality of the interpolation matters greatly. Some interpolators will use high order polynomial curves (with matching the order value to the signal being critical), which are more accurate but uses more resources and take more time. Some will do linear interpolation which is much faster but less accurate. Either way, there will be some error that exists, which is the difference between the estimated(interpolated) point value and what a real sample value should be. This error in value is also quantization error. A way to reduce this error is to oversample the original signal at some very high samplerate so that the quantization steps are very small at the cost of much higher processing and resource requirements. Once you get a higher accuracy signal that has been upsampled, you can then reduce the sampling rate once again. This works for A/D converters as well as software sample rate conversion. You then have to Low Pass Filter so that you remove the small amounts of harmonics that appear at higher frequencies from the resulting quantization steps. Downsampling is effectively the inverse. There's a dozen different ways to do it but the most common algorithm will decimate, or remove certain samples based on a ratio. This filters the signal then removes the samples required and may or may not apply more filtering. Neither one would realistically be "dirtier" than the other if properly done but upsampling would have more chances to introduce error since the process is more intensive and introduces "ghost" samples.
|
|
|
Post by Guitar on Jun 16, 2022 10:16:53 GMT -6
Wow, svart , really nice! Definitely mathy but not too mathy. This link gets really mathy but it's fascinating, in a way: www.sciencedirect.com/topics/engineering/quantization-step-sizeSection 4.2.2 is what I read. I like their little graph of quantization error, this funny Egyptian looking little curve. They describe it as a white noise with no mean, dependent both on bit depth and sample rate, in a way I don't fully grasp (and might not need to.) The "stair steps" doomsday analog cult were right, in a way. No, we don't play back stair steps. But yes, we hear the white noise of their residue.
|
|
|
Post by svart on Jun 16, 2022 11:06:10 GMT -6
Wow, svart , really nice! Definitely mathy but not too mathy. This link gets really mathy but it's fascinating, in a way: www.sciencedirect.com/topics/engineering/quantization-step-sizeSection 4.2.2 is what I read. I like their little graph of quantization error, this funny Egyptian looking little curve. They describe it as a white noise with no mean, dependent both on bit depth and sample rate, in a way I don't fully grasp (and might not need to.) The "stair steps" doomsday analog cult were right, in a way. No, we don't play back stair steps. But yes, we hear the white noise of their residue. But we don't because the noise would be in the form of harmonic content related to the quantization frequency step size, which should/would be much higher than the carrier wave being reconstructed. The digital filter(or analog in cases of a DAC) should be designed to filter off the residual harmonics.
|
|
|
Post by Guitar on Jun 16, 2022 11:09:03 GMT -6
Upsampling... there are no nothing bits, is what I realize. Any smaller bit, any smaller "sample" must still be placed in the "right" place. Inaudible silence is actually a perfection, that is hard for a machine to interpolate on its own, based on previous measurement (recordings.) Any deviation from inaudible perfect silence results in measurable, or hearable noise. Whether upsampling, or down sampling. You can hear the algorithm. Or sometimes the algorithm is so good it can reduce noise from other compontents!
|
|
|
Post by the other mark williams on Jun 16, 2022 11:40:03 GMT -6
So, the consensus is that it's a non-issue? It's starting to drift into 48 vs 96 territory, but I guess I'm more interested in any technical issues with 48 upsampled to 96? I plan on running these drum tracks through outboard, so there's going to be more conversion trips through my Aurora (n). Then again through the drum buss and master buss. Yes, this is the consensus. Sure, it would be better if he were able to record at the same sample rate the rest of the project is recorded at, but if you like the sounds and the performance, don't sweat it.
|
|
|
Post by jmoose on Jun 16, 2022 14:29:08 GMT -6
Yeah, I’m a bit bummed to have to deal with it. Shouldn’t seem like 96k would be a hassle these days. Its a buzzmer I get that and yeah... when even a $200 presonus interface can run at 96k you have to wonder? And FWIW I was running 88/96 back in like, 2005 on a pentium 4 box with a whopping 2 gigs of ram & pile of 250 gig drives. If it could be done then it can still be done now. As long as you dig the delivered sounds & performances that's all that matters... so do you like his tones & whatnot? What's the alternative? Throw 'em out & recut? Have you asked him specifically what happened & why he had problems? My thought is maybe possibly there's a teachable moment in here... He was tracking drums right? Not mixing so there shouldn't be many... if any plugs involved just microphones capturing sounds? How to put this..? Over the last few years across a handful of projects that have passed through... I've sort of noticed this trend where sometimes, the people hired to cut tracks on their instrument in their home studio... sometimes recording isn't really their thing. Meaning they aren't very technical about it. Everyone has a different journey with music right? Not everyone is in the same place & has the same background. Some people approach recording, cutting tracks more as a means to an end rather then being a focus. Like their more concerned with playing the instrument and recording is kind of secondary. Their musicians first, engineer maybe a distant third. Meaning in a case like this... maybe they don't really know how to do things like adjust buffer sizes & set up their rig? Seen it a hundred times. I have a friend who a few years ago started offering to cut tracks on fiver & his various circles. Started off fine but, at a certain point when he started getting better gigs he started to get comments about sound quality. He started asking questions about how to improve his process and... while he's a great musician his knowledge of recording was barely over the "place mic adjust gain hit record" threshold. Maybe its a case of something like this? Where if this is the first 88/96 thing he's ever gotten maybe he just didn't know how to adjust clocks & so on? I'd always love to imagine there's a cause and a solution. Maybe not on this song, too late... but going forward?
|
|
|
Post by Guitar on Jun 16, 2022 15:01:30 GMT -6
jmoose I encounter this all the time in local, physical conversations. Countless times. Lots of serious musicians I know, lots of amateurs who think they're serious, lots of recording and so on. The level of technical not-knowing-how is extraordinary! Some people might know "enough to be dangerous," or even "enough to do some really cool recordings, can you help me get these across the finish line, or just tell me if you like them." But it's mostly like you say: music first, recording some distance after that. This manifests frequently in terms of confusion, misunderstanding, even "telling you what," sometimes, that's a funny one to listen to. For me, the "know-it-all, be precise-specific," anyway. Most of these people (not all!) on the other hand have great musical ideas, worth being recorded, and that's a huge deal. You might find the reverse "syndrome" in recording-obsessives, occasionally. Over-emphasis on technique and lack of musical ideation. You might find that some self-professed "audio experts" are insufferable assholes, the "that guy" sound guy nightmares of the world that every female musician and most male musicians have encountered multiple times, that actually instill in some people a "fear" of recording! Some of these people who you might sometimes assume know "less" than you, if you pay attention, can expose you to new ideas, new ways of working! Some of these people might be literally impossible to work with, not knowing how to finish anything under any circumstances, saying they want things finished, and doing literally everything they can to destroy this from occurring, or doing nothing at all. So I think we can all learn from each other, and it's good to be understanding of misunderstanding, and answer it plainly without judgement. Or sometimes just to let things sliiiiide, live to fight another day. I think it's well understood that "making music with people" can be a huge hassle in terms of "personality management" or whatever you want to call it, and this can go both ways, of course, so try to be wise and try to choose who you work with wisely, when possible! And try to enjoy the people you're working with, otherwise what's the point? In my mind? The high sample rate thing is a hangup, a red-herring. No leading, no following. A non-issue. Just some desire to control the situation, to be the one who is in control.
|
|
|
Post by smashlord on Jun 17, 2022 10:00:33 GMT -6
So, the consensus is that it's a non-issue? It's starting to drift into 48 vs 96 territory, but I guess I'm more interested in any technical issues with 48 upsampled to 96? I plan on running these drum tracks through outboard, so there's going to be more conversion trips through my Aurora (n). Then again through the drum buss and master buss. Yes, a non-issue. Your Auroras are great converters and I would not worry about any potential undesirable artifacts from DA/AD roundtrips. I have also seen quite a few experiments where folks have run tracks 50+ times through DA/AD (with no processing) and there was no perceptible difference afterwards.
|
|
|
Post by smashlord on Jun 17, 2022 10:30:29 GMT -6
Yeah, I’m a bit bummed to have to deal with it. Shouldn’t seem like 96k would be a hassle these days. Its a buzzmer I get that and yeah... when even a $200 presonus interface can run at 96k you have to wonder? And FWIW I was running 88/96 back in like, 2005 on a pentium 4 box with a whopping 2 gigs of ram & pile of 250 gig drives. If it could be done then it can still be done now. There are practical considerations for running @ 48Khz. For one, I, and most of my colleagues who do this professionally have 7-10 active projects with at least 5-12 songs going at any one time. 96Khz files are huge and thats a lot of hard drive real estate when you consider you need it in 3 places at all times (2 physical and 1 cloud for me). Though others may disagree, I feel there are better ways to invest in my setup that make more of a tangible difference in the final product that I deliver to the artist. Also, my system is most stable @ 48Khz when running a large session with plug ins, MIDI, and H/W inserts. I completely understand what you are saying about your set up in 2005, but you also have to consider as hardware has gotten more powerful, software developers have been taking advantage of that and writing more and more resource intensive code to improve their products....."living up to your salary", if you will. If its a relatively small session and it's sent to me @96khz, I'll stay there. If its got 60 or more tracks (common) and has VIs and such, I convert. If the session is always hanging, I am going to loathe working on it and not approach it with the same enthusiasm, which does more of a detriment to the final product than any potential lack of frequencies that we all lost in our late teens anyways.
|
|
|
Post by notneeson on Jun 17, 2022 12:35:40 GMT -6
So, the consensus is that it's a non-issue? It's starting to drift into 48 vs 96 territory, but I guess I'm more interested in any technical issues with 48 upsampled to 96? I plan on running these drum tracks through outboard, so there's going to be more conversion trips through my Aurora (n). Then again through the drum buss and master buss. Yes, a non-issue. Your Auroras are great converters and I would not worry about any potential undesirable artifacts from DA/AD roundtrips. I have also seen quite a few experiments where folks have run tracks 50+ times through DA/AD (with no processing) and there was no perceptible difference afterwards. The drum sounds and performance should kick ass, that’s what you’re paying for. 48k is not a barrier to kick ass drum sounds. Round trip processing is a non issue even at 44.1 with good conversion, it takes a ridiculous number of passes to really degrade and the outboard is worth it or you wouldn’t bother anyway.
|
|