mhep
Full Member
Posts: 36
|
Post by mhep on Mar 11, 2018 12:01:33 GMT -6
There are a few simple rules of thumb:
• Properly designed converters should have no appreciable difference between sample rates between 44.1kHz and 96kHz
• There's zero mathematical benefit from recording at higher sample rates for audio work. It's simply Nyquist.
• Sample rate conversion does not benefit from doublings or halvings of math unless you have a synchronous sample rate converting hardware unit, which will likely sound worse than the software sample rate converters UNLESS you're doing a simple division conversion
• Software SRC is asynchronous and upsamples to the common multiple before dividing to the target sample rate, so all sample rates convert to all others equally
• 96kHz doesn't let the converter draw a more perfect waveform than 44.1kHz. They will be identical because there's only one mathematical possibility
• Comparing 44.1kHz and 96kHz cannot be done subjectively with your own voice or acoustic instrument because of the extreme color from acoustic comb filtering due the latency change. It can only be compared from double SRC opposite sample rate comparison.
• Many plugins already upsample and will receive no benefit from higher sample rates
• SOME plugins that do NOT upsample will benefit from processing at higher sample rates
• Some inexpensive or older (10–15 year old) converters and interfaces can benefit by recording at higher sample rates due to poor filter design that affects lower sample rates when it shouldn't
• The vast majority of professionals and professional studios operate at 48kHz or below.
|
|
|
Post by Martin John Butler on Mar 11, 2018 20:31:15 GMT -6
I'm sure some of what you say is true mhep, but "simple", I don't think so. If the plug-ins that you have do benefit from higher sample rates, than higher sample rates are better, for the time being at least.
Since the CD format came out at 44.1, even those who created and developed it stated they wanted higher sample rates and would have used them as standard if the bandwidth was cheaply available at that time.
Mathmatics are not the only thing affecting sound quality. I've always said that we can't measure everything yet, and that sometimes what we can't demonstrably hear, we feel. So I'm just saying the jury's still out. SACD sounds way better than CD, so it would seem the sample rate does matter, at least in that situation.
|
|
mhep
Full Member
Posts: 36
|
Post by mhep on Mar 11, 2018 21:37:40 GMT -6
SACD/HDCD/DVDA is usually a different master, or, at least different transfer.
The only way to know if your system benefits from higher sample rates is to beg or borrow a second, identical unit, split the signal to each and record simultaneously to both, with one at 44 and one at 96. THEN, you have to SRC one of each to the opposite sample rate and blind AB against the original and SRC'd version from the other sample rate. Most plugins that really benefit from higher sample rate already upsample.
|
|
|
Post by gouge on Mar 12, 2018 5:47:14 GMT -6
Plugins may benefit from higher sample rates but a good percentage of them won't work beyond 96k which is annoying.
It irks me that to mix itb I am forced to use 96k.
|
|
|
Post by Martin John Butler on Mar 12, 2018 8:19:03 GMT -6
I'd be glad if I could use 96k without my system getting the hiccups ;-)
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 12, 2018 9:34:25 GMT -6
There are a few simple rules of thumb: Ahem, • Properly designed converters should have no appreciable difference between sample rates between 44.1kHz and 96kHz Says who? Exactly what do you mean? A 'properly designed' converter will have enhanced frequency response and fewer issues near nyquist. When nyquist is an octave higher, the issues at 20K are significantly-reduced. We really have to get past 50-year-old research that sets a magical limit at 20K. Whether or not you can hear above 20K is only part of it. You have enhanced time resolution and improved imaging when sample rate goes up. You have few foldback artifacts to plague the rest of your mix. • There's zero mathematical benefit from recording at higher sample rates for audio work. It's simply Nyquist. Not at all sure what that means. At any given audio frequency you have double the points (if you're graphing) at double the sample rate. If you have 'perfect' filters, perhaps it's minimized. But there aren't any perfect filters. • Software SRC is asynchronous and upsamples to the common multiple before dividing to the target sample rate, so all sample rates convert to all others equally Almost. In any SRC conversion you still have a small amount of distortion because the word size is not infinite. Good SRC is very good indeed, but the resultant signal will be inferior to the signal that went in. • 96kHz doesn't let the converter draw a more perfect waveform than 44.1kHz. Of course it does. You have twice as many points and a smoother depiction of the waveform. While the anti-aliasing filter (if perfect) may make them equivalent at 20K, that filter will barely touch the signal if it's operating at double the frequency. Whether you will hear the difference is arguable, but it's nice to keep all that schmutz out until the very last pass. • Comparing 44.1kHz and 96kHz cannot be done subjectively with your own voice or acoustic instrument because of the extreme color from acoustic comb filtering due the latency change. It can only be compared from double SRC opposite sample rate comparison. You need to clarify that. Are you saying the only way to compare two signals is to pass them both through sample rate conversion? That's novel, to say the least. • Many plugins already upsample and will receive no benefit from higher sample rates • SOME plugins that do NOT upsample will benefit from processing at higher sample rates That depends on what the plugin actually does. Even an upsampling plugin will be smart enough not to upsample if it detects the DAW is already operating at high sample rate. • Some inexpensive or older (10–15 year old) converters and interfaces can benefit by recording at higher sample rates due to poor filter design that affects lower sample rates when it shouldn't Maybe yes, maybe no. But if you're running on a 15-year old converter it's really time to upgrade. Even current-day crap conversion is better than decade-old premium. • The vast majority of professionals and professional studios operate at 48kHz or below. Can't let that stand. It's true that a lot of home studios may operate at low sample rates. But that's a function of budget. All gear has to last a lot longer and it's more difficult to upgrade on a regular basis. But at the high-end, they've been operating at high sample rates for years. Capitol was regularly doing everything at 96K ten years ago (they told me that to my face, ten years ago). I'm sure the same was true of Air, Abbey Road and others. Wouldn't surprise me to know of many more. Most of the classical guys I work with record at 352.8K and have done so for several years. I know film score recordists and mixers who run at 192K, downsampling only at the very end. You'll find that an increasing number of BluRays offer high-resolution tracks at 24/96. It appears that a very large number of professionals know there's a benefit.
|
|
|
Post by Martin John Butler on Mar 12, 2018 12:49:08 GMT -6
I knew there was more to it ;-) Thanks Michael, that was interesting.
|
|
|
Post by popmann on Mar 12, 2018 14:22:18 GMT -6
SACD is NOT "24bit"....it's closer to 20bit from memory....or was it 18bit?....it IS however effectively 384khz in the time domain and 88.2 in the frequency domain. Of course the sample rate matters.
|
|
|
Post by adamjbrass on Mar 12, 2018 14:38:13 GMT -6
I thought SACD audio is stored as 1-Bit DSD
Of course, A Hybrid SACD also has a PCM layer
|
|
|
Post by popmann on Mar 12, 2018 15:32:16 GMT -6
It is. It also by nature has no "sample rate"....so all of the above is equivalents.
|
|
|
Post by adamjbrass on Mar 12, 2018 15:37:15 GMT -6
2.8224MHz
x4 times that of a Redbook CD
|
|
|
Post by popmann on Mar 12, 2018 15:39:53 GMT -6
Except it's not a PCM snapshot being taken at that rate....it's not the same thing. What's your point?
|
|
mhep
Full Member
Posts: 36
|
Post by mhep on Mar 12, 2018 16:28:42 GMT -6
Hi, Michael, I realize and recognize you have a lot of expertise to bring to the table. You are perpetuating a few myths, however. Doubling samples does not equate to doubling accuracy. Two sample points are all that are required to draw a *perfect* waveform with the reconstruction filter. Four sample points makes absolutely zero improvement, let alone eight. We need to dissociate the visual aspect of sampling. Simply sampling at twice the frequency of the sound is sufficient to draw a perfect waveform, given proper filtering to restrict the frequency range being sampled. Adding more samples within the audible limit only takes up more space and uses more bandwidth. While it may be true that 20kHz is an arbitrary cutoff most gear is manufactured and measured only to that frequency and lots of gear drops dramatically at both ends of the spectrum. As to my comment about properly designed converters, that would mean good, properly designed filters and, yes, there's no appreciable difference between the lower and higher sample rates on those systems. My SRC comment was targeted at a mythological benefit of recording at double the final delivery sample rate since down sampling would be simpler. That's not how our SRC works, for the most part. Again, 96kHz sampling does NOT draw a more perfect waveform than 44.1kHz between 20Hz and 20kHz. It simply allows sampling of frequencies beyond the range that lower sample rates can sample. It also allows sloppier filter design to stay out of the audible spectrum (going back to the properly designed converter comment). On the contrary to your statement about lower sample rates being more used by home studios than pro studios you'll find that the opposite is actually correct. 48kHz and below is still MUCH more widely used by professionals than any higher sample rate. Look at all the polls of top tier professionals. 90% use 48kHz and lower. Nearly ALL top tier mixers use 48kHz and lower. There are a few simple rules of thumb: Ahem, • Properly designed converters should have no appreciable difference between sample rates between 44.1kHz and 96kHz Says who? Exactly what do you mean? A 'properly designed' converter will have enhanced frequency response and fewer issues near nyquist. When nyquist is an octave higher, the issues at 20K are significantly-reduced. We really have to get past 50-year-old research that sets a magical limit at 20K. Whether or not you can hear above 20K is only part of it. You have enhanced time resolution and improved imaging when sample rate goes up. You have few foldback artifacts to plague the rest of your mix. • There's zero mathematical benefit from recording at higher sample rates for audio work. It's simply Nyquist. Not at all sure what that means. At any given audio frequency you have double the points (if you're graphing) at double the sample rate. If you have 'perfect' filters, perhaps it's minimized. But there aren't any perfect filters. • Software SRC is asynchronous and upsamples to the common multiple before dividing to the target sample rate, so all sample rates convert to all others equally Almost. In any SRC conversion you still have a small amount of distortion because the word size is not infinite. Good SRC is very good indeed, but the resultant signal will be inferior to the signal that went in. • 96kHz doesn't let the converter draw a more perfect waveform than 44.1kHz. Of course it does. You have twice as many points and a smoother depiction of the waveform. While the anti-aliasing filter (if perfect) may make them equivalent at 20K, that filter will barely touch the signal if it's operating at double the frequency. Whether you will hear the difference is arguable, but it's nice to keep all that schmutz out until the very last pass. • Comparing 44.1kHz and 96kHz cannot be done subjectively with your own voice or acoustic instrument because of the extreme color from acoustic comb filtering due the latency change. It can only be compared from double SRC opposite sample rate comparison. You need to clarify that. Are you saying the only way to compare two signals is to pass them both through sample rate conversion? That's novel, to say the least. • Many plugins already upsample and will receive no benefit from higher sample rates • SOME plugins that do NOT upsample will benefit from processing at higher sample rates That depends on what the plugin actually does. Even an upsampling plugin will be smart enough not to upsample if it detects the DAW is already operating at high sample rate. • Some inexpensive or older (10–15 year old) converters and interfaces can benefit by recording at higher sample rates due to poor filter design that affects lower sample rates when it shouldn't Maybe yes, maybe no. But if you're running on a 15-year old converter it's really time to upgrade. Even current-day crap conversion is better than decade-old premium. • The vast majority of professionals and professional studios operate at 48kHz or below. Can't let that stand. It's true that a lot of home studios may operate at low sample rates. But that's a function of budget. All gear has to last a lot longer and it's more difficult to upgrade on a regular basis. But at the high-end, they've been operating at high sample rates for years. Capitol was regularly doing everything at 96K ten years ago (they told me that to my face, ten years ago). I'm sure the same was true of Air, Abbey Road and others. Wouldn't surprise me to know of many more. Most of the classical guys I work with record at 352.8K and have done so for several years. I know film score recordists and mixers who run at 192K, downsampling only at the very end. You'll find that an increasing number of BluRays offer high-resolution tracks at 24/96. It appears that a very large number of professionals know there's a benefit. Edited for correcting autocorrect and for clarification.
|
|
|
Post by adamjbrass on Mar 12, 2018 16:29:15 GMT -6
My point is, "no sample rate" of DSD is not correct. There are three rates of DSD being used, and more on the way.
|
|
mhep
Full Member
Posts: 36
|
Post by mhep on Mar 12, 2018 16:32:55 GMT -6
SACD is NOT "24bit"....it's closer to 20bit from memory....or was it 18bit?....it IS however effectively 384khz in the time domain and 88.2 in the frequency domain. Of course the sample rate matters. Correct. I was 100% mistaken on SACD and was thinking of DVD-A.
|
|
|
Post by cowboycoalminer on Mar 12, 2018 16:36:08 GMT -6
There are a few simple rules of thumb: Ahem, • Properly designed converters should have no appreciable difference between sample rates between 44.1kHz and 96kHz Says who? Exactly what do you mean? A 'properly designed' converter will have enhanced frequency response and fewer issues near nyquist. When nyquist is an octave higher, the issues at 20K are significantly-reduced. We really have to get past 50-year-old research that sets a magical limit at 20K. Whether or not you can hear above 20K is only part of it. You have enhanced time resolution and improved imaging when sample rate goes up. You have few foldback artifacts to plague the rest of your mix. • There's zero mathematical benefit from recording at higher sample rates for audio work. It's simply Nyquist. Not at all sure what that means. At any given audio frequency you have double the points (if you're graphing) at double the sample rate. If you have 'perfect' filters, perhaps it's minimized. But there aren't any perfect filters. • Software SRC is asynchronous and upsamples to the common multiple before dividing to the target sample rate, so all sample rates convert to all others equally Almost. In any SRC conversion you still have a small amount of distortion because the word size is not infinite. Good SRC is very good indeed, but the resultant signal will be inferior to the signal that went in. • 96kHz doesn't let the converter draw a more perfect waveform than 44.1kHz. Of course it does. You have twice as many points and a smoother depiction of the waveform. While the anti-aliasing filter (if perfect) may make them equivalent at 20K, that filter will barely touch the signal if it's operating at double the frequency. Whether you will hear the difference is arguable, but it's nice to keep all that schmutz out until the very last pass. • Comparing 44.1kHz and 96kHz cannot be done subjectively with your own voice or acoustic instrument because of the extreme color from acoustic comb filtering due the latency change. It can only be compared from double SRC opposite sample rate comparison. You need to clarify that. Are you saying the only way to compare two signals is to pass them both through sample rate conversion? That's novel, to say the least. • Many plugins already upsample and will receive no benefit from higher sample rates • SOME plugins that do NOT upsample will benefit from processing at higher sample rates That depends on what the plugin actually does. Even an upsampling plugin will be smart enough not to upsample if it detects the DAW is already operating at high sample rate. • Some inexpensive or older (10–15 year old) converters and interfaces can benefit by recording at higher sample rates due to poor filter design that affects lower sample rates when it shouldn't Maybe yes, maybe no. But if you're running on a 15-year old converter it's really time to upgrade. Even current-day crap conversion is better than decade-old premium. • The vast majority of professionals and professional studios operate at 48kHz or below. Can't let that stand. It's true that a lot of home studios may operate at low sample rates. But that's a function of budget. All gear has to last a lot longer and it's more difficult to upgrade on a regular basis. But at the high-end, they've been operating at high sample rates for years. Capitol was regularly doing everything at 96K ten years ago (they told me that to my face, ten years ago). I'm sure the same was true of Air, Abbey Road and others. Wouldn't surprise me to know of many more. Most of the classical guys I work with record at 352.8K and have done so for several years. I know film score recordists and mixers who run at 192K, downsampling only at the very end. You'll find that an increasing number of BluRays offer high-resolution tracks at 24/96. It appears that a very large number of professionals know there's a benefit. This all makes sense to me. And why I believe with all my heart that tape sounds better. No math. Complete “it is what it is” so to speak. But alas, tape is gone and will never come back. The efficiencies are just too great with digital. This makes this topic a very important one for those of us who make music. Great topic. Keep it going!
|
|
|
Post by Martin John Butler on Mar 12, 2018 17:18:59 GMT -6
I do most of my work here in my NYC apartment at 48k. For my next projects I'd like to do it at 96k, but I don't think my computer will handle 96k if it's a large session. I use 48k because somewhere in my mind I think of syncing to video. I can't explain it, but after all the tracking, plug-ins and mixing, 48k seems to get me better sounding results than 44.1. I doubt I'd hear anything different in a blind test, but over time, track after track after track, something's better at 48k.
Correct me if I'm wrong, but I think Chris Stapleton's 1 album was done to tape as was one of Jason Isbell's. There are a few holdouts, but yes, it's basically gone. If I ever get to be satisfied with a home system, I'll look into tape again. Right now I'd like a Symphony, Bricasti, CLB1, Retro 176, a few new mics, and maybe a board, so I better win the lottery soon because music ain't paying that much lately.
|
|
mhep
Full Member
Posts: 36
|
Post by mhep on Mar 12, 2018 17:57:17 GMT -6
You need to clarify that. Are you saying the only way to compare two signals is to pass them both through sample rate conversion? That's novel, to say the least. No, please see post 78 for an explanation.
|
|
|
Post by adamjbrass on Mar 12, 2018 19:50:01 GMT -6
I do most of my work here in my NYC apartment at 48k. For my next projects I'd like to do it at 96k, but I don't think my computer will handle 96k if it's a large session. I use 48k because somewhere in my mind I think of syncing to video. I can't explain it, but after all the tracking, plug-ins and mixing, 48k seems to get me better sounding results than 44.1. I doubt I'd hear anything different in a blind test, but over time, track after track after track, something's better at 48k. Correct me if I'm wrong, but I think Chris Stapleton's 1 album was done to tape as was one of Jason Isbell's. There are a few holdouts, but yes, it's basically gone. If I ever get to be satisfied with a home system, I'll look into tape again. Right now I'd like a Symphony, Bricasti, CLB1, Retro 176, a few new mics, and maybe a board, so I better win the lottery soon because music ain't paying that much lately. 48k seems to "bump" the Filter just a wee bit out, 96k puts it at 48kHz, which is a bit more than a wee... This can be a cumulative effect.
|
|
|
Post by Quint on Mar 12, 2018 20:31:43 GMT -6
I'm 96k all the way, for most of the reasons mentioned in this thread already. I would consider trying 192k on multi-track if my system could hack it. That might be a reality for me to try in the near future, depending on how capable my new PC build ultimately proves to be. 96k and higher has a lot of benefits, beyond just the the sonic benefits, reduced latency being one of them. I'm also gearing up to try quad rate dsd for mixdown with a Tascam DA-3000. I'm really looking forward to giving that a go. At the very least, I think 96k sounds better than lower sample rates. I've compared them, and I prefer 96k over lower rates, even on different converters. Don't knock it until you try it. I'm all for the benefits of reduced latency and getting those supposedly "perfect" filters way the hell away from anything even remotely audible. A pretty great discussion with Dave Amels about "minimally acceptable" sample rates found here: www.gearslutz.com/board/gear-shoot-outs-sound-file-comparisons-audio-tests/654876-ampex-atr-102-anamod-ats-1-uad-waves-processed-files-11.html
|
|
|
Post by gouge on Mar 13, 2018 4:13:30 GMT -6
I'd be glad if I could use 96k without my system getting the hiccups ;-) I'm using 6-7 year old pc and have no issue loading 60 plugs at 96k
|
|
|
Post by Martin John Butler on Mar 13, 2018 8:09:11 GMT -6
I'm hesitant because I had way too many freezes and system overloads when I had an i5 mac, so I don't want to risk having issues later in the game after I've done 10 or 15 tracks. I have a 2011 iMac. It's a 3.4 Ghz i7 running Sierra with 32 G of memory.
|
|
|
Post by svart on Mar 13, 2018 9:05:33 GMT -6
I'd be glad if I could use 96k without my system getting the hiccups ;-) I'm using 6-7 year old pc and have no issue loading 60 plugs at 96k My "new" recording PC is about 3 years old and I can do that many at 88.2k/24 as well. I'm running reaper though, which doesn't have near the amount of overhead that other daws do.
|
|
|
Post by popmann on Mar 13, 2018 9:37:52 GMT -6
MJB,
You will continue to have issues with HD. NOT because of HD audio....but, because of your not not understanding where the bottleneck is in the system....Your (old) i5 is faster than anything I've mixed full 40 track 24/88 projects on.
No matter if you're using VIs or recording HD audio, or likely BOTH....the hard drive(s) is the bottleneck. Has been for as long as I've been using VIs (so 20 years)....now, back then, you ALSO had a 32bit RAM bottleneck at the extreme of usage....but, for a guy making folk rock what do you have a piano and drums kit? 32bit RAM (so 4gb) wasn't an issue until SD3.
Your BEST technical option is an internal 1TB SATA SSD to replace your magnetic....but, that's NOT something you can do yourself-I've done it, and it's super involved and not for the tech faint of heart....buying a Thunderbolt drive bay to put SATA SSDs in is.....but, buy a multi drive box, because SATA SSDs are not magic--I would GUESS one would be fine with your OS/Apps on the internal magnetic....but, I don't know how heavy your VI usage is and whether you have Internet in 192. It's better to buy the multi drive box....you want have room to grow. Since Apple won't LET you grow inside the case.
Update for Svart's post: Logic doesn't have tangibly more overhead than Reaper. IME. On the same Mac. Logic is actually running pretty bare metal for better AND worse. I mention that simply to point out that, while there's likely some overhead difference, it's not "the" reason he's having this trouble.
|
|
|
Post by svart on Mar 13, 2018 11:30:44 GMT -6
MJB, You will continue to have issues with HD. NOT because of HD audio....but, because of your not not understanding where the bottleneck is in the system....Your (old) i5 is faster than anything I've mixed full 40 track 24/88 projects on. No matter if you're using VIs or recording HD audio, or likely BOTH....the hard drive(s) is the bottleneck. Has been for as long as I've been using VIs (so 20 years)....now, back then, you ALSO had a 32bit RAM bottleneck at the extreme of usage....but, for a guy making folk rock what do you have a piano and drums kit? 32bit RAM (so 4gb) wasn't an issue until SD3. Your BEST technical option is an internal 1TB SATA SSD to replace your magnetic....but, that's NOT something you can do yourself-I've done it, and it's super involved and not for the tech faint of heart....buying a Thunderbolt drive bay to put SATA SSDs in is.....but, buy a multi drive box, because SATA SSDs are not magic--I would GUESS one would be fine with your OS/Apps on the internal magnetic....but, I don't know how heavy your VI usage is and whether you have Internet in 192. It's better to buy the multi drive box....you want have room to grow. Since Apple won't LET you grow inside the case. Update for Svart's post: Logic doesn't have tangibly more overhead than Reaper. IME. On the same Mac. Logic is actually running pretty bare metal for better AND worse. I mention that simply to point out that, while there's likely some overhead difference, it's not "the" reason he's having this trouble. Ok, but I'm using bone standard SATA spinning drives too. No issues here with speed, no SSD needed to get this performance.
|
|