The No.1 Website for Pro Audio
Why does everyone want 96khz ?
Old 18th November 2019
  #1
Lives for gear
 
TS-12's Avatar
Why does everyone want 96khz ?

didn't pay much attention to this much,
But now after Behringer wing announcement, almost in every post about it on Facebook, Instagram and forums there's at least one question "will it be 96khz".

Now reading on various audio forums and here on GS looks like everyone is recording at 96khz these days.


Besides lower latency and larger recorded file sizes,
Can you guys really hear the difference in 96khz over 44.1khz ?
Old 18th November 2019
  #2
Gear Guru
 
Brent Hahn's Avatar
 

The only people I know who do 96 and 192 are classical folks and sound designers. Sound designers do it because you can slow stuff way, way down without it getting weird. Like how you turn pigs and lions and horses into monsters and space aliens.
18
Share
Old 18th November 2019
  #3
Gear Nut
Generally working in Live sound you want higher resolution as it reduces latency on FX, Routing and Monitoring for IEMs and less to do with the frequency rage.

In the studio, I absolutely hear a difference, but it's more to do with how smooth the top end is at 96kHz.
19
Share
Old 18th November 2019
  #4
Lives for gear
 
Oldone's Avatar
Humans always want to go faster. It's in our genes.
17
Share
Old 18th November 2019
  #5
Lives for gear
 
TS-12's Avatar
Quote:
Originally Posted by Oldone View Post
Humans always want to go faster. It's in our genes.
how does 96khz make humans go faster ?
6
Share
Old 18th November 2019
  #6
Lives for gear
 
Mushy Mushy's Avatar
 

People just want to bitch and moan. They just look for a feature that is "missing" and focus relentlessly on it regardless if they actually need it.
32
Share
Old 18th November 2019
  #7
Lives for gear
 
mike vee's Avatar
because it sounds better and is lower latency...two of the most important things possible with dAW.
26
Share
Old 18th November 2019
  #8
Lives for gear
 
Papanate's Avatar
 

Quote:
Originally Posted by TS-12 View Post
didn't pay much attention to this much,
But now after Behringer wing announcement, almost in every post about it on Facebook, Instagram and forums there's at least one question "will it be 96khz".


Besides lower latency and larger recorded file sizes,
Can you guys really hear the difference in 96khz over 44.1khz ?
In the Studio the difference is readily apparent for me. Tighter low end and a smother clarity to everything in the upper end.
3
Share
Old 18th November 2019
  #9
Lives for gear
 

Last year, 2018, I recorded @ 96 but don’t have hardware to simultaneously convert to 48. My Aurora(n) can’t operate with different sample rates for input and output.

Supposing a great converter would do the best work, I dropped to 48 since I distribute @ 48.

Do you end up downsampling for distribution?

If so, do you use a hardware converter to downsample?

Does the difference hold up after downsampling?


Quote:
Originally Posted by Papanate View Post
In the Studio the difference is readily apparent for me. Tighter low end and a smother clarity to everything in the upper end.
Old 18th November 2019
  #10
Gear Guru
 
Brent Hahn's Avatar
 

Quote:
Originally Posted by StevenLMorgan View Post
... do you use a hardware converter to downsample?
By that do you mean, go 96k d/a out to the analog (hardware) world and directly into an a/d to get 48 (or 44.1)?
Old 18th November 2019
  #11
Lives for gear
 
kennybro's Avatar
Because they keep making bigger and bigger hard drives, and people need more information to fill them up.
24
Share
Old 18th November 2019
  #12
Lives for gear
 
bowzin's Avatar
Higher resolution does mean lower latency, but at the cost of increased CPU usage.

So pick your poison, more plugins and more CPU-intensive plugins open in a session, or pushing latency as low as you can.
8
Share
Old 18th November 2019
  #13
Quote:
Originally Posted by bowzin View Post
Higher resolution does mean lower latency, but at the cost of increased CPU usage.
This is interesting, why does A + B = C in this case? I didn't think latency was related to CPU usage at all. And why does higher res equal lower latency? One would guess that higher res = more info = higher latency.

The moment I realised I was working on pure assumption
Old 18th November 2019
  #14
Lives for gear
 
Space1999's Avatar
 

No 96kHz does equal lower latency. I was shocked and tried setting the session up at other sample rates, but there it was, lower latency with 96kHz.

I don’t know how or why, I just go with the flow. As far as 96kHz goes as sound quality I have this to say: The last time I was working in a studio we were using 20 bit 48kHz. That was a big deal. Fast forward to 2 years ago and I have an UAD system and Logic on a MacBook Pro and did some initial recordings.

I felt like Robin Williams in that movie “Coming to America” where he is a Russian who moves to America and enters a supermarket for the first time...just tears and disbelief. The overall sound quality improvements over the last 10 years including converter chips, plug-in quality, imaging and fidelity with no audible jitter. Just wow.

We fought very hard to get good sound through digital pre 2010. And the digital vs. analog debate actually had merit. Everything comes so easily and high quality now. You just have to do two things: Record at around -17dbfs or so and remember to curb the ultra high end digital offers so as not to have a terrible buildup of those frequencies.

I really couldn’t believe how natural and dare I say “analog” digital sounds here in 2019.
That has been my experience. And hell yeah I am going to record at the highest quality I can manage. I don’t settle for anything less than the best I can do.

Pat
10
Share
Old 18th November 2019
  #15
Lives for gear
 

Correct, if I A/D @ 96k (via Aurora(n)), when I am ready to render files for production @ 48k, Reaper 64 will resample to 48k when I specify that rate, prior to hitting the Aurora(n).

Preferring to handle conversion and rates via the Aurora(n), I just went back to recording @ 48k.



Quote:
Originally Posted by Brent Hahn View Post
By that do you mean, go 96k d/a out to the analog (hardware) world and directly into an a/d to get 48 (or 44.1)?
1
Share
Old 18th November 2019
  #16
Screw 96kHZ If your target medium is 44.1kHZ and if it is, then 88.2khZ is perfect. Because 44.1kHz is exacly halk of 88.2kHZ.
If your target medium is 48kHZ ,then i can see recoridng at 96kHz as its exactly half of 96kHz

Its all about better math 1101011101010100111001110010101101010101010101010110101010101010101101101010101
110001110010110110101011101011010110100001010101010011010101111111001010101010101010101111001101010
9
Share
Old 18th November 2019
  #17
Lives for gear
 
Xander's Avatar
Quote:
Originally Posted by soldat View Post
This is interesting, why does A + B = C in this case? I didn't think latency was related to CPU usage at all. And why does higher res equal lower latency? One would guess that higher res = more info = higher latency.

The moment I realised I was working on pure assumption
It's because latency is always X number of samples. If you have more samples per second (higher sample rate), then the actual latency time in milliseconds is less. Your sample rate and latency in seconds are inversely proportional.


Assuming 200 samples of latency:

@ 44.1 kHz sample rate:

200 samples/44,100 samples per second = .00453 seconds ~ 4.53 milliseconds

@ 96 kHz sample rate:

200 samples/96,000 samples per second = .00208 seconds ~ 2.08 milliseconds


However, it comes at the cost of higher CPU usage and data transfer rates because your computer has to process more data per second (the higher your sample rate, the more data per second you are processing).
22
Share
Old 18th November 2019
  #18
Lives for gear
 
bowzin's Avatar
Quote:
Originally Posted by soldat View Post
This is interesting, why does A + B = C in this case? I didn't think latency was related to CPU usage at all. And why does higher res equal lower latency? One would guess that higher res = more info = higher latency.

The moment I realised I was working on pure assumption
It's explained/confirmed more in the excellent stickied mega-thread by @ TAFKAT in the Music Computers sub-forum:

Audio Interface - Low Latency Performance Data Base

It ends up being a little more complicated, in that technically it still depends on the audio drivers, so it depends on the unit, and the drivers, what version of the drivers, etc. But all things being equal, yes higher sample rates = better Low Latency Performance (LLP), however at the cost of increased CPU usage (meaning less instances of plugins and Virtual Instruments (VI's) in a session before you max out your CPU specs and get underruns/freezes/etc.).

TAFKAT has maintained the exact same computer for years and years at this point as a "control," and tested tons of popular interfaces. The specs the manufacturers state are just a complete joke, so it's not possible to compare apples to apples... this thread is a huge leap forward in trying to compare apples-to-apples, because he's testing different interfaces on his unchanged "control" computer. So his experiments in that thread can compare the different interfaces in relation to each other, but cant for example tell you how many milliseconds of latency a certain interface will have on YOUR computer.

One of the important fields on his test results show how many instances of a set "control" plugin can be opened while remaining stable (I think it's like a compressor, it doesnt really what it is). So while some units have impressively low latency, if you can only open a handful of plugins at that super low latency setting, then it may not be useful after all. It's all a compromise.

For me I use 48kHz, because it maintains detail well above the frequency range I can hear (academically it should be perfect up to 24kHz, well beyond what I can hear), and I can utilize more plugins before my computer tops out, and I am always topping it out and having to freeze/bounce tracks, etc.
6
Share
Old 18th November 2019
  #19
Gear Guru
 
Brent Hahn's Avatar
 

Quote:
Originally Posted by CJ Mastering View Post
Screw 96kHZ If your target medium is 44.1kHZ and if it is, then 88.2khZ is perfect. Because 44.1kHz is exacly halk of 88.2kHZ.
If your target medium is 48kHZ ,then i can see recoridng at 96kHz as its exactly half of 96kHz

Its all about better math 1101011101010100111001110010101101010101010101010110101010101010101101101010101
110001110010110110101011101011010110100001010101010011010101111111001010101010101010101111001101010
No time to type a lot right now, but I'm not sure that's how the multiplying and dividing in sample-rate conversion works.
14
Share
Old 18th November 2019
  #20
Quote:
No time to type a lot right now, but I'm not sure that's how the multiplying and dividing in sample-rate conversion works.
Ive got time.
Its better math to go from 88.2 to 44.1 and to go from 96 to 48kHz.

44.1 + 44.1 = 88.2
48 +48 = 96
Its keeps the conversions simple
9
Share
Old 18th November 2019
  #21
Quote:
No time to type a lot right now, but I'm not sure that's how the multiplying and dividing in sample-rate conversion works.
Ive got time.
Its better math to go from 88.2 to 44.1 and to go from 96 to 48kHz.

44.1 + 44.1 = 88.2
48 +48 = 96
Its keeps the conversions simple. That means if you are recording at 88.2, its exactly half of 44.1 and if you are recording at 96, its exactly half of 48kHz.

I cannot explain any simpler
3
Share
Old 18th November 2019
  #22
Lives for gear
 

Quote:
Originally Posted by TS-12 View Post
didn't pay much attention to this much,
But now after Behringer wing announcement, almost in every post about it on Facebook, Instagram and forums there's at least one question "will it be 96khz".

Now reading on various audio forums and here on GS looks like everyone is recording at 96khz these days.


Besides lower latency and larger recorded file sizes,
Can you guys really hear the difference in 96khz over 44.1khz ?
i guess most folks did not once take part in a properly conducted blind listening test or care to find out about the results...

some gear is designed to sound equally good (or bad) at standard as at higher sampling rates.

not much love for dual or even quad speed (and even less for dsd/dxd), especially when recording huge channel counts (128+)!
1
Share
Old 18th November 2019
  #23
Lives for gear
 
kennybro's Avatar
Quote:
Originally Posted by CJ Mastering View Post
Screw 96kHZ If your target medium is 44.1kHZ and if it is, then 88.2khZ is perfect. Because 44.1kHz is exacly halk of 88.2kHZ.
If your target medium is 48kHZ ,then i can see recoridng at 96kHz as its exactly half of 96kHz

Its all about better math 1101011101010100111001110010101101010101010101010110101010101010101101101010101
110001110010110110101011101011010110100001010101010011010101111111001010101010101010101111001101010
Completely unrelated. The process does not care if it is half.

A good article that addresses more than this issue...
https://www.academia.edu/441305/Samp...z_Vs._88.2_KHz
27
Share
Old 19th November 2019
  #24
Another thing end users needn't worry about

Quote:
Originally Posted by CJ Mastering View Post
Its better math to go from 88.2 to 44.1 and to go from 96 to 48kHz.

44.1 + 44.1 = 88.2
48 +48 = 96
Its keeps the conversions simple
That really doesn't have anything to do with how the SRC math is actually done. Since the invention of polyphase filtering algorithms there's been no significant computational penalty in converting 96 -> 44.1 vs 96->48. The math for one isn't any more involved than it is for the other. The only thing that's more complex is "scheduling" the math. You'd only care about that if you were actually coding the SRC algorithm. I've done it, and it takes some careful indexing into the input and output buffers. But as the end user you just push a button, so why do you care?

David L. Rick
43
Share
Old 19th November 2019
  #25
Gear Guru
 
Brent Hahn's Avatar
 

Quote:
Originally Posted by CJ Mastering View Post
Ive got time.
Its better math to go from 88.2 to 44.1 and to go from 96 to 48kHz.

44.1 + 44.1 = 88.2
48 +48 = 96
Its keeps the conversions simple. That means if you are recording at 88.2, its exactly half of 44.1 and if you are recording at 96, its exactly half of 48kHz.

I cannot explain any simpler
Under difference circumstances, I'd feel like I was being patronized.
19
Share
Old 19th November 2019
  #26
Gear Nut
 

Quote:
Originally Posted by Brent Hahn View Post
Under difference circumstances, I'd feel like I was being patronized.
Look man, computers count using their fingers just like everyone else. Simple enough?
13
Share
Old 19th November 2019
  #27
Quote:
Originally Posted by CJ Mastering View Post
Screw 96kHZ If your target medium is 44.1kHZ and if it is, then 88.2khZ is perfect. Because 44.1kHz is exacly halk of 88.2kHZ.
If your target medium is 48kHZ ,then i can see recoridng at 96kHz as its exactly half of 96kHz

Its all about better math 1101011101010100111001110010101101010101010101010110101010101010101101101010101
110001110010110110101011101011010110100001010101010011010101111111001010101010101010101111001101010
As Dan Lavry explained to me politely but firmly in this forum many years ago, SRC doesn't really work like that. I had really thought it did. He made me go read his Sampling Theorem white paper (good thing, too, straightened out a lot of my thinking though it was a bit of a stretch to follow through the math processes, as I never took trig, but I followed through the logic of the procedures, I got that much out of Algebra II ).

clarification: I'm speaking of software SRC, above. I believe it's possible there may have been a time when some hardware SRC may have worked better on 'even multiples'... but it's not really my bailiwick. I only know the first bit because of my uncomfortable but ultimately helpful interchange with the redoubtable Mr Lavry.

Last edited by theblue1; 19th November 2019 at 03:21 AM..
10
Share
Old 19th November 2019
  #28
Lives for gear
 

Quote:
Originally Posted by CJ Mastering View Post
Screw 96kHZ If your target medium is 44.1kHZ and if it is, then 88.2khZ is perfect. Because 44.1kHz is exacly halk of 88.2kHZ.
If your target medium is 48kHZ ,then i can see recoridng at 96kHz as its exactly half of 96kHz

Its all about better math 1101011101010100111001110010101101010101010101010110101010101010101101101010101
110001110010110110101011101011010110100001010101010011010101111111001010101010101010101111001101010
Hasn’t this been disproven? In theory dividing by 2 is easier but, that’s not how computers calculate?

On top of that I’ve worked on some tracks that went to one of the hugest mixers in the game who said never send me anything at 88.1 again
3
Share
Old 19th November 2019
  #29
Quote:
Originally Posted by David Rick View Post
That really doesn't have anything to do with how the SRC math is actually done. Since the invention of polyphase filtering algorithms there's been no significant computational penalty in converting 96 -> 44.1 vs 96->48. The math for one isn't any more involved than it is for the other. The only thing that's more complex is "scheduling" the math. You'd only care about that if you were actually coding the SRC algorithm. I've done it, and it takes some careful indexing into the input and output buffers. But as the end user you just push a button, so why do you care?

David L. Rick



If I'd read the thread before posting, I could have saved myself (and the world) some blather.
3
Share
Old 19th November 2019
  #30
Lives for gear
 
12tone's Avatar
 

Quote:
Originally Posted by Dokterrock View Post
Look man, computers count using their fingers just like everyone else. Simple enough?
They actually only use one finger.
15
Share
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump