Go Back   Cockos Incorporated Forums > REAPER Forums > Recording Technologies and Techniques

Reply
 
Thread Tools Display Modes
Old 11-01-2015, 02:20 PM   #1
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default Sample Rates, Oversampling, Anti-Aliasing, Project Rate & Conversion, Nyquist Frequen

I have been involved with two very informative threads regarding these topics. I know this discussion pops up from time to time, so I thought I'd try to collect all the fantastic links into one post. Like a repository of sources for information regarding all the above. All of which are not so easy to get upon one googling of the topic. You can read all of the shenanigans from the original threads if you wish:
Oversampling vs Increased Samplerate
Nyquist Frequency... Why does this matter?

You may want to be extra vigilant while reading the linked articles because there is much that is specific to AD/DA Conversion and not-so-much about oversampling within a DAW. These are in no particular order, but I highly recommend the articles by Variety of Sound and Monty @ziph.org if you want a starting point.

TrustMeImAScientist article: The Science of Sample Rates (When Higher Is Better — And When It Isn’t) is also an excellent one to start with, maybe even the best!

In this link Indiana University claims that a sampling rate of more than 40 kHz should not cause aliasing.

Columbia University explains aliasing a little better and the need for filters. And, they have some example sound files.

Dan Lavry explains why any sample rate over 60 kHz is overkill for audio. He also has many graphs which portray the accuracy of digital sampling even at 44.1 kHz and the effect of anti-aliasing filters. Here is his other relevant paper The Optimal Sample Rate for Quality Audio.

Vlad explains EQ cramping near the Nyquist.

This MeldaProduction video explains and demonstrates aliasing.

This online tool by Infinite Wave Mastering allows you to side-by-side compare Sample Rate Conversion software such as real-time in DAWs and in off-line like R8 Brain or SoX.

Monty @xiph.org bashes Neil Young's Pono HD music distribution and explains in much detail why. This article really wraps up most all of what you need to know about digital audio.
Other Monty Videos:
https://www.youtube.com/watch?v=Ny7krNFAD1s
https://www.youtube.com/watch?v=d7kJdFGH-WI
https://www.youtube.com/watch?v=cIQ9...ature=youtu.be

TDR Ultrasonic Filter VST may be useful when working in projects with high sample-rates. Alternatively, you could also use Christian Budde's free VST, Rubber Filter, which is capable of 6-384dB/Oct.

AES Recommendation for delivery of recorded music projects (technical document)

A similar thread on the forum for Vienna Symphonic Library discussing why their libraries are distributed in 24bit/44.1kHz.

Goldmund White Paper: Does high-resolution audio sound better?

Tables on general acoustics giving instrument frequency ranges and SPLs of various sources.

There's Life Above 20 Kilohertz! A Survey of Musical Instrument Spectra to 102.4 KHz. This paper recognizes that instruments do, in fact, generate ultrasonics (which most microphones cannot accept and most ADCs will filter out).

The World Beyond 20kHz by David Blackmer, founder of Earthworks.

Wikipedia article: Hypersonic Effect. More about the fact that humans may be able to experience ultrasonics even though they cannot hear it.

Mastering for iTunes

EE Times article: Multirate DSP, part 1: Upsampling and downsampling
EE Times: Multirate DSP, part 2: Noninteger sampling factors
https://en.wikipedia.org/wiki/Upsampling
Ear Level Engineering; Sample Rate Conversion: Up

Some FREE VST wrappers you can use to oversample other VST that do not do have the ability:
VST Oversampler by ArkeCode oversamples by 2x any VST in it.
AntiAlias by Experimental Scene can oversample by 2, 4, 8, 16, or 32x!

Variety of Sound Articles:
"working ITB at higher sampling rates"
myths and facts about aliasing
the side effects of intermodulation in audio processors

EBU R128 - Loudness and Max Levels

Pipeline Audio video: Aliasing in DAW Amp Sims and Compressors

I hope this helps anyone wanting to know more about sample rates and such.

Conclusions:
Yes, you should mix at higher sample rates if you have the equipment to do it.
No, you do not need to record at higher than 24bit/44.1kHz unless you just want to avoid SRC before mixing. But, off-line SRCs like SoX,r8brain, and REAPER's own internal SRC do such an excellent job.
Compression will cause aliasing. Use at least 4x oversampling @44.1kHz to get aliasing below -50dBFS in ReaComp. Alternatively, you could mix at 192kHz with no oversampling (anti-aliasing).
Amp Sims, Saturation, and distortion plug-ins will all cause aliasing. Use the oversampling functions in these plug-ins.
Sometimes even 64x cannot reign in all aliasing. Accept it.
If possible, you may be able to reduce aliasing by Low-pass filtering prior to compression and distortion plug-ins since more aliasing occurs the higher you go in frequency. But, aliasing above -60dBFS began as low as 3500Hz with ReaComp on a sine wave sweep tone using 44.1kHz file with 2x Anti-Aliasing engaged, and progressively became worse as the frequency increased.

At least these were my observations... YMMV.
I did a test comparing REAPER's SRC vs r8brain. Also, includes instructions for converting entire project's samplerate.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000

Last edited by insub; 11-07-2015 at 04:33 PM.
insub is offline   Reply With Quote
Old 11-01-2015, 02:38 PM   #2
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

You clearly have too much time on your hands!

No, actually and truthfully: nice job, what a freakin' lot of work!

You know, I was thinking about all you've brought together here. Probably won't happen, but this info ought to be in a STICKY somewhere. Cheers!

Last edited by The Telenator; 11-02-2015 at 08:24 PM.
The Telenator is offline   Reply With Quote
Old 11-01-2015, 06:30 PM   #3
sostenuto
Human being with feelings
 
sostenuto's Avatar
 
Join Date: Apr 2011
Location: St George, UT _ USA
Posts: 2,880
Default

Super Reference List! Many thanks for major effort. Already checked a few not already seen. Hope many here take advantage.

Regards,
sostenuto is offline   Reply With Quote
Old 11-01-2015, 06:32 PM   #4
pipelineaudio
Mortal
 
pipelineaudio's Avatar
 
Join Date: Jan 2006
Location: Wickenburg, Arizona
Posts: 14,046
Default

Oh no! Actual science!!! Don't go to the Lounge or you will be lambasted for sure.

Excellent work BTW!
pipelineaudio is offline   Reply With Quote
Old 11-01-2015, 09:38 PM   #5
Mind Riot
Human being with feelings
 
Mind Riot's Avatar
 
Join Date: Mar 2008
Posts: 1,008
Default

Outstanding collection of resources you've compiled here!

I've been following the sample rate thread closely myself, it was largely excellent and answered some lingering questions I had.

But now this has raised a few more(uh oh)...

Quote:
Originally Posted by insub View Post
Yes, you should mix at higher saplerates if you have the equipment to do it.
No, you do not need to record at higher than 24bit/44.1kHz unless you just want to avoid SRC before mixing. But, off-line SRCs like SoX and R8 Brain do such an excellent job.
Bear in mind I'm just a hermit recording myself in a cave here, but are people really doing all of their recording and editing then exporting all of their .wav files in an entire project to SRC them in SoX or R8Brain only to re-import them back into their Reaper project to mix?

Not that it wouldn't be worth it if there's a demonstrable benefit. But that sounds like a lot of extra work.

Now I'm curious. When we're talking about mixing at a higher sample rate, are people talking about just changing the sample rate of their interface to run the project at that rate and auto upsample everything in Reaper or do people really go to the trouble to convert everything outside?

I've been recording and mixing at 24/44.1 for years and didn't see a compelling reason to change as long as my plugs allowed oversampling, but not ALL of them do, and that thread made me seriously consider trying mixing at a higher rate.

I just want to make sure I go about it correctly if I do.

EDIT: Finally got around to reading VOS' article on ITB higher sample rates; I had bookmarked it before. Looks like that is indeed what has to be done, unless one just goes ahead and records at a higher sampling rate even though there's no benefit at that point.

Here's a question though. Once 44.1k .wavs are upsampled to 96k or what have you, are they the same size as if they had been originally recorded at 96k anyway? In that case, shouldn't one just record at 96k with a view to mixing at that sample rate to avoid the SRC even though there's no sound benefit at the recording stage?

Thanks to all for any insight.
__________________
"Mah blahkinned sole izz daw-kaw thawn thah blahkissed nye-eeeet!!!"
SQUONK SQUONK SQUEE!!! SQUIDONK SQUIDONK DONK SQUEE!!!
"Thah daaahhhk of thah nye-eeeet izz lye-eeek my-eee sole-aaah!!!"

Last edited by Mind Riot; 11-02-2015 at 12:12 AM.
Mind Riot is offline   Reply With Quote
Old 11-02-2015, 07:41 AM   #6
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

Quote:
Originally Posted by Mind Riot View Post
Bear in mind I'm just a hermit recording myself in a cave here, but are people really doing all of their recording and editing then exporting all of their .wav files in an entire project to SRC them in SoX or R8Brain only to re-import them back into their Reaper project to mix?
I would think you would want to change the sample rate before any editing or mixing. There were several comments that time stretching is greatly affected by aliasing and REAPER does not give an option for oversampling prior to time-stretching which is something I use a lot.

Quote:
Originally Posted by Mind Riot View Post
Not that it wouldn't be worth it if there's a demonstrable benefit. But that sounds like a lot of extra work.

Now I'm curious. When we're talking about mixing at a higher sample rate, are people talking about just changing the sample rate of their interface to run the project at that rate and auto upsample everything in Reaper or do people really go to the trouble to convert everything outside?
I don't know if they do or not. Most of the discussions did not clearly identify who's using what method. However, this is an option I am considering due to sample rate limitations of my interfaces. Specifically, that I cannot utilize every input of the 2 MOTU units at their higher rates because the 8PRE is connected to the 828 via ADAT. So, I am stuck at 44.1/48kHz if I want to record all 20 tracks simultaneously.

Quote:
Originally Posted by Mind Riot View Post
Here's a question though. Once 44.1k .wavs are upsampled to 96k or what have you, are they the same size as if they had been originally recorded at 96k anyway?
Yes. File size is based on sample rate and bit depth for PCM wav files. I believe it becomes trickier when you get into variable bit rates of other file types.

Quote:
Originally Posted by Mind Riot View Post
In that case, shouldn't one just record at 96k with a view to mixing at that sample rate to avoid the SRC even though there's no sound benefit at the recording stage?
I would say, Yes! That is, if you have the interface to do so. But, also there were some comments that indicated, especially with budget interfaces, that recording at the higher sample rates generated much higher IMD (inter-modulation distortion). So, with those units there still may be some benefit to recording at the lower rates and using the off-line SRC afterwards.

At least this is how I interpreted the other threads and various above links.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000
insub is offline   Reply With Quote
Old 11-02-2015, 12:23 PM   #7
JamesPeters
Human being with feelings
 
Join Date: Aug 2011
Location: Near a big lake
Posts: 3,943
Default

Quote:
Originally Posted by insub View Post
I would say, Yes! That is, if you have the interface to do so. But, also there were some comments that indicated, especially with budget interfaces, that recording at the higher sample rates generated much higher IMD (inter-modulation distortion). So, with those units there still may be some benefit to recording at the lower rates and using the off-line SRC afterwards.
Everyone using higher sample rates should check their interface for IMD at the higher sample rates. Lots of interfaces claim to perform properly at those high rates but who knows if that's an honest claim or not. There is probably no "price guideline" that will let you know, but a quick test will resolve it.

https://xiph.org/~xiphmont/demo/neil-young.html

CTRL+F for "Intermod Tests:", read the info, download the files, and listen.

Then, consider that the IMD in those test files was purposely made such that you should hear it (normalized). There's still a tradeoff between aliasing in a project from a plugin (if you're hearing it) vs the IMD you'll hear from the interface at a low level if it doesn't perform well at the higher sample rates.
JamesPeters is offline   Reply With Quote
Old 11-02-2015, 12:46 PM   #8
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

Just keep in mind that test is for the DAC on the output, which gives no indication of the quality of the input ADC which should have an analog LPF to remove ultrasonic frequencies prior to conversion. I would assume that the LPF on the ADC is static (somewhere between 20-22kHz) regardless of the sample rate for recording.

I don't know how you would test the IMD of the ADC on the input.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000
insub is offline   Reply With Quote
Old 11-02-2015, 02:59 PM   #9
JamesPeters
Human being with feelings
 
Join Date: Aug 2011
Location: Near a big lake
Posts: 3,943
Default

Quote:
Originally Posted by insub View Post
Just keep in mind that test is for the DAC on the output, which gives no indication of the quality of the input ADC which should have an analog LPF to remove ultrasonic frequencies prior to conversion. I would assume that the LPF on the ADC is static (somewhere between 20-22kHz) regardless of the sample rate for recording.

I don't know how you would test the IMD of the ADC on the input.
True. I look at it this way though: if I'm working at a higher sample rate in general I'm probably going to be introducing something which will become part of what I hear while mixing.
JamesPeters is offline   Reply With Quote
Old 11-03-2015, 02:47 AM   #10
Mind Riot
Human being with feelings
 
Mind Riot's Avatar
 
Join Date: Mar 2008
Posts: 1,008
Default

Thank you both very much. I'll report back after I've listened to the files.

I'm not using a super low end interface, but not a MOTU or anything either. It's a Steinberg UR44 that's rated up to 192khz, made for Steinberg by Yamaha, but I have no way of knowing what level of quality the converters are that are in it.
__________________
"Mah blahkinned sole izz daw-kaw thawn thah blahkissed nye-eeeet!!!"
SQUONK SQUONK SQUEE!!! SQUIDONK SQUIDONK DONK SQUEE!!!
"Thah daaahhhk of thah nye-eeeet izz lye-eeek my-eee sole-aaah!!!"
Mind Riot is offline   Reply With Quote
Old 11-03-2015, 03:16 AM   #11
LightOfDay
Banned
 
Join Date: Jun 2015
Location: Lower Rhine Area, DE
Posts: 964
Default

outstanding excellent!!! all the things that are needed to be known but nobody cares. real science and maths and physics! I am excited!

the problem is: does Neil Young read that all? and if so, does he understand? I bet no. all the believers and stupid people turn will around and say: I dont think so. you cant make stupid uneducated people turning intelligent.

this all will not be read by the ones who should read it. so my applause to your effort is somehow pissing against the wind.
LightOfDay is offline   Reply With Quote
Old 11-03-2015, 04:56 AM   #12
Bouroki
Human being with feelings
 
Join Date: Jun 2013
Posts: 79
Default

Thank you!
Bouroki is offline   Reply With Quote
Old 11-03-2015, 06:02 AM   #13
Mind Riot
Human being with feelings
 
Mind Riot's Avatar
 
Join Date: Mar 2008
Posts: 1,008
Default

Hmmm...the audio was clearly audible on all the files at 96k. Not that surprising, I suppose; I'm not running high end stuff everywhere in the chain.

So then. Do I now still consider recording and mixing at 96k anyway given it's processing benefits and the assumption that a LPF will be present on the ADC?

Or...

Do I record at 44.1k, SRC and mix at 96k, assuming that no ultrasonic content will be present in the files even after conversion?

Or...

Do I just stick to 44.1k all around due to the IMD present in my system, assuming that it will outweigh the benefits of mixing at 96k?

Or...

Do I go down a bottle of painkillers and then beat my head in with a rock?


One of the articles in the links in the other thread posed an interesting question as well. That although sample rate itself has no demonstrable benefit beyond a certain point, some converters may in fact perform better at higher rates than lower ones.

A higher end converter at 44.1k may easily outperform a lower end one at 96k, but the lower end one may do better at 96 than it would do at 44.1.

(Sigh)

I see some A/B testing in my future.
__________________
"Mah blahkinned sole izz daw-kaw thawn thah blahkissed nye-eeeet!!!"
SQUONK SQUONK SQUEE!!! SQUIDONK SQUIDONK DONK SQUEE!!!
"Thah daaahhhk of thah nye-eeeet izz lye-eeek my-eee sole-aaah!!!"
Mind Riot is offline   Reply With Quote
Old 11-03-2015, 06:20 AM   #14
Judders
Human being with feelings
 
Join Date: Aug 2014
Posts: 11,044
Default

Honestly, I still think it's best to mix at the sample rate of the source files.

If you're using sampler instruments, then use whatever they were sampled at, to avoid on-the-fly SRC by the plugin.

If you receive 192kHz files, then consider downsampling to 96kHz for mixing.

If a plugin offers oversampling, use it. If it taxes your system, do an off-line freeze of the track.

Put the painkillers back in the cabinet and get on with making music
Judders is offline   Reply With Quote
Old 11-03-2015, 10:08 AM   #15
JamesPeters
Human being with feelings
 
Join Date: Aug 2011
Location: Near a big lake
Posts: 3,943
Default

Quote:
Originally Posted by Mind Riot View Post
Thank you both very much. I'll report back after I've listened to the files.

I'm not using a super low end interface, but not a MOTU or anything either. It's a Steinberg UR44 that's rated up to 192khz, made for Steinberg by Yamaha, but I have no way of knowing what level of quality the converters are that are in it.
Mine's a UR22.

I chose 48KHz as my default project sample rate mostly because I use some samples which are 48KHz, as Judders mentioned in the last post. I tested my UR22 for the IMD and heard it, so that became something to consider if I'm working in higher sample rates. Remember though: those files were edited such that the IMD would be noticeable if your hardware isn't working optimally for higher sample rates. Normally you wouldn't hear those artifacts nearly as loud in a mix (you may not notice them at all). But it becomes one of the factors to consider when you're trying to decide on a project sample rate.

For me, higher sample rates hold very little appeal. I'm not using 96KHz+ sample rate material imported into projects, and not using VSTi which can generate information beyond 22KHz (such that it might have aliasing if I don't use higher sample rates). The end result of my mix will always be 48KHz for "distribution" (putting up on Youtube mostly). So I get to cut out the middle man and stick with 48KHz. I'll rely on oversampling in plugins if/when needed, or choose my plugins carefully to avoid issues. If I get a wild hair to work at higher sample rates, RME here I come. (My friend's RME card played those sample files without any hint of sound.)

LightOfDay: As for Neil Young, I doubt reading is high on his list of priorities. I'm not concerned so much with the average music listener's opinion since I know I can make my files in 44.1KHz and they'll never hear a difference anyway. It annoys me that misinformation is spread about this sort of thing, and it makes things difficult in various ways when misinformation becomes "established knowledge". So, I like to help the conversation which provides proper information, promotes inquiry, that sort of thing. I don't know all the answers but at least I don't pretend to.
JamesPeters is offline   Reply With Quote
Old 11-03-2015, 10:14 AM   #16
Neenja
Human being with feelings
 
Join Date: Aug 2015
Posts: 222
Default

Quote:
Originally Posted by Mind Riot View Post
Hmmm...the audio was clearly audible on all the files at 96k. Not that surprising, I suppose; I'm not running high end stuff everywhere in the chain.

So then. Do I now still consider recording and mixing at 96k anyway given it's processing benefits and the assumption that a LPF will be present on the ADC?

Or...

Do I record at 44.1k, SRC and mix at 96k, assuming that no ultrasonic content will be present in the files even after conversion?

Or...

Do I just stick to 44.1k all around due to the IMD present in my system, assuming that it will outweigh the benefits of mixing at 96k?

Or...

Do I go down a bottle of painkillers and then beat my head in with a rock?


One of the articles in the links in the other thread posed an interesting question as well. That although sample rate itself has no demonstrable benefit beyond a certain point, some converters may in fact perform better at higher rates than lower ones.

A higher end converter at 44.1k may easily outperform a lower end one at 96k, but the lower end one may do better at 96 than it would do at 44.1.

(Sigh)

I see some A/B testing in my future.
If you heard a difference, something is broken with your DAC or your brain. There is no more information that you can hear in a 96khz file than there is in a 44.1khz file.
Neenja is offline   Reply With Quote
Old 11-03-2015, 10:38 AM   #17
LightOfDay
Banned
 
Join Date: Jun 2015
Location: Lower Rhine Area, DE
Posts: 964
Default

Quote:
Originally Posted by Mind Riot View Post
Do I just stick to 44.1k all around due to the IMD present in my system, assuming that it will outweigh the benefits of mixing at 96k?
what benefits??? there are none. except you are a manufacturer of harddisks.
LightOfDay is offline   Reply With Quote
Old 11-03-2015, 10:53 AM   #18
Judders
Human being with feelings
 
Join Date: Aug 2014
Posts: 11,044
Default

Quote:
Originally Posted by LightOfDay View Post
what benefits??? there are none. except you are a manufacturer of harddisks.
What about software monitoring latency?
Judders is offline   Reply With Quote
Old 11-03-2015, 04:01 PM   #19
Mind Riot
Human being with feelings
 
Mind Riot's Avatar
 
Join Date: Mar 2008
Posts: 1,008
Default

Quote:
Originally Posted by JamesPeters View Post
Mine's a UR22.

I chose 48KHz as my default project sample rate mostly because I use some samples which are 48KHz, as Judders mentioned in the last post. I tested my UR22 for the IMD and heard it, so that became something to consider if I'm working in higher sample rates. Remember though: those files were edited such that the IMD would be noticeable if your hardware isn't working optimally for higher sample rates. Normally you wouldn't hear those artifacts nearly as loud in a mix (you may not notice them at all). But it becomes one of the factors to consider when you're trying to decide on a project sample rate.

For me, higher sample rates hold very little appeal. I'm not using 96KHz+ sample rate material imported into projects, and not using VSTi which can generate information beyond 22KHz (such that it might have aliasing if I don't use higher sample rates). The end result of my mix will always be 48KHz for "distribution" (putting up on Youtube mostly). So I get to cut out the middle man and stick with 48KHz. I'll rely on oversampling in plugins if/when needed, or choose my plugins carefully to avoid issues. If I get a wild hair to work at higher sample rates, RME here I come. (My friend's RME card played those sample files without any hint of sound.)
I do plan on starting any new projects at 48k simply because the CD has become the exception rather than the rule, as opposed to the other way around.

But I'm only working with my own stuff here, writing and recording and mixing, so the source files are all up to me. My PC can easily handle working at 96k all the way through, but it can also easily handle running every plug I have with oversampling and running the ones without it in an oversampling wrapper.

So now I just have to weigh all of these factors against each other, the higher sampling rate versus per plug oversampling and wrappers and the potential problems with IMD the higher rate may bring with it.
__________________
"Mah blahkinned sole izz daw-kaw thawn thah blahkissed nye-eeeet!!!"
SQUONK SQUONK SQUEE!!! SQUIDONK SQUIDONK DONK SQUEE!!!
"Thah daaahhhk of thah nye-eeeet izz lye-eeek my-eee sole-aaah!!!"

Last edited by Mind Riot; 11-03-2015 at 08:26 PM.
Mind Riot is offline   Reply With Quote
Old 11-03-2015, 05:17 PM   #20
serr
Human being with feelings
 
Join Date: Sep 2010
Posts: 12,536
Default

Quote:
Originally Posted by Judders View Post
What about software monitoring latency?
If you're doing live sound and/or live performance work, this is the first thing you should be interested in. Find the best latency/performance.

It's 48k for my dual MOTU 828mk3 setup. 96k would be lower latency for the same buffer sizes of course but it chokes a lot of plugins at the low block size and restricts what you can do in this case.


Funny how charged up some people get about trying to "prove" that early the digital formats are as good as it gets or something.

When 24 bit formats came along, we were "in" and most analog formats sounded inferior (anything but a top of the line studio machine calibrated and run by a professionally trained operator - and even that's debatable).

When someone says "digital sounding" as a negative, they mean 16 bit recorded at a non ideal level and/or cheap converters where the analog inputs just don't deliver the clean full signal to the digitizer chips.

Anyone trying to attribute that level of degradation to lower sample rates is off base. Your root cause is something other than what you think is what I'm saying.

HD sample rates (88.2k or 96k) sound that tiny bit better. To the point that a recording sounds exactly like the original no matter what kind of revealing system you play it on. (Just in case 24 bit 44.1k or 48k didn't already on your system.)
Everything is relative in a system.
Probably really high end speakers and amps but then mid grade converters would reveal HD trumping lower rates the most. Add high end converters and it's only a subtle difference again.

If you hear something drastic - then something else is going on than what you thought.

Multiple generation copies are where ugly problems came up in the past and HD sample rates remove all trace of that. That doesn't suck.

Hard drives are already large to the point of a HD music library being trivial. It will soon be just as trivial as worrying about the size of a jpg as though you were still using a Commodore 64.

Portable digital players are the new exception. Or rather addition. Mp3's on a phone sound better than some boomboxes I remember friends having in high school. This is a good thing but that certainly doesn't make it a replacement for full quality home audio. (Some people seem to think that's the case apparently.) It's an addition to it. And it turns out that the CD format falls closer to a portable format nowadays. Masters are 24 bit now. Full 24 bit HD stereo and 5.1 surround audio is delivered on bluray and DVDA discs and digital downloads. That doesn't suck either.

Order of importance should be:
1. 24 bit everything.
2. Avoid unnecessary conversions.
3. System performance ability trumps HD sample rates.
4. May as well use the modern HD format whenever possible.

Microphones, Mic preamps, AD & DA converters, speakers, and amps.
This stuff hasn't changed in operation or importance. AD&DA converters are the new bit and just as big ticket an item. The hardware that gives the ability to capture sounds and then listen back to them still makes the biggest impact.

The funny part to me is that preserving the digital signals with HD formats is the easy part now. Nice hardware still requires work to obtain. Everyone is fighting over the easy part that you can do by just pushing a button.
serr is online now   Reply With Quote
Old 11-03-2015, 09:59 PM   #21
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

Quote:
Originally Posted by Mind Riot View Post
are people really doing all of their recording and editing then exporting all of their .wav files in an entire project to SRC them in SoX or R8Brain only to re-import them back into their Reaper project to mix?
I wanted to know if it was worthwhile to use an external SRC instead of letting REAPER do it. Long story short: According to my tests the answer is NO.

The fastest easiest way I know to convert your project to a higher sample rate is to use the Glue function. Here’s how to do it in Windows:

1. Change the project sample rate. Project Settings (Alt+Enter) > Project Sample Rate drop down. For my test I went from 44.1 to 96kHz.
Note: This is also where you set the quality of the SRC for Gluing. It is determined by the Playback Resample Mode.
2. Adjust the sample rate of your soundcard/audio interface. Preferences (Ctrl+P) > Audio > Device > Samplerate or you may need to do this from your interface’s dialog box accessible from here.
3. Select all (Ctrl+A) items from the Arrange Window.
4. Left-click any item and select Glue Items from the context-sensitive menu.
There will be progress meters as each file is resampled. I chose Glue because it places the new files exactly where they need to be. I did not try REAPER’s other available methods.

Now, for the test…

I used a normalized (peak=0dBFS) mono overhead drum mic recording that was 4:21 sec long, 24bit/44.1kHz original file.
I used a Playback Resample Mode of: Extreme HQ (768pt HQ sinc). R8brain quality was set to Very High. I converted from 44.1 – 96kHz (not 88.2 because my soundcard doesn’t include that option).

Conversion time, up-sample winner: r8brain 30sec, REAPER 52sec
Conversion time, down-sample winner: r8brain 22sec, REAPER 24sec
But… at REAPER resample mode of Good (192pt sinc) up-sample was 14.5sec
And, at Medium (64pt sinc) the time was about 5.5sec. More on these later.


Since I could not try to null at 96kHz I subsequently downsampled back to 44.1. This was the result that made REAPER the winner. In every case the files do not null. I used SPAN (spectrum type set to MAX) to check because while the Master Meter was indicating sound, I could not actually hear it as it was too low. Here are the results:

R8brain up/down-sampled file vs original
There was sound that ramped up in a linear fashion from 1Hz-10kHz with an amplitude of -169dBFS up to -59dBFS by 10kHZ. Then it dipped a little followed by a large bump from 18kHz-Nyquist (22.05kHz) that peaked at -42dB. This ultrasonic peak is what actually showed on the Master Meter. I actually could hear this one a little.

REAPER 768pt up/down-sampled file vs original
Same linear ramp in amplitude, but different values. From 20Hz-20kHz, amplitudes from -180dBFS to -110dBFS respectively. No dip, but rather increasing logarithmically upward to the bump that was from 20kHz-Nyquist that peaked at -42dB.

So, the bump was about the same peak level for both, but r8brain’s actually dipped into the audible spectrum and I could actually faintly hear the drum beat (mostly cymbals) when nulling with the r8brain file. In my test REAPER’s conversion file was a closer null than r8brain with all of it’s non-nulled, audible frequencies staying below -110dBFS while r8brains went as high as -59dBFS where I could hear it.

I compared the two 96kHz files to each other and their result was almost identical to comparing the r8brain render against the original file. So, the discrepancy is occurring on the up-sample. Not the down-sample.

I also compared REAPER’s other lower quality conversions @96kHz against the 768pt render with similar results. Their upsampled files nulled nearly identically 96 to 96 file as 44.1 vs original. I think it’s safe to say that the down-sample cycle is not the cause of null failure. Or, at least its contribution is insignificant.

The real kicker was that the default setting of Good (192pt sinc) still resulted in a better null than the r8brain. Time for conversion was 14.5sec up to 96kHz and 7sec back down to 44.1kHz. The remaining “null” signal ramped from 1Hz-10kHz, amplitude -180dBFS to -86dBFS with a bump from 19kHz-Nyquist with a peak of -47dBFS. Medium (64pt sinc) performed similarly. Ramp 1Hz-10kHz, -178dBFS to -87dBFS with a bump from 16kHz-Nyquist peaking at -48dBFS.

I don’t know how this compares to the real-time oversampling within plugins, but I see no advantage to leaving REAPER and using a 3rd party SRC. Even the Good setting outperformed r8brain in half the time. Good still kept its filter bump nearly all the way above audible while Medium dipped down to 16kHz. So, I’d stick with the default setting of Good (192pt sinc). I did not test the Better (or any other) resample mode, but I suspect that it may be the best compromise between conversion time and audible range discrepancy. I may just uninstall r8brain now that my test is complete.

The next test will be to convert a previous complete mix and see if I can hear any difference. Perhaps I won't, but I don't think I always activated oversampling for compressors. I believe there is some tangible advantage to working this way: Doing the first up-sample in higher quality off-line vs letting all the oversampling occur in real-time. Besides, there are DAW specific functions such as time-stretching to consider which do not offer an option for oversampling.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000

Last edited by insub; 11-04-2015 at 06:19 AM.
insub is offline   Reply With Quote
Old 11-04-2015, 06:17 PM   #22
The Telenator
Banned
 
Join Date: Dec 2011
Location: Oud West, NL
Posts: 2,335
Default

Insub: "The real kicker was that the default setting of Good (192pt sinc) still resulted in a better null than the r8brain. Time for conversion was 14.5sec up to 96kHz and 7sec back down to 44.1kHz. The remaining “null” signal ramped from 1Hz-10kHz, amplitude -180dBFS to -86dBFS with a bump from 19kHz-Nyquist with a peak of -47dBFS. Medium (64pt sinc) performed similarly. Ramp 1Hz-10kHz, -178dBFS to -87dBFS with a bump from 16kHz-Nyquist peaking at -48dBFS.

I don’t know how this compares to the real-time oversampling within plugins, but I see no advantage to leaving REAPER and using a 3rd party SRC."


I was hoping to come away from all these latest sample rate threads with some sort of 'golden nugget' of info. Sure enough, this is it. Thanks so much! I have little time for running tests of this sort lately, so this is greatly appreciated!
The Telenator is offline   Reply With Quote
Old 11-04-2015, 06:41 PM   #23
Mind Riot
Human being with feelings
 
Mind Riot's Avatar
 
Join Date: Mar 2008
Posts: 1,008
Default

Quote:
Originally Posted by insub View Post
I wanted to know if it was worthwhile to use an external SRC instead of letting REAPER do it. Long story short: According to my tests the answer is NO.
Intriguing results! But this is actually unrelated.

I was just going to thank you for your other thread on essential freebies. I've picked up most of the best free ones over time, but the No.6 limiter and especially the Anti-Alias filter VST were most welcome additions to my toolbox. Thank you for putting that thread together and keeping it in your sig, or I might never have discovered either of them.
__________________
"Mah blahkinned sole izz daw-kaw thawn thah blahkissed nye-eeeet!!!"
SQUONK SQUONK SQUEE!!! SQUIDONK SQUIDONK DONK SQUEE!!!
"Thah daaahhhk of thah nye-eeeet izz lye-eeek my-eee sole-aaah!!!"
Mind Riot is offline   Reply With Quote
Old 11-05-2015, 09:41 AM   #24
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

Quote:
Originally Posted by The Telenator View Post
I was hoping to come away from all these latest sample rate threads with some sort of 'golden nugget' of info. Sure enough, this is it. Thanks so much! I have little time for running tests of this sort lately, so this is greatly appreciated!
Quote:
Originally Posted by Mind Riot View Post
I was just going to thank you for your other thread on essential freebies. I've picked up most of the best free ones over time, but the No.6 limiter and especially the Anti-Alias filter VST were most welcome additions to my toolbox. Thank you for putting that thread together and keeping it in your sig, or I might never have discovered either of them.
You're welcome!
It's my pleasure, really. To be able to help someone else with my discoveries.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000
insub is offline   Reply With Quote
Old 11-05-2015, 08:08 PM   #25
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

Quote:
Originally Posted by LightOfDay View Post
what benefits??? there are none. except you are a manufacturer of harddisks.
I have to disagree. I think there are several good reasons for mixing at higher sample-rates. Even when recording at standard 44.1/48kHz.

1. Some plug-ins that could benefit from oversampling do not have it built-in. ReaXComp and Softube Saturation Knob are both good examples.
2. Some actions like time-stretching are native to the DAW and have no option for oversampling.
3. You can convert your project to a higher samplerate using off-line SRC which will be of greater quality than the real-time SRC of oversampling.
4. You can remove the first multiple of the plugins that do need oversampling. For example, when using ReaComp @44.1kHz you should probably be using at least an AA setting of 4x (based on my tests), but @88.2/96kHz you should be able to get by with 2x. And, the real-time oversampling would now be running from your higher-quality off-line SRC, or better yet, you recorded at 88.2/96kHz so no SRC was used yet.

I didn't think that clean EQ's benefited from oversampling. But, this thread shows how oversampling corrects ReaEQ's Nyquist filter cramping. I assume that running the project at a higher rate accomplishes the same thing, but I have not tested this theory.

Whether you appreciate this thread or not, I believe that I have learned something valuable at least. And, maybe it will help someone else out that cares.

Still, don't take my word for it. In the Variety of Sound article linked above the author promotes mixing at higher sample-rates. This coming from a plug-in coder, so I imagine he knows a lot more about it than me. This list is just my take away from all the reading and testing.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000
insub is offline   Reply With Quote
Old 11-06-2015, 01:53 AM   #26
Mind Riot
Human being with feelings
 
Mind Riot's Avatar
 
Join Date: Mar 2008
Posts: 1,008
Default

Indeed, the benefit of oversampling/higher sample rates to EQs is well established, and is by no means limited to ReaEQ. Many EQ plugs suffer the same problem up in the high end and benefit from using a higher sample rate or an oversampling wrapper/anti-alias filter.

I've been using the modified versions of ReaEQ and ReaXcomp with the OS wrapper from that thread for a while now, at least until I bought a better EQ. I submitted a feature request to add native oversampling to ReaEQ, but so far no joy.
__________________
"Mah blahkinned sole izz daw-kaw thawn thah blahkissed nye-eeeet!!!"
SQUONK SQUONK SQUEE!!! SQUIDONK SQUIDONK DONK SQUEE!!!
"Thah daaahhhk of thah nye-eeeet izz lye-eeek my-eee sole-aaah!!!"
Mind Riot is offline   Reply With Quote
Old 11-06-2015, 02:44 PM   #27
JHughes
Banned
 
Join Date: Aug 2007
Location: Too close to Charlotte, NC
Posts: 3,554
Default

Thanks for this thread insub. Thanks to ReaDave too for bringing this up in context with Reaper plugins. Awesome stuff!
JHughes is offline   Reply With Quote
Old 11-06-2015, 03:41 PM   #28
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

It's my pleasure!

I got on this quest partly because the excellent REAPER User Guide doesn't offer much in explanation of oversampling.

For instance, it tells you that there is an optional setting for the Playback Resample Mode. But, it does not describe which setting you should use. And, more importantly, WHY. There is also never a mention that the glue SRC is determined by this setting. I gleaned that tidbit from another thread on here where they seemed uncertain if that were the case. But, I tested it and it is true.

I can understand not wanting to try explaining all this in the manual. But, I think there should be some recommended settings and short explanation as to why. The average PC today is much more powerful than when REAPER first came out. I think more emphasize should be made for people to convert to mixing at 96kHz.

And, if they did then ReaEQ and the like wouldn't ever need an oversampling option!
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000
insub is offline   Reply With Quote
Old 11-07-2015, 08:13 AM   #29
JHughes
Banned
 
Join Date: Aug 2007
Location: Too close to Charlotte, NC
Posts: 3,554
Default

The KVR Audio link is no good, it should be "http://www.kvraudio.com/product/vst-oversampler-by-arkecode" I think.
JHughes is offline   Reply With Quote
Old 11-07-2015, 04:16 PM   #30
insub
Human being with feelings
 
insub's Avatar
 
Join Date: Mar 2014
Location: Louisville, KY, USA
Posts: 1,075
Default

Thanks, J!

I updated post #1 and included a short description this time for those two VST wrappers.
__________________
Everything you need to know about samplerates and oversampling... maybe!
My Essential FREE 64bit VST Effects, ReaEQ Presets for Instruments
Windows 10 64 bit; MOTU 828 MKII, Audio Express, & 8PRE; Behringer ADA8000
insub is offline   Reply With Quote
Old 06-20-2020, 01:54 PM   #31
sightlessness
Human being with feelings
 
sightlessness's Avatar
 
Join Date: Mar 2010
Posts: 1,002
Default

Does anyone have the files to test our interfaces for intermodulation distortion?
It's mentioned in this video:
https://www.youtube.com/watch?v=rXl9QIKFfA4

which links to:

https://productionadvice.co.uk/high-...c-sound-worse/

which links to:

http://people.xiph.org/~xiphmont/demo/neil-young.html

Which is 404 (not found).
__________________
I want to live PEACEFULLY PLEASE WORLD "LEADERS" GET THIS DONE/LET IT BE FOR GOOD AND MAKE HISTORYYYYYYY! Thanks.
sightlessness is offline   Reply With Quote
Old 06-20-2020, 02:00 PM   #32
clepsydrae
Human being with feelings
 
clepsydrae's Avatar
 
Join Date: Nov 2011
Posts: 3,408
Default

Internet archive has your back (donations are always needed/welcome there): http://web.archive.org/web/20200426202431/http://people.xiph.org/~xiphmont/demo/neil-young.html
clepsydrae is offline   Reply With Quote
Old 06-20-2020, 05:18 PM   #33
sightlessness
Human being with feelings
 
sightlessness's Avatar
 
Join Date: Mar 2010
Posts: 1,002
Default

Thanks so much.

So pretty much any interface has intermodulation distortion?

My Duet (latest gen) has it, my BF Pro FS has it, TAscam 16x08, Behringer 404 HD...
Scarlett 2i2... sigh...
__________________
I want to live PEACEFULLY PLEASE WORLD "LEADERS" GET THIS DONE/LET IT BE FOR GOOD AND MAKE HISTORYYYYYYY! Thanks.
sightlessness is offline   Reply With Quote
Old 06-20-2020, 06:34 PM   #34
clepsydrae
Human being with feelings
 
clepsydrae's Avatar
 
Join Date: Nov 2011
Posts: 3,408
Default

Quote:
Originally Posted by sightlessness View Post
So pretty much any interface has intermodulation distortion?
Don't quote me, but AFAIK it's not possible to build a real-world device with 0% IMD. Whether those distortion products outweigh whatever purported benefits of high frequency content is a separate question. (I am personally on board the "44.1/16 is the correct choice for reproduction" bandwagon.)
clepsydrae is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 09:02 PM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.