R/E/P Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: 1 2 [All]   Go Down

Author Topic: 96k or 88k  (Read 11118 times)

Pingu

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1196
96k or 88k
« on: June 25, 2006, 03:56:26 PM »

I know this has probably been done and done so please forgive me.


Down sampling to 44 from
88 or 96k.

Is the SRC from 88 always going to sound closer to the source than 96k because of it being an exact half.

Logically one would think so.

But, are the results evading logic.

What have you guys found.


The reason i ask is that i have always recorded at 44.1 but the other day tried a recording at 88 and it sounded sweet to me.

My destination is cd and i am thinking of going to 96k.

Will the downsample be much different from the 88k audio.



Cheers








Logged
If I defend myself I am attacked. But in defenselessness I will be strong, and I will learn what my defenses hide.

blueintheface

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
Re: 96k or 88k
« Reply #1 on: June 25, 2006, 06:43:31 PM »

Before we even get into higher sample-rates, what SRC will you use?

Some are pretty ropey

And you say 88.2 sounded sweet. Sweeter than 44.1?

Logged

Revolution

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 364
Re: 96k or 88k
« Reply #2 on: June 25, 2006, 09:38:32 PM »

FWIW I have recorded at 44,88 and 96 and always found the higher sample rates (88,96)to have sounded better even after the sample rate conversion to 44.1K.

I just don't buy the argument that when bouncing down from 88 to 44 that by taking out every second sample means that the end product is going to sound the same.It doesn't according to my experience.

It's incredible how often I here this from guys who only have 44,48k capability or limited storage.

After trying all of these sample rates even 192 I am now sold at 96K.

Spock

  • Newbie
  • *
  • Offline Offline
  • Posts: 10
Re: 96k or 88k
« Reply #3 on: June 26, 2006, 12:25:52 AM »

One would think that going from 44.1 to 88.2 could be done by just repeating each sample, or making a average between two samples.  Going the other way you would think that you could just drop every other sample, or average each two samples to make the results.

However, I don't think you would like the results.

Read Dan's paper on sampling.  You have to treat each sample as a sinc function, then add up the contributions from the sinc functions of several samples on each side to figure out what your new value should be.

Once you have a SRC program that does things right, making it take care of non related in and out rates, shouldn't be that much harder.
Logged
Spock: Are you sure it isn't time for a "colourful metaphor"?
The Voyage Home

Pingu

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1196
Re: 96k or 88k
« Reply #4 on: June 26, 2006, 03:36:31 AM »

Revolution wrote on Mon, 26 June 2006 09:38

FWIW I have recorded at 44,88 and 96 and always found the higher sample rates (88,96)to have sounded better even after the sample rate conversion to 44.1K.

I just don't buy the argument that when bouncing down from 88 to 44 that by taking out every second sample means that the end product is going to sound the same.It doesn't according to my experience.

It's incredible how often I here this from guys who only have 44,48k capability or limited storage.

After trying all of these sample rates even 192 I am now sold at 96K.



This is how im feeling.
Logged
If I defend myself I am attacked. But in defenselessness I will be strong, and I will learn what my defenses hide.

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: 96k or 88k
« Reply #5 on: June 26, 2006, 04:17:43 AM »

Revolution wrote on Mon, 26 June 2006 02:38

FWIW I have recorded at 44,88 and 96 and always found the higher sample rates (88,96)to have sounded better even after the sample rate conversion to 44.1K.

I just don't buy the argument that when bouncing down from 88 to 44 that by taking out every second sample means that the end product is going to sound the same.It doesn't according to my experience.

It's incredible how often I here this from guys who only have 44,48k capability or limited storage.

After trying all of these sample rates even 192 I am now sold at 96K.


Actually it's not surprising it sounds different.

At the very least you're exchanging the last stage of downsampling in the converter for one in your DAW, which has the potential to be much better since you have more DSP and memory to apply to it.

If you do any non-linear processing then that can make a difference too (depending on how the algorithms are implemented).
Logged

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: 96k or 88k
« Reply #6 on: June 26, 2006, 04:21:49 AM »

Pingu wrote on Sun, 25 June 2006 20:56

I know this has probably been done and done so please forgive me.


Down sampling to 44 from
88 or 96k.

Is the SRC from 88 always going to sound closer to the source than 96k because of it being an exact half.

Logically one would think so.

But, are the results evading logic.

What have you guys found.


The reason i ask is that i have always recorded at 44.1 but the other day tried a recording at 88 and it sounded sweet to me.

My destination is cd and i am thinking of going to 96k.

Will the downsample be much different from the 88k audio.



Cheers











You could make an SRC that sounded as good doing 96kHz->44kHz as 88kHz->44kHz

However it is easier to make a good divide by two SRC.

Check out the SRC though. I saw a page where they'd actually done some tests and was rather shocked at the lack of quality of many of them, including the standard ones in several DAWs.
Logged

blueintheface

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
Re: 96k or 88k
« Reply #7 on: June 26, 2006, 06:57:00 AM »

I don't think Pingu wants to discuss SRC . . .  Twisted Evil

It's a slippery slope - someone might suggest blind listening tests next . . .
Logged

kraster

  • Full Member
  • ***
  • Offline Offline
  • Posts: 199
Re: 96k or 88k
« Reply #8 on: June 26, 2006, 07:24:36 AM »

blueintheface wrote on Mon, 26 June 2006 11:57

I don't think Pingu wants to discuss SRC . . .  Twisted Evil

It's a slippery slope - someone might suggest blind listening tests next . . .



Well Pingu did say that his ultimate destination is Cd so SRC is pretty relevant.  Wink

Personally I think that recording at the Sample rate of the ultimate destination is the best way. I try to avoid SRC when at all possible. And, in my experience, the benefits of recording at high sample rates are lost/altered when downsampling from a higher rate.

88.2 has two advantages over 96k if you decide to use higher sample rate.

1. As Jon pointed out the maths involved in Src are probably easier to implement when doing a synchronous conversion. I fyour ultimate destination is CD then 88.2 is an obvious choice.

2. 88.2 is closer to Dan's optimal sampling rate than 96k. And you still get 44.1k of bandwidth.
Logged

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: 96k or 88k
« Reply #9 on: June 26, 2006, 07:59:29 AM »

blueintheface wrote on Mon, 26 June 2006 11:57

I don't think Pingu wants to discuss SRC . . .  Twisted Evil

It's a slippery slope - someone might suggest blind listening tests next . . .


And how can you possibly avoid SRC when talking about sample rates at the converter that differ from sample rates in the delivery medium?

Logged

blueintheface

  • Newbie
  • *
  • Offline Offline
  • Posts: 24
Re: 96k or 88k
« Reply #10 on: June 26, 2006, 08:01:02 AM »

kraster wrote on Mon, 26 June 2006 12:24



Well Pingu did say that his ultimate destination is Cd so SRC is pretty relevant.  Wink

Personally I think that recording at the Sample rate of the ultimate destination is the best way. I try to avoid SRC when at all possible. And, in my experience, the benefits of recording at high sample rates are lost/altered when downsampling from a higher rate.



I guess the irony didn't translate - I agree with you that good SRC is vital, and generally best avoided altogether.

I'm also a fan of blind listening tests.

Been there too many times twiddling a knob that wasn't connected to anything - until it was in the position that just 'felt right'.

People in general seem to have a hard time accepting that psychology plays a huge part in aural perception - and it's so easy to prove.
Logged

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: 96k or 88k
« Reply #11 on: June 26, 2006, 08:09:38 AM »

kraster wrote on Mon, 26 June 2006 12:24


Personally I think that recording at the Sample rate of the ultimate destination is the best way. I try to avoid SRC when at all possible.


If you look at the architecture of a modern ADC, they actually do the front end sampling at the same high rate whether they are outputing at 176, 88 or 44. They then have a different number of downconversion steps before outputing a digital word. Therefore when you make the decision as to whether you should set your converter to 44kHz, or set it to 88kHz and then downconvert in your DAW, you are NOT making a decision as to whether or not to employ SRC. Your decision is as to whether the SRC should occur in the converter, or in your DAW.

If both are properly implemented then you are right that you should not hear a difference. However due to constraints of chip size, or processing power, either may be compromised.


Logged

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: 96k or 88k
« Reply #12 on: June 26, 2006, 08:12:39 AM »

blueintheface wrote on Mon, 26 June 2006 13:01

Been there too many times twiddling a knob that wasn't connected to anything - until it was in the position that just 'felt right'.

People in general seem to have a hard time accepting that psychology plays a huge part in aural perception - and it's so easy to prove.



Agreed!!!

In fact I note that quite often the people who are the most adamant that psychology isn't affecting their perception, only their objective listening skills, are the ones whos comments and conclusions SCREAM "psychosomatic"!!
Logged

Pingu

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1196
Re: 96k or 88k
« Reply #13 on: June 26, 2006, 08:20:07 AM »

Jeezus.
I love how all of a sudden this turns into placebo.
Logged
If I defend myself I am attacked. But in defenselessness I will be strong, and I will learn what my defenses hide.

Jon Hodgson

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1854
Re: 96k or 88k
« Reply #14 on: June 26, 2006, 08:50:01 AM »

Pingu wrote on Mon, 26 June 2006 13:20

Jeezus.
I love how all of a sudden this turns into placebo.


We're not saying it is placebo in your specific case, or at least I'm not.

We're saying that when you're trying to judge what should be (assuming good implementation) very small differences then blind tests are important. It was a general comment and not one aimed specifically at you or your findings.

The answer to your question is that 88kHz should be a better choice for you than 96KHz, for the two reasons given. The sampling itself should be more accurate (though the difference is probably below human perception anyway) and the downconversion will quite possibly be better, or at worst equal.

It might also benefit you to test some different SRC options to do that conversion, because not all are created equal.
Logged

cerberus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2651
Re: 96k or 88k
« Reply #15 on: June 29, 2006, 12:13:37 PM »

Revolution wrote on Sun, 25 June 2006 21:38

After trying all of these sample rates even 192 I am now sold at 96K.


i was mastering last week at 88.2kHz and i switched waves lnmb from native to the apa hardware version (5.2) which does 2x internal upsampling at the moment (wheras the native version (5.0) does not internally upsample. )  apparently most people don't notice, but suddenly in a certain context, the difference was like night and day.  

dan has a theory i understand.. so i don't know why i heard what i did ... and i am not in the mood to go 192 at the moment, in fact i can't afford to. so this is not what  i want to observe.  but it  seems reasonable to me that 384khz internal processing (around 4x my current working rate) could sound even truer to the source.

jeff dinces

danlavry

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 997
Re: 96k or 88k
« Reply #16 on: June 29, 2006, 01:35:55 PM »

cerberus wrote on Thu, 29 June 2006 17:13

Revolution wrote on Sun, 25 June 2006 21:38

After trying all of these sample rates even 192 I am now sold at 96K.


i was mastering last week at 88.2kHz and i switched waves lnmb from native to the apa hardware version (5.2) which does 2x internal upsampling at the moment (wheras the native version (5.0) does not internally upsample. )  apparently most people don't notice, but suddenly in a certain context, the difference was like night and day.  

dan has a theory i understand.. so i don't know why i heard what i did ... and i am not in the mood to go 192 at the moment, in fact i can't afford to. so this is not what  i want to observe.  but it  seems reasonable to me that 384khz internal processing (around 4x my current working rate) could sound even truer to the source.

jeff dinces



When your material at some sample rate, then upsample it, the audio bandwidth content does not increase, nor is there "new music" or "new higher harmonics"...

So in theory, you gain nothing.

In practice you may gain or loose, depending on the specific gear. Remember that 99% of DA's will do upsampling to a much higher rate.

So when you allow the software to upsample by say X2, the DA does not have to do the first upsampling stage (which is the more demanding stage).

There are at least 2 places to look at:

1. Which sample rate converter is better? The software or the DA internal SRC.
2. How does the DA perform when fed various sample rates?

So it is all about the implementation - how does the specific audio chain work at different settings.

Regards
Dan Lavry
http://www.lavryengineering.com
Logged

cerberus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2651
Re: 96k or 88k
« Reply #17 on: June 29, 2006, 05:17:35 PM »

danlavry wrote on Thu, 29 June 2006 13:35


When your material at some sample rate, then upsample it, the audio bandwidth content does not increase, nor is there "new music" or "new higher harmonics"...
i understand that there should be nothing wrong with 44.1 source files assuming the converter was properly designed. but each generation of converters seems to raise the bar for this. i notice you are still making improvements in your designs and seem to offer at least two quality levels (blue,black). so that suggests to me that converters are not nearly perfected yet; and thus  perhaps it's safer to record at higher rates at this moment, even with the best converter made.  is that a fair assessment?

jeff dinces

Pingu

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1196
Re: 96k or 88k
« Reply #18 on: July 04, 2006, 04:44:02 PM »

What does the AD have to do with a plugin upsampling for processing?



Logged
If I defend myself I am attacked. But in defenselessness I will be strong, and I will learn what my defenses hide.

mpd

  • Newbie
  • *
  • Offline Offline
  • Posts: 15
Re: 96k or 88k
« Reply #19 on: July 05, 2006, 10:26:13 AM »

cerberus wrote on Thu, 29 June 2006 12:13


i was mastering last week at 88.2kHz and i switched waves lnmb from native to the apa hardware version (5.2) which does 2x internal upsampling at the moment (wheras the native version (5.0) does not internally upsample. )  apparently most people don't notice, but suddenly in a certain context, the difference was like night and day.  

dan has a theory i understand.. so i don't know why i heard what i did ... and i am not in the mood to go 192 at the moment, in fact i can't afford to. so this is not what  i want to observe.  but it  seems reasonable to me that 384khz internal processing (around 4x my current working rate) could sound even truer to the source.



What were you doing when you noticed the difference?  I have been meaning to work out the math, but I can think of two situations where internal interpolation can make difference.  Both have to do with implementation of IIR filters.  Basically, the same filter parameters will result in different tranfser functions in the z-domain at different sampling rates, which could result in audible differences.  This is different than running the converters at a higher rate, though.
Logged

cerberus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2651
Re: 96k or 88k
« Reply #20 on: July 05, 2006, 01:59:12 PM »

mpd wrote on Wed, 05 July 2006 10:26

cerberus wrote on Thu, 29 June 2006 12:13


i was mastering last week at 88.2kHz and i switched waves lnmb from native to the apa hardware version (5.2) which does 2x internal upsampling at the moment (wheras the native version (5.0) does not internally upsample. )  apparently most people don't notice, but suddenly in a certain context, the difference was like night and day.  

dan has a theory i understand.. so i don't know why i heard what i did ... and i am not in the mood to go 192 at the moment, in fact i can't afford to. so this is not what  i want to observe.  but it  seems reasonable to me that 384khz internal processing (around 4x my current working rate) could sound even truer to the source.



What were you doing when you noticed the difference?  I have been meaning to work out the math, but I can think of two situations where internal interpolation can make difference.  Both have to do with implementation of IIR filters.  Basically, the same filter parameters will result in different tranfser functions in the z-domain at different sampling rates, which could result in audible differences.  This is different than running the converters at a higher rate, though.

i was literally attempting to free up some cpu on the computer...i was not listening for nor expecting any audible difference, which is why the event was so striking to me.   the plug is a multiband compressor/ expander with adustable response profiles (e.g. knee) and f.i.r. (linear phase) crossovers.  on that project i  used it as a  low level expander in an m/s array.

jeff dinces

Ronny

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 2739
Re: 96k or 88k
« Reply #21 on: July 06, 2006, 09:18:01 PM »

blueintheface wrote on Mon, 26 June 2006 08:01

kraster wrote on Mon, 26 June 2006 12:24



Well Pingu did say that his ultimate destination is Cd so SRC is pretty relevant.  Wink

Personally I think that recording at the Sample rate of the ultimate destination is the best way. I try to avoid SRC when at all possible. And, in my experience, the benefits of recording at high sample rates are lost/altered when downsampling from a higher rate.



I guess the irony didn't translate - I agree with you that good SRC is vital, and generally best avoided altogether.

I'm also a fan of blind listening tests.

Been there too many times twiddling a knob that wasn't connected to anything - until it was in the position that just 'felt right'.

People in general seem to have a hard time accepting that psychology plays a huge part in aural perception - and it's so easy to prove.




You won't find me in that camp, 90% of the misinformation that you hear on the internet is because people trust their ears too much and don't take into consideration sensory inhibition. Basically they don't confirm with the science, what they think their ears are telling them. Our perception can change from day to day and even hour to hour and our senses cross reference each other. For example the olfactory gland (nose) plays a huge part in how good or bad different types of foods taste, take that away with a cold for example and all food tastes bland. The taste buds haven't changed but the nose is clogged so it doesn't do it's normal cross sensory job. ITR, the eyes affect the way the auditory cortex receives audio information. Human elements such as mood, how fresh the ears are or how fatiqued they are, how much coffee you drank that day, how loud or low the music was that you were listening to previous to the test, how much cerumen is present in the ear canal, are you presently happy or pissed off at the boss, fought with the wife that morning, IOW emotion and a host of other physical factors that sway validity and have people coming up with all kinds of stuff that can be easily proven in the blind tests.

I guarantee you that if you don't confirm what your ears hear with blind tests and start performing them, you'll realize in a short time that the ear may be a great device, but the auditory cortex fails at being consistent and to top it all off, on the long term tests, the effectiveness of human memory and recall has a whole lot to be desired. Most blind tests bring up more questions than they answer, but it's the best that we have when you are audibly comparing sample rates, gear etc. and trying to eliminate psychoacoustic and psychosematic factors. The sighted tests always have some amount of expectations, these expectations can lead to imagination. Our brains are not perfect at consistent audio evaluation and when differences are very minute, the mind has a way of having some people hear things that aren't there. It's no surprise to me because I've been there done that and learned that my ears aren't perfect. I'm not afraid to admit it. There is really no such thing as a golden ear, all day, every day, all of the time. The science doesn't change from day to day, the human mind does, it's really that simple, confirm what your ears are telling you with blind, double blind when possible and null testing. You have nothing to lose and only more insight to gain, as to what is "really" happening.  
Logged
------Ronny Morris - Digitak Mastering------
---------http://digitakmastering.com---------
----------Powered By Experience-------------
-------------Driven To Perfection---------------

Pages: 1 2 [All]   Go Up
 

Site Hosted By Ashdown Technologies, Inc.

Page created in 0.036 seconds with 19 queries.