Audio Asylum Thread Printer

Get a view of an entire thread on one page

Computer Audio Asylum

the dubious pleasures of resampling

89.178.26.206

In Reply to: RE: cPlay 2.0b26 Released posted by cics on May 28, 2009 at 01:45:15

Hello, everybody! I know that many people like resampling to 96, or even 192. As I work in graphic design sphere, I often come across images, resampled by other people, who do it to conform to printshop standards. As our wave files can easily be represented, and they are (in professional programs like samplitude), as a black&white image, it is very easy to understand with the above picture, what resampling does to original image (wave). On the picture You see a detail of my cat. To the left - non resampled original (16 bit 44,1 kHz), to the right the same image (resampled 400 % - read to 96 kHz) The arrows show artefacts of resampling. Take a close look. Resampling was done in Photoshop, the king of professional image processing, so it can be compared to Sox or SRC.
If you watch carefully you will certainly find more. To better see the difference, save the image and zoom in on it in picture viewer program.
This is by no means to discourage You from resampling, I thought it might be interesting to represent the fact this way
Serge.

Follow Ups:

If it looks or sounds better then - audiozorro 10:07:23 06/01/09 (0)
IMO the pleasures are not dubious. Similarly, if something looks or sounds worse then IMO the shortfalls are not dubious. Only if one cannot discern whether something looks or sounds better is anything dubious.

The good thing about cPlay, as I understand it, is that upsampling is an option that can be bypassed. You may find that your DAC has better playback at 24/96 than 16/44.1 or 24/192. You may also find that software upsampling is superior to hardware upsampling and perhaps you prefer software upsampling with a NOS DAC or no resampling with a NOS DAC. On the other hand do you really want to resample in both software and hardware and does it sound better?

It always seems to come back to implementation is everything.
Analogy should be like this - cics 00:21:59 05/31/09 (0)
Take a low-res pic which would be our 44.1k. Then take a hi-res pic (@ ~4x) to be our 192k version. Yes the upsampled results are as per the hi-res picture if strict bandlimited interpolation is done (SRC does this; I'm not able to verify SoX but their site suggests so).

At issue with the upsampled hi-res pic would be added artefacts ("ringing" or more correctly "Gibbs phenomenon"). New information is added wherever we go from low to high intensity (i.e. discontinuities). This means at these boundaries the low intensity area now contains information that's NOT visible to us. This would be in the UltraViolet (UV) spectrum. Aside: Radio, TV, Microwave & Infrared are lower down the spectrum to visible light. Beyond UV are harmful X-rays and Gamma-rays.

So does this invisible artefact affect the visual? Experimentation with SoX confirms this to be true. SRC offers no settings but has a redeeming feature in that the pre-ringing (same for post-ringing) is just ~0.3ms in duration. SoX offers settings that reduce the pre-ringing but at the expense of changing some aspects of the visual (phase shift). Although the pre-ringing is short, the levels are relatively high (some 36db lower than transient peak) bringing into question how well your speaker's tweeter can handle this ultra-sonic load. Should this be at or near the break-up or resonance frequency, your tweeter is most likely to loose the plot...
RE: the dubious pleasures of resampling - Dawnrazor 22:39:34 05/30/09 (4)
Hey serge,

Forgive my ignorance, but I thought all this depends on the dac. Some handle different rates better dont they?
- RE: the dubious pleasures of resampling - steppe 03:05:49 05/31/09 (3)
  Hello! Yes, I think that generally resampling takes place within dac but not in case of cMp-cPlay, I believe, where it is done by src or sox and then fed to the dac to be used for conversion unaltered. The process for resampling can be different depending on the math. In graphics, which, I repeat, is very similiar to wave, the math can be bicubic, bilinear and some others, also different dithering (even the terms are the same) processes are involved like bayer, fatting, floyd-steinberg math, diffusion, dispersion... Next, if I remember correctly, some DAC manufacturers employ tracing - the process that converts our wave into scaleable peak which is closest to analogue (according to their decision and or understanding). IMHO one of the best but hardest ways to achieve the best sound is to try to read and convert to analog every bit to the last one, but I am not an engineer nor am I a program developer, so I may be mistaken, but resampling does NOTHING to bring us closer to truth. I will prepare more pictures with dithering and vectorization (bezier) examples, used in graphics that closely resemble what is happening to the wav within the dac or resampling program (if it is interesting). Also, once again I want to say that I am not trying to discourage anybody from resampling, nor am I ungratefull or trying to say that Sics is doing something wrong.
  Serge.
  - The TRUTH? - DHT 4 ME 06:15:25 05/31/09 (2)
    what the heck does that have to do with anything. You are listening to a live event through speakers , in a different room, with different amplification, at different volume, at a different distance, with different cabling, in digital, with different DAC"s, and different computer chips, different versions of cPlay (or other players), oversampling DAC's....etc.
    It's a friggin representation of an event.
    You should know that with your photographic endeavors. Fidelity is a goal but it is more than pure specifications.
    
    If every system sounds different what is TRUTH?
    
    Now remember that I am sensitive to this being a professional photographer. I hear comments like "is that the way it really looked?" all the time. I try to explain that the slide film I use responds differently than the eye to a scene and the Ilfochrome has it's own character which is dependent on the illumination etc etc. Even the most accurate photograph is a representation and fine art is an "Artist's" representation.
    
    So everything is but a representation and higher bit depth and sampling rates are the way of the futuure. Right now the playback hardware is ahead of the media. Non oversampling DAC's will probably never be made again in light of HD audio. So we live with it and find better ways.
    Bob
    www.PlateauLight.com
    - RE: The TRUTH? - Dawnrazor 21:50:42 06/01/09 (0)
      D,
      
      I try to explain that the slide film I use responds differently than the eye to a scene and the Ilfochrome has it's own character which is dependent on the illumination etc etc.
      
      I think that is what I was getting at. I thought some dacs like getting info at say 24/96 while others might be better off at 24/192.
    - RE: The TRUTH? - steppe 08:10:58 05/31/09 (0)
      I mean the initial wav file. Nothing more.
      Serge
Photoshop analogy - DHT 4 ME 16:34:33 05/30/09 (1)
I am a graphics pro also.I agree with your assessment however I combat the fuzzy uprezzing by grossly over rezzing and then downsampling.
It is the equivalent of sampling to 352.8 then do a DC offset and downsample to 192.
Of course this is in Photoshop and I have no way of doing it for audio.
Just and interesting idea
Bob
www.PlateauLight.com
- RE: Photoshop analogy - kalimasada 21:22:15 05/30/09 (0)
  Yep, I agree, by upsampling to 400% and then viewing the result in 25% image size, Serge has in fact do two resampling processes. I can usually accept images downsampled using Photoshop internal function, but for upsampling I prefer it done with external plugin such as Genuine Fractal Print Pro. I prefer editing wave file with Wavelab software, but for upsampling, I prefer SRC.
  
  Sopian
RE: the dubious pleasures of resampling - theob 13:42:34 05/30/09 (2)
yes I agree upsampling de-focuses the sound just like it does in the picture. but I recall what j gordon holt said when somebody told him that solid state devices sound more accurate than tubes to which he said then I like tubes better than accuracy. I go through periods where I like 16/44 better than 24/192. Its all an illusion.
- ILLUSION was the important word - rickmcinnis@dogwoodfabrics.com 09:34:50 05/31/09 (0)
  My schoolboy's limited understanding of the vastness of Hindu culture, I do wish I could take the time to truly understand, but the Hindu concept of ILLUSION is one of the most important concepts in coming to "understand" the world.
  
  All is illusion and one makes of it what they will. This hobby is about illusions within an illusion, no one knows what is really TRUTH, it is illusive and its discovery will do no one any good, or harm for that matter, for it is meaningless. Especially in this context.
  
  We try to find something that allows our consciousness to experience an event, an event we most likely did not attend in a hall we most likely have never entered, through microphones and recording equipment we will never own, blah, blah, blah.
  
  I think steppe had a clever insight, but not clever enough. I know there is no "bad" intention on his part, it was interesting, just not very useful. DHT for ME's analogy was far more accurate, to my mind.
  
  This is an interesting discussion but it leads nowhere. With such bare bones as RED BOOK we have to find a way to fill in the blanks and a best version of the "blank filler" will come. Just as with the LP, no one heard what we are able to hear now with LPs at the beginning of digital era, some amazing refinement occurred after what was the end of the LP's commercial life. Who knows what will happen with the CD, but all of us have lots of them and we would like to hear what is on them as best we can, as it was for those with large LP collections twenty years ago.
  
  My fear is that the future, except for a few specialist exceptions, is moving towards even less information, not more. In ten years we might be pining for the CD to return. I would bet someone will find a way to upsample/extrapolate this information into something much better than we would have ever thought possible.
  
  Bye,
  
  Rick McInnis
- You have found the right word - steppe 14:16:28 05/30/09 (0)
  Hello, Theo! You have found exactly the right word - de-focuses, it softens, washes away the edge. Unfortunately I cannot post a bigger image - or tiff image, where everything is much more explicite - it won't fit on the screen here. Also an old CRT display (which I cherish) is much better at such subtle differences. If we zoom to highier percentages the "96 kHz" right side picture looks awfully washed-out.
  I personally prefer no resampling .
  Serge.

Post a Followup:

FAQ

Post a Message!

Forgot Password?

Moniker (Username):
Password (Optional):
	Remember my Moniker & Password (What's this?) Eat Me
E-Mail (Optional):
Subject:

Message:	(Posts are subject to Content Rules)

Optional Link URL:
Optional Link Title:
Optional Image URL:
Upload Image:
E-mail Replies:	Automagically notify you when someone responds.

To view your new posting or follow-up, click on the RELOAD or REFRESH button on your browser.