WaveEditor Upsampling

  • Thread starter Thread starter jtwrace
  • Start date Start date
You only need to add dither when you reduce the bit depth.
 
1) Who else would say that you need to dither when you're upsampling?

2) WHY are you upsampling? You realize that you're not making anything better, right...?
 
Massive Master said:
1) Who else would say that you need to dither when you're upsampling?

WaveEditor recommends it.

2) WHY are you upsampling? You realize that you're not making anything better, right...?
Click to expand...

LOL. Being an audiophile we are crazy. :) There has been some talk about how great it sounded on one of the forums that I frequent so I figured I would give it a try as WaveEditor is free for 15 days. Purchase price is only $79. Why is it not making anything better? WOW. I wish you would join the talk on the other forum to explain this...
 
Dithering only affects bit depth/word length values. When you increase your bit depth, say, from 16 bit to 24 bit, all you're doing is adding 8 zeros onto the end of the original 16-bit value. Hence, even though the theoretical precision changes, the actual specific value does not change at all. Therefore there is no need for dither, and in fact, dither would actually technically make the value less precise since it would actually change the value (though only slightly.)

Dither is used when you go in reverse, when you hack off bits from, say, 24 bit to 16 bit. There you are losing both precision and changing overall value (unless those last 8 bits that were hacked off just so happened to be all zeros, but that is going to be an extremely rare occurrence.) There dither is used to "fuzz" or "smooth off" the lowest significant digit.

Think of it this way, like we were talking regular base 10 numbers. Take the value of PI out to 16 decimal places and you have ~3.1415926535897932. If you simply increase the word length to 24 decimal places without resampling, you have ~3.141592653589793300000000. No need to dither because the value remains exactly the same.

If you do resample, you'll get ~3.141592653589793238462643, with no need to dither because the resample is already at the limit of precision, and you really only have a 50/50 chance of dither making the value any more "accurate" than it already is at that last bit, with an equal chance of making it less accurate. So there's really no point when it's a toss up between making the value "better" or making it "worse" when you dither.

When you chop off bits however, you are indeed loosing accuracy - guaranteed - whether you re-sample or not; that last bit has a huge chance of being less accurate. There the randomization of dither actually "smooths off" the effects of simple bit truncation over the long haul and is where dither is most desired.

G.
 
joining talks on forums to explain things like that are generally pretty useless. Usually, if someone is convinced that upampling makes their audio sound better, no amount of explanation can convince them otherwise.

Truth is, it doesn't make it sound better to upsample. Any cheap D/A converter upsamples on the way out anyways. And you do not need to dither when converting up to 24 bits. In fact, you don't need to convert up t0 24 bits because it's not changing anything. Six chicken nuggets in a small box is the same thing as six chicken nuggets in a large box.
 
Dither and noise shaping are related to reductions in bit depth, not sample rate.

What do you expect to achieve by upsampling? The interpolation used when upsampling is going to be the same as the wave reconstruction occurring in your DAC anyway. All you are doing is leaving yourself with unnecessarially larger files. Any high-frequency content over the nyquist frequency that was filtered when the audio was sampled originally is not going to be magically summoned back, and you would probably have to be a dog to appreciate that anyway! (...yes, I call BS on the 'sensing of ultrasonic sounds' argument)

As for changing to 24-bit... again, all you are doing there is making the files larger. The audio will still be 16-bit, only contained in a 24-bit file and with the bottom 8 bits filled with zeros.
 
jtwrace said:
LOL. Being an audiophile we are crazy. :) There has been some talk about how great it sounded on one of the forums that I frequent so I figured I would give it a try as WaveEditor is free for 15 days. Purchase price is only $79. Why is it not making anything better? WOW. I wish you would join the talk on the other forum to explain this...
Click to expand...

If you have CD sound it's at 16/44.1. You can't add data that's not there. Changing the sample rate and bit depth can't change the quality of the sound it's locked once it's rendered to it's final form. There are no hidden unused samples and bits that will magically appear and make things sound "Better" if you up the bit depth and sample rate

If the file is 24/96 then it's already playing back at 24/96 in whatever .WAV player you're using unless you burn it to a CD in which case your CD burning software will shorten the wordlength and sample rate to 16/44.1 (you may need dither here)
 
bozmillar said:
joining talks on forums to explain things like that are generally pretty useless. Usually, if someone is convinced that upampling makes their audio sound better, no amount of explanation can convince them otherwise.

Truth is, it doesn't make it sound better to upsample. Any cheap D/A converter upsamples on the way out anyways. And you do not need to dither when converting up to 24 bits. In fact, you don't need to convert up t0 24 bits because it's not changing anything. Six chicken nuggets in a small box is the same thing as six chicken nuggets in a large box.
Click to expand...

No need to be rude about it. I'm here to ask and learn.
 
Ha ha. Sorry, didn't mean to sound rude. I'm just saying people will tell you all sorts of crazy things that they do to improve their sound, a majority of which don't do anything.

As explained better by glen, you really aren't adding anything to the sound by upconverting or upsampling. If that was what you were going to use waveeditor for, then I would suggest saving your money for something else.
 
bozmillar said:
Ha ha. Sorry, didn't mean to sound rude. I'm just saying people will tell you all sorts of crazy things that they do to improve their sound, a majority of which don't do anything.

As explained better by glen, you really aren't adding anything to the sound by upconverting or upsampling. If that was what you were going to use waveeditor for, then I would suggest saving your money for something else.
Click to expand...

OK. Thanks for being nice. :)
 
Oh god, audiophiles... next they'll be trying to sell us digital interconnects that go one step better than bit-perfect, and optical cables that operate faster than the speed of light :p

Ok, its possible there might be some negligible differences between interpolation occurring in the realtime reconstruction in a DAC and offline SRC, but we're talking differences that should be absolutely impossible to percieve, unquestionably so if we're talking high end, well clocked DACs. Its literally just filling the gaps with samples that don't actually need to be there... its a simplistic analogy, but if you have to draw a straight line from A to B, adding more points between them isn't going to be of any help when it comes to drawing that line because you already know the path from A to B.

There is still no way that the interpolation can create any content that was not there originally, and I'd love to hear more about these magic algorithms that were being claimed could do this. The iZotope SRC stuff is good, but not that good! Anyway, its main intended use is in studios for accurate downsampling with good filtering.

I'd also love to see those audiophiles do some proper scientific testing to confirm what they can and can't actually hear. Foobar ABX time anyone? ...oh wait, where did you all go? We're ya'll scared off by the prospect of having to face up to the fact that you can't actually hear a difference? :D


jtwrace, please don't take this the wrong way; I have nothing personally against you - its just a bit of light banter!
 
Last edited:
Oh man, I just read the first few posts in the link you put up here. At first I was 100% sure he was being sarcastic, but then I realized he was actually being serious. Trying to explain basic audio concepts to these people is like trying to explain to a two year old why he can't eat cookies for dinner.
 
mattr said:
Oh god, audiophiles... next they'll be trying to sell us digital interconnects that go one step better than bit-perfect, and optical cables that operate faster than the speed of light :p

Ok, its possible there might be some negligible differences between interpolation occurring in the realtime reconstruction in a DAC and offline SRC, but we're talking differences that should be absolutely impossible to percieve, unquestionably so if we're talking high end, well clocked DACs. Its literally just filling the gaps with samples that don't actually need to be there... its a simplistic analogy, but if you have to draw a straight line from A to B, adding more points between them isn't going to be of any help when it comes to drawing that line because you already know the path from A to B.

There is still no way that the interpolation can create any content that was not there originally, and I'd love to hear more about these magic algorithms that were being claimed could do this. The iZotope SRC stuff is good, but not that good! Anyway, its main intended use is in studios for accurate downsampling with good filtering.

I'd also love to see those audiophiles do some proper scientific testing to confirm what they can and can't actually hear. Foobar ABX time anyone? ...oh wait, where did you all go? We're ya'll scared off by the prospect of having to face up to the fact that you can't actually hear a difference? :D


jtwrace, please don't take this the wrong way; I have nothing personally against you - its just a bit of light banter!
Click to expand...

I understand what you are saying....I"m not that crazy now. I was at one time though. Then some double blind testing and it all went away. I completeley understand...

bozmillar said:
Oh man, I just read the first few posts in the link you put up here. At first I was 100% sure he was being sarcastic, but then I realized he was actually being serious. Trying to explain basic audio concepts to these people is like trying to explain to a two year old why he can't eat cookies for dinner.
Click to expand...

Oh come on. Everyone is entitled to have an opinion just as you do. For me personally I don't know enough about all this so that's why I'm here. To get insight from people that do this for a living. If you or anyone else would post on www.computeraudiophile.com it would be interesting as there are many DAC designers on that forum.

If not, I understand.
 
I read the first post also -

Going sideways here - Specifically, Norah Jones -- "Come Away With Me" is a great sounding album with a lot of contrast. If there's something "wrong" with it, that contrast is so great that I could see someone adding dither noise to it and considering it "better" sounding. Personally? The only way I listen to it is on 1/2 speed, 200g vinyl for basically the same reason.

I'm a WaveEditor user myself (even before it was priced at the freakishly low price that you can get it for now... Cough...) but I've never read the manual and have idea what it says about dithering. But if it does say something like that, it'd be the first I've heard of it...

But back to the point with that specific recording -- You can "hear the ladders" on that recording. The 'quiet' is SO quiet' and the 'loud' is SO clear that it feels almost unnatural when compared to a typical pop recording. In a rare occasion such as that, I could see where dither noise might make it feel 'smoother' (as mentioned, the sound of it on heavy vinyl is very smooth indeed).

[/SIDE NOTE]
 
Just dug out my copy of Nika Aldriges "digital Audio Explained" ....

( I usually keep it buried far away from view because ,quite frankly , reading it for too long tends to make my head hurt !!!:mad:)


So here goes .


1.) If stuff sounds better on your system when you up sample a mix and compare from the original, lower sample rate , then it's your converters .
They are benefitting from the lesser slope needed for the anti-alias and re-construction filters.
Most converters made recently are already using oversampling ( and lots of other DSP tricks ) to avoid the problems of phase shifting caused by very steep anti-aliasing and re-construction filters .



2.) the main DSP processing that can benefit from higher sample rates is the non-linear processes like dynamics (compressors , limiters and plugs that are adding analog goodies like HARMONIC ( as opposed to IN-Harmonic !!) distortion ) .

Most comp pluggs are internally upsampling and lots of others are using internal 64 bit engines. Even the free rea comp from cuckos alows you to set it to oversample X 8 or more . PSP has processors that use the "FAT" process to up sample internally . Voxegno's newer versions allow you to set them to as much as 8X over sampling internally .

Limiters can use the extra samples to take better advantage of the lookahead capabilities.


3.) It is the programming behind the processor that makes the sound .It is not neccessarily refelected in it's cost . Demo and use your ears .


From the Voxegno primary users manual
Oversampling Selector
This selector allows you to select a “quality factor” for the plug-in. An oversampling
allows a plug-in to run on a higher internal sample rate thus offering a better overall
sound quality. Almost all types of audio processes benefit from an oversampling:
probably, only gain adjustment, panning and convolution plug-ins have no real use
for it. An oversampling helps plug-ins to create more precise filters with minimized
warping at highest frequencies, to reduce spectral aliasing artifacts in compressors
and saturators, to improve a level detection precision in peak compressors. The
higher the oversampling setting is, the more CPU resources plug-in will require since
a CPU load is increased proportionally to the specified oversampling setting: at the
“8x” oversampling setting plug-in uses exactly 8 times more CPU time (and that is
excluding the time necessary to perform the oversampling itself).
Click to expand...



and
If not specified otherwise, Voxengo plug-ins are using poly-phase IIR low-pass filters
with at least 106 dB stop-band attenuation and 6% transition band width (which
starts at Fs/2) for the oversampling. Please note that these poly-phase filters impose
a phase coloration which sounds slightly different on various working sample rates.
Click to expand...


4.) One possible reason to upsample when mastering is because you enjoy the effect of the filters of the algorithum you use . I would expect that just as with dithering , a good and dedicated M.E. would have several different choices with which to proceed with up sampling if he decided to do so ... ( more arrows in the proverbial quiver ) .


Maybe the soft comps you have will sound a bit better after you upsample ; ... maybe they won't ... no guarntee .. no magic bullet , no one stop shopping , no easy scroll through the presets to sonic nirvana ...


Cheers
 
Last edited:
You must log in or register to reply here.
Back
Top