Dubstepforum.com

aliasing is good, they wouldn't make bitcrusher plugins if it wasn't.

I really wouldnt worry about it too much, unless your using loads of high pitched tones etc, getting a decent mix together will make 100x more impact. Ive done a load of tests and its barely noticeable even on individual instruments.

Still the science is interesting... to a degree

24 bit is far more important than >44khz Imo.

cheers

AJGR wrote:aliasing is good if you want it, they wouldn't make bitcrusher plugins if it wasn't.

fxd

paradigm x wrote:getting a decent mix together will make 100x more impact.

Agree 10,000%, and especially for me , not everything I do requires mixing though, need just raw clean hifi sound.

Aliasing does my head in. It's my one proper bone to pick with digital dj'ing. I hate hate hate it when laptop/serato/traktor djs use the pitchlock function on anything with a solid sine sub (ie: every single one of my tunes), you can instantly hear the program of choice struggling to recreate the smooth wave sound without any artifacts on indeed in the right pitch. However, from my reading of the post, as the majority of sub will sit between 30-80 hz, doesn't that mean that having a sampling rate of only 1.6khz or above would provide artifact free resampling/pitch correction? I guess within the remit of dj'ing you have so many other frequencies taking up sample points it becomes a mess to timestretch and preserve pitch regardless of my SL3 doing it in 48khz?

Surely a good dither will deal with aliasing?

Or does that just soften the harsh frequencies?

I dont find it usefull to to spend a lot of time on alliasing. There might be a view open source things and freeware stuff that does not have correect oversampling but really.. meh.

Oversampling:
http://www.savioursofsoul.de/Christian/programs/
http://sites.google.com/site/chrisrwalton/oversampler

gnome wrote:Surely a good dither will deal with aliasing?

Or does that just soften the harsh frequencies?

http://www.3daudioinc.com/3db/showthrea ... jitter-A-D
http://www.gearslutz.com/board/attachme ... ldrich.pdf

pete bubonic wrote:Aliasing does my head in. It's my one proper bone to pick with digital dj'ing. I hate hate hate it when laptop/serato/traktor djs use the pitchlock function on anything with a solid sine sub (ie: every single one of my tunes), you can instantly hear the program of choice struggling to recreate the smooth wave sound without any artifacts on indeed in the right pitch. However, from my reading of the post, as the majority of sub will sit between 30-80 hz, doesn't that mean that having a sampling rate of only 1.6khz or above would provide artifact free resampling/pitch correction? I guess within the remit of dj'ing you have so many other frequencies taking up sample points it becomes a mess to timestretch and preserve pitch regardless of my SL3 doing it in 48khz?

cant be aliasing on sub... not at those frequencies. Timestretching artefacts. Realtime timestretching isnt going to sound that good at the moment.

paradigm x wrote:

pete bubonic wrote:Aliasing does my head in. It's my one proper bone to pick with digital dj'ing. I hate hate hate it when laptop/serato/traktor djs use the pitchlock function on anything with a solid sine sub (ie: every single one of my tunes), you can instantly hear the program of choice struggling to recreate the smooth wave sound without any artifacts on indeed in the right pitch. However, from my reading of the post, as the majority of sub will sit between 30-80 hz, doesn't that mean that having a sampling rate of only 1.6khz or above would provide artifact free resampling/pitch correction? I guess within the remit of dj'ing you have so many other frequencies taking up sample points it becomes a mess to timestretch and preserve pitch regardless of my SL3 doing it in 48khz?

cant be aliasing on sub... not at those frequencies. Timestretching artefacts. Realtime timestretching isnt going to sound that good at the moment.

I don't think that's true. I saw it demonstrated in Wavelab a few weeks ago that a low-frequency sine wave can cause aliasing or imaging or somesuch, I forget. Anyone have any details on this?

im not sure... wiling to be proved otherwise but i cant see how a LF sine could cause aliasing...

Occams razor suggests that the realtime time stretching is whats flipping his shizzle up...

paradigm x wrote:im not sure... wiling to be proved otherwise but i cant see how a LF sine could cause aliasing...

Occams razor suggests that the realtime time stretching is whats flipping his shizzle up...

I agree with you, I'm just pointing it out because this started as a theoretical discussion. I remember the lecturer sweeping a sine wave around the bass and subbass frequencies while we were watching the frequency graph, and you could clearly see two or three small harmonics mirroring around the Nyquist frequency. I can't remember exactly what he said, but I think it was an aliasing-related issue.

well, thats weird. Never heard of it myself.

gnome wrote:Surely a good dither will deal with aliasing?

Or does that just soften the harsh frequencies?

Differ helps to prevent quatization errors when rendering, it's more to do with bit depth than sampling rate.

nowaysj wrote:Looking at these graphs, can someone shed some light on quantization noise? I notice that flstudio 8 is relatively clean (and I mean relatively!) in terms of aliasing, but shows some signs of quantization noise. I'm pulling up heady papers, best so far is by Ethan Winer, but I really don't totally trust him.

When you record an analogue waveform (such as the electric signal created by moving air hitting a microphone), you are converting it to digital. This means the computer has to measure the voltage every fraction of a second.

But voltage can vary infinitely, whereas digital numbers have discrete steps.

So if you're using a 16-bit recording bit depth (which can describe 65536 different numbers), but the incoming voltage is 0.7325567842345985767563984753945875 volts, what number does the computer choose?!

Obviously, it's impossible to choose an exact value because none of the values the computer has available are exactly that of the incoming voltage.

And now for something completely different: have you ever used a shaper in your music? A shaper is usually represented on a two-dimensional graph describing its transfer function. A linear transfer function looks like this:



A linear transfer function means the input is exactly equal to the output.

But the transfer function of an analogue-to-digital converter is not linear. Say we're using a 3-bit system (which can describe 8 different numbers) with a range of 0-0.8 volts (just for simplicity). If the incoming voltage is anything between 0 and 0.0999999999..., the computer records a 0. If it's anything between 0.1 and 0.19999999999..., the computer records a 1. And so on. What this basically means is the transfer function looks like a staircase.

And as we all know, a shaper with a staircasey function is going to cause some filthy, filthy distortion

So basically, that's quantization noise/distortion/error: it's the distortion resulting from the digital conversion of analogue signals. Obviously it depends on exactly how a given system rounds off the numbers and so on. And bit depth has a lot to do with it: a system running at a higher bit depth will have smaller "steps" on the "staircase", and therefore the resulting distortion is a lot softer.

Dither is a system of injecting random noise into the system so that instead of distortion, you end up with noise which is a lot gentler on the ol' ears (because it basically turns the staircase back into a line).

Does that help?

henrebotha wrote:Does that help?

yes it does, thank you!

Okay, so when looking at this link:

http://src.infinitewave.ca/

Logic 9 seems to have a shit tun (technical term, sorry) of quantization noise, while Reaper 3.52 seems to be far cleaner. What is Reaper doing that Logic isn't? And how big of an issue is quantization noise? How much impact does it have on the final sound, on a sound that has been resampled 10 times? In logic, it looks like the q noise is around -130 db shouldn't be too big of a problem?

i remember about myself, bugged over these SRC comparisons for ages before settling for a daw...

sometimes i really think i can hear the difference in daw sound qualities

but then i hear some work from some people using a said daw.. and i dont know anymore

anyone else feels the same ?

this discussion is lovely.