The aliasing thread

[Urs]

The tests may not accurately show the difference between the quality of the playback and pitch engines in the sampler. But I believe that it does show there at least is a difference.

Yes, I totally agree.

Anyhow, the site is made up in a way that the "innocent visitor" should assume that certain samplers are better than others, which might not be the case. It could also be the other way round...

Cheers,

Urs

[WilliamK]

..

[ericj23]

The BOTTOM image is a 15khz sine waveform done with a 192.000hz samplerate. Notice that it is "almost" smooth.

Well here's where good old Nyquist comes into play. A digital signal does not sound like it looks. 44.1kHz samplerate can play a perfect sine at 20kHz, meaning it can most definately play one at 15kHz. However looking at the the samples isn't gonna say much about how it sounds. When it's filtered and played back through the DA converter we're gonna hear it like a perfect sine.

The tests may not accurately show the difference between the quality of the playback and pitch engines in the sampler. But I believe that it does show there at least is a difference.

Just like you said though, more samples pitched up and down and compared would be interesting.

/Majken

just a little point - a cd player at 44.1 khz cannot play back a sine wave perfectly at 20,00 khz - even vaguely - it has two samples per complete sine cycle - so you play join the 2 dots and get a sine wave right ? - but its only got two samples to do a sqaure wave or a sawtooth - so how does it tell them apart - how does it know how to filter them to get the desired result

it doesn't

[george]

Just like you said though, more samples pitched up and down and compared would be interesting.

Pure sine waves are not capable of producing harmonics, that's why it is easier to "see" generated alias.

If you ask me, you are totally right, and I would REMOVE both sites. Unless they change how they test the programs.

Anyhow, the site is made up in a way that the "innocent visitor" should assume that certain samplers are better than others, which might not be the case. It could also be the other way round...

The site is providing resampling engine information, which is IMHO not biased and informative since they are graphics and not forum marketing hype. There are enough external links there to make someone aware of what aliasing is and make his own conclusions.

I don't think someone will stick or drop a sampler for the resampling quality only. As I have previously said, it is a feature I consider people should look, that's why I traced those graphics. What I do find surprising is how other developers have jumped here with arguments I won't comment.

[WilliamK]

..

[Majken]

just a little point - a cd player at 44.1 khz cannot play back a sine wave perfectly at 20,00 khz - even vaguely - it has two samples per complete sine cycle - so you play join the 2 dots and get a sine wave right ? - but its only got two samples to do a sqaure wave or a sawtooth - so how does it tell them apart - how does it know how to filter them to get the desired result it doesn't

A cd can theoretically play a perfect 22kHz sinewave since that's half of the samplerate. And yes this seems odd, cause like you say there's only 2 samples. But that's all it needs, 2 samples i sufficent to play back a perfect sine at that frequency. Which is what the Nyquist Theorem is all about.

When you're saying squarewave or sawwave you're beyond 20kHz and that information gets filtered out. A saw wave can be created using a series of sines (additive synthesis) however in this case the only audible of the sines would be the 20kHz base note. All others would get filtered out.

http://www.musicdsp.org/phpWiki/index.php/Nyquist

Here's a lil picture:
http://www.geocities.com/bioelectrochem ... t_wave.gif

/Majken

[Z3R0T0N1N]

Well...
I almost bought Jeskola XS1 because of the page. I am glad that I learned a little bit more about this stuff from reading discenting views, and following the different points of view and reading about this stuff; I found out that what I have is MORE than good enough, which would not have been the case if every developer just clammed up and didn't comment.
so, maybe if i had posted the site I would be defensive but as a user who had not paid much attention to the whole aliasing debate, I do feel like it has at least the potential to mislead. Not intentionally, and from a strictly logical point of view one can say that if someone is misled by it then they are ignorant. but in the real world....

anyway...

regards
s:b

[Majken]

I think you're both right actually. The tests are very honest. They're not fabricated to show XS-1 as a much better sampler than the rest or anything like that. On the contrary there are a bunch of samplers that look good and some that don't. It's based on measuring and not someone listening to audio files which also makes it likelier to be correct.

However on the other hand we have to ask what we're really testing here. Is pitching sinewaves beyond 15kHz ever going to be of any interest?

It would be nice to see the test complemented with various samples that are not just sinewaves pitched at lower octaves with audio examples so you can compare the soundfiles to see if there is more aliasing or not.

However making a proper test is always hard, it's always gonna be possible to argue that the test was specifically made to show a certain thing.

/Majken

[george]

However making a proper test is always hard, it's always gonna be possible to argue that the test was specifically made to show a certain thing.

It shows aliasing and nothing else, and despite William words, they are correctly made. I have had some people sending waves too but afaik they have been following the same guidelines stated at the site.

I have also noticed some spectral graphing done using wusikstation at 192khz and for your information, using such high frequency rate makes clean disavantages over all the other samplers because they were all working at 44.1khz.

Comparing all the sampler features will make it much more interesting though, but that should be a Sound On Sound article job since my research on the aliasing subject is nearly done.

[WilliamK]

..

[george]

Again, you haven't commented some simple statements I did. Like for example, pitching up a 15Khz, who would ever do that?

To see how proper filtering is made. There shouldn't be any alias if the pitched note root goes beyond nyquist.

The tests I did with Wusikstation were all done at 48khz, and showed almost no-aliasing at very high frequencies.

I'm now interested to make a test under the page guidelines, so we can properly compare with other engines using a 15khz sine wave at 44.1khz and running at the same frequency rate. Is it very difficuly to use a .wav only in wusikstation demo?

[WilliamK]

..

[george]

I think you have overlooked my question:

Is it very difficuly to use a .wav only in wusikstation demo?

I will be downloading the demo in the meanwhile.

Wusikstation sounds good just like any other sampler

I'm sorry to say this, but your statement is false. Check the site again and see how much developers actually care about resampling qualities, with special mention to VSampler developers who passed from being one of the worst to one of the better ones.

Linear interpolation compared to 64 point SINC is not the same. Wusikstation may or may not excel in other features, but definitely not "as good as any other sampler" regarding resampling.

Lucky us that people care of "how it sounds"

Let me express my dissapointment about how self contradictory is, from a developer side, to take care about "the sound" and using linear interpolation instead something better like 8 point SINC who has been around for ages now (IIRC was used in SB Live for the soundfont engine, and I'd say EMU samplers probably do too).

[schlsdwow]

george and william

both of you excel in producing wonderful instruments, capable of beautiful sounds and fantastic capabilities. both of you have great strengths in some areas, and could improve in others.

the very same can be said of rgc:audio's Rene, who's astounding and breakthrough sfz resampling engine really kicked this debate into gear (iirc), as well as Ohm Force's Laurent de Soras (sp?) whose research and writing have gotten many of us thinking....

the point that I am slowly getting at is that if this debate, emotional or otherwise, gets you thinking more about your dsp code, getting you (indirectly or not) to learn a new thing or two about resampling, dsp design, programming, etc., then -

all of us as developers and consumers of this technology are the winners, and better off for your time spent arguing here

it's nice to see, please keep it up & sorry to interrupt

[WilliamK]

..