Do hardware VA's alias as badly as software?

[Aroused by JarJar]

Just wondering from the people that have experience using them - do hardware VA's like Nord, Alesis, Access Virus, Roland etc... tend to suffer form aliasing in the upper register similarly to VSTi's that aren't oversampled?

Thanks,

NO for a very good reason. All software is coded differently. They are not the same. As Urs pointed out some alias badly, some not at all, some little etc.etc.

Yes, the software in hardware synths is coded differently. The hardware is usually very similar though (typically Freescale/Motorola DSP chips). In theory, you could plug the Virus C code into a Supernova II...of course it would only use one of the ~9 or so DSP chips since that's all the C had.

...but different languages don't change the universal realities of DSP such as Nyquist and aliasing.

Of course not in the same way as they don't limit particular DSP developer to produce own clever algorithm to fight with this limitation

Whether you're powered by chips or by general purpose CPUs, you can write good or bad algorithms for either. The processessing advantages of dedicated DSPs doesn't automagically translate to less aliasing.

[foosnark]

People often talk about going to 96k to prevent aliasing, but it really only gives you a little more than an octave at the high end, compared to 44.1. Maybe that's enough, or maybe it doesn't matter. I usually lean toward the "doesn't matter" camp but I fill my stuff with noise and grit anyway.

And in software, it's not necessarily the synth plugin that is aliasing. With FLStudio for example, some high notes with 3xOSC with triangle or square wave will alias... and if you stick a steep lowpass filter after it, the lower-frequency sound of the aliasing goes away along with the high frequencies.

[Kees49]

Nord Lead 2 here, no aliasing...

[EvilDragon]

Or so you'd like to think. Try a hardsync sound and play it in upper octaves. Guaranteed to alias. FM aliases as well. I had my play with NLs and Electors, they all alias at some point.

[chk071]

Can someone describe exactly what aliasing does to the sound, preferably with some sound examples? Read so much about it, but never really got around what it actually does to the sound.

[VariKusBrainZ]

Seems some people here dont quite understand what aliasing is and the methods used to defeat it.

The most basic and easy to implement solution is band limiting I believe.
ie you band limit the frequencies, usually of the oscillators, in order that any aliasing that occurs outside of the band will be filtered out (made inaudible).

Any synth that is able to create frequencies that fall outside its internal sampling rate will alias at some stage in its architecture. It all depends on the methods used to counteract this.

Get some education on the principles of digital audio and then come back and discuss.

Alot of n00bs seem to think that digital should perfectly reproduce any signal.
Have read up on why CDs are are 44100hz.

[VariKusBrainZ]

Can someone describe exactly what aliasing does to the sound, preferably with some sound examples? Read so much about it, but never really got around what it actually does to the sound.

Google

[Music Engineer]

Can someone describe exactly what aliasing does to the sound, preferably with some sound examples? Read so much about it, but never really got around what it actually does to the sound.

i posted an audio example in this old thread (warning: turn down the volume, when listening to those):

http://www.kvraudio.com/forum/viewtopic.php?t=231711

and in the tutorials section of my website, i have this article:

http://rs-met.com/documents/tutorials/D ... ignals.pdf

which explains the phenomenon.

[Aroused by JarJar]

Can someone describe exactly what aliasing does to the sound, preferably with some sound examples? Read so much about it, but never really got around what it actually does to the sound.

Aliasing manifests itself as new frequencies which are almost inevitably inharmonic barring occaisional unusual concidences. The effect is, as mentioned before, somewhat similar to ring modulation (in which usually inharmonic frequencies appear).

It's very easy to show how this happens with a pencil and paper, LOL.

At its worst (or best if you're a lo-fi guy) aliasing sounds like a mass of crunchy metallic noise. Most aliasing you'll find nowadays is subtle, unless you're really getting carried away- but many modulations which are perfectly normal for analog modulars will cause loads of aliasing in all but a few digital synths.

The Nord G2 isn't prone to aliasing, it's running at 24/96 with control and audio rates almost certainly identical etc., but it does alias. The very first thing I ever did on a G2 was check its aliasing, as I was comparing it to Reaktor (which was a joke of aliasing at that time). The G2 is very nice, but it's not real "digital smooooth".

A little high-frequency aliasing here and there doesn't mean much in real life. Synth-heavy music with lots of aliasing is definitely degraded though.

[cron]

Can someone describe exactly what aliasing does to the sound, preferably with some sound examples? Read so much about it, but never really got around what it actually does to the sound.

I'm not a programmer, but my understanding is that aliasing happens when a synth isn't prevented from attempting to generate audio content at frequencies above the sample rate. You end up with something that sounds kind of like the frequencies above the sample rate being 'reflected' back into the audible range.

Here's a (very) extreme example using the freeware Magical 8-bit plug, which I'm assuming from the sounds of things makes absolutely no attempt to avoid aliasing. As the square wave rises in pitch, its harmonics that go above 44.1kHz (the sampling rate I used in this example) are reflected back down into the audible spectrum instead of disappearing. What you should hear, and what you will hear with the majority of nicely coded synths when doing extreme pitchbends like this, is a smooth pitchbend upwards that eventually disappears when it hits 20.5kHz, sounding indistinguishable from a sine wave sweep once it goes above 10.25kHz (because all of the squarewave's harmonics are above 20.5kHz from here, only the fundamental frequency is in the audible range above 10.25kHz). Instead you get this total mess.

http://soundcloud.com/chqtestsubjects/aliasing-example

[chk071]

Thanks guys, also for the sound examples. So i figure aliasing can result in different symptoms? Sounds like grit on Robin's sound example, while it sounds like some kind of LFO modulation and noise on cron's. I remember also listening to some aliasing example on the Roland Gaia's supersaw, which sounded like a bit reduction, if i got that right.

[cron]

I believe that it generally sounds the same. The difference between Robin's example and mine is just that the pitch of the sound being aliased is moving in my example whereas it's just playing one note in his. When the pitch of the note being aliased isn't moving, you just get a static 'snapshot' of the aliased frequencies. I did a long pitchbend up in my example just so you can hear how aliasing is related to the original sound a little better, but I bet you'd get something closer to Robin's example if I were to just play individual notes with it.

Also, bear in mind that my example is really extreme. I'd say the stuff you hear between around 0:04 and 0:07 is pretty much getting on for the worst aliasing you'd ever have to deal with in the 'normal' use of an instrument that makes no attempt to anti-alias. The note being played is already very high by 0:07, and it wouldn't surprise me if it's gone above the range of human hearing only a few seconds later than that.

[@midnight]

Just wondering from the people that have experience using them - do hardware VA's like Nord, Alesis, Access Virus, Roland etc... tend to suffer form aliasing in the upper register similarly to VSTi's that aren't oversampled?

Thanks,

NO for a very good reason. All software is coded differently. They are not the same. As Urs pointed out some alias badly, some not at all, some little etc.etc.

I am pretty sure all plugin synths that don't oversample suffer from aliasing. Notice I used the word oversampled in my original post.

I've never tried Diva but I am guessing it aliases too when oversampling is off at 44.1khz.

I have synth squad, it audibly aliases even at 2x oversampling when pitchbending the highest notes. I've made patches using heavy waveshaping and modulation that actually still have audible aliasing all the way up to 8x oversampling, in those cases I have to bump it up to 16x or even 32x oversampling.

[EvilDragon]

Diva doesn't have an option for oversampling. It already oversamples internally many times per sample (because of the zero-feedback-delay filter prediction). You only get "Accuracy" parameter, which boils down to the amount of iterations of predictions for each sample...

[IncarnateX]

Aliasing is to modern VAs what drifting was to analog vintage synth. When they were popular everybody hated it. When analogs became vintage, everyone tries to emulate it. One of the problems with making a clone of Jp8000 is actually to emulate the way it aliases, which is a part of it's special sound character. Ironically, Synapse audio has just released the Antidote RE synth. One of its waveforms is called "digital saw" and is a special aliasing waveform, which is meant to emulate the string and pad sounds of older digital VAs. There you go, aliasing is becoming a vintage feature.

Cheers