Aliasing and hardware synths

[meatballfulton]

Why does this happen anyway?

Because there is no over-sampling and anti-aliasing filter in the code. A.K.A. bad maths

Aliasing causes frequency shifting.

With aliasing for frequency A and Nyquist limit B the output is A-B if A > B.

With ring mod for two frequencies A and B the output is A+B and A-B.

Simple example:

Sample rate is 32khz so the Nyquist limit is 16khz. Harmonics above 16khz will be shifted down by 16khz. For example a harmonic of 19 khz will result in a 3khz harmonic instead. So like ring modulation, you end up with dissonant harmonics.

[knolan]

The following synths are good on the aliasing front:

Yamaha AN1x (very clean even in exposed solo lines at the upper end of the keyboard). Very good for a Virtual Analogue synth from the late '90s. And well respected too.

Korg OASYS and Kronos both use ultra-low aliasing oscillators in their various synth engines, again rendering them very clean even at high frequencies; and once again a joy to play.

[madtheory]

I seem to recall a regular here complaining that, with heavy audio rate modulation, the AN1X had too much aliasing for his tastes? Not sure... never had the pleasure. But if it's well regarded by knolan and Paul Wiffen, that's good enough for me

Spurred on by a Gearslutz discussion about whether the different versions of the Novation synth engines (A-Station, X-Station, KS and V-Station) are actually the same or not, I did a side by side comparison of KSR and V. Interesting results.
1. Encoders on my KSR were a little dusty.
2. Pitch bend wheel on keyboard also dusty.
3. Had to manually initialise V-Station patch because I can't remember the trick for getting a patch sysex file from KSR into V-Station. There is a change needed in the header code. Maybe I should've scrolled through the V-Station patches to fin the Init Patch... so there is a little more envelope mod on the KSR. Soz, was rushing.
4. The notes don't trigger at precisely the same time on the external synth and the plugin. MIDI jitter

So what you hear in the recording is each synth in mono, panned hard L and R. One sawtooth, "VCO drift" set to 0, no modulation of any kind. First a simple chord progression, then C in each octave low to high, pitch bending the high notes, then some filter and amplifier modulation on one note, C2. I use the KSR to change envelopes and filter on both synths simultaneously. Also did some pitch bend testing. So you might hear a little glitching on those moves as first, reduces as the dust clears LOL. I really must open up my master keyboard and service all pots, sockets and the AT strip.

Anyway, here is the recording, have at it:
https://www.dropbox.com/sh/c0r7u50e9...1fKuzFbua?dl=0

They sound very very very very very close. But there is a difference in the code. You can hear why, with the very high notes in the middle of the recording

[knolan]

At the end of the day with all digital gear - you can push it over the limit of where it behaved cleanly and where "aliasing" or "aliasing like" behaviour kicks in. So you could well be right that you read someone finding that limit with the AN1x - perhaps when it's implementing Ring Modulation, Hard Sync or FM.

But - for straight forward oscillators - it's beautiful (I have one).

Indeed - over the past year, on and off, I've been "configuring" a personal sound library for it - as in - trawling though all of the online available sounds for it and selecting the best there are along lead, pad, basses, pianos - the usual suspect categories, to built into a personal "optimised" library.

I 'm using the excellent AN.xfactory software by Derek Cooke to rapidly load the various sound banks onto the AN1x where I can then personally audition each sound, and if I like it, take note of it in a spreadsheet, and then build a more optimised library using only the xfactory software.

I've finished auditioning the available programs online - perhaps 3000 - 4000 programs over countless dozens of bank files - and have indeed selected the very best and created the spreadsheet - it contains about 700 - 800 sounds. So the remaining job is to build the optimised sound banks.

Why I'm saying all of that is - having spent all this time goung through essentially every available sound for it - I am -honestly - blown away by the quality of it. It really is amazing. And I dont' say that likely. The quality of moog like leads and basses, the breath and depth of pads and the sheer variety of electric pianos, organ sounds - is quite remarkable.

But its the underlying quality of the synth engine that underpins it. This synth is not remarkable from an onboard effects stand point - its strength is its synth generation system - and boy is it rock solid. I really love it.

When I bought the AN1x I was not that enamoured with it because it looks quite ordinary and it's front panel is quite tricky to navigate because of the sheer number of features (and the manual keeps annoyingly saying how easy every thing is - but it's not (to me!)) - but the strength of the instrument is quite something else. It's Hard Sync is incredibly strong, sharp, clean and convincing - it give my Mono Poly a total run for its money.

and then there are all the other fabulous features of the AN1x - Free EG's (control surface knob recording); amazing arpeggiator and step sequencer features and playability and the Scene Morphing - but to get back to the original point - if others have managed to push it to its limit I'm sure it can be done - but - to me - it's incredibly pristine - even in territory where even modern VA's struggle.

The AN1x has an earnest reputation - and is used by a fair number of the bigger "Tribute Bands" to do convincing emulations of early synth classics. It's really worth checking out (as are the AN engines on the EX5 and EX5R and the AN200 for the same reason). AN is an impressive VA technology by Yamaha.

[briandc]

Hi everyone,
I was just re-reading this thread about aliasing..

I've read that, the aliasing from a "single" audio signal aside, it would seem that aliasing is also an issue when there is "system overload." From SoundOnSound:

Aliasing can occur either because the anti-alias filter in the A-D converter (or in a sample-rate converter) isn't very good, or because the system has been overloaded. The latter case is the most common source of aliasing, because overloads result in the generation of high-frequency harmonics within the digital system itself (and after the anti-aliasing filter). ...Consider what happens if our 10kHz [sine] tone is cranked up too loud and overloads the A-D converter's quantising stage. If you clip a sine wave, you end up with something approximating a square wave, and the resulting distortion means that a chain of odd harmonics will be generated above the fundamental. So our original 10kHz sine wave has now acquired an unwanted series of strong harmonics at 30kHz, 50kHz and so on.

source link

I imagine that it is not only the case when one wave is at a distorting level, but also when several notes are played together, where the "summation" of all the tones could cause interaction that could produce signals going above the Nyquist Frequency.

Furthermore, Miller Puckette, in his very interesting writing, "Pure Data Synthesis," would seem to suggest that the aliasing problem in digital audio is also due to the actual waveforms themselves:

.... The salient features of classical waveforms are either discontinuous jumps (changes in value) or corners (changes in slope). ...the sawtooth and rectangle waves have jumps (once per cycle for the sawtooth, and twice for the rectangle), and constant slope elsewhere (negative for the sawtooth wave, zero for the rectangle wave). The triangle wave has no discontinuous jumps, but the slope changes discontinuously twice per cycle.
To use classical waveforms effectively, it is useful to understand how the shape of the waveform is reflected in its Fourier series.... These waveforms prove to be much more susceptible to foldover problems than any we have treated before, so we will have to pay especially close attention to its control.

His view is that additive synthesis is "safer" than other synthesis techniques, as far as avoiding aliasing is concerned:

most techniques other than additive synthesis don’t lead to neat, band-limited signals.

brian

[Ashe37]

If your waveform is a good analog-sounding sawtooth or pulse wave, the was does not 'jump' back to zero, there is a transition involved. Apparently Puckette likes dealing with mathematically pure waveforms, which most people would label 'sterile'.

However, as a note, i was able to make the Novation Peak alias, but I had to purposefully do some pretty strong audio-rate modulation to make it happen.