Korg Forums

Howdy, I stumbled upon a rather interesting take on the Radias MMT synthesis:

http://blog.zacharcher.com/2011/03/04/d ... rg-radias/

of course this is not a proof that radias uses samples instead of real oscillators but if it does it's quite disappointing

Edit: of course i know that samples can be used as oscillators but in the case of such stuff as square/triangle/sine wave etc you'd expect to find an algorithm generating the sound.

Anyone has ideas what could be causing the cutoff and aliasing effect in radias?

omfg this article again, its total horseshit.

Previous discussion here:
http://www.korgforums.com/forum/phpBB2/ ... cf621b2383

Wow X-Trade...I searched for that discussion but I couldn't find it...

Even if the oscillators were samples, what would be the difference to a mathematic formula? Even the first organs oscillators, that had a set of sine-shaped gears rotating in front of magnetic pick-ups are a kind of samples, the mellotron is also a sample player (magnetic tapes).

Any method to bring something to oscillate is valid, even samples. Important is what the oscillator outputs.

I personally would prefer sampled oscillators (provided it has a good interpolation algorythm) to mathematically or physically generated oscillators because they provide the possibility to create any waveform you want providing more variety in what is the essence of every synth.

axxim wrote:Wow X-Trade...I searched for that discussion but I couldn't find it...

Same here, that's why I started the thread. Sorry, if I had found it I would have just commented on the old thread.

axxim wrote:Even if the oscillators were samples, what would be the difference to a mathematic formula? Even the first organs oscillators, that had a set of sine-shaped gears rotating in front of magnetic pick-ups are a kind of samples

The difference would be that a mathematical formula can calculate the output at equal resolution (sampling rate) for all pitches, whereas a sample will start to quantize when it's being stretched to lower frequencies (and the loss in the high end will start to become visible as there was no data above the sampling rate to begin with). Also when compressing the sample to higher frequencies it can cause unwanted artifacts unlike an algorithm which will produce a "pure tone" calculated specifically for each requested frequency.

axxim wrote:I personally would prefer sampled oscillators (provided it has a good interpolation algorythm) to mathematically or physically generated oscillators because they provide the possibility to create any waveform you want providing more variety in what is the essence of every synth.

I would prefer to have both. Samples are extremely nice thing to have as oscillators (especially when using external input as an oscillator) but equally i'd love to use interesting formulas as oscillators without the possible artifacts introduced by compressing and stretching samples.

I'm actually pondering on purchasing Radias as it seemed like a very versatile synth, I just became curious when i stumbled upon this blog post.

Another thing worth pointing out is that the Radias/R3 engine lets you alter and modulate the shape of the waveform for oscillator one. Things such as pulse width modulation would be hard to replicate with a sample based waveform, which is why you don't really see this feature on ROMplers.

hermanni wrote:...The difference would be that a mathematical formula can calculate the output at equal resolution (sampling rate) for all pitches, whereas a sample will start to quantize when it's being stretched to lower frequencies (and the loss in the high end will start to become visible as there was no data above the sampling rate to begin with). Also when compressing the sample to higher frequencies it can cause unwanted artifacts unlike an algorithm which will produce a "pure tone" calculated specifically for each requested frequency...

That is why I wrote "(provided it has a good interpolation algorythm)" in my post.

Surely I wouldn't make such an oscillator only playing the sample at the desired frequency. It rather is to be used as a look-up table with the respective interpolation which involves additional calculations that also take processing time, but works as clean as a mathematical formula. Of course an oscillator at about 22kHz would have only two outputs, the same as the formula will if the sample frequency is 44.1kHz

hermanni wrote:I'm actually pondering on purchasing Radias as it seemed like a very versatile synth, I just became curious when i stumbled upon this blog post.

I highly recommend it--excellent synth. I've had one for a couple years and just started really digging into more extensive programming with it.

So far, I've been quite happy with the results.

I'd like to point out that the author of that blog post has never acknowledged or discussed the major errors he made that we revealed, nor has he posted any follow-up entries to his blog or here. If he's not "man enough" to even acknowledge his errors, then he's the last person you should trust to evaluate the Radias. He was merely trying to find a way to justify his own incompetence in getting the Radias to sound the way he wanted, and he never demonstrated any skill or knowledge about synth programming, whatsoever. It's total BS wrapped in enough techno lingo to sound credible on the outside, but it's like crap wrapped in a Christmas box.

He had posted the same thing over in a real music forum and a bunch of people pointed out that he was wrong, but unfortunately, his personal blog post about it pops up easier on a Google search which contains no corrections. Probably just deleted whoever tried to debate it.