KVR Audio

itoa · Post by **itoa** » Wed Nov 26, 2014 7:36 pm

Yes and the conclusion was,that analytical solution it's not a problem with 2 linear waves.

mystran · Post by **mystran** » Wed Nov 26, 2014 7:48 pm

itoa wrote:Yes and the conclusion was,that analytical solution it's not a problem with 2 linear waves.

Well, yeah .. that's the "general" part because "2 linear waves" isn't exactly very flexible.

itoa · Post by **itoa** » Wed Nov 26, 2014 8:06 pm

Well, saw, triangle, pulse and 'segmented sine'

.

But referring to the subject.. most probably we (Ernstm

) are talking about non-analytic FM. Most probably these polynomials produce low pass filtered signals. And this reduces aliasing during x-modulation. But is this method better? I'm not sure, I want this star-dust over 10khz!

ernestm · Post by **ernestm** » Thu Nov 27, 2014 3:03 pm

Well yes, there are two parts to the FM issue, as I understand; first that conventional bleeps add dissonance, because when the output frequency is changed, they no longer antialias but instead create harmonics of the wrong frequency.

The second part is antiialiasing the result, and as observed, this isn't so big a problem with AM, but for FM yes the results can be significantly aliased if you are not using a sine wave, and linear oscillators are ok, but this technique does work much better with FM for pulses because the transition in the resulting signal is also nicely smoothed.

http://www.yofiel.com/software/cycling- ... scillators

ernestm · Post by **ernestm** » Thu Nov 27, 2014 3:05 pm

...and the transition shaping is essentially not frequency dependent, except for adding oversampling at higher frequencies

itoa · Post by **itoa** » Thu Nov 27, 2014 3:16 pm

Did I get you right?
So, you generate band limited signal for modulator osc and then use it to modulate 2nd band limited oscillator?

ernestm · Post by **ernestm** » Thu Nov 27, 2014 3:59 pm

Well that was why I was particularly pleased with this method, yes, an EPTR antialiased pulse oscillator can be a modulator without creating audible alias artifacts. The resulting aliasing is well above Nyquist.

itoa · Post by **itoa** » Thu Nov 27, 2014 4:01 pm

This attitude is just wrong.. (I had problems with this a time ago)

The problem is inter-blep modulation - you use "band limiting artefacts" as a modulation source. You may use a "darker" step, like polynomials, that cuts frequency earlier,to suppress aliasing (and loose high frequencies) but imo this is solving the wrong problem. This is exactly the same as you low pass modulator before modulating target freq

Forget about samples and quantisation and compute this analytically like continuous geometry.. When you detect discontinuity in modulator or destination signal, compute blep position - but... add the blep _only_ to destination signal.

ernestm · Post by **ernestm** » Thu Nov 27, 2014 4:04 pm

ernestm wrote:Well that was why I was particularly pleased with this method, yes, an EPTR antialiased pulse oscillator can be a modulator without creating audible alias artifacts. The resulting dissonant aliasing is well above Nyquist.

ernestm · Post by **ernestm** » Thu Nov 27, 2014 4:09 pm

I should add, the current published code cannot actually modulate itself. for the generic example, I assumed the modulator is an external sine, and it therefore uses spline interpolation to determine the modulator frequency. The spline interpolation would incorrectly guess the modulator frequency during FM modulation by another EPTR pulse.

However, if it was modulating itself, or another osicllator was modulating it, I would already know the modulator frequency and not need to interpolate it. It would be a change to a single line of code and adding an input parameter for the modulator frequency for it to work correctly, in fact it would be better, because the modulator frequency would not be guessed by interpolation.

itoa · Post by **itoa** » Thu Nov 27, 2014 4:15 pm

ernestm wrote:
ernestm wrote:Well that was why I was particularly pleased with this method, yes, an EPTR antialiased pulse oscillator can be a modulator without creating audible alias artifacts. The resulting dissonant aliasing is well above Nyquist.

Answer yourself why

I know that you are excited, I'm the same (our fuel

).. but nothing new was invented here and there are better methods for this. Just spend some time on browsing this forum.

ernestm · Post by **ernestm** » Thu Nov 27, 2014 6:30 pm

I am perfectly well aware I didn't invent anything new. Davey asked for help because he couldn't get EPTR to work as suggested it could in professional IEEE and MIT documents. Peter figured out the tanh on sin from an article in Computer Music Journal, but her hadn't made an EPTR, just a PTR. t tried it, and it worked really well except at high frequencies, so I added my own resampling method which I learned from Oki Electtic 20 years ago. It's very nice you folks figured out something similar, and Im verity happy for you, but the fact is, there is no other working code demo of EPTR in the public domain. That's what i said, and if you want to install the Max demo and try it, you can see it yourself without spending any money. You are welcome and happy thanksgiving.

tonfilm · Post by **tonfilm** » Fri Nov 28, 2014 4:56 am

thanks, this thread was very informative. but why is the author banned!?

mystran · Post by **mystran** » Fri Nov 28, 2014 5:12 am

tonfilm wrote:thanks, this thread was very informative. but why is the author banned!?

I was actually wondering the same.

tonfilm · Post by **tonfilm** » Fri Nov 28, 2014 5:16 am

I am currently working on a C# implementation of the EPTR method described here. It seems that it is a quite low on CPU method to get some decent oscillators...

What other methods are there which have better audio quality and are also low on CPU? any terms i could search?

EPTR Pulse and 'LTR' Antialiasing Oscillators