Friday, August 28, 2009

The kind of screenshot I fancy



I rechecked my DMG waveform synthesis yesterday, and made sure I had the range and AC Coupling emulation right according to the lowest 32Hz freq of the DMG Wave Channel.
(i dont have a pro mod DMG, I personnaly like the sound of the phones out of mine)

EDIT: For fairness to the scientific community, here are the actual samples.
1)chipsounds
2)DMG phone out

Slight tonal difference, but as herr_prof mentionned it can sound different from DMG to DMG.

More: Spectrum of DMG:


More: Spectrum of chipsounds:


You can save the files and animate them, as you will see the real DMG has some broadband noise at -90db, while chipsounds is pretty flat. Should i add extra noise samples to mix in? I seriously thought about adding them as extras

10 comments:

  1. Ears are probably better than scopes.. the sound can vary from dmg to dmg!

    ReplyDelete
  2. True, i edited my post with samples.

    ReplyDelete
  3. This reminds me of the screenshots i saw when I was reading up on Arturia's emulation of the minimoog. You had the two waveforms layered on top of each other, with the red one being the synth and the blue one being the plug. Damn near perfect.

    Davie-boy, keep doing what you do!

    ReplyDelete
  4. Hi

    in all fairness that type of picture doesnt tell the whole story. But the title of the blog stands, its juts nice when you compare waveforms and they match. Theres lots more than static waveforms tough, pitch accuracy, and amplitude fluctuation accuracy which, surprisingly is not that easy to emulate.

    ReplyDelete
  5. Very cool. I'd like to see the FFTs as well.

    ReplyDelete
  6. You can do the analysis yourself since the samples are provided, but i made some extra shots myself.

    ReplyDelete
  7. Wow. Pretty wild set of partials in that sound. :-)

    Just a few off-the-top-pf-my-head observations. They could be bullshit, as I'm not in the trenches with you.

    1. Your noise floor up through 300Hz is rather high. Maybe a high-pass filter is indicated? Although this would affect phases and any DC bias you're modeling. So maybe a bad idea. Maybe it's modeling error. And I kinda doubt it's audible anyway. And as for your question about the noise floor at -90db, yeah you could add some there, but again I don't know if it's audible. This is just the osc, right? So a lot of the real answer comes from what happens in the filters as a result of these differences in the oscs.

    2. You're missing some action around 70 Hz. MAYBE that's audible.

    3. This is the biggie to me, and I bet it's audible. The amplitude of your partials (especially the first 7, which will really make up the character of the sound) descend almost in a straight line in your synth (yeah, I know we're not on a linear scale here).

    But the hardware partials seem to be grouped in pairs after the fundamental. The same with a bunch of the upper partials. You've done a wonderful job, so I feel like I'm nitpicking, but if you're going to try to improve the modeling in an audible way, I'd try to mimic those paired amplitudes.

    Not meant as criticism--I think chipsounds is the bomb.

    ReplyDelete
  8. About the audibility of differences, the wav files are linked to the post, so you an try it out.

    Ill investigate the 70Hz peaks, and try to record other tones for more details. Maybe its some (phase)/distortion in the analog circuitry.

    Thanks

    ReplyDelete
  9. Yeah, I don't have the best ears for that. Remember, I worked at COMPUTE!'s Gazette. I'm old and busted. Can't even hear the highs anymore. Plus I'm listening through cheap computer speakers.

    Chip sounds are really complex with a lot of crazy stuff--not-harmonic partials and weird noise. A lot gets masked. You can hear a difference and not be sure what it is.

    Plus, me looking at a DFT doesn't tell me everything. Like you said you need to see how the amplitudes move through time. But I know from experience that the relationship between the first few amplitudes are absolutely crucial to the timbre of the sound. That's where you can make your best improvements, IMO.

    You're pretty close to shipping, aren't you? You should just go for the big bang-for-the-buck items now and save insane details for the next version. You've got great sounds already.

    Emulating a classic is a tricky thing. People will love you and hate you. They WANT it, but they are likely to pick it apart.

    But that's so much better than working on a VSTi that no one notices. :-)

    Anyway, don't mind me. I don't want to distract you from shipping.

    I wouldn't mind playing with the beta or taking the conversation offline. Let me know if you want my email and I'll figure some way to get it to you.

    ReplyDelete
  10. yes we are pretty close to shipping, and there will be more intense nit pickers once its out, im sure :) Ive got very constructive comments by some of them in the beta team.

    You can send me a private email if you want to pursue this. email info at plogue com and tell its anonymous from CG compute gazette.

    Cheers

    ReplyDelete