View Single Post
Old 14th September 2020
  #59
Here for the gear
 

There is a lot of amp sims that are just waveshapers or clippers of some kind and they all sound different. In the end of it even tube amp is a very advanced yet kind of waveshaper.

I have looked at Artem's samples that were claimed to be made of the real amp and I gotta say that after my peek at spectrogram it shows that there is nothing of useful real amp alike content after 11kHz - this is the way Kemper Profiling Amplifier behaves if a profile was made with a miced up amp. It literally mirrors the signal that is there since 0-11kHz and mirrors it vertically so it ranges from 11kHz to 22kHz, then applies a low pass filter of some sort. What a gimmick

This usualy happens when no interpolation present or the Kemper developers wanted to mask the fact their thing will work at the sampling rate of 22050Hz.

Well, even despite all this Kemper indeed sounds pretty good. What's strange is that Kemper does produce correct result only if no cabinet and mic were present when profiling was conducted, e.g. the full range of 0-44100Hz is present. After so many years that this thing is being around you guys could not hear it - this is why it is so simple to fool you.

Anyway, the real amps indeed have interesting harmonics fadeoff what I call it. The point here is that the harmonics cut off abrubtly only in one case - the hard threshold was set or a hard clipper used. Replace it even with an even a bit more soft one and the problem disappears.

What else is important to mimic an analog distortion - filtering and "even/odd harmonics" behaviour. I specially enlcosed that term into brackets to demonstrate that every developer have their own way of dealing with those.

I do it by dynamic offset, e.g. biasing the working point of my "circuit" and also there is another module that can do something else.

Filtering must not be plain flatty, otherwise distortion could only sound tight and lose correct low end behaviour - important for palm mutes. Or it might sound too loose, e.g sounding somewhat correct only for a specific guitar.

If you do it my way then signal dependencies must be discovered and a target "model" configured in a way to make it sound as close as spectrogram and osciloscope could confirm this inside of some margin of error.

The real physical modeling, e.g. white box approach has all this stuff calculated according to a "model" used. Different developers also tend to optimize their stuff in various ways. Someone creates so called look-up tabels (waveshapers again), someone like Mercuriall trains a neural network solver - not really sure what they use now but they claimed to do this in the very beginning of their work, in the early free amp sims like JCM800, JCM800 HotMod, Harlequin and so on.

Now, let's take a look on Amped 800 HG.mp3 - there is no clear sign of harmonics abruptly gone! But there is a lot of mud in the high end - the harmonics of the amp sim. The question I'd ask if the real amp tend to create this mud too? Soldano SLO 100 does, Peavey 5150 does it too. Marshalls as far as I seen this and heard - not really but as soon as a power amp cranked up this mud appears, here is the clip of Johan Segeborn playing JCM800: https://youtu.be/X-C8kDKoisI?t=169

So my opinion would be here that the end result matters more than the tools used, perhaps indeed Mikko will enhance the amp sim in the next update if he is sure it's required of course.
Attached Thumbnails
ML Sound Lab ML-800 amp sim...-kemperexplanation.jpg   ML Sound Lab ML-800 amp sim...-kemperhg800.jpg   ML Sound Lab ML-800 amp sim...-amped_800hg.jpg   ML Sound Lab ML-800 amp sim...-amped_800hg_explanation.jpg  
Attached Files

real 800 HG (1).mp3 (1.24 MB, 588 views)

amped 800 HG.mp3 (1.24 MB, 582 views)