And back in the 80’s we tried really hard to minimize those thumps!
1 Like
Would you spend $15 on a pristine OTA model with zero thump, zero control signal feed through, and zero 2nd order harmonics? I could certainly deliver that sort of product.
2 Likes
I personally would prefer to be able to turn off the analog badness and only have the analog goodness. But the only stuff I’ve bought in VCV have been a lot of VCV stuff, because I think Andrew deserves a little payback for all the work.
1 Like
Ok, I’ve added support for x2 and x4 oversampling. I up sample not just the audio input but also the drive, freq, and res signals that go to the filter, and down sample for all of the outputs. It is included with CF100 v2.0.3, which I’ve submitted to the Library.
Here is how things look now for a 2khz input signal with the CF100 at x4 oversampling of 44.1khz, and the E440 at whatever internal oversampling it uses by default which you can’t change:
5 Likes
May you provide a listening example of both modules?
nice! So the aliasing went from -48 db to below -96db? very impressive!
2 Likes
So E440 bass boost seems buggy since as you said at zero reso it should not impact right?
Yes, that’s according to the JH circuit. The bass boost is just adding some of the input to the resonance feedback signal (from the LP4 output stage), and then apply the resonance gain to the sum of those two signals, so at zero resonance you should get zero chain in what you hear.
1 Like
Check out the videos I posted earlier in the thread, they have audio too!
1 Like
Here are some fantastic sounds from Omri Cohen putting the CF100 through its paces:
11 Likes
This is interesting analysis. The Surge xt filters dont seem to show any harmonics in default state except vintage ladder type 1. The others you need to add saturation. Alright Devices also no harmonics.
1 Like
Try A 1khz sin with 2khz cutoff no res and increasing input signal level to check for filter core non-lineraities
Nice! That’s a hell of a demo. Probably sounds hokey but “it really does sound like an analog filter”. Good stuff!
3 Likes
I appreciate the comment! I made sure to cover all the major non-linearities and behaviour of the circuit so it does sound just like the circuit - but with so much marketing BS and blatant lies in the industry saying such stuff is almost pointless since all the boys have cried wolf so much no one believes anything.
This is one of the most detailed analog modelled filters ever made to date, I listen to a lot of them and I know when others are cutting corners or just plain getting it wrong.
Ultimately it really needs to be about creativity. When you turn the knobs does the sound inspire you? That’s what top notch kit, analog or digital, is all about, every setting being the sweet spot, and it’s quick and fun to dial in the kinda sweet spot you’re after.
16 Likes
Couldn’t agree more!
I just went through all the sound demos on VCV Rack Modules – Cytomic. All delicious and top-notch quality sounds. Really PRO stuff. I can’t get over that “woody” character of the NPN2 model. That alone is worth the price of admission IMHO. Bought!
1 Like
If you don’t mind me asking, how many hours do you reckon you have spent on this filter Andy? (I realise that this does not included the years spent gaining the expertise to take the task on).
Would be happy to hear more too on what the broad workflow is to analog model something like this, and to this level of detail.
1 Like
The core DSP took around 5 days full time work all up, but spread over a week or two. I have solved similar circuits before, but not this exact arrangement.
The only reason I could pull off something with so much accuracy and detail so quickly is I’ve spent the last 10 years doing R&D automating this process, which is part of the reason I haven’t released many products. I’ve written a fully symbolic matrix solver and code generator with symbolic optimisations (symbolic just being algebra type stuff where you deal with the name of something, not its value, eg x = 2, the symbolic thing is x the value is 2). I specify a list of all the devices in the circuit, which includes how they are connected to each other, and my program goes to work, first solving the matrix equations symbolically, then inserting all the specific device non-linearities and how to linearise them, and then uses a multi-pass symbolic optimiser to make it all run super fast by pulling out any common sub-expressions and pre-evaluating them, as well as cancelling out anything that can be simplified.
I’ve outline the process on how these systems can be solved, but automating the process takes some work. I gave a talk on it here (sorry for all the umms, I gave this talk at around 4am my time):
You can also find more detail in some technical papers I’ve written, which you can access here:
25 Likes
Many thanks for that. I watched the talk, whose maths went mostly over my head, but I learnt a little and found it fascinating. This stuff is hyper complex and I’m not surprised that there are so few people on the planet who can do DSP at this level. What drives you more here - (a) the research into modelling methods, (b) the production of tools to automate the modelling as far as possible, or (c) the production of the end models? Is it a case of ‘I’ve largely done what I want to do with (a) and (b) and now I can focus more on (c)’ or are you still driven more by (a) and/or (b)?
I’m not clear to what role the hardware plays here, as opposed to the circuit diagram? If I understand what you have said correctly, your tools attempt to solve the entire circuit in one go, rather than breaking it down into parts, solving those individually then amalgamating those solutions? My guess then is you are not poring over a physical circuit measuring voltages and probing signals at various points? Or are you?
Did you build the Haible circuit to compare the output with your model? I feel that the answer to this question should be obvious! (I do recall you physically modifying your A-124 to enable self-oscillation).
Sorry for these pretty ignorant questions but analog modelling is a dark art that is just about indistinguishable from magic for me!
1 Like
The VCV community is very fortunate to benefit from the results of all that hard-won research.
4 Likes