Wow, cool demo!! That really is impressive.
1 Like
I actually just found a little bit I had wrong with the Maca matching I did. The JH version has a q-bias voltage of around 0.7 V to increase the headroom of the resonance before asymmetric clipping, and the Maca / Datasheet doesn’t. I thought it was the OTA buffer being different, but it wasn’t. Point is it’s close enough anyway that this is what is going on in the circuit that it really doesn’t matter - and I’ll keep to the JH version anyway since I prefer the sound.
2 Likes
I’ve submitted v2.0.4 to the VCV Library, so keep an eye out. I’ve updated the product page with the new info: VCV Rack Modules – Cytomic
CF100 v2.0.4
- DC Block toggle to allow for audio or CV processing. Please note the cv processing isn’t very accurate, but it is fun - I did double check against the Maca filter and it’s spot on for the results. For example if you increase Drive you offset the low pass output DC output level a lot as well.
- Added adjustable noise depth from Off to 200%, with 100% being accurate to the analog circuit, and 200% being around x1000 boost over 100% to it can be used as a polyphonic noise source.
- Added a yellow LED to the dots of the knobs so if the room light is lower it is still easy to see the positions of the knobs
- Fixed an issue where right clicking and selecting Initialize didn’t work
2 Likes
I like the design of the panel
1 Like
Thanks! I gave Pyer some design instructions and he came up with the design, which I’m really happy with. Here is his web page:
2 Likes
I’ve submitted v2.0.5 of the CF100. I decided to add the SSI2140 Datasheet design, which is what is used in the LA67 Maca Filter and what I have been testing against. I changed the DC Block right click menu to a Circuit menu, with three choices:
Circuit:
- AC JH (original circuit)
- DC JH (original circuit with DC blocking removed)
- DC Datasheet (from the SSI2140 datasheet)
Both DC versions can be used for CV processing as well as Audio. Also the DC versions have extended cutoff frequency range down to around 0.01Hz
10 Likes
CF100 2.0.5 has been accepted and is ready for download.
5 Likes
Here is the re-done noise of the CF100 vs the Softube Prophet models. The Softube is clearly just adding noise to the input and there isn’t any constant hiss like an analog circuit with the cutoff frequency down low:
https://cytomic.com/files/forums/cf100-vs-softube-prophet-noise.mp3
2 Likes
The LA67 Maca Eurorack filter arrived, which uses the SSI2140, which is the filter circuit the CF100 is based on. I’ve now been able to make a comparison video like I usually do, where the CF100 is switched to the DC coupled datasheet circuit model:
11 Likes
its just incredibly awesome ! Thanks for adding this mode also !
No problem! If the modelling is done properly and actually models the analog circuit accurately then it’s super easy to make some modifications to match other variations of the design.
As a side note, if a developer ever says something like “my model doesn’t match the analog circuit because I didn’t like the sound / behaviour of the original” just ask them to include it anyway as a mode, since if they have accurately modelled the circuit this will be easy for them. Oh, and get them to also post an A/B comparison video like I’ve done with a scope and FFT display. Most likely they won’t be able to because when a dev says that sort of thing, what I hear them saying loud and clear is: “The circuit was too hard for me to model properly in detail, so I did what I could manage, which is hopefully still useful even if it doesn’t properly match the analog circuit in all modes of operation.”
5 Likes
How many people on planet Earth are doing the level of modelling you’re discussing? You, Urs Heckman, and how many others? A dozen?
2 Likes
Apparently thousands, if you go on the level of detail devs / their web pages say they’ve achieved.
1 Like
Vult, obviously. It’s pretty esoteric (I can’t do it!), but ima say a lot more than dozens?
I’ve not seen the Vult stuff match any analog circuit accurately. If you could please point me to any video Leonardo has done that matches like I’ve shown for the A-124 or the CF100 I’d love to see it. I’ve watched all his youtube vids comparing his models to circuits and none of them match, there are always major differences in sound, and always a bunch of explanations as to why that is.
4 Likes
ok, u the master!
Let"s hope Urs gets into VCV as well!
1 Like
I don’t know about master, there are plenty of smarter people around than me, but I try and be as honest as possible with the level of detail and accuracy of my models. For example I’ve always pointed out my models haven’t included analog modelled noise, and I’m glad I’ve finally got some basic modelling of this in the CF100!
5 Likes
Sorry for the off topic. Taken from the u-he FB page at superbooth this spring. I see at least a Repro1 OSC. So there’s hope.
And then Dieter has just to give Andy an A-110 module… 
…oh and I see they’ve just released the Filterscape 1.5 beta. 
4 Likes
I have had some email conversations with @andy-cytomic regarding matching the results of my filters against the analog counterparts. I was going to write some of my process here, but I’m in a hurry and I’ll just paste my latest answer to Andrew.
For most of the released filters, which I modelled almost 5-6 years ago, I have very little interest on creating comparisons against hardware filters. As I mentioned, I do not have the original circuits, I have tweaked the models (the circuits and the code) to my linking, and I have performed simplifications to make them run efficiently. The biggest reason for not doing it, is that I don’t have time to replicate them since I’m focusing on other tasks nowadays.
For me the most important aspect is to make the filter sound good and not accurate to the analog counterpart. Because sometimes the analog circuit may have some limitations and undesired behaviours. To give you an example, my 303 filter model has level compensation when the resonance is increased and I reduce the gain when the resonance is zero so the filter does not saturate as much as the original. In my opinion, my changes make it better and more pleasant to my ears, but it will not match the original filter.
The second most important aspect is the performance. Since most of my filters run in the Freak hardware module, they need to be efficient. The module has limited resources and some of the most accurate models I have simply do not run on it. As I mentioned before, I have some ideas to improve my Polivoks model and make it more accurate and efficient so I can add a better model to the Freak module.
I understand the fun on making models that perfectly match reality. That’s what I do for living (in addition to creating the simulation tools). But I also like the freedom of customising the models and make my own versions.
In the future I have plans to make some accurate models of some analog filters I’m designing for an upcoming hardware (analog) module. In that case, I will take the time to make the virtual version match my analog design because the analog version will be already designed to my liking.
10 Likes