Cytomic "The Drop" Resonant Filter

[vitocorleone123]

Will you add the VCV filters to this in the future, as well as the clean SVF one like in Ableton? Those VCV filters are state of the art.

Also, what is the use case for the "Clean SVF"? Do you want a filter with lower CPU?

What I meant was a clean low cpu digital mode like Ableton's. Sometimes the non linearity is not needed, and yes I can just pull up a clean eq plugin but adding it here is just more convenient and I might want a clean highpass but an analog lowpass for example.

The Drop is all about detailed analog models with characterful non-linearities. Having a filter without non-linearities would mean that the filter is the only one in the entire set of 10 that would behave completely differently when comparing between circuits. A linear filter cannot self oscillate without infinite growth to the self oscillating resonance amplitude and no matter the input signal level there is no change in sound apart from the equivalent increase in output signal level. The clean filters would not only need a very clear way to differentiate it from the others, but also features like the Drive knob, > 100% resonance, and resonance trim knobs all would need to be disabled as they wouldn't do anything.

The easiest workaround right now is to use a low cpu filter to begin with, say the KSM, and switch off HD mode for it, and then adjust the input pad to -12 dB (which boost the output by +12 dB), and then adjust the pre to -12 dB and post to 0 to +12 dB. Switch off the Boost button, and use the Drive knob on the filter you want to have drive. This leads to the loudest harmonic of around -80 dBFS, so a sine wave at 0dBFS has a very subtle -80 dBFS amount of drive added. If you are using drive on the other filter you will not hear this drive on the "clean" filter.

So a completely linear filter is not really an option for The Drop. One option could be to have both an SVF and an SKF with only a single resonance non-linearity and an input op-amp soft-clip non-linearity (other filters in The Drop have a minimum of 6 non-linearities). Then at least then the drive knob, and the > 100% resonance knob, and resonance trim knobs would do something, but only if you want them to. If you kept the resonance at 0%, and didn't send in loud signals then I could construct the filters to have a completely clean sound with no drive added. As soon as you increase the resonance then you would again need to adjust the gain staging as listed above. Does that sort of filter sound useful?

I see, I forgot about the fact that the GUI would also need to adapt to the clean mode. That's too much work for a one man request, I'll stick with the low gain staging MD mode method. Unrelated but what have you been up to? It's been a minute.

I'm sure there's lots more, but Andy was finishing The Scream.

[andy-cytomic]

... Unrelated but what have you been up to? It's been a minute.

I've been working on some more R&D type tasks, switching over my circuit solver to be able to take the netlist output of LTSpice as its input. This means I can simulate a circuit using LTSpice, then with a few additional Cytomic specific directives that I add like other spice directives inside the LTSpice schematic, I can also generate c++ code to simulate the same circuit. This has taken pretty much all of the last month, and the work is only just completed. This means I can input much larger circuits more easily without error, since I can quickly simulate them in LTSpice to make sure things are working before then running my solver on the same netlist and checking my results back to LTSpice.

I've also had a breakthrough in sorting out a much more stable algorithm to guide the convergence steps when solving pn-junctions. My old method was a bit brittle and required lots of fine tuning to get it to work, and it couldn't handle the full input voltage range. Now I've got a method which is looking very stable and hopefully very general that will hopefully work without fine tuning for different circuits. Every diode and every transistor has pn-junctions, so they are really important to solve well. This is also critical for a good discrete Blackmer gain cell for the VCA of The Glue v2:
https://en.wikipedia.org/wiki/Blackmer_gain_cell

[jtsterays]

... Unrelated but what have you been up to? It's been a minute.

I've been working on some more R&D type tasks, switching over my circuit solver to be able to take the netlist output of LTSpice as its input. This means I can simulate a circuit using LTSpice, then with a few additional Cytomic specific directives that I add like other spice directives inside the LTSpice schematic, I can also generate c++ code to simulate the same circuit. This has taken pretty much all of the last month, and the work is only just completed. This means I can input much larger circuits more easily without error, since I can quickly simulate them in LTSpice to make sure things are working before then running my solver on the same netlist and checking my results back to LTSpice.

I've also had a breakthrough in sorting out a much more stable algorithm to guide the convergence steps when solving pn-junctions. My old method was a bit brittle and required lots of fine tuning to get it to work, and it couldn't handle the full input voltage range. Now I've got a method which is looking very stable and hopefully very general that will hopefully work without fine tuning for different circuits. Every diode and every transistor has pn-junctions, so they are really important to solve well. This is also critical for a good discrete Blackmer gain cell for the VCA of The Glue v2:
https://en.wikipedia.org/wiki/Blackmer_gain_cell

Might as well do a dbx too while you're at it.

[andy-cytomic]

I'm doing an update to The Drop to include the CF100 SSM2040 based filter. I have a feature request to add a keyboard modifier to allow for switching both HP + LP to the same circuit upon clicking the arrow or menu, which should be easy enough to add. Are there any other suggestions people have?

[Funkybot's Evil Twin]

On a similar note: maybe another modifier to randomize the circuit for both? Or even a menu item/icon. Sometimes I just want to hear how different filters sound with the same settings, and a randomized selection option might spark some interesting combinations I wouldn't have tried otherwise.

Unrelated to features: any tips for getting a good envelope-triggered wah for guitar? I always seem to struggle with dialing in the sensitivity and filter cutoffs on that.

[solidtrax]

Yeah, something like parameter locking (whilst scrolling through patches) could be really a nice addition.

[ericj23]

I forget, can the LFO speed track midi input and have glide between notes?

[Erik_Lucas]

I'm doing an update to The Drop to include the CF100 SSM2040 based filter. I have a feature request to add a keyboard modifier to allow for switching both HP + LP to the same circuit upon clicking the arrow or menu, which should be easy enough to add. Are there any other suggestions people have?

Woah. Awesome. Love the VCV Rack module.

[andy-cytomic]

I forget, can the LFO speed track midi input and have glide between notes?

There is no pitch tracking of LFO 1 or LFO 2 from midi note. I don't know exactly what you mean by glide on an LFO, but you can definitely have legato triggering of the envelopes and LFOs, ie if you play an overlapped note there will not be a trigger, only when you let go of all notes, then play a new one will a trigger occur.

[jtsterays]

When tweaking knobs, it seems that The Drop reduces the GUI fps of FL Studio by half (~72 to ~35). VST2, latest FL, W10, 5800x. Glue and Scream are normal.

[andy-cytomic]

When tweaking knobs, it seems that The Drop reduces the GUI fps of FL Studio by half (~72 to ~35). VST2, latest FL, W10, 5800x. Glue and Scream are normal.

I'm using Juce v8 at the moment, and they are in a transition towards hardware accelerated GUI drawing, but it's all a bit broken at the moment sorry. I'll add an option to switch between D2D and the software GUI drawing in the next build, and this will hopefully help, but it may also cause a blank interface.

(edit: also due to the small LED meters showing drive being spread around the user interface the entire interface gets redrawn every time they update due to how draw regions are aggregated by the OS and Juce, not much I can do about it sorry, but hopefully D2D will be quicker)

[andy-cytomic]

I've finished adding the CF100 filter to The Drop, which I've called the PT1 (Prophet 5 revision 1). I've also designed a dedicated 4 pole high pass filter, which preserves the same sort of asymmetric clipping that the low pass filter has, but unlike the CF100 which generates the HP from the LP, this is a dedicated HP filter, so when pushed into clipping it still high passes the signal, unlike the CF100 version which breaks down. This matches how all the other cascade / ladder filters are done in The Drop.

I'll hopefully get a release done in the next few days after I've added a few other things to The Drop, but until then you can have a listen to the new PT1 HP and LP filter circuit models in the new demo on the web page.

New circuit model PT1 demo to listen to here:
https://cytomic.com/product/drop

[jtsterays]

When tweaking knobs, it seems that The Drop reduces the GUI fps of FL Studio by half (~72 to ~35). VST2, latest FL, W10, 5800x. Glue and Scream are normal.

I'm using Juce v8 at the moment, and they are in a transition towards hardware accelerated GUI drawing, but it's all a bit broken at the moment sorry. I'll add an option to switch between D2D and the software GUI drawing in the next build, and this will hopefully help, but it may also cause a blank interface.

(edit: also due to the small LED meters showing drive being spread around the user interface the entire interface gets redrawn every time they update due to how draw regions are aggregated by the OS and Juce, not much I can do about it sorry, but hopefully D2D will be quicker)

I see, if that's the case then I don't think it's necessary to add it. It's not really a big problem, we can just wait until the transition is finished. Also one feature request, I just tried Ableton Auto Filter recently and I noticed they have a feature to detune the LFO speed to offset the L & R (Spin) instead of phase offset which is quite unique, I don't know other plugins that do that, could you consider implementing it to Drop?

[EsmileZ]

N*I*c*e I love the ssm filter. Can't wait to try it out

[andy-cytomic]

I see, if that's the case then I don't think it's necessary to add it. It's not really a big problem, we can just wait until the transition is finished. Also one feature request, I just tried Ableton Auto Filter recently and I noticed they have a feature to detune the LFO speed to offset the L & R (Spin) instead of phase offset which is quite unique, I don't know other plugins that do that, could you consider implementing it to Drop?

I've already added the setting, it wasn't much work, and seeing the number of bug fixes to the D2D code in Juce 8 for almost every point release I think it will be required to workaround continuing issues.

I won't implement a spin control for the LFOs in The Drop. I've implemented a chasing type system for various aspects which won't work unless the left and right are within two cycles of each other (2 cycles because of the swing). For example the random numbers generated the "leading" LFO side will have a new random number each step, and the trailing side will then repeat that random number at the phase delay. With spin this would be broken as the cycles would be out of sync with each other, and it would also break various other aspects of what makes the LFOs in The Drop unique and musical. Feel free to modulate the cutoff of the left and right signals yourself at audio rate with the external sidechain input and FM.