Cytomic 'The Glue' Compressor

[Aloysius]

You sound like a mad scientist (in a good way). Next step will be world domination.

[andy-cytomic]

I accidentally uploaded one of the files with a +2dB gain, I noticed that clipped so then decided to just upload the files without any gain, but something went wrong. I've re-uploaded the original files with a -b.wav / -b.mp3 extension so as to make it clear of the change, and editing my original post with the new links with the correctly matched volumes.

[andy-cytomic]

A quick news update: I've soldered together the DIYRE G-Bus comp after measuring each component along the way and it's sounding great. I've entered all the component values to my model and double checking all the voltages are matching at the moment and will post some audio examples once that's done:

[Puduku]

As time passes by I find myself using The Glue more and more often. It has finally become my go-to compressor without me even realizing it.

Always good results, no weird behaviors, no strange parameters that you can overdo, pristine sound... A perfectly well-thought plugin.

[andy-cytomic]

I've got a great model of my diyre g-bus compressor where I've matched the components really closely. There are some subtle differences when pushed with really hot signals (+-4V and above) with ratio 10 and the fastest attack and release, but all up it's very good. I've matched the DBX202 circuit to be as close to the That 2181 as possible, but there are some very minor differences there too. Here are some audio examples:

(edit: I've updated the .mp3 links to .wav for the .wav versions, edit2: added dry links)
fjolle doof slammed as hard as possible, fast attack, fast release, ratio 10:
fjolledoof-dry.mp3
fjolledoof-dry.wav
fjolledoof-att0.1-rel0.1-rat10-diyregbus.mp3
fjolledoof-att0.1-rel0.1-rat10-diyregbus.wav
fjolledoof-att0.1-rel0.1-rat10-glue2.mp3
fjolledoof-att0.1-rel0.1-rat10-glue2.wav

loud kick loop slammed as hard as possible, fast attack, fast release, ratio 10:
loudkick-dry.mp3
loudkick-dry.wav
loudkick-att0.1-rel0.1-rat10-diyregbus.mp3
loudkick-att0.1-rel0.1-rat10-diyregbus.wav
loudkick-att0.1-rel0.1-rat10-glue2.mp3
loudkick-att0.1-rel0.1-rat10-glue2.wav

loud kick loop hard compression, slow attack, fast release, ratio 10:
loudkick-dry.mp3
loudkick-dry.wav
loudkick-att30-rel0.1-rat10-diyregbus.mp3
loudkick-att30-rel0.1-rat10-diyregbus.wav
loudkick-att30-rel0.1-rat10-glue2.mp3
loudkick-att30-rel0.1-rat10-glue2.wav

If anyone has dry loops they would like compressed as examples please email some download links to andy@cytomic.com , including which sort of compression settings you'd like used and I'll process them through both the diyre bgus and the glue 2.

[electro]

A quick news update: I've soldered together the DIYRE G-Bus comp after measuring each component along the way and it's sounding great. I've entered all the component values to my model and double checking all the voltages are matching at the moment and will post some audio examples once that's done:

Behringer cloned Mutron III for under $100. Consider modeling that next.
https://www.behringer.com/product.html? ... e=0709-ALB

[dayjob]

Behringer cloned Mutron III for under $100. Consider modeling that next.
https://www.behringer.com/product.html? ... e=0709-ALB

Yawn.

[joshmakesmusic]

Great compressor, one of my favs!

[andy-cytomic]

I've been working away on The Glue v2, implementing the pnp and pjfet input op-amp macro models, getting a that2181 model working etc. There was one thing that kept on escaping me until yesterday. There was an extra little wobble with ultra fast attacks and releases with high ratio that I didn't know the what was causing, and I finally found it after just over a month of work. I had an error in my op-amp modelling of the gbw (gain bandwidth product) parameter, and it was putting the stability capacitor low pass cutoff way too high in frequency. The extra low passing is in the feedback loop of the sidechain / envelope follower section and with that phase shift with feedback gain creates a small resonance like a resonant low pass filter. Here are the plots to show the impact. The left is what I had with the cutoff too high, the middle is matched properly, the right is slightly exaggerating the ringing:



The blue plot is the diy-re g-bus and the yellow is glue 2, pushing things as hard as possible (for the diy-re) with 0.1mS attack, 0.1S release, and ratio 10. Now I've tracked that down I can start simplifying and optimising the model getting it ready for production use!

[DupliClap]

Impressive results, excited for the beta.
What's the current CPU usage and your target?

[trmupstage]

Great work Andy!

[vitocorleone123]

Impressive results, excited for the beta.
What's the current CPU usage and your target?

Not sure if it was here or another forum, but, if I remember right, a bit more than what The Scream is using. Glue v2 being built with the same tech, but a lot more parts being modeled than The Scream.

[andy-cytomic]

Impressive results, excited for the beta.
What's the current CPU usage and your target?

There will be multiple modes, MD (Medium Detail) and HD (High Detail), and then within those you'll be able to select the VCA type from Ideal, That2181, and DBX202. CPU usage of a single core on an M1 chip I'm aiming around 2% for the most efficient, up to around 10% for the most detailed, then oversampling is optional on top of that.

[Yorrrrrr]

Good stuff! Excited for The Glue 2.

[jahpooh]

How's the progress coming Andy, beta anytime soon ?