Cytomic 'The Glue' Compressor

[andy-cytomic]

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

I'm getting closer to finishing the beta with each day. I've got a great model of the That2181 chip now, which took a bit longer than expected. It is more accurate to the DIYRE than using the DBX202 model. I've got the mono version matched and I'm working on the stereo version now for the audio examples.

The main things still to do for the beta are:
- Optimising and making the DSP more robust to model parameter changes
- Schematic editor that supports swapping of VCAs.
- User interface layout, texturing, rendering, and slicing
- Basic testing in various DAWs

I'll post the audio examples in an article soon, and will continue to post updates as they come. As I finish each task it will be one thing less to do for the beta.

[andy-cytomic]

Well it was yesterday, but kvr was down when I went to post this!

https://cytomic.com/happy-17th-birthday-to-the-glue/

[Yorrrrrr]

Congrats! The Glue is awesome, can't wait for The Glue 2.

[Weasel-Boy]

Well it was yesterday, but kvr was down when I went to post this!

The 14th is also Albert Einstein's birthday, so The Glue is in good company.

[jahpooh]

Any updates Andy ?

[andy-cytomic]

Any updates Andy ?

Things are going very well!

Over the past month or so I've been working on the website, filling out the products info pages with more details, and generally making it more useful, and I'm very happy with the results, but I'm back to full time dev on The Glue v2.

There were a few gotchas I ran into with matching the model to the diyre bgus which took a while. I measured all the resistors before soldering the circuit together to save time on guessing, but this kinda backfired since I forgot to double check the resistance of the multi-meter probes by themselves. Lesson learned. Not to self: before any critical resistor measurements first measuring a direct connection of the probes to themselves, then measure across the resistor, then subtract the former from the latter.

I didn't want to change any of the values of components I'd measured to get a match, since they should be right. In a last ditch effort I did end up changing one of them, a single resistor down from 52 ohms down a little - which was on the final voltage divider to the VCA I tried around 50 ohms, and the match was better, then tried 47 ohms and the circuit suddenly matched. That's a 10% change, which I wasn't comfortable making, but the schematic did list the component should be 47 ohms. I realised I should double check the multi-meter, since it may account for 1 ohm or so, but it ended up being the whole 5 ohms I was out!

I'm now happy with the match of the analog model to the circuit, everything is super close to my diyre gbus. I can probe the circuit pretty much anywhere, and output that same voltage from my model and they match really well. There was some additional fine tuning of the voltage offsets generated by each individual op-amp in the ratio shaper, they are all slightly different (anywhere from 6-12 mV), and have a large impact on the voltages generated because of the large gains and non-linearities involved.

I'm now working on basic optimisations, splitting up the circuit into three main chunks, and connecting them together: Input Stage, Sidechain Stage, Main Stage. The Input is shared between the two with very clean buffering. The Sidechain Stage includes the sidechain filtering, VCA, ratio shaper, envelope follower, and signal conditioning of the main audio cv and feedback sidechain cv. Then the Main Stage VCA takes the main audio cv and does the compression, followed by some signal conditioning and the output stage.

[Aiynzahev]

I’m looking forward to this. What effect do you think these changes will make to the action of the compressor?

[meloco_go]

Andy, SSL promotes their BUS+ compressor and among the marketing thing is this statement:

"4K Mode on THE BUS+ authentically reconfigures the VCA from balanced to unbalanced operation, introducing a controllable amount of harmonic distortion that can be precisely dialled in to suit your source and creative intent."

How does GLUE2 operate? Is changing from balanced to unbalanced mode a possibility?

[jahpooh]

Can't remember the last time I was so eagerly anticipating a compressor plugin. Whats your eta on the beta?

[andy-cytomic]

I’m looking forward to this. What effect do you think these changes will make to the action of the compressor?

When comparing The Glue 1 to The Glue 2's full detail model the main difference is the subtle distortion always present, and the more asymmetric response in the left and right sides widening the stereo image a little. The compression action itself is pretty similar for small gain reductions amounts, but when pushed hard v2 sounds more natural and "bends" more smoothly, especially for slower attack times with fast release.

The Glue v2 will have the option for a linear main signal path as well, which will sound closer to The Glue v1 and use less cpu.

[andy-cytomic]

Andy, SSL promotes their BUS+ compressor and among the marketing thing is this statement:

"4K Mode on THE BUS+ authentically reconfigures the VCA from balanced to unbalanced operation, introducing a controllable amount of harmonic distortion that can be precisely dialled in to suit your source and creative intent."

How does GLUE2 operate? Is changing from balanced to unbalanced mode a possibility?

Differential operation is critical for analog hardware if you want to cancel noise and distortion. For software running a full differential path means modelling twice the components, which means twice the cpu load. In hardware there are physical limits to the components being used which can't be bypassed, so doing differential processing makes sense to lower distortion. In software I can have idealised aspects of components, or completely ideal components to avoid some of the distortion introduced in the first place, which makes sense to keep the cpu down.

For The Glue v2 I'm keeping the regular unbalanced signal path, and will offer a perfect exponential VCA as well as a full component level model of the DBX 202. For the DBX 202 there is a symmetry control to adjust the amount of distortion. At this stage I'm keen to finish the product and get it released instead of adding new features, but it would be interesting to see if a differential VCA is possible within cpu constraints.

[DiodeLadder]

Hello Andy,

Do you think you can make a simple VCA saturator plugin to simulate the channel VCA fader coloration, and bundle it with the Glue 2? SL4000 series consoles had CV faders controlling VCAs for automation (except the more expensive Ultimation system, which could switch between VCA and non-VCA mode), and this was a big part of the reason why they sounded the way they did. I thought I'd ask since you already have the VCA models done. If it isn't too much extra work, please consider it.

Also, I saw that the DIYRE kit comes with switchable output transformers, and wondering if that mod is part of Glue2 as well? I don't know what that sounds like, but just curious.