Cytomic 'The Glue' Compressor

[andy-cytomic]

Unlinked R-L is the only thing I have yet to notice needed for mastering.

This is definitely on the way, I currently plan to have 5 positions of cross feed from 100% stereo link (like it currently is) through to 0% which is dual mono.

[andy-cytomic]

That’s a great idea! One thing that turned me off from all-in-one solutions like Compassion was that it was easy to destroy a great compressor setting because there were too many unintuitive options. Even DC8C can have this issue. But having a few non-standard, yet well-thought-out tweak-able components under the hood is brilliant.

With analog models I've founding adjusting the component values leads to good sounding results pretty much all the time. I'm not so sure that adjusting component values is that intuitive unless you know a lot of theory, but at least most of the time you'll end up with decent sounding results, even if it's not what you expected

[meloco_go]

When this thread was started Bush Jr. was still president

Classic!

Today marks 10 years since the first beta of The Glue was released! A huge thanks to everyone that has supported Cytomic. I'm looking forward to getting the HD version of The Glue out this year and to the next 10 years of plugin making.

Wow!
Congrats
The Glue was actually one of the first if not The first plugin I purchased when I decided I gotta be more serious about doing audio. And I still use it after all those years.

[Funkybot's Evil Twin]

When this thread was started Bush Jr. was still president

Classic!

Today marks 10 years since the first beta of The Glue was released! A huge thanks to everyone that has supported Cytomic. I'm looking forward to getting the HD version of The Glue out this year and to the next 10 years of plugin making.

Wow!
Congrats
The Glue was actually one of the first if not The first plugin I purchased when I decided I gotta be more serious about doing audio. And I still use it after all those years.

It really is amazing how well it's held up. I still use it regularly. Even with a few newer SSL Bus Comp plugins in my folder, The Glue remains in heavy use because it sounds awesome and offers a flavor I can't get from the others. I don't think there are any other 10 year old effects in my folder still getting use outside of maybe the first UAD mk. 2 plugins which probably came out around that time.

Thanks Andy! Look forward to you wrapping up The Scream soon and hearing the HD version of The Glue.

[andy-cytomic]

As many probably have already heard, there is a company called Distopik (don't know why the name, they seem quite utopian to me) which apparently develops interfaces which provide digital control over analog equipment. And this company runs a service called Mixanalog where one can upload tracks to their server and run through a HW gear while controlling and streaming results online; when a user is happy with the results, he/she may bounce the tracks to high-quality wavs.
One of the pieces of gear currently online is HW compressor based on SSL quad with Gold-Can VCAs.
Of course, I jumped at the possibility and spent quite a lot of time testing it, bouncing some of my tracks, and comparing it to the other plugins including The Glue.
The behavior of this HW comp and The Glue is very, very similar. There is a slight difference, however, and if I had to describe what it is I'd say the HW is more "emotional." I am a techy guy myself, and I have such vague terms as much as anyone, but I can't better describe it any other way. While trying to analyze and dissect this quality, I feel like it has something to do with how the compressors act on transients. I got somewhat similar results (not sonically but in this direction) by delaying the sidechain into the Glue. But this effect becomes apparent very soon. Trying to use some coloration before or after the plugin does not provide this quality, so I don't think this is due to the output amp.
Anyway, The Glue is very good and holds against the hardware very favorably. There is a slight edge to the HW still, though.
I am looking forward to The Glue 2.0, and if it catches up with the HW, I'd be a very happy man. And it would be very interesting to know what exactly provides this sensation which the HW has.

Hi meloco_go! Thanks for posting this, I really appreciate it. It's great to hear that The Glue is generally holding up to a vintage hardware SSL with DBX VCA. There will be a lot more character in the hardware unit since there are multiple levels of non-linearities, so I expect The Glue not to sound as exciting since it will be doing less to the stereo field, and adding less distortion. The HD mode will definitely make up a lot of ground in this regard, and hopefully match the results but without the hiss.

[MogwaiBoy]

There will be a lot more character in the hardware unit since there are multiple levels of non-linearities, so I expect The Glue not to sound as exciting since it will be doing less to the stereo field, and adding less distortion. The HD mode will definitely make up a lot of ground in this regard, and hopefully match the results but without the hiss.

So exciting. Keep up the amazing work, Andy. You must be proud of what you've achieved and your continued quest for even more excellence says a lot about your commitment and passion in what you do. Happy 10 year anniversary for your wonderful creation!

[meloco_go]

There will be a lot more character in the hardware unit since there are multiple levels of non-linearities, so I expect The Glue not to sound as exciting since it will be doing less to the stereo field, and adding less distortion.

It would be very exciting to play around with the parameters and find what exactly contributes to the "euphonic" character of the hardware.

The HD mode will definitely make up a lot of ground in this regard, and hopefully match the results but without the hiss.

Indeed, and AD/DA step would also be avoided.
Also, I am eagerly awaiting the possibility to tweak the model and get more mileage out of it as far as times and thresholds!

[mrj1nx]

Tying in to this discussion, Ive been intrigued to whats still left in hardware that we cant replicate in plugins. Is it just a feeling or an actual measurable thing? And if so, why cant we replicate it? Is it a diminishing returns equation? Tons and tons of cost for a totally negligable difference?

(Asking this as a purely in the box producer that are curiously looking at hardware, also amazed by the huge difference in cost of hardware)

[4damind]

I use the Glue a lot for sidechaining which is a bit more complicated in Cubase (needs a quadro group and some routing). So my question: I know VST3 is planned to be implemented in all Cytomic plugins.. but will we see this in the near future (with Glue2)?

[BDeep]

Congratulations on 10 years!

Also... I’d pay good money to be able to buy an upgrade to Live’s included Drop filters and the Glue. The SH model in Wavetable... yummy

[andy-cytomic]

Tying in to this discussion, Ive been intrigued to whats still left in hardware that we cant replicate in plugins. Is it just a feeling or an actual measurable thing? And if so, why cant we replicate it? Is it a diminishing returns equation? Tons and tons of cost for a totally negligable difference?

(Asking this as a purely in the box producer that are curiously looking at hardware, also amazed by the huge difference in cost of hardware)

Most analog modelling plugins, despite the hype, don't contain that much detail, so there is loads of room to improve things and bring them closer to the sound of hardware. The problem is most people are not interested in using more than around 10%-20% of a single core for a single plugin, even considering that a powerful multiple core cpu costs a fraction of hardware devices they want modelled. So it's a balancing act between how much detail is reasonable to include vs how much cpu that will take.

My plugin The Scream models a very basic overdrive pedal circuit which contains only two op-amps (two transistors at the input stage and two voltage limiting diodes at the output), two transistor buffers (one at the input to the pedal, one at the output), two clipping diodes (in the feedback of the first op-amp), and a bunch of resistors and capacitors. The Scream takes around 6%-20% cpu at 44.1khz for the full detailed model, so is already pushing what most people will put up with in terms of cpu load for a single plugin. Clearly for more complicated circuits something has to give, and it is usually the detail and accuracy of the model to save cpu.

I've found the best way is to offer both a low detail low cpu mode and an option to switch on a higher detailed mode. I call these MD (medium definition) and HD (high definition), although there is just a single button on the interface called "HD". I've got this in The Drop, and I'm adding this to The Scream and The Glue. The MD mode for The Glue will be the current version of The Glue, and the HD mode will model a lot more detail, and take a lot more cpu.

[andy-cytomic]

I use the Glue a lot for sidechaining which is a bit more complicated in Cubase (needs a quadro group and some routing). So my question: I know VST3 is planned to be implemented in all Cytomic plugins.. but will we see this in the near future (with Glue2)?

I sympathise with the issue you have using Cubase, but also would like to point out that Steinberg could easily offer you the functionality you want with VST2.4 plugins just like every other major VST supporting DAW on the planet can eg: Ableton, Reaper, Tracktion, Bitwig, Studio One, Reason, FL-Studio etc.

I plan to update to the latest version of the Juce framework some time this year, which now supports VST3, but since that is a large amount of work I have postponed it to at least after I have released The Scream.

[mrj1nx]

Tying in to this discussion, Ive been intrigued to whats still left in hardware that we cant replicate in plugins. Is it just a feeling or an actual measurable thing? And if so, why cant we replicate it? Is it a diminishing returns equation? Tons and tons of cost for a totally negligable difference?

(Asking this as a purely in the box producer that are curiously looking at hardware, also amazed by the huge difference in cost of hardware)

Most analog modelling plugins, despite the hype, don't contain that much detail, so there is loads of room to improve things and bring them closer to the sound of hardware. The problem is most people are not interested in using more than around 10%-20% of a single core for a single plugin, even considering that a powerful multiple core cpu costs a fraction of hardware devices they want modelled. So it's a balancing act between how much detail is reasonable to include vs how much cpu that will take.

My plugin The Scream models a very basic overdrive pedal circuit which contains only two op-amps (two transistors at the input stage and two voltage limiting diodes at the output), two transistor buffers (one at the input to the pedal, one at the output), two clipping diodes (in the feedback of the first op-amp), and a bunch of resistors and capacitors. The Scream takes around 6%-20% cpu at 44.1khz for the full detailed model, so is already pushing what most people will put up with in terms of cpu load for a single plugin. Clearly for more complicated circuits something has to give, and it is usually the detail and accuracy of the model to save cpu.

I've found the best way is to offer both a low detail low cpu mode and an option to switch on a higher detailed mode. I call these MD (medium definition) and HD (high definition), although there is just a single button on the interface called "HD". I've got this in The Drop, and I'm adding this to The Scream and The Glue. The MD mode for The Glue will be the current version of The Glue, and the HD mode will model a lot more detail, and take a lot more cpu.

Fascinating. How much of that detail is the moving of the needle vs saturation vs x?

[andy-cytomic]

Fascinating. How much of that detail is the moving of the needle vs saturation vs x?

I'm just talking about the audible circuit model, not the user interface and metering. I'm not really sure what you're asking here, the detail is mostly in how deep the modelling goes, how many components are included, and how many of their ideal and non-deal behaviour is included - basically the detail!

Ok, perhaps a basic example. Let's consider one diode. The usual model model for a diode looks like this:
https://en.wikipedia.org/wiki/Diode#Sho ... e_equation
id = is * (exp (vd / (n * vt) - 1)

Good plugins model this stage properly, but there is more doing on in a typical diode, and the next level up of complexity is this:
http://qucs.sourceforge.net/tech/node68.html
id = is * (exp (vd / (n * vt) - 1) + isr * (exp (vd / (nr * vt) - 1) + icd

where icd is the non-linear capacitance comprised of the diffusion and junction capacitances:
cd = cp + tor*gd + cj0*(1 - vd/vj)^-M

where gd is the derivative of id with respect to vd and I've left out the temperature dependance terms.

This second form is what the diodes in The Scream are, normal and recombination currents, and non-linear capacitance as well as series resistance, and maximum forward current. I'm not sure any other plugin bothers to model diodes in that detail, and as far as I'm aware The Scream is the only plugin that allows you to change all the details of the diode model or even import your own SPICE diode models into the plugin and use them, for example:

.model 1N4148 D(type=silicon is=0.7p n=1 isr=4.8n nr=2 cp=0 tt=3.5n cj0=0.7p vj=0.869 fc=0.5 m=0.33 rs=0.6)

[mrj1nx]

Right, what I really meant with the needle is the volume changing component of how the plugin acts on the sound. But the way its modeled in your case on a component by component basis, that means the volume changing aspect is just in the sum of all the parts put together. I was imagining it being applied like layers: changing the volume, adding hum, analog saturation etc. Im assuming many plugins do it that way.