Why is SSL style compression so popular? [general question]

[soapfoot]

If someone truly wanted to make an innovative compressor in 2023, they probably wouldn't be messing with capacitors, resistors, and semiconductors.

They'd be working with lines of code!

 

[Matt Syson]

One good thing about VCA compressors (dBX block/THAT units )is that the VCA element(s) have a very closely defined control 'law' that extends to something like 140dB which is mostly differing between units by absolute temperature (once some basic trims have been applied) so tracking between channels in stereo or multiple channels (Quad/to 8.1 ) is pretty easy (if you follow the application sheet recommendations). Using pulse width 'chopper' designs can also achieve good stereo tracking but have quite a few other issues to get to grips with.

 

[JMan]

I would have recommended Heikki's Vari-Mu compressor since it sounds great, is reasonably simple and inexpensive. But it's not really ideally suited for beginners with high voltage, tubes (needing calibration, too) and quite a lot of wiring. The GSSL is really as simple and cheap as possible for a high end stereo compressor.

Without a doubt. I guess my post wasn't meant as a recommendation so much as a response to the idea of "new and/or different" designs in DIY in present times.

 

[living sounds]

Without a doubt. I guess my post wasn't meant as a recommendation so much as a response to the idea of "new and/or different" designs in DIY in present times.

Understood and agreed.

 

[amplexus]

The success of the SSL circuit in the audio world comes down to what it does well: Stereo buss compression that is mostly transparent and clean and has a reputation for 'glueing' a mix together without sounding overly grabby, but still gives you the 'punch' you want from a nice tight comp. It's flexible as hell and there are very few if any genres it can't handle gracefully. It's not overly coloured, and it helps give your mix a 'finish' so you can use it everywhere and anywhere.

In the DIY world it's just as easy to understand: The circuit is very simple- the signal path is basically a couple NE553*s and a VCA chip. Input buffer, VCA, output buffer. It's basically entirely cheap IC's and low voltage caps. So ideal for adaptation to DIY, and a perfect starter for most people to get into their first 'big' build. And in a world where a massive cohort of producers and artists are working mostly or entirely inside the box, a stereo bus comp with the features and sound of the SSL is a great choice if you're going to have limited money and rack space for outboard gear.

Piggybacking on what everyone else is saying re: compression being a mature tech- There are only so many ways to achieve audio gain reduction and they really all boil down to either manipulating the amplification of a signal or variably shunting some of the signal to ground. And inside all of those circuits there are really only a handful of ways to meaningfully modify the core of the circuit and have it still function in a useful and predictable way (Excluding the intentionally weird or glitchy efforts of circuit benders). All of the useful ways to do this in the analogue domain have been explored and refined and the mechanisms are well understood. So it stands to reason that you're really only seeing variations on a few themes and almost none of them are "new" in any meaningful sense of the word.

The SSL style comp is a very successful and popular topology, and the list of compressors specifically built for and excelling at stereo-only duties is pretty small to begin with, so it would make sense that it seems over-represented in the DIY community and the wider audio world.

I agree fully with @soapfoot - if you want to see weird cool and new things done with compression, look to plugins. Code can open up worlds of processing that we could never achieve practically in the analogue domain.

 

[cpsmusic]

Thanks for the feedback(feedforward), I was not trying to piss on anyone's parade, just thinking aloud.

OK, let me rephrase my train-of-thought in a more pragmatic question:

Are there are any similar similar designs worth building, derivate of (G)SSL design or not, that are:

-"...designed to be stereo from the ground up..."
-using off the shelf/easy-to-get components (VCA etc).
-beginner-friendly to build
-proven to work
-work in a more RMS/feedback/opto/soft knee/levelling fashion.

Thanks!

​

I can only comment from an audio production perspective. The SSL Bus Compressor is a classic for all the reasons pointed out above. However, there are mix engineers who use others. Off the top of my head, the API 2500, the Neve 33609 in particular. Also, Chris Lord-Alge uses a Focusrite Red 3 (early model with transformers) and sometimes a Shadow Hills Mastering Compressor as well as the SSL Bus Comp.

Also, when referring to the "SSL Compressor" note that SSL consoles also had channel strip compressors that were very different. This is from GearSpace:

The channel comp was based on an RMS detector, dbx-style "over-easy" soft knee, and a program dependent attack. The bus comp is a peak compressor with fixed attack settings and a variable knee according to the ratio. Based on that I wouldn't say they shared a similar design at all.

Cheers!

 

[JMan]

That’s a good point — the ssl channel compressor and bus comp are not the same thing. I think we don’t bat an eyelash when someone says “ssl style comp” because clones and variants of the bus compressor might well be the most ubiquitous diy project out there. Sort of like saying “Band-Aid” or “Kleenex.”

 

[soapfoot]

I can’t imagine using a 33609 on mix bus. They’re so aggressive!

But I do love them on a drum submix

 

[altfidelity]

If someone truly wanted to make an innovative compressor in 2023, they probably wouldn't be messing with capacitors, resistors, and semiconductors.

They'd be working with lines of code!

I do code though, mostly within JSFX and REAKTOR environment. Otherwise is too much hassle bothering with VST2, VST3, AU protocols and various compatibility issues, that are ever-changing on a daily basis.

I code mostly for fun and me, but I did put out some of my work: Inga Naïve Leveller

That's the problem I guess, in DSP land you are more flexible with rearranging code/circuit design, instant troubleshooting and evaluating.

I honestly like the digital domain, but it has its shortfalls.
It's inherently bad at emulating non-linearities, especialy electro-magnetic ones (tape, traffos, inductors ...)

Something as simple as transformer (which is basically two wires coiled near each together) is nearly impossible to model fully.
Hysteresis, eddy current loss, non-linear phase response, core over-saturation, it's just so much going on at the same time and being interconnected.
You have to seriously "dumb-down" models to use FX/plugins without latency and in real time.
That's why plugins always sound 85% there (which is good enough) but it would be straight-up-lying to say it's a 100% authentic model of a specific hardware unit.

The Golden Age 54 for instance is just a chinese made clone, more or less, of the vintage Neve

It's impossible cheap though, which is an achievement on itself, and much more interesting than anything else in its price range.

 

[gyraf]

As simple and straight-forward as the GSSL is, I know of at least a couple of recent/upcoming hardware companies showing very creative and ingenious whole-new designs by iterating the basic circuit beyond recognizably. For me, this is where it gets interesting

I think the basic design forms a simple minimalist explanation of what a compressor might be. And don't underestimate the intrinsic value of having been legitimized as "the pro's tool of choice" - this adds the magic dust that is an important motivator/aim for many.

(..) There are only so many ways to achieve audio gain reduction and they really all boil down to either manipulating the amplification of a signal or variably shunting some of the signal to ground. (..)

..we actually came up with kinda' a third way for our passive compressor: Wanting a fully-passive, still floating balanced audio path, we couldn't just shunt signal to ground and still have everything floating (this would require an active re-balancing circuit). Instead we shunt between the two balanced input signals with an array of four matched LDR's: +L to -R, -L to +R, +L to +R and -L to -R. The beauty of this topology is that if you look closely, you'll find that now we have two LDR's that control M component compression and two that controls the S component, i.e. independent M/S control - and just like that the G24 turned into a M/S compressor also

/Jakob E.

 

[altfidelity]

...whole-new designs by iterating the basic circuit beyond recognizably.

...therefor creating something new. Creating something new by improving on a copy/clone of something else.

Jakob, I fully regret If I came off as negative/disrespectful towards you and whole GSSL affair. That was never my intention.

If I am completely honest, I conceptually understand circuits but I don't know **** about actually designing circuits.

My attempt may be naive, but how would I start translating attached block diagram to a circuit diagram?
the sketch is a 30 seconds draft while eating lunch, so don't expect to much it's just a rough design that seems to yield nice results in DSP.

 

[gyraf]

Your threshold is the AC gain before your rectifier, and ratio is your DC gain on the control voltage after the rectifier.

If you have a rudimentary threshold function in the rectifier, e.g. a diode drop at its output, this 0.6V or so drop scales by the AC gain making variable threshold

you already have first-order integrator in the attack/release, no reason for further?

let Ratio be DC gain after attack/release timing

Why the exp conversion function? What VCA's are you planning on using?

 

[altfidelity]

you already have first-order integrator in the attack/release, no reason for further?

Lone peaks above threshold from peak detector wouldn't charge the second integrator much at all, but peaks following each other continuously, would start charging the slow averaging integrator (capacitor?), basically increasing ratio (I guess ?) on heavy gain reduction cycles. I may be totally wrong.
It makes more sense if you have peak detector's release set very short (around 100ms), compared to second average detector (let's say 500ms to 1sec).
I dubbed this property "INERTIA" in my dsp designs, it has a nice ring to it

Your threshold is the AC gain before your rectifier, and ratio is your DC gain on the control voltage after the rectifier.

If you have a rudimentary threshold function in the rectifier, e.g. a diode drop at its output, this 0.6V or so drop scales by the AC gain making variable threshold

Ok.
So if I understand correctly, AC output1+ AC output2 summed to gainstage (threshold), then diode rectifier with 0,6V drop, then DC gainstage again (ratio).

let Ratio be DC gain after attack/release timing

Wouldn't then the ratio affect attack & release timing? (increased ratio - shortened attack etc...)

Why the exp conversion function? What VCA's are you planning on using?

I have no idea (yet), whats the conventional way/expected CV signal for controlling VCA?


Thanks Jakob, it means a lot.

 

[living sounds]

It's impossible cheap though, which is an achievement on itself, and much more interesting than anything else in its price range.

Yes, I think it's a re-packaged Alctron CP540, it has a great vibey sound at a small price. Often smaller than DIY. ;-)

 

[Matt Syson]

The sidechain rectification should be 'dual mono' then summed into the attack/threshold etc to reduce the risk of the stereo image falling apart and issues with strong somponents that are centre and out of phase which leads to under compressing, an issue with compressor limiters used for broadcast where they have to be 'tolerant' of whatever the incoming stereo may present. Some of the issues are about what you CAN control using the technology you have chosen. Opto (resistive) devices are always slow acting and have wide tolerances, making stereo or 'ganged' operation near impossible if you need repeatability and tracking. A 'rectifier' can be made as fast and accurate as you wish and then implement attack and 'release functions afterwards which COULD be ganged (stereo) OR made using combined 'DC feds from the rectifier stages. Also deciding if you want RMS, peak or combination or even dual processing of the (fast) rectified audio. VCA chips (modules) have a few foibles such as control voltage feedthrough if the rising/falling edge is particularly fast because of the diffeent collector -base capacitance of the input PNP/NPN transistors used. Fortunately most comp/lims shy away from really fast 'attack' because it actually sounds bad on most audio signals (being a 'Fourier click') where an abrupt change of level 'fools' the ear/brain into thinking the signal has a momentary increase in pitch (the signal cycle is cut) which is interpreted quite commonly as a 'click'.
Fets can also suffer from feedthrough but have more problems with distortion and noise that largely overshadow this. Opto resistor devices simply can't operate sufficiently fast to have this problem.
The lA2A comps may be so popular because the selected optos do the bulk of the compression in an 'ear friendly' way but transients they miss are 'caught' by the saturation/soft clipping' of the output stage/transformer as it's output is not always capable of a particularly high level of output before clipping. Not played with one for a few years but when feeding it's expected 600 Ohm termination the output is certainly 'rounding off' at +18 dBu or thereabouts, obviously dependent on the condition (cathode emission) of the output valve. Once you have all this figured out you then have to design the board so that there is minimal (or at least defined) crosstalk from the rectifier/sidechain processing into the audio paths.
A 'failing' of the Gain Brain' (Valley people) where the switching of LED (bargraph) current intrudes into the audio at some fairly specific levels (when one of a handful of the display LEDS get switched on.

 

[altfidelity]

should be 'dual mono' then summed

Ok, makes sense. If sidechain signals would be perfectly out of phase the detector would "see" nothing. Makes perfect sense, thanks!

OR made using combined 'DC feds

That's the way it makes sense to me, rectified and then summed signals feeding one attack/release stage, which drives both VCAs . It just seems like less components and complexity.

I really believe in concept of secondary envelope, which would integrate peak detector.
It's basically integrational component component of PID feedback control loop, while peak detector would be the proportional component.

PID controller - Wikipedia
PID “Proportional, Integral, and Derivative” Control Theory

Concept being, peak detector is handling microdynamics, while secondary integrator is handling macrodynamics - basically perceived overall level.
In my experience, you can also get away with using faster attack and release times, without getting weird artefacts.
The variable amplitude of second integrated CV, would allow to tune PID system in a different ways, therefor changing the character of compressor.
The tuning of PID control loop is a science in itself. PID Tuning | How to Tune a PID Controller - RealPars

FETS can also suffer from feedthrough but have more problems with distortion and noise

I would rather use FET gain stage, mainly because FETs seems funkier and more nonlinear (which is cool to some degree), but it seems like a more complicated/more expensive way of doing the same thing. Are FET based compressors recommended for first timers like me?

Edit: nevermind, I'l start with something like THAT1510/2180.

.

 

[altfidelity]

This is literally the first circuit sketch i've done ever. It's more guessing than designing, but I will give it a try, since I'm doing this just for fun. Can anybody comment, which component would short first

 

[richiyobs]

This is literally the first circuit sketch i've done ever. It's more guessing than designing, but I will give it a try, since I'm doing this just for fun. Can anybody comment, which component would short first
View attachment 105601

Hi
From what I learned you need to add a log amplifier in front of your detector for a better behavior when using That2180
Best

 

[gyraf]

..you'll need a threshold diode after sum before attack pot (the rectifier diode drops are canceled being inside opamp feedback loop)

and you'll need to take your "peak" signal from the output of the timing buffer amp, NOT from it's input (taking from input would act as a load parallel to release pot)

other than that - just try it, and see (listen)?

/Jakob E.

 

[altfidelity]

and you'll need to take your "peak" signal from the output of the timing buffer amp, NOT from it's input

I know this is a trivial question for you, but that's the first fundamental concept I don't understand.
How do I know, when a certain stage needs a buffer amp, and when not?