Different types of compression, SSL vs 1176 vs 1a2a vs DBX160 ect...

[DaveGamble]

I love talking about dynamics. Can you tell?

Dude. PLEASE feel free to continue!

Sean Costello

Heh, this is how I feel when you start telling me about reverbs. <3

Dave.

[DaveGamble]

Found this via Google

http://srtalumni.com/SRT_Manuals/Valley ... namics.pdf

Also Softube made a emulation
http://www.softube.com/index.php?id=dmite

Hope that is of interest!

*gasp*! Via google? Crikey. That's some excellent googling!

Ok, page 17 (labelled as 16) IS the schematic! It's low-res, but hey, I can see what's going on, roughly!

Ok, top left, TL074 input (I'd prefer to see NE5532s, but hey).
U1 +appears+ to be a pretty decent logging amp (you could just use a THAT/dbx chip for the job, but no shame in rolling your own).
Hard to make out, but U5 looks like it's the threshold... would be awesome if the res were higher!
U9A/B are the rectifier... thus far, this is pretty much what you'd expect from a feed-forward dbx style design.

Waaaaaait. Ok, just checked the parameter list... this makes more sense now (because there's so much on the schematic I can't read).

Right,
Top-left, U1, audio in, ext in. You see U2 there? That's an analogue switch. This is the "Int/DS/Ext" switch.
But where's the de-essing circuit, you say?
It's C3. It's just a 1st order high shelf! Amazing! (ok, it's R10,C3 in parallel with R9, to be exact).
The API2500 does the same sorta thing, in fairness (though I'd say with a fair bit more finesse!), but it's a sidechain EQ.

Ooohh, ok, so, U1 (After U2, reading left to right) IS indeed a logging amp, calibrated.
U5 is acting as a comparator for the threshold, but then U8 acts to, I think, tweak some gain to compensate for something.
Now, I'm not 100% sure, but it looks like U5 is an RMS detector kinda job. My /guess/ is that either C9 or C10 is the RMS cap, but without a really close look, that's how it looks.
Actually, it must be C9, because C10 is the release capacitor... I think U5+U7 is actually doing a literal squaring.

GR signal comes out of U9A, and at that point, we're thresholded, either limiting or gating, with release.
U6 again does a "do we want to be +ve or -ve" trick, picking whether you go into U4 +ve or -ve.
C11 is likely the only smoothing before the VCA, so I'd say this is attack.

The TA101 is the VCA, which takes that signal and uses that to determine gain to apply. Then U10 is an unbalanced line driver, and we're out.

It's entirely feed-forward, from what I can tell. Any feedback is localised and without any capacitative elements, so this would have been awesome fun to model (and quite easy, with a legible schematic!).

It'd be ok to think of this as a loosely expanded dbx. Except that actually, those dbx chips are REALLY accurate, and this won't likely do as well, unless they're really going to town matching the transistors up.
It has some cute features, like the sidechain EQ, and the range thing (which I didn't explain, but basically, from what I can see, depending on whether you're in limit or gate mode, it injects some voltage into an appropriate part of the circuit to clamp your GR signal).

Looks kinda fun. If you told me that the Softube model phase-cancelled the original (noise floor excepted), I would fully believe you. This is the kind of circuit that makes modelling great fun, because you CAN get it exact.

The use of analogue switches is a nice touch (avoids pops when switching them), but it's fairly unexceptional design. Of course, I'm biased, but it's nowhere near as sophisticated as a Stride design (SSL/Neve/Focusrite).

It's cute, and I like it, but I don't know why I'd want to use one especially. I have a hunch that a decent-spec dbx would outperform it for noise/accuracy. There's really nothing in there to add "character" - unless it's been poorly calibrated at the factory, but really all that would do is break the thresholding. You could certainly replace it with a Compassion Mod. Decoupled AR, log curve law, release to signal, attack time to zero (use VCA smoothing instead), and just use a shelf to boost the s/c EQ if you want the DS function.

Hope you enjoy! (and if anyone finds a more clear schematic, I'm happy to go in further!)

Dave.

[DaveGamble]

Please continue Dave

Any more questions?

Dave.

[schnapsglas]

You should write a book. I would buy a copy.
Time to start Kickstarter! Give this man every schematics and two Fairchild units!

[SWAN808]

Please continue Dave

Any more questions?

Dave.

Hi Dave - can you explain why the vari mu design is different to the others?

And as well as attack, release and knee characteristics - what is it that makes one compressor grab hard like an SSL whereas a vari mu design less so? Is that also timings or something else to do with the gain reduction element (FET, VCA etc)?

Why is it that software compressors seem to handle bass frequencies differently to hardware? I tend to always hear a more solid low-mid in hardware.

Very interested in this stuff...

[dalor]

Interesting read!
If you had all of these units and a drum loop could you immideately tell which unit would do the best job to meet your clients vision?

[DaveGamble]

Hi Dave - can you explain why the vari mu design is different to the others?

Because it's a complete hack, and it's a very loose approximation of what it's trying to achieve?
It's like a compressor that has a go, and then fails badly!

It's not different to the others, per-se, it's just massively underspecced, and working exclusively with valves, it's practically (key word there) impossible to prevent interaction between sidechain stages.

You could emulate vari-mu (and people have) with solid state gear; probably by linking A/R circuits, and clamping a bunch of the control signals.

And as well as attack, release and knee characteristics - what is it that makes one compressor grab hard like an SSL whereas a vari mu design less so? Is that also timings or something else to do with the gain reduction element (FET, VCA etc)?

Literally ratio, and the depth of gain reduction it can apply.

See, over time there's improvement in circuit designs, to the extent that the designer (and this is absolutely true by the time you get to a dbx) can design the circuit in stages, and achieve precisely what's desired.

I suppose the answer to your question very specifically is that the maximum amount of gain reduction that can be applied has been improved with time. But further, it's that a great deal more finesse can be applied with newer (higher spec) components.

Building a high-precision high-range logging amp WAS impossible, but by the time you get to dbx, they're selling ICs that have something like 80dB of range, which is fantastic!

But now suppose you've only got 20dB of maximal GR (as a result of only being able to measure 20dB excursion from the threshold)... well, once you REALLY need to kick in, because the signal is a long way over the threshold... what do you do? Nothing. That's what you do, because you can't. I'd call this "ceiling"; and it's exactly analogous to gate range.

Feed-back designs were actually an attempt to ameliorate this problem, which they did, with varying degrees of success.

Why is it that software compressors seem to handle bass frequencies differently to hardware? I tend to always hear a more solid low-mid in hardware.

Very interested in this stuff...

There is no theoretical reason that would make that likely to be the case - in fact quite the opposite; the theory suggests that if anything will differ between the two, it's handling of top-end.

If you can demonstrate with some examples, we can (and should!) evaluate this, but I suspect the reality is that when you're monitoring software, your head is in the sweetspot, while when you're monitoring the outboard, your head is outside the sweetspot, and you're adjusting timings accordingly. I support this claim with the observation that half a wavelength of low mid is going to be in the order of a foot or two, which is probably how far you have to move your head to look at a unit you're adjusting.

Dave.

[DaveGamble]

Interesting read!
If you had all of these units and a drum loop could you immideately tell which unit would do the best job to meet your clients vision?

I think you vastly overestimate how precise a client's vision can be, how accurately they can describe to you what that vision is, and actually how much of the finished result you provide. And how much of the client having the sense of having achieved their vision is a consequence of the emotional state you project onto them while engineering.

But as an answer to your question:
Lots of these units have fun controls, which are great, and sometimes you can find something that really gets you what you want, but I got bored of the idea of auditioning a dozen processors, and built one that includes all the neat controls, so you can zoom in.

And as another unsolicited reply: There IS a school of thought that would argue that you ought to try hard to achieve your desired dynamic behaviour BEFORE processing the drum buss (i.e. in the mix of the individually processed channels). Of course this thinking doesn't apply at all to working with sampled drums, nor with the idea of bus compressing your drums to add character... I only mention it in the sense that if the client gives you access to the individual drum parts, you might well find that you can get great results by being much less precise with each individual channel, and some love at the end, rather than trying to work very hard on a stem.

Dave.

[DaveGamble]

You should write a book. I would buy a copy.
Time to start Kickstarter! Give this man every schematics and two Fairchild units!

Thanks.

I do want to, but there are so many things to do.

Dave.

[Debutante]

I wish someone would compile a compendium of all the different types with example plugins of what does what

Peak vs RMS - some can hybrid

Valve/tube
Opto -- I'd like to try more of these, for example
feed forward vs(backward = opto)
FET
VCA
Vari-Mu
... i've heard of some Class A or U somethings ...

now THAT would make a great sticky.

[manducator]

Any more questions?

Dave.

Can you make more presets (hardware emulations) for COMPassion?

[meloco_go]

Building a high-precision high-range logging amp WAS impossible, but by the time you get to dbx, they're selling ICs that have something like 80dB of range, which is fantastic!

What's a logging amp?

If you can demonstrate with some examples, we can (and should!) evaluate this, but I suspect the reality is that when you're monitoring software, your head is in the sweetspot, while when you're monitoring the outboard, your head is outside the sweetspot, and you're adjusting timings accordingly. I support this claim with the observation that half a wavelength of low mid is going to be in the order of a foot or two, which is probably how far you have to move your head to look at a unit you're adjusting.

What if you use headphones or test a render?=)

[DaveGamble]

Building a high-precision high-range logging amp WAS impossible, but by the time you get to dbx, they're selling ICs that have something like 80dB of range, which is fantastic!

What's a logging amp?

If you can demonstrate with some examples, we can (and should!) evaluate this, but I suspect the reality is that when you're monitoring software, your head is in the sweetspot, while when you're monitoring the outboard, your head is outside the sweetspot, and you're adjusting timings accordingly. I support this claim with the observation that half a wavelength of low mid is going to be in the order of a foot or two, which is probably how far you have to move your head to look at a unit you're adjusting.

What if you use headphones or test a render?=)

A logging amp is a waveshaper which generates y=log(abs(x)). You use it to convert a signal (which is a voltage) to a dB measurement.

I'd love to know if you find the same thing in cans, but further, if you find it with renders, we can delay compensate the renders, and phase cancel them!

Dave.

[DaveGamble]

I wish someone would compile a compendium of all the different types with example plugins of what does what

Peak vs RMS - some can hybrid

Valve/tube
Opto -- I'd like to try more of these, for example
feed forward vs(backward = opto)
FET
VCA
Vari-Mu
... i've heard of some Class A or U somethings ...

now THAT would make a great sticky.

If someone did this, I'd be happy to enter a bunch of info about the hardware.
Could even try to turn it into a set of Mods.

Dave.

[DaveGamble]

Any more questions?

Dave.

Can you make more presets (hardware emulations) for COMPassion?

Yes. What of?

Dave.