The problem has never been a single VCA device, but when you start stacking VCAs plus Opamps becomes the problem. Stacking devices in front of one another, usually a combination of VCA's and Opamps increases the summed noise floor logarithmically.
Actually, if the two noise sources are random and of precisely equal level, they sum with a resulting increase of 3dB. IF they're equal. And they're usually not, the higher noise floor dominates, and the increase from it plus the second source is less that 3dB. In the case of the processors I cited that already have a total DR equal to 16 bit PCM, this is simply not a problem.
Noisy opamps hasn't been a problem for many decades now. We got the 5534 in the mid 1970s. In a 60dB mic preamp gain circuit equivalent input noise is -129dB...about 2dB off theoretical thermal noise. And that's perhaps the most common opamp in broadcast gear of the 1980s on, until we got the all that PMI/SSM stuff which nudged the noise floor down farther. There were some earlier noisey opamps, like the 709, 301, and (ugh...for audio??) 741. But those haven't been in audio gear for over 45 years, certainly not in the audio path of a processor.
But that's the thing, forget the published specifications. Take a good quality audio measurement system of the day like a Potomac Instruments generator and analyzer, or an Audio Precision and measure the s/n of an audio chain of old that uses VCA stacked processing devices, versus even an older digital Optimod, like the 8200. The difference is quite obvious.
This is a fairly pointless argument, as neither of us can prove it, and we're not throwing actual measurements into the mix. But I'll just disagree with "quite obvious" for now. The noise increase, related to processors, has to do with the increase in gain required to get them above threshold. If they're all set for unity gain, you won't notice any noise increase. So, we crank up the input to the pre processor so it's doing 10dB of GR. What happens when it releases? You now amplify whatever noise is coming in by 10dB. Do that again in the next processor...etc, etc. But it's not the processor itself, it's the noise in the system that's being amplified. You could tell by terminating the input to the first processor, vs connecting it to whatever you're driving it with. In those analog days we had noisey telco lines, analog STLs, and analog source gear. If you didn't pre-process before the biggest noise maker (lets say, an analog STL), you'll be amping up its noise at the TX end. A lot. Again, not strictly the processor at fault.
What's changed is the system. We have digital systems end to end. That means we now have complete knowlege and control of where 0dBFS is, and how much DR we have below that. But, we still have to be careful with the above situation. If we're sliging 16 bit audio around, we "only" have 96dB, and really, more like 93dB of DR. If we allow 15 for headroom in the system (before processing), our noise floor is now "only" down 78dB from average signal. Now you go into a processor with that, and push it to 10dB of GR. And then let it fully release. Now your noise floor is 68dB and everything is working just fine, nothing's broken, nothing has "noisey VCAs" or "noisey opamps".
Now before we got saying "yeah, but I run 24 bit audio", yeah, no you don't. You have 24 bit words. Your actual audio is 20 bits of DR (because ADCs can't do noise floors lower than thermal noise), and your source devices are more than likely 16 bits, probably compressed. And then there's the mic, preamp, and the elephant in the room...the room. And we're right back to where we were.
Every dB of noise being generated by stacked devices is wasted modulation and adds to diminishing the overall listening experience.
I need to clarify here. I'm not taking ANY issue with the need and goal of low noise. Not at all. Low noise is good. But your sentence above isn't making sense. Lets say you somehow manage a noise floor at 78dB below 100%. And then you add another 3dB of noise, to put it at 75dB. What modulation has that wasted? Your -78dB was 0.0125% mod. You've bumped that up to a whopping 0.0175% "wasting" an additional 0.005% of your total mod. Pick your noise levels, run the calculations, it never wastes even 0.1% of modulation. Diminish the listener experience? The average acoustic noise level in a quiet living room is 30dB SPL. To hear the -75 noise floor above that, they have to play your station at 105dB SPL, which is darned loud. While possible, there are few formats that would be tollerable that loud for very long. And that assumes they could actually receive your -75dB noise floor (which they can't). Car listeners? Nope. The average interior noise in a car is between 60 and 70dB SPL, covering any broadcast noise you may have quite well.
You are literally "amplifying" the "noise issue".
I will say, just so everyone knows I DO care about noise, I spend many years working for a station where noise mattered. A lot. The transmitted noise floor could be as low as 82dB below 100%, that's until a source was played, even a digital one. We found that nothing on earth could receive that low a noise floor. Nothing. Low 70s in practical receive situations. There's one thing HD Radio improved on.
Of course, those who want to compress and clip the crap out of their audio chain to make it appear loud, don't care about things like listening experience other than their own skewed perception of what sounds good.
Different issue. We can discuss, but there's literally no point right now.
And also too, old Compellors are, well, old. Caps dry out and are rarely replaced until the piece of equipment fails, if at all.
If you were able to toss a new Compellor on the test gear and one that's thirty years old, of course, there will be a noticeable difference in noise and performance.
All true, and I've never (hopefully!) implied that anyone should just toss an old piece of gear into the air chain. But...funny...I've had the opportunity to pop the top on several Compellors lately. None new, of course, and yes a couple were in their 30s. They'd been sitting for years, unpowered. My goal was to rehab them then find them a new owner. The first step, after making sure nothing is smoking, is to run some audio tests. You are correct, there was a noticable difference between their noise performance and the original specs. They were better! Yeah, better. No power supply ripple, no hiss, boom rattle, nothing unusual other than to beat their published specs. That doesn't mean they didn't get recapped a bit, because hey, fair is fair. But it does say a lot about how Aphex designed stuff. And the new caps didn't improve anything. I had a dirty pot and a couple of dirty contacts to clean.
I did mention the Audio Chameleon, right? There's your "new Compellor". Well, not exactly. Probably much better, but I haven't tried one personally.