I noticed that Tube Compressor by default reduces the incoming signal by 3 dB even if the MIX knob is set to 0% (e.g. fully dry).
To me this makes no sense and feels like a bug.
I tested also other compressors (Vintage Compressor, Black Valve, Compressor, VoxComp) and they all behave as expected eg. they do not reduce the signal at all when MIX knob is set to 0% (=dry).
Confirmed. Looks like a bug to me, too.
Is it downmixing stereo to mono? That might make sense.
EDIT: In reference to pan law, I mean.
Yeah, pan law was the first thing that crossed my mind, too. I checked before I confirmed - it’s not pan law related.
I don’t think it is a bug. The clue is in the name: “Tube” compressor. This comp has a tube simulation (perhaps at the output amp stage) that cannot be completely turned off (drive has a minimum of 1dB). Essentially, you always have a soft clipper in the signal path, which will a) generate harmonics, which you can see in an analyzer after the plugin, and b) will clip (or “shave off”) the peaks a bit.
No downmixing and Pan Law settings make no difference.
And as I wrote other compressors do not behave like this (I tried them all as insert plugins right after the Test Generator)
I also tested the channel strip version of the Tube Compressor = it also reduces the signal by 3 dB when fully dry.
Even when the mix knob is set to dry?? That IMHO should essentially bypass the plugin completely no matter what especially when all settings incl. INPUT and OUTPUT are set to 0.
(Black Valve is a tube type compressor too)
Ah, sorry, I’ve somehow completely missed the “dry” bit 
Agree, in that case it seems not correct, with 0% wet output is effectively quieter, I checked that with a meter before and after the compressor just to be sure
Mix = 0%: -3dB
Mix = 50%: +3dB
Mix = 100%: -3dB
Would this be expected?
Not sure if it’s “expected,” as I personally wouldn’t expect a 100% dry signal to be attenuated; not intuitively, anyway. That said, it’s probably “by design” as 3dB can’t be a coincidence 
- seems like an explicit attenuation in accordance with some manner of summing rules.
Even if it is a bug - it will not get fixed. A fix at this time might destroy the sound of old projects from users, so it will stay as it is.
Thank you for the info to make all of us aware of the situation. We’ll have to learn to live with it.
Yeah I do realize that most probably there is no way of fixing it WITHOUT destroying levels in existing projects using this plugin…
Good point!
How about adding a 3db correction switch with a user guide explanation?
Satisfies both prior and new users.
OK, so I searched a bit on how crossfading between two signals is usually implemented, and there are basically two ways that are used: a linear (or constant gain) or an equal power crossfade.
It seems that Tube Compressor uses the second approach, which applies a non-linear function to the signals.
See here for a rather nice, visual explanation:
So it isn’t actually a bug, but simply a different implementation of a crossfade (although one that - from what I read - is a bit unusual when crossfading two correlated signals)
