Please keep in mind, whatever is written in the o-ring manufacturers catalogues or ISO/DIN/JIS or whatever standard you want to look at, those are recommended values, for all sorts of applications and all sorts of conditions. 5.7% might be way lower than the recommended value, however, that does not have to mean it will be leaking in this particular application. In theory, as soon as you have compression along the whole circumference, you have a seal. For anything to slip by, it has to overcome that "compression pressure". Now, if you add pressure from one side, the rubber will actually press against the sealing surfaces with a higher pressure, too, which is the "compression pressure" plus the equal amount of pressure added from the side, so in sum the o-ring will always keep pressing against the sealing surface with a higher pressure than it is exposed to (as long as it stays intact, of course).
In other words, if you go deeper, yes, the o-ring will be pushing against the sealing surfaces with a higher pressure, but it also has to withstand and equally higher pressure from the outside. But the difference will stay the same.
But if I had the choice, I'd go with a design that has 20% compression than with one that has 5%. There are reasons why 20-30% are recommended. By the way, depending on which standard you look at, those values vary. In fact, in the latest ISO (3601 if you want to know more), the recommended range varies by cross section, though if I remember correctly it's always above 10% for static applications.
In the end, it's the field testing that determines if it's sealing properly or not. Sadly, sometimes the customer is the guinea pig.
[...]
My 5D housing uses a standard 1/8" ring. The terminology is unfortunate because a 1/8" o-ring has an actual thickness of 0.139 +-0.004. The guidelines specify that the groove should be between 0.101 and 0.107 deep, which would give a compression ratio of 20% to 30%. The width for a standard o-ring should be 0.177 and 0.187 to give room for the o-ring to expand laterally when compressed.
The groove in the housing measures 0.100 deep and 0.136 wide. The depth conforms to the standard depth assuming enough clamping force to compress the two halves until they touch. However the cam latches do not provide enough force to do this. When closed, there is a gap of 0.032 so that the o-ring compression is only 0.008 or 5.7% which is seriously under-compressed.
[...]
Do I read this right, your findings are that the groove width of your housing is actually narrower than the thickness of your o-ring?
And if you were to close your housing so the housing parts touch, which you say it does when you vacuum your housing, the groove height will be reduced to 0.100?
That sounds too small. Rubber is incompressible, so when you squeeze it from 0.139 down to 0.100, that rubber has to move somewhere. That's why you should have a wider groove. Granted, it will change it's shape to look more like the groove, so going from circle to square cross section will help, but it feels like something is off. A closed groove with 0.1 height and 0.136 width won't fit an o-ring with 0.139 cross section diameter. Unless I understood you wrong?
And P.S.: Both my housings (Canon) use a radial seal design, so, the force which both parts of the housings are pressed together with won't change the compression ratio of the o-ring.