depth test to allow VR < 0

[carltape]

At present we are plotting ln(VRmax / VR) for the depth test. This doesn't work if VR is negative.
VR = 100(1 - E) = 100(1 - int(d-s) / int(d))
Let's instead plot ln(E/Emin) -- I think this will solve the problem, since E is always positive.