MCX returns zero detected photons when detector radius equals or less than 0.5

[fangq]

Joe Prosser reported the below issue in a mailing list post
https://groups.google.com/g/mcx-users/c/5wAuk4G7n2o/m/k-2keouAAQAJ

Problem: when a detector's radius is set to 0.5 or less, it does not detect any photon even issrcfrom0 is set properly.

Subject: | [mcx-users] Re: issue with detector radius in mcxlab?
Wed, 26 Aug 2020 13:48:33 -0700 (PDT)
Joe Prosser <joepro6 at ...>

The issue also happens when issrcfrom0=1, only at the detector radius of 0.5 instead of ~1.0 (see below). Sorry for the confusion I had removed that line, trying to minimize the cfg but didn't realize I caused the loss of photons in a different way.

Changed cfg.detpos(1,4) to 0.5
added cfg.tstart=0
added cfg.issrcfrom0 = 1

Same issue. Interestingly the XY exit positions of detected photons when the radius is 0.51 actually forms a 0.51 radius circle. If the circle was only tangentially intersecting the volume, I would expect the radius of XY exit positions to be smaller in that case. 
 
When cfg.detpos(1,4) is set to 0.51:
figure; scatter(detphoton.p(:,1),detphoton.p(:,2),'.')

> On Wednesday, August 26, 2020 at 12:35:49 PM UTC-4 Joe Prosser wrote:
> Hi Dr. Fang,
> 
> I seem to be having issues with detector radius. When I run the below in MCXLAB I get 5190 detected photon packets. If I change the detector radius detpos(1,4) from 1.2 to 1.1, I get zero. I realize there is a work-around to save the exit positions and filter them but the number of detected photons seems to drop off precipitously as the detector radius decreases from 1.2 to 1.1. For a source/detector distance of 20mm the radius threshold is around 0.5mm. Script below. Thanks for taking a look!

> Joe
> Joseph Prosser, PhD
> 
> Revision info
> $Rev::f9ad72$2019.4 $Date::2019-09-06 18:50:12 -04$ by $Author::Qianqian Fang $
> ###############################################################################
> - variant name: [Fermi] compiled by nvcc [7.5] with CUDA [7050]
> - compiled with: RNG [xorshift128+] with Seed Length [4]

This can be reproduced using v2019.4 or latest github version with the following test script

%% set up the parameters and S/D locations
clear all;
cfg.srcpos = [24.5 40 0];
cfg.tstart = 0;
cfg.tend = 5e-9;
cfg.tstep = 2e-11;
cfg.nphoton = 1e7;
cfg.vol = ones(49,120,41);
cfg.srcdir = [0 0 1];
cfg.autopilot = 1;
cfg.detpos = [24.5000 50.0000 0 0.5];
cfg.prop = [0 0 1.0000 1.0000; 0.0264 2.8684 0.6200 1.5111];
cfg.issrcfrom0 = 1;
cfg.issaveexit=1;

[fluence,detphoton,vol]=mcxlab(cfg);