I'm using a ~60pF photodiode, 3x3mm, directly hooked up on the input of the DDC112, in photovoltaic mode (no bias), as indicated in the datasheet. The second input is left floating.
I'm seeing a bias current input of about -0.15pA on both inputs. The dark current from the photodiode is probably below this level.
The datasheet specifies a noise of max 6ppm of FSR. At 350pC, this means 3.5fC. I'm seeing a fluctuation of about 50fC, no matter the range used. This points out to some external noise of unknown origin. I guess I could add more shielding around the DDC112 to form a better faraday cage, use a better insulator material for the PCB, and isolate furthermore the analog power supply. Maybe the fact that the 5V digital power comes from the USB port doesn't help. Anyway, finding and adressing the exact source of this extra noise is another project in itself.
Per specification, at the lowest range (50pC), we should be able to get a noise below 1fC, or in the order of 3000 electrons. It's very small in the electronic world, but very big in regard to camera sensors (read noises in order of 1e-)
I'm using a white source of known power, settable between 1fW to 10pW. Light is sent directly onto the photodiode with a 1mm fiber optic. The entire circuit is put in a black box to avoid surrounding lighting getting in. With everything in place, I'm seeing about 0.3pA with an input of 1pW. This is well in accordance with sensitivity of about 0.3A/W for 500nm. It's great that I can show that the relationship still holds for this particular circuit at low light levels.
With a noise of 0.05pA, this means that I can measure light in the range of 1pW to 1nW with a resolution of 0,1pW. Pretty impressive, considering the fact that i'm doing this with a 30$ chip.