Great article ! Every manufacturer puts all kinds of tricks to give spectacular demos at their boot, but once you test the cameras in the lab, you see the real performance ! And there's a big difference in behaviour, also according to the lenses you use. Same camera with same settings, but varying brands and types of lenses also give a wide variety of performance. Maybe this can be an interesting test as well.
With all of the trickery that manufacturer’s use to hide their products’ true low light level capabilities, I have given up specifying low light performance. Even using performance recommendations like you have suggested are challenging because the manufactures don’t publish all of the detail needed to properly analyze the camera’s performance. Until there are some meaningful standards and the products are tested by an independent testing agency (like IPVM), I will stick to specifying a few cameras that I have confidence in.
We put out those performance recommendations because some organizations are not allowed to specify specific models (procurement rules, etc.). Our goal with our recommendation is primarily to stop unrealistic lux rating specs (like camera must be 0.00001 lux) that reward the bad actors and screw good products.
John, this is vey helpful in situations like ours. I work for a university and everything over a certain price has to go through the procurement (bid) process or be gotten from a recognized "sole source vendor". Specific language is important. Thanks.
Usually when i need to value cameras sensitivity i check the illumination level necessary to form the full signal at Gain=1, Texp=1/50, if possible turn off gamma correction and auto-iris just in case. mostly it is in the range 10 - 40 lux (F1.4) easy to measure. the behaviour in other conditions is very predictable.
All above is valid but there are additional factors that also need to be considered.
1) As shown at the beginning of this article video noise is a major factor in the usability of the picture . The A, B, and C show this clearly with "C" having the best signal to noise ratio ( S/N). Manufacturers again distort their specs based on S/N by giving the sensitivity spec with "AGC ON" and/or "Sense Up ON" then the S/N with "AGC OFF" and/ or "Sense Up OFF" obvious that specs. should be done with the same settings.
2) The lux rating for the scene is given but critically no manufacturer speaks about the reflectivity of the scene. So they measure with a white card or other highly reflective object. Then the installer puts the camera viewing a parking lot that is black asphalt with virtually no reflectivity.
3) The old IRE was a valid measurement for surveillance and broadcast. It, combined with the S/N values, gave a more accurate picture of how a camera performed. The 30% number with a high S/N would indicate a good minimum usable picture. It would be nice if there was way to duplicate this with IP cameras and manufacturers using this as a standard ( We can dream can't we).
4) In actuality the best way to standardize light measurement and take the lens "f" stop and reflectivity out of the equation is to use a measurement of the light at the faceplate of the sensor as the specification. A number of manufactures tried to implement this many years ago ( analog cameras) and it fell to the wayside because this would keep all honest and many did not want to be honest.
5) At the end of the day the user is the victim of all the specification manipulation as many times they buy product that does not work in their application. The single best way for the user to protect themselves is to have an onsite demo with competitors doing side by side before they buy the system.
4) In actuality the best way to standardize light measurement and take the lens "f" stop and reflectivity out of the equation is to use a measurement of the light at the faceplate of the sensor as the specification.
I agree, measurement at the sensor takes the reflectivity out of the equation. How do you take the lens out of the equation, though? Measuring uncollimated light doesn't seem accurate.
a measurement of the light at the faceplate of the sensor as the specification.
Some manufacturers will just pass along the minimum illumination from the imager supplier. But that does not help since those numbers are not any more realistic or helpful for an end user buyer nor will it help compare to others who use different approaches.
You of course are correct about manufactures using the sensor manufacturers specs. In fact this goes even further. Much like Brian pointed out the other day about the OEM China's makers specs being used by the seller verbatim, some camera manufactures just copy specs from the higher class manufacturers and put them on their spec. sheets so they are comparable.