To make the signal stronger, we can register not just the color of one sub-pixel (the red, green or blue dot), but of the whole triad, that is, draw a white line on a black background under the photosensor’s window. The three different colors of the triad are created by means of external color filters, while the three sub-pixels are actually identical, so we won’t get less accurate results measuring light-up time of the three sub-pixels rather than one. Meanwhile, the total signal intensity will be doubled in comparison with the signal from red sub-pixels only (doubled, not tripled, because the sensitivity maximum of the photodiode lies on an infrared range of 950 micrometers, and the sensitivity to green or blue colors is lower than to red).
The photosensor was made from a Siemens BPX90 photodiode and an Analog Devices AD8604AR precision amplifier. The photodiode’s shunt resistance was set to 10 megaohm to achieve required sensitivity. The amplifier was powered directly by the computer. In order to reduce the noise from the switching mode PSU, power was sent via an LC-filter and compensation regulator based on the 7805 chip.
Amplifier with the photodiode
The whole contraption was put into an aluminum case to avoid noise and external light. So, without any special measures (like a separate PSU, precision regulators and the like) we reached the noise level of about 10mV. It allows performing measurement tests without any problems, considering the maximum signal level is about 4.5V (the operational amplifier limits it as it has unipolar +5V supply).
Assembled sensor, top view
The bottom of the case was covered with black mat film to reduce noise from side lighting. An aperture was drilled against the photodiode. Its diameter was about 3mm.
Assembled sensor, bottom view
During the measurements the sensor was installed onto a horizontal panel and a small program was drawing a horizontal white line, one pixel wide, on the screen just against its light-sensitive window. The sensor’s output was connected to the oscilloscope. The trigger of the oscilloscope was set to switch on along with the fall edge of the signal (when measuring pixel fadeout time) or its rise edge (when measuring pixel light-up time). Thus, the oscilloscope’s scanning is switched on when the white line appears and disappears, fixing the signal from the photodiode.
We tested the system on an ordinary CRT-display (Samsung SyncMaster 750s and CTX PR705F). That’s how the phosphor lights up and fades out at 85Hz screen refresh rate (the oscillogram was taken against a simple white background, but the phosphor under the photosensor was highlighted by the electron beam at a refresh frequency, so there were periodical “flares”):
First of all, we checked the reaction time of the photodiode. The rise edge is about 400 microseconds long, which complies with the calculated reaction time (the maximum operational frequency is 5…10kHz, considering the resistance of the shunt and capacities of the photodiode, wiring and the input of the operational amplifier). Evidently, this is enough to register processes taking several milliseconds.
After testing the system on CRT-displays, we put tried it on an LCD display from NEC - LCD 1525V. The display worked in 1024x768@60Hz mode; its brightness and contrast were set to maximum.
Pixel light-up time
Pixel fadeout time
In the oscillograms above the levels of minimum and maximum brightness are marked with horizontal lines, and the time of transition across 10% and 90% brightness – with verticals. Thus, the light-up time for NEC LCD 1525V was 48ms, the fadeout time – 10ms, making the total of 58ms.
Now, that we have discussed in detail our methodology, let’s go over to the tests. All the tested displays worked in 1024x768@60Hz mode with the maximum brightness and contrast. Ten oscillograms were taken for each display (five – with pixel light-up and five – with pixel fadeout). The results were then averaged.