Example of my design solving a complex accuracy problem

Akos Szoboszlay

(c) 2014 Akos Szoboszlay

The “given” was to use an Avalanche Photo Diode (APD) because the light signal intensity was very low, and APDs have very high gain in a small size, but that gain varies greatly with temperature and voltage. The traditional way to solve the temperature variation is to use a thermister at the APD, read that by an ADC and software, and then adjust the high voltage across the APD using a look-up table and DAC controlling the HV Supply voltage. Another “given” was over-all system gain accuracy. I did an error budget analysis using a spreadsheet, and found the gain spec could not be met using a thermister, which is an open-loop method, because the signal goes through several more stages with inaccuracies, and the inaccuracy of the HV supply, all adding errors, primarily due to tempco.

Therefore, I came up with this closed-loop method: 

Shine light from an LED of a constant brightness for frequent automatic calibration of gain of both APD and follow-on stages. Gain is known — measured — and is both adjusted and, if not precisely set, calculated out by software. LEDs, however, have their own inaccuracies of brightness versus current, due to aging and temperature, but silicon photodiodes are very stable. So, I designed another servo loop, with the LED and photodiode. This inner servo loop was analog only.  I selected a photodiode whose tempo became zero at the wavelength of light corresponding to the LED’s wavelength. This forced the LED light to be constant. This light went through the APD during calibration, which converted light to current and eventually the signal went to a 16-bit ADC and software. As a result, the gain of the APD and the follow-on gain stages met spec. The HV supply inaccuracies, which effect APD gain, were eliminated due to the outer closed loop. 

In addition to the electronics described above, partial mirrors / beam splitters were added to the optics.