Silicon photo-sensors have been in electronic circuits since the inception of the era of silicon electronics. More than likely the photo-sensing characteristics of silicon were quickly discovered in the lab., as the scientist worked from daylight hours into the evening. More than likely the first generation silicon IC chips were designed to respond to electrical signals, not light. Imagine the surprise of the designer when it was discovered that the silicon chip also responded to light. To this day IC designers regularly cover their wafers-under-test to shield out extraneous light. This is done so that the real performance and characteristics of the chip can be assessed. Although the light sensitivity of silicon is viewed by most IC designers as an undesirable by-product of the silicon, systems designers have exploited this transfer of light into electrical energy in various applications. Consequently, there are a wide variety of applications that use silicon to sense the intensity and characteristics of light.

In these systems a silicon sensor converts light into charge, which can be defined as an electrical current in the time domain. These silicon sensors are some of the electronic world's "eyes" which can be used to analyze blood, non-invasively search for tumors, detect smoke, position equipment, or perform chromatography; to name a few applications. The phenomena of converting light into charge is fairly well understood. The real challenge put before the systems designer is how to convert the low-level currents from the photo-sensor into a useful electrical representation. To even further exasperate the difficulty of the design, the level of accuracy in these applications is continually increasing.

Several analog front-end circuits have been created in an effort to effectively capture the small-level signal that is generated from the photo detector in these applications. The classical design topology is a hybrid solution that starts with the transimpedance amplifier and uses a high-value resistor in the feedback loop of the operational amplifier. This circuit design uses resistance to provide a real-time linear representation of the light source. Another alternative, that was developed early on, used a logarithmic amplifier. This solution is best suited for light sources that have an extremely wide dynamic range of luminance. In contrast, other devices such as the integrating amplifier or a sample-and-hold type topology have been used. Both of these solutions convert the output signal, whether it is described in amperes or pico coulombs, of the photo detector into a voltage represented by the charge on a capacitor.

The signal at the outputs of these analog, front-end circuits usually requires a multi-pole analog filter and/or an analog gain. In this manner, the combination of the input stage and filtering stage separates the signal of interest from the noise floor. After the analog filter/gain, the signal is digitized, usually by a sampling ADC.

A third design approach that has been developed digitizes the photo-sensor's charge directly. This is done with a monolithic-charge-digitizing ADC. Internally, this monolithic ADC topology is not much different than the design architecture of switched-integrator followed by a discrete ADC. Given these various topologies for photodiode signal conditioning circuits, the objectives of the overall design still prevail. These high precision circuits must be designed with ample bandwidth, low noise (to give good digital accuracy) and high system stability.

Transimpedance Amplifier

Besides the inverting and non-inverting voltage-gain type configurations around an operational amplifier, the transimpedance amplifier circuit is second-to-none in terms of popularity. Most typically it is used to solve precision, wide-band current-to-voltage conversion problems. In this circuit, the photodiode is placed across the inputs of an operational amplifier. A high value resistor (500 k to 100 MOhms) connects the inverting input of the amplifier to the output. Finally, the non-inverting input is referenced to ground (see Fig. 1 below.) If the photo-sensor across the differential input pins of the amplifier is excited by light, electrical charge is generated. The only path of escape for this charge is through the high value resistor in the feedback loop of the amplifier.

Fig. 1

This simplistic approach is not without its design challenges. First, the operational amplifier must have a relatively low input-bias current, in the pico-ampere range. An appropriate amplifier for this circuit would have a FET- or CMOS-input stage with low-voltage noise and offset-voltage specifications. In the end, the transimpedance amplifier circuit must be optimized for stability, bandwidth and low-noise performance. A great deal has been written about the nature of this type of design. The final design solution is not always intuitively obvious. The combination of the photo-sensor, operational amplifier and amplifier-feedback element (plus their respective parasitics) create quite a rat's nest of formulae for consideration. However, it could generally be said that these types of circuits are best suited for photo-sensors with relatively high-level output signals and for applications that require wider-bandwidth performance.

Switched Integrating Amplifier

The switched integrating amplifier circuit offers another alternative to this application problem. In contrast to the transimpedance design, it reduces the analog circuitry and brings the digitization processor closer to the photodiode. The switched integrator, as shown in Fig. 2 below, uses an operational amplifier with a fairly small capacitor (10 to 500 pF) in the feedback loop. This is combined with an array of closely-matched analog switches. The analog switches can be implemented discretely, however, mismatches can cause severe errors as a result of charge injection. Consequently, the monolithic version of the switched integrator is more desirable. In terms of the general operation of the device, the photodiode output charge is collected on the integration capacitor (Cf) while integrating switch, S1, is closed. When S1 is opened the output of the amplifier stops at a dc voltage level that represents the total charge that has been collected on the integration capacitor. Once the output of the switched integrator is digitized with a discrete ADC, the reset switch, S2, is used to discharge the integration capacitor bringing the output of the amplifier to zero volts. When this is accomplished S2 is opened and the integrating switch (S1) is closed starting the process over again. The analog gain of this input stage is changed with values of the integration capacitor or increase in integration time.

Fig. 2

Charge Digitizing ADC

In today's market, it would be reasonable for the system designer to ask for a monolithic solution to this problem. This type of IC chip should have a photo-sensor-ready input, digital output (20-bits would be OK) and software-gain capability. The good news is that this sensing technology has not escaped the trend towards higher chip-integration. The complete monolithic solution, by definition, converts the photo-sensor's output charge to a voltage, filters that signal and gives the user a final digital representation (see Fig. 3 below.) This type of monolithic device combines the analog integration, oversampling, correlated-double-sampling, and digital filtering producing precision results.

Fig. 3

With this circuit, the photo-sensor is excited by light. Initially, the charge is steered towards the top integrating amplifier (by S1) for a predetermined period of time, tint. At the end of the integration time, S1 changes so that the bottom integrating amplifier is directly connected to the photo-sensor. In conjunction S3 connects the top integrating amplifier to the delta-sigma ADC. While the bottom integrating amplifier is collecting charge from the photo detector, the dc output of the top integrating amplifier is being converted into a 20-bit digital word. Once this dc signal is converted, the reset switch (S2a) of the top integrating amplifier closes, bringing the output of that integrator to zero. Meanwhile, the bottom integrator has been collecting charge across the capacitor in the amplifier's feedback loop. At the end of a predetermined amount of integration time, tint, S1 is toggled to connect the photo detector to the top integrating amplifier. Concurrently, S3 is toggled to connect the bottom integrating amplifier to the ADC. At this point, the dc output of the bottom integrating amplifier is digitized while the top amplifier is integrating the photo detector's charge. Once the dc output of the bottom integrating amplifier is digitized, S2b is closed in order to discharge the capacitor (Cf2) in the feedback loop of the amplifier. Meanwhile, the top integrating amplifier has been collecting charge. And so the cycle is repeated. This system is very effective as long as the integration times are longer than the ADC conversion times.

There are advantages to using this type of solution: First, the analog input stage is already optimized for the photodiode sensor. Secondly, the ADC following the front-end analog stage has a specialized interface. There are no compromises made with this interface as a result of making it a solution for many applications. This ADC only services one application, the internal output of the analog stage on the chip. Thirdly, there is no lost charge from the photo detector. This is accomplished with the coordination of two integrating amplifiers. Another distinct advantage is that this charge digitizing ADC has lower susceptibility to EM interference, and the absence of external PCB traces between the analog stage and the ADC converter considerably lowers the risk of interference.

Watching Technology Change

Photo-sensing circuits have changed over the years. The first solution was pure analog, using the transimpedance amplifier followed by filters to reduce the broadband noise. Once the analog signal is duly conditioned in this type of circuit, it is digitized. From the classical transimpedance amplifier, the switched integrator has gained favor. The switched integrator was the first step toward bringing the digital portion of the circuit closer to the signal source. And so the migration of the photo-sensing application solution has moved on to totally integrated IC systems, such as the charged digitizing ADC. What's the next step towards integration? Will the photo-sensor be brought into the IC or will the micro-C be tacked on the back? Or will we jump on yet another paradigm that takes the eyes of the electronic system to a new level?