Filter-on-diode technology reaches maturity

In many applications, a wavelength-selective filter is used in conjunction with a photodiode to discriminate the signal of interest from background light. A typical silicon photodiode has a broad spectral sensitivity from approximately 200 to 1100 nm. For example, in a barcode-scanning system intended for use in a brightly lit store, a narrowband filter might be used to distinguish the laser light from the ambient radiation. However, including a separate filter in the design can increase system

Th Lfw25425 31

Filter-on-diode technology reaches maturity

Depositing a

multilayer filter directly onto a

photodiode results

in a more compact and economical device for high-

volume applications.

Roger Forrest

In many applications, a wavelength-selective filter is used in conjunction with a photodiode to discriminate the signal of interest from background light. A typical silicon photodiode has a broad spectral sensitivity from approximately 200 to 1100 nm. For example, in a barcode-scanning system intended for use in a brightly lit store, a narrowband filter might be used to distinguish the laser light from the ambient radiation. However, including a separate filter in the design can increase system size, weight, cost, and mechanical complexity, as well as allow ambient light leakage.

For this reason, manufacturers have attempted for some time to produce integrated devices in which the optical coating is deposited directly onto the detector surface. Advanced Photonix Inc. (Camarillo, CA) developed the Filtrode technique for mass-producing filter-on-diode devices (see photo). It delivers the performance necessary for the majority of detection applications, as well as the economy required by high-volume users.

Traditional photovoltaic operation

To understand the problem, it is useful to briefly review photodiode operation by looking at a standard photodiode (see Fig. 1, top). Typically, a relatively thin region of p-type silicon is diffused into the top of an n-type region to create a p-n junction. Excess holes from the p region and excess electrons from the n region diffuse across the junction until an internal potential, or natural bias, is set up in the device, which counteracts further charge movement. The resulting zone, in which electron-hole recombination has substantially reduced the number of free carriers, is the depletion region.

When light of sufficient energy is absorbed in a photodiode, electron-hole pairs are created. These charge carriers reach the depletion region by random diffusion and are quickly swept across by the internal field. If the two sides of the device are electrically connected, this charge movement can be sensed as a current that is proportional to the incident light intensity. This response is referred to as photovoltaic operation.

The size and placement of the depletion region determines the speed of the device. Any carriers created directly in the depletion region are quickly swept across, producing the fastest possible signal. Conversely, carriers created in the bulk device must first diffuse into this region. For high-speed applications, the device is operated in the photoconductive mode. Here, an external reverse-bias voltage is applied to the device. This bias combines with the natural internal bias to produce a much larger depletion region. In conjunction with a very thin p+ layer, this design ensures that almost all the carriers are directly created in the depletion region, resulting in the fastest possible response. Eliminating diffusion losses of carriers also increases device linearity. However, the trade-off is that the higher bias causes some current leakage across the junction and increased dark noise.

Photodiodes are manufactured using conventional semiconductor-device-fabrication technology. A silicon wafer is polished on one side, and lithography is used in conjunction with diffusion to build up the required structure. Insulating regions of silicon dioxide and a single-layer antireflection coating are then applied. The antireflection coating is necessary because the polished silicon is naturally very reflective due to its high index. Part of the antireflection coating is subsequently etched away, and a metallic front contact is deposited to enable electrical connection. The backside of the wafer also is metallized to form the second electrical contact.

Unfortunately, the techniques used for depositing the single-layer antireflection coating on standard photodiodes are not capable of producing complex, multilayer films that are required for spectrally selective coatings, such as bandpass, long-wavelength pass, and short-wavelength pass designs. As a result, these types of filters are incorporated as separate, discrete components in systems where they are needed. Obviously, this limitation makes overall package size larger and requires the inclusion of mounting hardware. Therefore system complexity and costs are greater, and additional steps are needed in assembly to insert and align the filter. This limitation also makes the final product more sensitive to shocks and vibrations that can cause misalignment between the filter and photodiode.

Photodiode manufacturers have attempted to deposit evaporated multilayer dielectric coatings onto photodiodes in the past. But hard coatings, which have the most-desirable environmental characteristics (such as thermal and humidity resistance), are nearly impossible to etch away to clear space for the subsequent front-contact metallization. Masking off the electrode area during coating deposition is not practical because filter performance degrades around the mask edges. Conversely, soft coatings, which can be etched, do not provide the physical durability and environmental attributes that are needed for most real-world applications.

Filter-on-diode design

One novel solution to this problem is to literally turn the device upside down in the Filtrode configuration. In this filter-on-diode design, light enters the device from what is normally the rear side (see Fig. 1, bottom). The other side of the photodiode (traditionally the front surface) has been altered slightly so that both electrical contacts can be made from this side. This alteration eliminates the need to etch through the coating to deposit the electrodes and thus allows the freedom to use sophisticated coating-deposition technology.

While the geometry of rear illumination greatly facilitates production of the interference filter, it also has an impact on photodiode speed. If the photodiode is merely inverted, most of the light will be absorbed in the thick n-type layer, with the resultant charge carriers having some distance to travel to the depletion region. Furthermore, n-layer carriers are holes, which diffuse at one-third the rate of electrons. While high speed is not a requirement in all applications, the goal was to develop a device with the broadest possible application.

The speed problem was addressed in three ways. First, a high-resistivity silicon is used, along with a reverse-bias (photoconductive) operation. This approach ensures a deep depletion region. At the same time, these photodiodes are designed with a thin n layer. As a result, most photogenerated carriers are produced in or very near the depletion region, maximizing speed.

However, there is a subtle trade-off in choosing just how thin to make the n layer. A thinner device has higher capacitance and, hence, a larger resistance-capacitance time constant, which slows it down. Thus, there is an optimum thickness that is about one-half that of a typical photodiode.


Besides enabling the practical production of filter-on-diode devices, this configuration also has a significant positive impact on the assembly process and utility of the final, packaged device. Instead of using the high-end TO can package with an integral or separately mounted filter--which is traditional for low-volume production photodiodes--the filter-on-diode chip can be mounted in a lead frame with a hole in it. The lead frame is a thin piece of Kovar to which the chip is bonded using optically clear epoxy and then wirebonded, thus allowing both rearside illumination and frontside electrical connections. This arrangement in turn is encapsulated into a clear mold to allow light transmission through the final package (see Fig. 2). Because this same technology is used for packaging many other types of electronic components, automated assembly equipment for economic volume production is readily available.

With both electrical contacts on the opposite side from the illumination, the photodiode dies also can be mounted in a flip-chip configuration on lead frames with encapsulation. The popular flip-chip package is useful in minimizing the overall size of an assembled circuit board and components and makes these devices attractive to designers concerned about overall package size and assembly cost. In other flip-chip configurations, the dies can be mounted directly to printed-circuit boards, ceramics, and flex cables with conductive epoxy or solder.

Coating performance

The coatings used to date are ion-assisted, evaporated multilayer films. These coatings are dense structures that are mechanically robust and exhibit very high resistance to cycles of temperature and humidity.

Using this technology, bandpass, long-wavelength pass, and short-wavelength pass filters can be made over the spectral range 340 to 1080 nm, a good match for the response of the silicon itself. Designs produced so far exhibit in-band transmissions of greater than 80% and out-of-band rejections of about 99.9%. The narrow re sponse curve of a photodiode with a 670-nm laser-diode bandpass-filter response would allow laser detection in the presence of background light (see Fig. 3). Furthermore, coatings for 0° angle of incidence can usually be used at up to 40° incidence with only minor shifting of the central wavelength.


Potential uses for photodiodes with a directly deposited multilayer filter include virtually any medium-speed, volume application where filtering is required and size or ruggedness are a consideration. A prime example of this type of application is the hand-held barcode scanner, in which the filter is a bandpass design. These photodiodes also are well suited for use in infrared communications, rangefinding, and a variety of biomedical instruments, such as those used for urinalysis and blood-chemistry analysis. The filter-on-diode configuration even enables bidirectional detection or the detection of two different signals simultaneously (on the front and rear sides).

Overall, the market for this type of device can be expected to grow rapidly because these components satisfy two important OEM design trends: they reduce system size and complexity, and they integrate seamlessly with the latest in electronics-miniaturization technology. o

Th Lfw25425 31
Click here to enlarge image

The filter-on-diode design integrates an optical coating directly onto the diode surface. Encapsulating the photodiode device into a clear mold allows light transmission through the final package.

Th Lfw25425 32
Click here to enlarge image

FIGURE 1. In a standard photodiode, a single antireflection coating is deposited above the active region (top). Turning the diode upside down in the Filtrode configuration permits a multilayer filter coating to be applied (bottom).

Th Lfw25425 33
Click here to enlarge image

FIGURE 2. Rearside illumination and frontside electrical connections are allowed by mounting the filter-on-diode chip in a lead frame with a hole in it (top). The flip-chip configuration minimizes the overall package size and assembly cost (bottom).

Th Lfw25425 34
Click here to enlarge image

FIGURE 3. With a 670-nm bandpass filter, the Filtrode photodiode shows a narrow response that would enable it to detect a 670-nm laser diode in the presence of background light.

ROGER FORREST is the engineering manager at Advanced Photonix Inc., 1240 Avenida Acaso, Camarillo, CA 93012.

More in Detectors & Imaging