Zero bias detector diodes for the RF/ID market - HP's HSMS-285x - includes related article on backscatter RF/ID systems - Product Announcement

Hewlett-Packard Journal, Dec, 1995 by Rolando R. Buted

Using the various equations for voltage sensitivity, it is common to plot [gamma]2 as a function of the saturation current [I.sub.s], as shown in Fig. 5, for given values of [C.sub.j], [R.sub.s], and [R.sub.L]. Since [C.sub.j], [R.sub.s], [R.sub.v], and [I.sub.s] interact with one another, it is not simple to lower [C.sub.j], say, without increasing [R.sub.s]. By using the model, we were able to select the best combination of these parameters to maximize the voltage sensitivity at a given frequency. The process model ensured that our design was within the limits of our existing manufacturing capability. In this way, we were able to minimize development costs and time to market.

The fabrication process is relatively simple. Using a heavily doped silicon wafer substrate (to keep [R.sub.s] low), an epitaxial layer is grown with tight controls on the doping level, thickness, and doping transition width. After silicon dioxide and nitride passivation, photolithography is used to define a contact window. A well-controlled metal process is used to deposit the metal, which defines many of the critical parameters. The metal is etched to an appropriate size for bonding in the plastic package. The wafer is cut into individual the and attached to a leadframe using a silver epoxy. It is then molded into the final plastic configuration. Fig. 6 shows the device cross section and the layout.

The packaged device can be 100% tested for various dc parameters such as forward voltage bias [V.sub.f] and breakdown voltage [V.sub.br] Many of the dc parameters have been correlated with high-frequency parameters, thus ensuring the performance of each part and eliminating the high costs associated with high-frequency tests.

Performance

The initial lots that were processed after being designed in the process and device simulator performed very closely to the predicted values. Minimal model changes were necessary. In fact, the results were sufficiently good that no design iterations were necessary and the data sheet specifications were set using those lots. Although we did not experience the normal kinds of process variation that we would expect over a long period of time, our confidence in the model accuracy allowed us to simulate these variations to show that the specification would still be met. In addition, we could use the model to determine what process and device parameters could be changed for future improvements to the diode. Figs. 7 and 8 show actual voltage sensitivity compared to the calculated values.

For comparison with the HSCH-3486 glass package diode, Fig. 9 shows [gamma]2 as a function of frequency. The different values of [C.sub.j], [R.sub.s], and [I.sub.s] of the two diodes cause the HSMS-2850 to provide greater performance at frequencies less than 3 GHz while the HSCH-3486 is superior above 3 GHz. Because of its simpler packaging and testing, the HSMS-2850 is much lower in cost than the HSCH-3486.

Conclusion

Hewlett-Packard's newest silicon zero bias detector diode has one of the best price/performance ratios on the market. We feel that these diodes will become an integral part of many tag applications being designed today and will be considered in future designs and technology. They provide excellent voltage sensitivity for many of the frequency ranges being used in the RF/ID industry at a very low cost.