During 1997-1999 period I was working for Philips Consumer Communications. It was a joint venture between Lucent and Philips and we made telephones. We had the sole right to sell phones under the AT&T logo. I had a magnificent apartment on the ocean in Long Branch, NJ. This was at the Sea Verge apartments.

I worked with a great team of firmware engineers. My managers were Ken Kasiske and Trey Weaver. I was on the 'Ocelot' team. We were tasked to make their very first 900Mhz all digital cordless telephone. What a son-of-a-gun this project was. The heart of the project was an 8051 based base-band chip called the Phox by AMD. We took their DECT reference design and modified it for our project. The reference software did not come fully debugged. With a little sweat my team, Grady Brosier, Bassam Chamoun, Ken Bocan and I, got it working! I was the lead on this team, I was in charge of the firmware for the phone base station, Caller ID, and communications stack. I was also the only one with significant prior 8051 experience.

The most interesting thing I remember about this job is that I found the most difficult bug I have ever encountered there in my life. Briefly I will describe what I found. The DECT protocol is an acknowledged protocol between handset and base. The state machine that controls when the transceiver re-transmits or acknowledges was getting into a wrong state. This state was represented by several variables so first we had to recognize this part of the problem. The transceiver would get in this state only intermittently. If you used the phone heavily for 15-20 minutes, it would lock up. We would set breakpoints in the software, but it took forever to reproduce the problem.

Finally, we found how the transceiver state machine could get into this state. Out of tens of thousands of lines of code an interrupt had to occur only between two sequential (tiny) 'C' statements that were setting two semaphore flags. It took us weeks to understand the state machine and weeks to trap this problem. My firmware partner and I were pressing buttons on the phone for hours to reproduce the problem.  Like I said, we finally got the phone working 100% to spec, and on time, with some hard work! The problem had come in on the reference design.

Cindy Price (our project manager) had some challenges to work through toward the end of the project on the hardware front as well. The hardware guys were led by Brian Kim and Paul Newland. Paul came out of Lucent's Bell Labs. The were (are) two of the best engineers I have ever met. I had a lot of respect for them. It seems the hardware guys found that the front end of the radio was being saturated by RF near the 900 Mhz frequency. This was an issue because if the customer lived near a paging tower his cordless telephone would not work. The solution was to include a SAW filter on the front end. A SAW is a narrow band filter.

Then end of this story is that we all got the phone out, but VTech wound up purchasing our division a couple years later. They bought the whole division so they could use the AT&T brand name and logo!