Page 16 - Integrated Wireless Propagation Models

XIV P r e f a c e

deviation of prediction error is within ±8 dB. In small cells, the prediction error should
be ±5 dB.
Measuring signal strengths in a large area is costly. Therefore, we may depend on
many prediction charts on a specified coverage generated roughly by area-to-area
models. Each chart represents a specified environment. Then the Lee model derives a
point-to-point model using a terrain map and the antenna effective height to fine-tune
the signal strength. The point-to-point prediction model is useful when cell size is large.
When cells become small, measuring signal strength in a small area is considerably
easier and less expensive; we can collect more empirical data in a particular confined
area. Thus, the prediction model becomes more accurate based on the empirical data.
How was the Lee model created? In 1976, engineers at Bell Labs were discussing the
deployment of a cellular system in an area without doing a massive measurement in
that area. Then I wrote an internal memorandum about my point-to-point prediction
based on Okumura et al.'s legendary paper in 1968, which was the first paper to provide
the ground work for an area-to-area model. At that time, I had to prove to Bell Labs
engineers that the tool was working. Initially, in 1976, I had requested engineers from
the Tri-State Telephone Team (which consisted of New York Bell, New Jersey Bell, and
Connecticut Bell) to assist me because they could provide me with the measured data
that I wanted in their territory. Also, they had the manpower to generate the coarse
terrain charts I wanted. First, I had used the commercial 50,000:1 scale topographical
maps printed by the U.S. Defense Mapping Agency. Each map was roughly 5 by 8 miles
(8 by 13 km). The contour lines were in 20-ft increments. To make plotting a terrain
contour along a mobile path on a map easier, we generated grid maps. Each grid was
about one-half of a square kilometer. Each map had 18 by 24 grids. The Tri-State
engineers used an eyeball average of the terrain altitudes in each grid. Each grid had
one value. Each map had 432 (18 by 24) values, and one in each grid. It was a tedious job.
After the grid maps were generated, we started to pick the routes in the Tri-State
territory for getting the measured data. Every planned route on a particular map was
drawn, crossing the grids where the route laid to make the signal strength measurements
on that route. The altitudes along the route were plotted according to the eyeball
average altitudes. Once the transmitter antenna was set up with its known location,
height, power, and gain, as well as the height and gain of the mobile unit, the Tri-State
engineers passed information on to me. I was the one to use hand calculations according
to the eyeball values in the corresponding grids on the terrain contour map to plot the
predicted signal strength curve along the mobile path and gave that to the Tri-State
engineers before the measurement. The next day, the Tri-State engineers took the
measured data on the particular planned route and compared these with my predicted
values. I never went out and participated in taking measurements. Therefore, it was a
totally independent process between the measurement and prediction.
The process went on for more than a year, and the predicted values were very good
compared with the measured data in general. Not only was my prediction tool good,
but the eyeball average of altitudes also turned out to be a good method. After more
than a year of doing hand calculations to predict signal strengths in many different
Tri-State areas, I was like a computer. At that time, I was young and had a lot of energy,
and I was not afraid of my colleagues teasing me.
Finally, Bell Labs decided to use my tool in 1978. The eyeball average proved that
the coarse grid contour maps were adequate. Then the 1 o by 1 o tape scaling 250,000:1
)
was purchased from the Defense Mapping Agency Topographic Center (DMATC .