Hi everone.
I think the same that Peter.
And Sirinivas in the original post don't says anything about extenal ROM then i supose that not exist.
Sirinivas don´t explain the sice of him chips and I supose is an unique chip.
In other terms if i do an entire decode i need more than two or three '138. It is the same if repeats the adresses in 0000h and 0020h or 32K/4 times together.

Other Connection with a single 138

CS- to gnd
CS- to gnd
CS to psend (ok Phillip it is nicer)
A to A15
B to Px.x1 (the function of A16)
C to Px.x2 (the function of IO/M)
Q0..3 to CS- of four 8255 or another chips
Q4..7 to CS- of 32k RAMs

The unique problem is in this conection you must to change A16 to select the different 8255

Félix.