If all 100 boards had 10K pullup resistors on collectors of the transistors the net equivalent resistance of the wired or line back to the master would be about 100 ohms. Might work at a slow baud rate. How long can you afford to spend having the master do comms at 300 or 600 baud,,,,to 100 devices.

If you only had one 10K pull resistor at the master the rise time of the signal would be very long as the 10K tries to charge all the capacitance along the wiring to 100 devices. Might work at a very slow baud rate.

Without a buffer from the output of the master how in the world is the Tx line ever going to drive 100 loads that may extend quite a long distance. Even if all boards were packed neatly into a rack with 3 cm spacing your looking at 300 cm of distance....and any real cabling would be at least 2 times longer. Have you computed the actual capacitance of such a lash up and anticipated if the MCU chip could drive it???

100 total boards implies either a rather good sized power supply with quite a power distribution system ..OR.. a whole bunch of separate power supplies. Have you even begun to evaluate the common mode ground variations you will experiece along a buss wire that makes its way to 100 stations? Can you keep this common mode for all 100 connections to less than 0.2 volts or so so that the logic levels at the unbuffered RxD pins of each station will be reliably near GND when the signal is low?

No !! Its not a case of wanting to every time design a system to standards and using flashy parts. It is a need out of necessity of what it takes to get a communications link to WORK.

Why do you think you can kludge together a micro-controller to microcontroller serial buss for 100 devices without buffering and get it to work. I think it is your lack of real analysis of the problem and its design requirements that make you think it is as easy as hooking two things together on the lab bench.

Why do you think that RS422 parts (which can be used in tri-statable multidrop) or RS485 exist in the market place. Are they simply some chip marketers wet dream? No way,,,,they exist because this is what is required to make high node count networks work!!

So what if you have to provide for three or four RS485 repeaters to get a 100 node network to function. The cost of these is miniscule when ammortized across 100 stations.

You quipped about using Notebooks with TCP/IP. Even here you would need some form of "repeater". If the network was CAT-5 Ethernet I could anticipate that you would have on the order of six or seven buffered 16-Port Ethernet hubs or switches to make all the connections!!

Michael Karas