Thread: RS232 master/slave unit

So basically, you want to re-create the 6 channel multiplex/demultiplxer in the green box???

Why not use RS485.

Do you intend to create your own switching protocal? If so, what is it?

I'm sorry if I'm not getting it. To switch 6 circuits, you'll need 3 bits minimum. The protocal can be something like this:

Slave 1 - 000
Slave 2 - 001
Slave 3 - 010
etc.

Just switching to each slave using the first 3 bits of a packet would be pretty simple. You simple need to capture the first 3 bits of each packet and feed them into a de-multiplexer. The next bit can be the control bit, and then the receive can just ignor the rest. It can be done with simple digital chips, you just have to work out the logic.

Hi BrownOut, does the attached drawing help?
It would make decoding at the other end very easy

Oh! Most Important, data comms is only one way i.e. Master to Slave.
No reply path is required.

Then I think you just need a register to hold each of the slave bits, and connect slave to it's respective register. If you use a UART, just connect each slave to one of the 8 data i/o's. If your using something else, just do a serial to parallel conversion, and connect the slaves.

Don't feel bad. I've spent most of the last 20 years writing HDL code. I've forgotten most of what I used to know about real electronics

I was thinking about your project last night. I'm interested that you want to try to make something yourself instead of using a URAT. I've thought about that too, because I think you learn more by doing it yourself. Conceptually, it's a simple problem, you only need to sense the start pulse and then use a linear shift register to capture the serial data and do the serial to parallel conversion. The only trick seems to be to get the clock correct for capturing and shifting the data. But I think I have a solution.

The idea is to start the clock when a start pulse is received, but the transition sould occure in the middle of each bit period. That allows the maximum tolerance for drift. In the attached diagram, two one-shots are connected back-to-back for clock generation. The first is triggered by the start pulse, and produces 1/2 the data period before firing the second one. 52uS was chosen for 1/2 the period for 9600BPS. The clock continues to be generated for the duration of the transaction. You'll have to somehow keep track of the bits received so you can disable the clock at the completion of the transaction. That should not be hard to do. Then, all you need is a shift register, and you're done!

The connection to the first one-shot should be a pulse that is properly oriented to 'fire' it. If you're careful how you connect it, you could theoritically resychronize the clock at every data transmission, although you shouldn't need to.

Good luck!

Top Cat,

Do I understand your circuit correctly? You have a master transmitter that encodes 8 signals into a byte to be transmitted via a network to a (one) slave unit that will decode this byte into the respective 8 signals, each driving a load?

It’s not a RS485 network scenario where you have one network master commanding 8 individual nodes scattered over the network?

I’ve just completed the latter system. I wrote my own protocol to talk to max 8 (eight) slave nodes. The master in turns commands a node and waits for the feedback from that node, before commencing to the next node, always repeating for all 8 nodes.

Sorry, but buy a programmer and use PIC's, it will be a LOT simpler, a LOT cheaper, and a LOT more functional - it's almost 2010, why use 1970's technology?.

How can it be alot cheaper? 4000 series logic goes for about a dime a throw
How can it be simpier? It's already about as simple as it can get.