Getting started with USB

blinky465 · Joined: 03 Jan 2006 Posts: 8

I'm planning a usb-driven 240v switching device, which is to eventually be software programmable (a bit like a really simple time-based PLC).
I'd like to make the device(s) link together similar to a daisy chain - where one device can either plug into a PC (as a master device) or another device of the same time (so it acts as a slave device).

I've put together some ideas at http://www.multiedge-net.co.uk/usb_outline_brief.zip about the daisy chain effect and how it would be handled in software.

The theory has held up in the past, when I've linked Siemens Logo PLC units together via RS232. Has anyone any experience of something similar, or does anyone have any comments?

Thanks, Chris

asmallri · Joined: 12 Aug 2004 Posts: 1635 Location: Perth, Australia

This is not likely to work. There are a range of problems such as two diode volt drops per slave device, USB max cable length, loss of signal integrity with each PCB / Cable layout.

There are a range of alternatives such as RS485, RS422 that will provide you a bus based solution. My preferred approach would be to use Ethernet. This is a relatively low cost but extremely scalable technology.
_________________
Regards, Andrew

http://www.brushelectronics.com/software
Home of Ethernet, SD card and Encrypted Serial Bootloaders for PICs!!

blinky465 · Joined: 03 Jan 2006 Posts: 8

I'd heard that there was a limit to the maximum length to USB connections, and assumed there would be signal loss of distance. I like the idea of reusing the USB sockets on each device, instead of using a "hybrid" connection method; my original plan was to include RS232 serial connectors as well, but then the USB socket would be redundant on all except the "master" device.

Would the same design work if a signal booster or USB repeater or similar was included in every device, on the outgoing signal ?

asmallri · Joined: 12 Aug 2004 Posts: 1635 Location: Perth, Australia

blinky465 · Joined: 03 Jan 2006 Posts: 8

Sorry if I've oversimplified things - I'm just getting used to miniaturing things and replacing relay switches with transistors, and PLCs with microchips!

Would introducing a couple of transistors (as you would relay switches on a "scaled up" model) allow the signal to be boosted? Would transistors be quick enough for low-speed USB1.0 signalling?
(updated document http://www.multiedge-net.co.uk/usb_outline_brief.zip)

Or should I just bite the bullet and stick to RS232/serial comms between the devices and have the first device in the chain convert this data to send it back to the PC via USB?

(the second, third and subsequent devices use their USB ports as "pass-through" connectors for sending data back to the PC).

Ttelmah · Guest

No.
USB, is bidirectional. Each wire can be carrying signals in both directions, according to where in the 'transaction' you are. Hence you cannot just add transistors.
Swtching to RS232 or 485 (preferable to my mind), would be a sensible solution.
You won't be using the USB ports at all on the latter devices in this case.

Best Wishes

blinky465 · Joined: 03 Jan 2006 Posts: 8

Looks like it's back to the drawing board!
Thanks for the help - I'll post back here when I've something else drawn up....

hillcraft · Posted: Wed Jan 04, 2006 10:08 am

USB is a pain.

Why don't you use CAN-BUS. I have created an task specific Industial PLC that communicates with the PC via RS232. The controller then communicates with 1 - many slave units. One can of course change the RS232 interface to a USB interface. My system does not need to be connected to a PC all the time so using RS232 is fine.

blinky465 · Joined: 03 Jan 2006 Posts: 8

Guest · Posted: Wed Jan 04, 2006 11:13 am

RS232 RX and TX lines are unidirectional. You must remember that you shouldn't try to use USB in real time via HID. Windows services HID when it feels like it.

Yes RS232 via a laptop is a problem. USB to RS232 bridge controller also have their issues. They cannot support non-standard baud rates and they tend to fall asleep. I have ended up purchasing a PCMCIA RS232 card to get around these issues.

You must also remember that CAN-BUS has proved to be very robust in Automotive and Industrial applications. Microchip / CCS support for CAN-BUS is superb. Another thing is that CAN-BUS already has the worst of the signalling / protocol issues sorted.

My vote is without a doubt for CAN-BUS.

blinky465 · Joined: 03 Jan 2006 Posts: 8

kender · Joined: 09 Aug 2004 Posts: 768 Location: Silicon Valley

Speaking of FTDI, about a month ago they cme ot with a new chip. It has an on-board EEPROM and oscillator. It costs $4.50, if i remember correctly. The package is 28-TSSOP. I haven't tried it out yet, but chips ank EV-kits are available.

blinky465 · Joined: 03 Jan 2006 Posts: 8

When I've used signal lines for low voltage control, "echo and ignore" is nice and simple for non-critical apps. If I go back to the idea of using serial communication (forgetting about USB for now) would the design at http://www.multiedge-net.co.uk/rs232_outline_brief.zip hold up?

kender · Joined: 09 Aug 2004 Posts: 768 Location: Silicon Valley

The system you are describing in the text sounds like a ring, where secondary the node just re-transmits the message, unless this secondary itself is an addressee. The drawings, you show, look like a linear bus. So, I might be somewhat confused. However,..

The RS232 system (bus) you have depicted will work fine for the one-way communication from the master to the secondaries. You will not even need the diodes for this. However, is several secondaries try to talk to the master at the same time, you will get a bus collision. Unfortunately, RS232 itself doesn't have the methods for the arbitration of collisions, because it was designed for point-to-point communication.

My advice would be to go with CAN. You can even avoid writing the firmware for the secondaries, because you could use SLIO chips such as MCP25050.

asmallri · Joined: 12 Aug 2004 Posts: 1635 Location: Perth, Australia