I have to create a test harness to allow automatic testing of a system that I'm working on. It needs several hundred digital inputs and several hundred digital outputs. They don't need to be isolated or anything - just logic input/outputs. I don't mind much what it looks like or what the form factor is, or how it interfaces to the PC (USB, PCI, ISA etc.) - I just want to minimise the cost. Can anyone suggest the lowest cost way of adding a lot of digital I/O to a PC?
Thanks - Rowan
Lowest hardware cost? Probably a bunch of shift registers connected to the parallel port and bit-bang. It will be slow and you need to do your own software, but the hardware will be cheap.
-- Stef (remove caps, dashes and .invalid from e-mail address to reply by mail) "Roman Polanski makes his own blood. He's smart -- that's why his movies work." -- A brilliant director at "Frank's Place"
at
ed
or
of
Like Stef said, if you want to make your own boards, you can use a variety of chips. I think I would consider I2C or SPI controlled I/O chips. There are a number of them available up to at least 16 I/Os per chip with control over the output state as well as being able to tristate them. Probably the cheapest way to get I/Os is to use MCUs with a high pin count and low cost. They could be treated like "smart shift registers" but with a lower cost per I/O. Selecting a device that has a lot of I/Os and little memory should minimize the cost per I/O.
FPGAs could be an option, but I don't think they will be as cheap per I/O as MCUs. But they might be worth a look. The low end members of the newest families could be in the running. With an FPGA you get a lot of flexibility for specialized interfacing.
Rick
at
ed
or
of
Dunno it it would be in your budget, but you could get one of these;
then stack some of these on it;
but probably $$
at
ed
or
of
We are doing something similar, perhaps we can work together. We are using a 40 pins multiplexer with 26 I/Os driven by 4 lines: serial output data, serial output clock, serial input data and serial input clock. We only need one chip in each unit, but you can drive multiple chips from the same micro. You can tie the data lines together, so each additional chip only need 2 clock lines.
email: linnix at live dot com
USBmicro.com USB to SPI devices connected to a lot of shift registers. Chips like the 74HC595 and 74HC165. The SPI devices connect in seties and banks of them can be selected with different chip selects.
The device also interfaces to I2C (and LCD, stepper motor, 1-wire devices) and is for sale here: circuitgizmos.com
Nowadays, I wold opt for tying into an IDE card to get off the bus and into "user land". No, don't create a "disk device" but, rather, just write to those I/O ports and hang your device off that 40pin ribbon.
Printer ports are going away (gone?).
IDE is as well -- but you can still find PCI IDE cards. A tiny bit of steering logic and a kludgey protocol to talk to the IDE addresses should be a weekend project.
Supporting it on ________ operating system will be the bigger problem (you may be best served by using an open source "DOS" and dedicating that machine to that purpose.
at
ed
or
of
You forgot to mention a) The R/W bandwidth needed b) The physical separation of the nodes, and the physical grouping.
A simple and effective expansion, is a serial port in a 'twisted ring' loop. The master sends with the 9th (parity bit) set 99% of the time, and the slaves look for a parity-rise, and then remove XX bytes, and echo XX bytes with the parity cleared, else they simply echo. As you propagate around the loop, parity-low widens, and you can insert/remove nodes and they self-locate.
If you need more bandwidth, then flip to something like a FT2232H in SPI/JTAG mode, and use a chained SPI block.
This can be HCMOS, or CPLD, with CPLD being more flexible.
-jg
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.