Backplane Switching

[MLK18]

I am trying to find a way to switch between boards connected to a backplane.

Here's what I'm trying to do:

I am automating the test for a certain board that is attached across a backplane to a CPU board. Because of the simplicity of the test, I'd like to be able to attach multiple boards to the backplane at once, but be able to switch between them. Each board has a fifty-pin connector that plugs in to the backplane.

Is this possible? I realize I could use 50 latching relays per connector, but they take up quite a bit of real estate and at 3+ dollars apiece it gets rather expensive. Is there a better way to switch all fifty signals?

Thanks for any advice!!

 

[VEBill]

With modern standardized 'backplanes', the signals are sufficiently high frequency and with optimised propagation, that you can't easily switch beyond any board-enables inherent in the spec.

If it is just some work-a-day homebrew 'backplane' then it could be a lot easier.

For production level functional test, one might simulate the whole backplane from scratch so that a complere function test could be performed.

What is the backplane? VME? ISA? S-100? ;-) ...

 

[VEBill]

typo: 'complere' := complete

 

[MikeHalloran]

If the test is simple, why do you need a 50 pin backplane and a motherboard?

Couldn't you just put a high cycle 50 pin test socket on a generic embedded computer, and run the test on that? Make a couple of them, and just clip them onto the boards to be tested.







Mike Halloran
Pembroke Pines, FL, USA

 

[BrianG]

You don't say what kind of product or system these particular boards are for but I would guess that if you have a separate CPU board and a backplane you probably have other board types that it "talks" to over some sort of common bus on the backplane.

If the various board(s) use a common bus system to connect to the CPU then they would need to use tri-state drivers on the outputs so that different boards do not "fight" each other. This also means that the boards must be individually addressable so that only the desired board "connects" at a time. This means that you should not need a whole load of connections to be switched between them since this is what the tri-state drivers are for.

I suggest you investigate the boards / bus architecture and also see what facilities are available within the CPU software and automated test system for configuring different addresses.

 

[IRstuff]

As a general rule, if you put and hold a particular board in RESET in a VME chassis, it should go tri-state.

TTFN

 

[MLK18]

A little clarification:

When I said this is a simple test, I was comparing it to some other boards that I am also resposible for testing. A big portion of this test involves downloading firmware to the circuit board, which requires it to be linked to the CPU. However, due to the nature of the this particular board, only one board can be linked to the CPU at a time.

Originally I thought I would be able to just switch the power pins on the connectors on and off, and since the board wouldn't have power, the CPU wouldn't recognize it. A potential problem with that is then there would be a bunch of "dead" circutry in between the next board and the CPU, which could possibly lower the signal integrity enough that the CPU and the second (or third) board wouldn't be able to communicate.

It seems that I really would have to switch all fifty pins.

Also, I'm not sure exactly what you mean by "tri-state" but I don't think these boards have that functionality. This is no "homebrew" backplane, but it does have signal termination and other more sophisticated things than just fifty wires running to fifty other wires.

What'd be really nice is a fifty-pull single-throw relay... Not that that's realisic!!

 

[VEBill]

"...downloading firmware to the circuit board, which requires it to be linked to the CPU."

Modern system typically are using JTAG for this purpose. The JTAG interface can be used during production, during test, and even in the field.

 

[IRstuff]

A commercial board should be using a standard VME interface, which should tristate, e.g., go to a high-impedance condition, which should allow a different board to take control of the bus.




TTFN

 

[geekEE]

Is this a standard or a proprietary bus?

If it's a standard bus, there may be extender boards available that already do what you want.

Assuming it's a proprietary bus, do you want to hot-swap the devices under test?

If it's not hot-swap, you might just be able to ignore the lines that are inputs and power pins and just switch the outputs and bi-directional signals. Maybe some analog switch chips would work.

If it's hot-swap, then that's a whole different can of worms. Sequencing the power makes it more complicated.

 

[VEBill]

What would be cheaper?
1) Engineering and building a complicated Nx50 switching system.
2) Or obtaining 'N' additional CPU cards and backplanes.

Answer depends on 'N' (the number of parallel boards). If 'N' is low, probably cheaper to parallel the entire chassis.