Backplanes, the good, the bad and the ugly? or, when to not use a backplane?

[magnus0re]

Hi.
I was looking at the design of some boards (they are already shipped). Produced in 100s, some expensive FPGA on them, made to be mounted in racks. 6U in height. All in the same installation.
The boards are not using a backplane.
Instead of using a backplane, there are separate power connections, I2C connections and CAN connections, all on the front.
All that really needs to be on the front is the main I/O, basically some fast serial links, and a jtag.

And I was thinking, why is it that nobody thought to use a backplane? It would make setting up the system simpler, take less time, reduce the BOM of the boards and require less cables.
It could also potentially reduce the board area required, which would make the large boards cheaper. All good things, as far as I'm thinking.

Or have I stumbled upon something? Is there a case to be made for not using a backplane? Have there been cases when using a backplane was a definite disadvantage?
For this I turn to your collective wisdom. 

[wizard69]

I have not idea what the original designers where thinking but the number one issue is that once you design a backplane interface you are stuck with it.   These can get old real quick.  Contrast this with using a CAN bus which has a set of standards to build to.  Even if the CAN bus is on a non standard connector the signals will not change over time.

Also there are other design issues.  Can you use the same connector for all of your power requirements and signal requirements.  If not you have to use multiple connectors and get all of your mechanicals right.   By the way good backplane connectors are not cheap and in the case you have noted it isn't even clear that you need a lot of pins to connect the boards.   In the end the team might have just thought that a backplane was a waste of time especially consider yet another board to design and validate.

There are literally hundreds of dead backplane designs out there.   If the designer could map his requirements onto an old backplane standard there might be some good in that case.   In this case resuse hardware for a defunct or hardly used backplane, to build a custom installation on.    But then you get into expenses again and likely many unused I/O lines.     For example use an old VME backplane for your voltage sources and I/O lines.

Hi.
I was looking at the design of some boards (they are already shipped). Produced in 100s, some expensive FPGA on them, made to be mounted in racks. 6U in height. All in the same installation.
The boards are not using a backplane.
Instead of using a backplane, there are separate power connections, I2C connections and CAN connections, all on the front.
All that really needs to be on the front is the main I/O, basically some fast serial links, and a jtag.

And I was thinking, why is it that nobody thought to use a backplane? It would make setting up the system simpler, take less time, reduce the BOM of the boards and require less cables.
It could also potentially reduce the board area required, which would make the large boards cheaper. All good things, as far as I'm thinking.

Or have I stumbled upon something? Is there a case to be made for not using a backplane? Have there been cases when using a backplane was a definite disadvantage?
For this I turn to your collective wisdom. 

[magnus0re]

Exactly the kind of answer I was looking for! Thank you very much.

[bson]

Another option is a bus ribbon cable and board fingers.