The case of flexing your backplane muscle

By Gary Crowell

In a past life, I worked with a group on an in-house piece of test equipment designed to be built into a standard 6U card cage for a 19″ rack.  My job was to decide on all of the interconnects in the box, the card and backplane connectors, and to design and layout the backplane itself (This box had over 300 connectors and connector segments that fit together like a 3D puzzle box). A conventional backplane would have been at least 0.25″ thick, and probably 20 layers or more.

Because the power portion of the backplane had to distribute 13 different voltages, some at very high currents (this was before POL was practical), I decided to use two backplanes stacked together. One, a power backplane PCB, would essentially be a complex array of buss bars implemented in heavy copper layers. The other was primarily only a signal backplane, with limited local heavy plane areas to carry current between the card connectors and the points where the power backplane was bolted on. This way, they could each be only 0.125″ thick, and they formed a “box-beam” structure when bolted together on 0.25″ standoffs. This also allowed connectors to be placed on the back of the power backplane for converter cards to be plugged in from the rear of the card cage.

All was well; both board layouts were nearly complete, when the manager on the project raised doubts about the strength of the backplane structure. The cards and backplane mated 1000-pin connectors, and the insertion force was expected to be considerable. The backplanes were supported primarily at the sides, so the concern was that the backplane would flex in the center under the pressure of a card insertion. His suggestion was to add more mounting holes to the boards, and leave room for a supporting structure behind the backplanes.  This would have really messed up my nearly complete layouts, so I was looking for some kind of guidance to determine if this late change was really necessary.

These backplanes were actually larger and thicker than any boards we had designed before, so we didn’t have prior experience, and there were no materials readily available to mock up a test model. Time was short. We didn’t have the time or expertise to do a structural analysis. But I did have a good variety of blank PCB’s in smaller sizes and thicknesses. So one evening I stayed late and rigged up some tests on the boards I had. Using some six-packs of soda as “standard” weights, and spanning boards between some video cassettes, I started measuring deflections.

Measuring with similar boards of different thickness, gave me an idea of how thicker boards would respond. Using different weights (# of six-packs) gave a force/deflection factor. Similarly, I tested the effect of different widths of boards, and the span between support points. Finally I was able to bolt two boards together to see the effect of the box structure. By combining all of the deflection factors I had learned, adding a big safety factor and extrapolating to the size of the real backplane boards, I came up with an estimate of a maximum deflection of about 30 mils for the maximum force we expected.  Since I had been very conservative in my estimates, I sent the boards off for fabrication with confidence, and without the extra support area or holes.

The moment of truth arrived when we received the backplanes and assembled and installed them in the card cage. We slid the first card into the center slot of the cage. The connector-to-connector alignment was perfect. As we jacked the card into place with the ejector latches, it was apparent we had a problem. There was too much flex, and the card wouldn’t insert completely. But there was no flex in the backplane; it was solid as a slab of granite. The rails of the metal card cage were flexing! Adding some structural support to the card cage made it at least as strong as my backplane, and the cards inserted happily ever after.