RoHS CompliantAll Small Tree cards, switches and storage products are RoHS compliant!

Your dishwasher is broken…

December 26, 2012 by Steve Modica

Steve ModicaI started in the computer industry around 1988 as a computer engineering co-op student with a small company called Herstal Automation. Herstal built memory boards for very old HP 1000 A600 and A900 computers.  These computers were some of the very first real-time computers ever made.  Auto companies and the medical industry used these to monitor real-time processes like engine performance or patient vital signs.

To give you some idea how old these were, the machine had toggle switches on the front so you could hand enter machine code instructions one at a time. This was a good way to enter tiny little test programs or force the machine to boot when you were stuck.

Typically, I would build up memory boards and boot the machine to test. Sometimes, a board would fail.  When that happened, I would use the front panel toggle switches to put in a small assembly instruction program that would write all 1’s into memory. Then I would dump the memory and see if there were any single zeros (bad data line) or entire words that were 0 (bad address line).  I could almost zero in on the exact pin or chip that was failing and replace it.  I could unsolder a bad chip and replace it so cleanly that you couldn’t tell it wasn’t machine soldered.

One day, I recall a board that was behaving very strangely. I couldn’t seem to get it to power on.  There simply were no lights.

I pulled the board out and did a visual inspection. This is tricky because if you look at a pattern (like pins on a board) the eye will see a clean pattern.  It’s very easy to miss a bent pin.  Our brains fool us into “seeing” the pin even when it’s not there.  One has to intentionally look at each and every pin.

Even with all that inspection, I could not see a problem. Finally, I pulled out my voltmeter and started looking for a bad trace.  Maybe one of the caps or resistors on the board was simply bad and was not passing current through.

I touched the power pin and the power lines on the downstream chips and could not read a connection. Hmmm…I started working backwards, closer to the pin, but still no connection.  Eventually, I had both leads on the pin. Still no connection! The gold pin on the edge of the board was not conducting electricity.

I have to admit, I was not a good physics student. I hated physics. However I did very well in electronics and I know that gold conducts electricity.  In fact, gold is great at conducting electricity. So I did the scratch test. I took a tiny little screwdriver and gently slid it against the pin (scratching gold pins on these boards was a definite no no).  A thin film of plastic bubbled up.

Then I knew what was wrong.

In those days, electronic boards were assembled by hand by installing parts and bending the pins down to hold them in place.  The pins were then nipped off and the boards were sent to a “wave solder” facility.  These places would pass the board over a flowing “wave” of molten solder and all of the pins and pads would be soldered at once in a very uniform and reliable manner.

There was one problem with wave soldering in that if you had gold connector pins on your card (like a PCI card), the solder would stick to the gold and ruin it.

To avoid this, wave solder companies would use a special water-soluble tape on these edge connectors. After the wave solder was complete, the boards were put into an industrial dishwasher and the tape would dissolve. (This would also clean the boards and make them look nice and new).

So the reason I had this layer of plastic was that our wave solder company’s dishwasher was broken.  It was not heating the water enough to completely dissolve the protective tape layer. I confirmed with a quick phone call and used a tooth brush and some hot water to fix the rest of the batch.

Obviously, in this day and age, this is a pretty rare problem.  Boards you buy for your edit stations are built in large quantities and Quality Control Tested on pin grids and test rigs to quickly rule out any obvious problems.

That being said, there are things you need to consider when dealing with large PCIE boards that you might have to plug into an older machine:

  1. Make sure you aren’t carrying around any excess static that might zap the board. If the vendor provides one, put on an appropriate grounding strap when installing a board.
  2. When installing, be very careful not to knock loose any surface mount components.  In the old days, things were soldered right through the board, but today, they are only surface soldered.  If you bump one of those little chips too hard, you will knock it off. Depending on what it is, your board may not work at all, or will be intermittent.
  3. Watch carefully for interference.  Often, large graphics boards have huge heat sinks or cases and fans on them. Make sure none of these devices is touching a neighboring board. This could lead to electrical shorts or overheating.
  4. Make sure all wires and cables are strapped or tied down! If you leave unused connectors hanging, one day they are going to end up hitting a fan or a hot component and melt.
  5. The 5Volt and 12Volt supply wires inside most computers aren’t going to kill you. However, imagine what might happen if you shorted a ring or watch against a 12V line. It would get extremely hot very quickly and may even melt (while touching your skin!). So don’t take these relatively low voltage supplies lightly.

When working with complex electronic assemblies, following these steps can save you a lot of time and frustration down the road.


No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.