Building

May 20 2016

I now have everything I need to build the boards: parts, PCB, and stencil. I need to preflight the build, then it is time to manufacture!

Preflight

1. Put the solder paste in the warmer, set to 105°F/41C
2. Eyeball the stencil on the PCB, make sure everything lines up. Make sure there are not stencil apertures over the through-hole pads and JTAG probe pads.
3. Reload the tray
   Make sure all the parts are oriented correctly.
4. Make sure all the parts are loaded in the PnP feeders.
   Learn Pick for all the tray components
5. Mount the PCB in the machine and learn the fiducials and reference points.
6. Learn PCB
   Detect the PCB height using auto-detect, with a finger on the ESC key. If the machine over-shoots the board, recalibrate the nozzle 1 open/close suction parameters for both WHITE and BLACK nozzles. Auto-detection of the board height is very unreliable, if it fails more than twice abandon it in favor of the eyeball method.
7. Run through the tape pick list
   Auto-advance OFF
   Learn Pick > Learn all by camera (small camera icon)
   Make sure there is a component at the pick location.
8. Run through the tray pick list
   Print a picture of the loaded tray to use as a guide. Reset the tray X, Y, and No entries.
   Learn by camera each tray component.
   Verify the X and Y values match the actual tray.

   

9. Do a practice pick to verify the tray height.
   Select a tray component, click Modify LIB.
10. Run through the place list. Production > Vision before production.
11. Test the reflow oven.

L302 and L501(Ferrite bead) were a late addition that didn't make it onto the tray but is present in the component list. I created a new tray feeder (Q23/223), assigned the feeder to L302 and L501 in the place list, and un-skipped them.

Q204 is missing from the pick list. This is CAP-0805-1uF which was moved/merged to Q22/222; 204 was a duplicate of the same component. FeederID 204 is no longer used.

Build: PnP

1. Mount the PCB in the stencil jig.
2. Find my paste squeegee.
3. Apply solder paste.
4. Mount the PCB in the PnP work area.
5. Run production.
6. Visual check: Production > Vision before production
7. Mount extra SMT components
   L501 (Feeder Q223)
9. Place the board in the oven.
10. Run the reflow profile.

The first production run was epic FAIL. I did not apply the paste smoothly and there was too much. The bigger problem was that even though I thought I learned the fiducials (F2,F3) the place was offset by a wide margin and the machine was dropping pieces outside the board. I ran the vision check and it lined up, not sure what went wrong.

Some of my patterns need to be tweaked to make the paste apertures smaller.

I will clean the board and stencil, reload the tray, and try again. The paste can be removed by gently wiping with cotton balls lightly soaked with rubbing alcohol. Take extreme care not to wrinkle the stencil during cleaning. Always wipe away from the held edge, never back. Thoroughly clean my hands afterward, do not leave solder paste on the skin or it will begin absorbing the lead.

The second run was on target, but the results were very poor, with 16 missing or bounced components. I believe the problem was not in the pick or place, but in the "stick" to the board. Nor is the problem due to too much paste, which helps the parts to stick better but then causes bridges during reflow.

I think three things went wrong:

• I applied the paste before I verified the tray picks and re-learned the fiducials and reference points. This let the paste dry out (oxidize?) and lose some of its tackiness.
• The board was poorly secured to the work area. I was still using the 4 posts to support the board with one post too short. This allowed the board to bounce when placing components near the short support. This configuration was left over from the paper board, where the laminated boards were too thick for the edge clamp. I need to remove the two posts, resume using the edge clamp, and re-learn the board height.
• Less likely, but still a possibility, is that the paste is too old and is not working well. This is my original tub from 2012 and has not been stored well. I will open up my fresh jar for the next run.

Since this build ran to completion, an entire board was consumed from the tray. I recovered the major components. I hope the paste that is still adhering to the legs does not create more problems.

The paste apertures are definitely too large. I need to go back and review the patterns.

0402 resistors and caps are definitely too small, much more trouble than they are worth. All future designs should use 0603 (or larger) unless there is a compelling reason to go smaller. I can handle 0402, but everything has to be just right -- most importantly the paste apertures in the stencil. For tiny parts, the apertures need to be even smaller than the pads.

I need to reset the tray, reload the depleted components from stock, clean the stencil, board, and squeegee, and try again.

It is now 1950, time to take a break for dinner. It appears the build will extend to a second day. :(

May 21, 14:40

I need to refill some of the tray components:

• 207: CAP-0603-22uF, 1276-2867-1-ND, Qty 2
• 212: CAP-0402-220nF, 445-4972-1-ND, Qty 8
• 214: CAP-0603-3300pF, 1276-2275-1-ND, Qty 2

I need to order more 220nF caps, preferrably upgrading to 0603 (Digikey 1276-1112-1-ND).

I ran out of D13, D-0402-GRN, the green leds. I have some 0402 yellow leds I can use.

The fresh paste seemed to apply to the board better. The auto build went well until it failed to recognize some of the larger tray components. I think it was being fooled bits of leftover paste on the undersides, which I will need to clean. I skipped these and continued.

Then Autotronik hung in mid-stroke. The application locked up and I had to Ctrl-Alt-Del to kill the process and restart. This is the first time I have experienced an application hang during a production run.

Reviewing the partially built board, good news and bad news. Good news: the parts that were well placed and stuck. Bad news: I was a bit off with the paste, and the lower left corner was bad -- the board would have failed QA even without the crash.

This is progress. I will need to clean, reset, restock, and restart -- for the third time.

I spent quite a bit of time wiping solder paste off even the tiniest of parts... I'm looking at you, TPS62237! (Note to self: TPS62237 is for use in watches and nanomotors). I restocked the tray and swapped the empty green led autofeeder for yellows.

The thorough cleaning created a different problem, 4 parts stuck to their compartments and the machine was unable to pick them. I reshuffled them and they worked on the second pass. The pick height on the yellow leds was off and required a third run. For some unknown reason, L301 was missing and not present in the error list. I created a place list with just that part and it placed.

1910> With a lot of help, the third run made it to the finish line. The visual check looked good, all the parts were present and in good shape. I am more concerned about the amount and position of the paste. I am praying to the Wicking Gods that the solder sorts itself out in the oven...

2000> The board looks good out of the oven, PnP visual inspection shows the components have centered themselves. I am worried about shorts, especially between the pins of the MSP430.

Build: Hand Solder

1. 7 2-Post headers:
   J509, VDD-MTR, BSL, RESET, VCC-HUB, VCC-EZFET,CTRL-3V
2. 3 3-Post headers:
   Pwr:JTAG, Pwr:VCC-5V, Pwr:3.3V
3. 3 4-Wire leads:
   OUT-X1, OUT-Y1, OUT-Z1
4. 1 Terminal block:
   J301
5. Q202(32KHz crystal
   The crystal is not polarized, either way is OK.

   

6. C304(100uF radial cap)
   Square pad is ground (short leg), round pad is positive (long leg)
   

   

7. 1 20-Post headers:
   J641

Test

2115> I found the old MSP430-FET430UIF and the TagConnect adapter, plugged it in... "Unknown device". This is the message when the MSP fails to respond to the FET. I looked at the MSP430 under the scope and it looks like the legs are all fused together into one giant solder ball.

I have to write off this board and everything on it. This is about $30 in parts, $35 for the stencil, and another 2 days of layout and 4 days for the new stencil to arrive. This is demoralizing, but it is progress. The fresh solder paste made a huge difference, I believe I could place all the parts in a single run next time. The big mistake was the stencil: I failed to check the shrink on the solder paste apertures, especially for the MSP430s. This is the second time () I have learned this lesson the hard way.

I wrote up a guide for setting the paste shrink: . The problem is in the stencil, not the boards, so I can continue to build with the remaining 3 boards and a new stencil. I must create the new stencil from the archived layout for the 20160511 boards, editing a new copy of the archived layout, and using the "expedient" method of tweaking the pads directly in the layout. I cannot regenerate the board from the schematic, as I have made changes to it since ordering the boards.