Finally with the prototyping complete and the board functioning as required. A few final touch ups to the design were made. The most significant included changing the inductors from through hole to surface mounted and changing the power connector from a five pin to a two pin model. The PCB board itself was not the made with quality in mind and this was evident with certain routes having faults in them. With the design complete and Robert Reed back in South Africa the next goal was to manufacture production quality boards in larger numbers using South African industry from start to finish.
The image to the right shows the prototype in its completed stage. The Ultra bright blue LED indicates that it is in fact producing the high voltage that is needed. For more information regarding the prototyping (click here).
Even though we had tested the design and knew it would work we still had to make a production prototype to test whether the South African industry could produce the boards fault free and with the required quality. The first thing we needed to do was contact various companies getting quotes, time scales and whether they could actually produce what we needed. This process turned out to be quite painless and we settled on a company which manufactured the PCB boards quite quickly and with high quality.
At the top we have a close up of the etching and we can clearly see the higher quality compared to prototype on the right. Looking closely at the through pins we can see that they have soldered through the hole and not simply drilled and placed a copper tube. The latter has a chance that the tube gets pulled out if you need to replace components (this happened a few times).
The two pictures on the bottom show the overall quality of the boards. On the left we have the top of the board and the bottom is on the right. The biggest difference between these boards and the prototype is the extra layers that were added. The green is a solder mask which is essentially a "paint" that forms a thin layer of non conducting material over all areas that are not going to be soldered. This is much preferred since it stops accidental contact and shorting of components and also helps with the soldering since stray flux makes no contact with the etch. The silk screen (lines in white) were also added which helps with component placement and also adds a more professional look and feel to the boards. The last major improvement was the removal of "floating" pieces. Looking at the photo on the right we can see that all that's left is the etch routes and nothing else. The prototype had the entire layer left behind (all the dark green areas were floating copper planes). This is a much cleaner design and less copper also means less chance for accidental contact.
The mounting procedure was very similar to that of the prototyping. Before mounting could begin the components had to be organized according to their values. This proved quite tedious as the supplier wrote the order code on the packages and not the values. This meant searching the order numbers to find the value of the components as they are so small that printing the values on the chip is not practical.
As before, the smaller components were mounted first. There were some issues with getting the SMD inductors to stick to the board due to their weight. In the end quite a large amount of soldering was required to form a strong and clean connection. The whole process went extremely quickly as it was the second time it was being done. In just four to five hours almost the entire board was populated.
The picture of the top right shows the sorting process with the green board waiting to be soldered. The picture below and to the left is the board with all components mounted. There is only the power connector that is missing as it had not arrived at the time of this photo.
The board was tested for faults and none were found. The board performed exactly as expected. All LEDs that indicate correct voltages and power were illuminated. The potentiometer which adjusts the input voltage going into the Ultravolt component needed slight tuning to get the output exactly at -830V. This was just a matter of slowly turning the screw until the correct reading on the multimeter was observed. The manufacturing prototype was a huge success. The quality was very good and the price of the PCB and components was very affordable. The next five boards have finished being populated by one of the companies we sourced. They have been tested and are working as expected. The boards must now be shipped to CERN where they will be used in the next MobiDICK units that are currently being manufactured. Below are some pictures of the final board being tested and connected up to the power supply.