This article follows the previous article on this topic, this time with a video after the break...
For soldering components, we also use a fully automated process. The PCB is put in our autoplacer robot. First, the robot gently injects one or more drops of solder paste on each pad where a component has to be placed, using a small motorized syringe driven by an Archimede screw. Then, it picks all board components one by one from a reel or from a tube, and places them at the right place and in the right position.
Of course, this magic only works because we have fed into the autoplacer software the exact shape and footprint of each component. This information is extracted directly from our component database, from the component 3D model, using tools that we developed in-house. This fast prototyping method, based on dispensing, works amazingly well, including for tiny QFN components with very small pitch.
Once all components are placed on the board, it only has to go through the reflow oven to complete the soldering. In order to get good solder pads, the paste needs to undergo a specific temperature profile. The profile is specified by the solder paste manufacturer, but it must also respect temperature constraints of every component on the board. This is not always easy to achieve with our 3-zone oven for the fine-grained lead-free solder paste that we need to use for syringe dispensing. Luckily, in this case, the profile does not need to be perfect, since we only use this solder paste for short-lived prototypes.
After a minute or so, the prototype comes out of the oven. At this stage, the only missing feature is the through-hole contacts (vias) for PCB tracks that move from one side to the others. For complex circuit boards with many vias, we use a metallic paste that can be applied through the holes just after the PCB milling, before placing components. But for small devices, it is usually faster to simply solder a 0.25mm wire through the holes. To be on the safe side, we also take this opportunity to brush the CPU pins and the USB connector with a soldering iron and flux to remove any undesired bridge.
The circuit is now ready for programming. Since we always use PIC24F processors, we created a dedicated tool for binding logical functions to each processor pin and automatically get an embedded project created with full USB and/or network stack support. This makes it possible to instantly test the prototype base functions. Once the USB link is confirmed to work, we only have to add the lines of code to handle the specific sensor or output peripheral. When we notice a design mistake in the scheme or in the PCB routing, we need to fix it and start over the prototyping. The good news is, it only takes about 2 hours to fully rebuild a new prototype since everything is automated. Magic, isn't it ?
The whole process described in this article and the previous one is illustrated in the video below:
The next step is to move from prototyping to production. But this is another story to come...