For the Zaunkoenig M2K we infamously used a STM32F7 microcontroller unit (MCU) so powerful that it could be used for military drones, in order to pull off 8,000 hertz. For our two upcoming gaming mice, the Zaunkoenig M2S as well as the Zaunkoenig M3K, this military drone MCU was no longer an option. In 2023 the world market for this MCU is even more dried up than it was back in 2021. And the prices are even higher. Back in 2021 we paid $10 for one of these MCUs. Today that price is at a staggering $120 per MCU.
If you are building a military drone worth $100,000 you probably do not care about that price hike. Dropping $120 on the MCU for a gaming mouse however would be a bit silly.
So we started looking for alternatives. And we found exactly one other STM MCU that could do 8,000 hertz and that was in stock (the global chip shortage that started in 2020 is still not over, and many MCUs simply are not in stock). This one alternative MCU has the same 216 MHz Cortex-M7 core that the M2K MCU also has. In other words: for a gaming mouse this is more than enough overkill.
This alternative MCU also is much smaller. With a size of just 4.166 mm × 4.628 mm this MCU is only 19.3 percent the size of the M2K MCU, which was a whopping 10 × 10 mm. When you put the M2S/M3K MCU on top of the MCU for the M2K the difference in size becomes quite apparent:
Such a small MCU enables a much smaller printed circuit board (PCB) design. And with a smaller PCB the bottom shell on which that PCB sits can be smaller as well. This is great news if you want to make the worlds lightest gaming mouse even lighter.
Metal balls attached
The extreme miniaturization of the M2S/M3K MCU comes with strings attached, however. Or metal balls, to be more precise, as the following picture of the underside of the M2S/M3K MCU shows:
In the above picture you can see 100 of these attached metal balls. Each of these metal balls is a connection to an important part inside the MCU. There are power connections, ground connections, or input/output (I/O) connections. These I/O connections are needed for the switches and the sensor, for example.
With 100 balls the M2S/M3K MCU has more than enough connections for a gaming mouse. Sure, the M2K MCU had 176 balls, but over 100 of those were unused in the M2K, so having less connections is inconsequential.
Besides the beefy 216 MHz core of the M2K MCU this high amount of connections is another important attribute that made the M2K MCU a hot candidate for being used in drones (in a drone you simply have lots of stuff to connect to the MCU; accelerometer, gyroscope, inertia measurement unit, compass, barometer, GPS, the motors; the list is long).
Back to the metal balls however: these metal balls are made from a tin-silver alloy; put differently they are solder balls. If you place this MCU onto a PCB and then put that PCB into an oven these metal balls will melt and form a mechanical and electrical connection with the PCB. Electronic parts with metal balls on their bottom side are called «ball grid array», or BGA for short.
Back in the 1990s, when BGA parts first became common, two of these metal balls had a distance of 1.27 mm from each other (measured from center to center); these BGA parts hence were called «1.27 mm pitch BGAs». Soldering those 1.27 mm pitch BGAs was relatively easy.
Over time however the pitches of BGAs got smaller and smaller: 1.0 mm, 0.8 mm and 0.65 mm (which is what the M2K MCU is using). Soldering the 0.65 mm pitch BGA of the M2K MCU was still relatively easy, though obviously already much harder than soldering 1.27 mm. The smaller the pitch, the closer those metal balls are to each other, resulting in higher risk for a solder bridge (an unwanted electrical connection). One single solder bridge can be enough to ruin an otherwise fine PCB.
In the case of the M2S/M3K MCU the pitch is so small that it is not even called «BGA» any more: it is called micro BGA instead. The pitch of the M2S/M3K MCU is just 0.4 mm. That is so small that if you accidentally got a single hair between two metal balls the chance for a solder bridge would be substantial. When I spoke to our assembly house in the Swabian Alps they described soldering a 100 ball 0.4 mm pitch micro BGA as «eine Hausnummer», which loosely translates to «a tall order».
Not only is micro BGA difficult to solder though. It is hard to design a PCB for micro BGAs as well. With Micro BGAs many of the traditional PCB design techniques that can be comfortably used for traditional BGAs cannot be used anymore (more on that in our upcoming Blog article about the M2S PCB).
I guess this is the explanation for why the M2S/M3K MCU was the only STM MCU capable of 8,000 hertz that you can currently find in stock: it is extremely painful to design a PCB that works with this MCU and it is extremely painful to solder that MCU.
Could the M2S/M3K MCU be used for a military drone as well?
When we ordered the M2S/M3K MCU from a Chinese supplier we were not asked what kind of product we wanted to build with those MCUs.
The more I think about this the more I suspect that the M2S/M3K MCU is not a commonly used MCU for military drones. Even if you design the perfect PCB for such a small MCU, and you have the soldering done by an assembly house specialized on small and intricate PCBs, one big issue remains: micro BGAs first became available in 2009 so there simply is not a lot of data regarding their long term reliability. And long term reliability is of extreme importance for military equipment.
It is similar for PCBs for space rockets: when reliability is of paramount importance, old technology is oftentimes favored over new technology; simply because this old technology has a proven record of reliability. When there are millions of dollars in the air you tend to play it safe.
That, or the M2S/M3K MCU is a common MCU for military drones, but the Chinese supplier that we bought it from just did not care at all about what we were going to use those MCUs for.