Time to end the year with a new clock.
I listed a few issues in the previous log of the project, the biggest issue was the accuracy of the clock. To solve this and the other problems a new hardware revision was needed. In revision 0.2 an external crystal has been added, as the 20 pins package of the STm32F042 does not have pins for an 32Khz RTC crystal a normal 8Mhz crystal was added. The RTC in the STM can also use the high speed crystal as input. The slight difference in distance between the numitrons and the annoying pin twist have also been fixed.
The new PCB.
This also means some changes in the software, and here the downside of using the MBED environment shows, changing something low level as the clock settings for the RTC is near impossible. In the end I had to add the MBED-dev source instead of using the precompiled libraries and in the file "targets/TARGET_STM/TARGET_STM32F0/rtc_api.c" change the RTC settings.
The clock has been running for about a week now and is still accurate, a big improvement over the few minutes a day in drift it had with the internal crystal.