Homebrew Metrology The CERN Way

We won’t pretend to fully grok everything going on with this open-source 8.5-digit voltmeter that [Marco Reps] built. After all, the design came from the wizards at CERN, the European Organization for Nuclear Research, home to the Large Hadron Collider and other implements of Big Science. But we will admit to finding the level of this build quality absolutely gobsmacking, and totally worth watching the video for.

As [Marco] relates, an upcoming experiment at CERN will demand a large number of precision voltmeters, the expense of which led to a homebrew design that was released on the Open Hardware Repository. “Homebrew” perhaps undersells the build a bit, though. The design calls for a consistent thermal environment for the ADC, so there’s a mezzanine level on the board with an intricately designed Peltier thermal control system, including a custom-machined heat spreader blocker. There’s also a fascinatingly complex PCB dedicated solely to provide a solid ground between the analog input connector — itself a work of electromechanical art — and the chassis ground.

The real gem of this whole build, though, is the vapor-phase reflow soldering technique [Marco] used. Rather than a more-typical infrared process, vapor-phase reflow uses a perfluropolyether (PFPE) solution with a well-defined boiling point. PCBs suspended above a bath of heated PFPE get bathed in inert vapors at a specific temperature. [Marco]’s somewhat janky setup worked almost perfectly — just a few tombstones and bridges to fix. It’s a great technique to keep in mind for that special build.

The last [Marco Reps] video we featured was a teardown of a powerful fiber laser. It’s good to see a metrology build like this one, though, and we have a feeling we’ll be going over the details for a long time.

18 thoughts on “Homebrew Metrology The CERN Way

    1. In my my misspent youth I hung about with a lot of kids who were into grafitti. For a while there was a trend to fill a shoe polish applicator or paint pen with HF to tag glass. Man as a bit of a geek and knowing the dangers I couldn’t shout loud enough about how stupid that was.. not only to the user but subsequent public/cleaners etc. It seemed to die out fast once folk realised . But still makes me uneasy.

    2. No worries on that front, as long as you don’t create a pressure vessel, you can’t overheat the PFPE. I won’t be say there’s absolutely no risk involved, but a Teflon-coated frying pan is a lot more dangerous.

      1. I do agree that it would require all the liquid PFPE to fully boil so that it could then be heated up to 350°C (662°F) and decompose. But it is not like that is an impossible mode of failure.

        A Teflon-coated pan is dangerous, but it requires that you heat the frying pan, up above 350 and optimally pour cold water on it (search for “Polymer fume fever”).

        1. It’s not impossible to get it to decompose, but I doubt one could do it by accident. Like even if the hotplate went berserk and you somehow forgot you where trying to reflow a PCB in the last two minutes, the biggest health risk is going to be the smoldering epoxy of the PCB itself…

    1. Technically if everything is operating within expected parameters, there should be no HF. There is the possibility of a fault causing the overheating PFPE above 350°C (662°F) and decomposing. If that happens then fluorine will break down most materials, leak as a gas because it’s boiling point is −188.11 °C, (​−306.60 °F), and instantly form HF with the natural water vapour in the air. Then you are messing with HF.

  1. If only that vapor reflow used a liquid with less risk, The idea of accidental HFl generation is enough to put me off. One bad hotplate away from a bad day.

    I’m left speechless by the project though. He doesn’t do anything by half measures.

    1. He is. That’s what the Fluke calibrator mentioned at 31:40 is for. At the top of every NIST-traceable calibration is a primary standard. This instrument is not that standard. So even though it’s crazy precise, it’s not enough. I think when you get that many digits, it’s not about absolute accuracy, but repeatability and being able to compare two different measurements. But just the same, here’s an article from NIST about a primary voltage standard with 1 ppb accuracy and no need to calibrate: https://www.nist.gov/news-events/news/2013/04/primary-voltage-standard-whole-world. Maybe they will release that as open-source hardware, so Marco can build that as well.

      1. Messing with RF and cryogenic superconductors together, what could possibly go wrong.

        But joking aside, this thing will not reach 8.5 digits, period. There are certain inherent issues in this ADC topology which prevents it from reaching that uncertainty level on an absolute measurement. It’s very good, but not *that* good. This measurement module is clearly a custom design for CERN’s specific application – where it probably excels. I also doubt he went through the trouble of burning-in the circuit properly, etc. So the Fluke calibrator is plenty in this case, it’s mostly annoying how this is called a 8.5 digit multimeter while in reality this is more akin to a 5.5/6.5 digit from Keysight or Keithley performance wise.

  2. Amazing ! Would there be any measurable benefit of also using an ovenized crystal for the ADC ? I am looking into making a slightly less expensive version to experiment with on my own using an LM399 (and maybe an OCXO ??)

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.