Here I am, after days of "radio silence." During this time, Kiriakos must have imagined all sorts of things.
Lately, I’ve been hitting the books again, reviewing university-level applied thermodynamics and thermal physics. I wanted to refresh my knowledge base so I could redesign this project with better expertise and awareness. More importantly, I’ve re-evaluated everything: requirements, causes, consequences, solutions, and expected performance.
So, I wiped the slate clean and restarted from scratch, as if I had never encountered the problem before.
I’ve always blamed the recessed, non-binding-post banana jacks for making it so difficult to build a proper shorting plug—and unfortunately, I still think that's the case.
In the HP 3457A multimeter service manual, it states: "Equipment Required. A low thermal short (copper wire) is required for this procedure." Essentially, the challenge lies in finding pure copper wire and scraping off the surface oxide before using it. Fluke, for its 8588A, uses a very simple PCB designed to be clamped directly by the binding posts.
However, with multimeters like the Keithley 2000/2001 series, HP 34401A, or Fluke 8845A/8846A, that's not an option. You are forced to use 4mm banana plugs and connect them somehow.
Fluke and Keithley chose a PCB approach; HP went with a wire. Nothing new so far—these are well-known, tedious facts. But what actually generated oceans of forum discussions and solutions ranging from the highly imaginative to the deeply practical is the PDF written by Martin L. Kidd, titled "Watch Out for Those Thermoelectric Voltages!" My only critique of his work is that he tackles the issue with a bit too much simplicity.
Unfortunately, in my opinion, anyone owning these multimeters just has to live with Seebeck coefficients. There is no magic solution, unless you bend a single solid copper rod to fit all the jacks (some have actually done this, but they ended up welding four copper segments together).
Therefore, the correct approach is to minimize the effects, or even calculate the exact error value—which is doable if you know the composition and thickness of all the layers involved.
Eliminating thermal EMF entirely means using pure copper and constantly fighting its oxidation.
Again, nothing new here. In fact, when I decided to DIY this back in January 2025, I thought the PCB should adapt to the Staubli plugs, not the other way around, because that's where the issue originates.
The intuition was right, but I targeted the wrong variable. I shouldn't have started from the Staubli banana plugs, but rather from the multimeter's jacks themselves—or better yet, from the internal construction of the DMM.
It is self-evident that the input jacks are warmer than the ambient room temperature. Martin L. Kidd mentions this as well: "It should be noted that the terminals of the DMM are at a higher temperature than room temperature, so there will always be a thermal condition to be aware of. In this case, the input thermals measured about 29 ˚C."
So, in our case, the real issue isn't external drafts, a warm hand touching the shorting plug, or anything like that. It’s the instrument's jacks themselves heating up the banana plugs, which in turn heat up the PCB, the copper traces, the nickel, and the gold plating, creating unwanted thermal EMFs. As long as the internal temperature of the instrument is higher than the ambient temperature (which is pretty much always the case), there will be a heat flux flowing from the inside of the DMM out into the shorting plug and everything connected to it. Unfortunately, if my thermal cameras aren't lying, these temperatures are not uniform across all four jacks.
On my HP 34401A, the bottom-right jack is slightly cooler than the others, especially compared to the Sense HI terminal (top-left).
This single observation brings us to our first crossroads: should we use plugs with low thermal conductivity, like brass, or beryllium copper plugs, which have higher thermal conductivity? Choosing one over the other leads to completely different design choices for the PCB and its traces.
p.s. I used an AI to translate this post from my native language, as it's objectively much easier to write down complex, long thoughts in one's own tongue before sharing.
Topic: Fluke 884X-SHORT VS Keithley 8610 Shorting Plug 4-Wire (multimeter calibration) (Read 23618 times)
Home
Help
Login
Register