So if you ever took apart a precise meter you would see that they use an odd resistor which looks like a ceramic tile with metal laid and laser cut on top of it.
Could this kind of performance be reached by using discrete components, like maybe foil SMD resistors mounted on a heat sink with the super thermally conductive sheet I saw mentioned in another thread (having thermal conductive properties similar to diamond)? (pyrolytic graphite sheet)
Are those laser cut resistor networks used for cost reasons or strictly for performance reasons? I imagine that it might be cheaper to create your own laser cut resistor networks if you invest in the machinery and decide to build 10,000 meters.
Is it possible to get resistors like that made? I want to gut a parts meter now.......
Well, in DMMs, there are different applications for those resistor arrays or assemblies.
The first type contains a set of stable reference resistors, for the different OHM ranges ( 100 Ohm, 1k, 10k,...), dividers for DCV amplification (x10, x 100, etc.), and perhaps also shunt resistors for current ranges.
Low T.C. and low T.C. matching in the case of dividers is required.
As most modern instruments have SW calibration, the absolute value of resistors, or their ratio, has not to be that accurate.
The advantage is, that the complete set of stable resistors is produced in one run, and their stability figures are identical. Therefore, no external T.C. matching of discrete resistors is required.
Additionally, the component count on the PCB is smaller.
The 2nd application type are the DCV input divider for 1kV and 100V.
Here, a fixed divider ratio of 100:1 is needed.
For more simple DMM designs, mutiple ratios of 1000:1, 100:1, 10:1 are required.
This divider design is very delicate, if you think of 1kV input voltage.
The reason is the high power dissipation of typically 100mW (1kV^2/10MOhm) , which creates non linear behaviour of the divider over increasing voltage, which leads to an additional uncertainty of many ppm in the region between 100V to 1kV.
This non linear effect originates from the different power dissipations over the 100kOhm and the 9.9MOhm part of the divider, especially if you design that divider with discrete resistors.
From 100V to 1kV, the 100k is loaded with 0.01 to 1mW, but the 9.9M with 1 to 99mW.
Therefore, the 9.9M resistor will heat up over that input range, typically 10°C, but the 100k nearly stays at the same temperature.
Well, the metal foil resistors have low T.C. of <2ppm/K,therefore 10°C difference for the 9.9M creates an additional mismatch of the ratio at 1kV of 20ppm (10 K * 2ppm/K), compared to the ratio at 100V.
The Caddock dividers have low T.C. of 10ppm/K and T.C. matching of 2ppm/K, but by design, they are additionally thermally matched:
The 9.9M is physically bigger than the 100k, so the heating is lesser.
And both resistors are thermally coupled, i.e. always nearly on the same temperature, because they are either on the same substrate or their substrates are bonded together, so that the 100k resistors is heated by the power dissipation of the 9.9M resistor.
By their T.C. matching (2ppm/K), and their nearly equal temperature, their divider ratio will also stay nearly constant. This is specified by the voltage coefficient of 0.02ppm/V, giving 20ppm at 1kV.
That's similiar to what you can achieve with metal foil resistors.
If you inspect the specification of the famous HP 3458A , you will find an additional 12ppm uncertainty for 1kV, which arouses exactly from the mentioned thermal effect. HP also used such a Caddock like 100:1 divider. (That's RP7, the blue component in the picture below)
This is very mediocre for such an instrument (0.01ppm resolution!).
Other manufacturers can do better, although all of them also use those dividers.
Their trick is, to measure this power dissipation effect once (at 100V, 500V, 1kV), and then make a linearization calculation by the quadratic power dissipation formula.
So they attain uncertainties of around 1ppm, even at 1kV.
Well, I hope my - a bit longer - explanation has shown, that this special case of resistor dividers arrays has got some non trivial 'magic' inside.
Frank
Home
Help
Search
Login
Register
