Reference Resistor Decade 0.1%
In the design of measuring instruments, a resistor decade is needed to test and calibrate the measuring instrument. Initially, I was thinking of a processor controlled decade, which would involve switching resistors using relays (arbitrary voltages cannot be applied to solid state switches). It would be sufficient to switch 4 resistors in each decade with values of 1, 1.8, 3.3 and 5.6 (i.e. an interval of 1.8). That is, rather than finding exact resistors 1, 2, 4 and 8 (or assembling them from multiple resistors), it is easier to use resistors from a series (even with low accuracy) and use a reference resistor to measure their exact value. The processor will then take care of compiling the correct value, it is able to compensate for variations in the value of the resistors (that is why the interval is 1.8 and not 2.0). The disadvantage of the relay solution - the resistance of the contacts of each relay adds up (can be improved by adding relays that bridge multiple decades at once) and the long path through the relay contacts forms a coil. It would be possible to improve the situation by switching the resistors in parallel instead of connecting them in series (suitable for low-value resistors), but it's still not quite enough. Then I was inspired by a resistor decade from China, with resistors 1.0x in series, switched by jumpers - it had only 1% accuracy, it doesn't set the value automatically, but as I found out, for amateur practice such a solution is fully sufficient. I just used 0.1% resistors. But I haven't completely abandoned the idea of automatic decade, mainly because of the prospect of achieving even higher accuracy.
Wiring diagram is simple - in a row of DIL switches you move the jumper to the appropriate position indicating how many of the same resistors will be used in the row.
Click on the image to enlarge...
The surface joint is designed as a one-sided joint. Originally I wanted to solder the resistors on the same side as the jumpers, but then I preferred to solder them from the bottom so I could stick the top side with a label (it is very important here). To do this, I modified the PCB - I blacked out the holes for the resistors to make them easier to solder from the side of the jumpers (the holes remained on the Eagle board). I used resistors with 0.1% accuracy. I didn't get the lowest resistors from the 1R range in 0.1% accuracy, I had to use 1%, but it doesn't matter - this range is only supplementary, to set lower orders of higher resistors, so lower accuracy is not a problem there.
Bottom side (joint side):
Bottom side modified, with holes for resistors removed:
Plate mounting (top view with jumpers):
Printing of the top side (I printed the print on paper on an inkjet printer, glued it to the circuit board, taped it with adhesive and re-drilled the holes):
I screwed the circuit board to the bottom of the box KM78.
It's one of the few things where you don't need to program the processor. :-) It's good to check the value on at least a reasonably sized ohmmeter. Even if it's not very accurate, you can at least compare the resistances of the higher decade with the sum of the resistances of the lower decade. Example:
During these comparison tests I found that some resistances were not only close to the stated tolerance of 0.1%, but some exceeded it. So you need to either allow for a slightly worse accuracy than the stated one, or choose resistors that better match the desired value.
When selecting resistors, one could use either an accurate bridge or an admittedly inaccurate ohmmeter, but with a stable reading, to compare the match. 10 identical 0.1% resistors (preferably from different factory scales) would be combined and compared to a resistor with 10 times the value - in which case it would be quite likely that the deviation of one of the resistors in the series would not show up as much, due to averaging. Alternatively, a bridge that could accurately compare values at a 1:10 ratio would still help.
Wiring diagram in Eagle Free
Graphic documents (schematic and circuit board)
Complete download of RBox documents
The parts selection was tailored to a GM Electronics store near me. At the moment I can see that there are no accurate resistors in stock with 1M and 10M values - so I need to use 1% accuracy (buy more of them) and select more accurate values by comparing with lower ranges (on lower ranges set the value to 9999... and compare on ohmmeter or bridge).
The total price for everything is 504 Kc (including box and photocuprextit).