Digital decade resistance box

[kxenos]

I was interested in having a decade resistance box because I find it convenient when working with op-amps, filters etc. Then I shaw the prices of used and new ones and thought that I'm in a crisis  and they are too expensive for me. Then I had a look on prices of 10 throw switches and well I thought I design a box without those expensive switches.
I need 100R to 10M range so that means 5 decades, so I thought I make it modular and use the prototyping service of itead or similar. I chose Coto 9007-05 reed relay because it's cheap, reliable and small. The drawback is it's switching current which is 0.5A and it's switching power at 10W. Since I plan to use the box in op-amp feedback circuits and other low current applications I thought it's adequate.
I didn't want to use a uC so my thought was to implement a counter. I used the CD40110 BCD up/down counter IC that has direct output to CK 7-segment display and the 74HC192 BCD up/down counter that has 4bit BCD output. I then convert those 4bit words to switch the correct relays using a quad NOR and a quad XNOR IC. I use a 1-2-2-4 scheme for resistors because this way I can use only 200R, 2k etc. resistors and I can even the usage of the relays for the two 2 positions. There are 3 switches per decade, Up, Down and Zero. After debouncing, Up and Down signals pulse the corresponding inputs of the counters and Zero resets both counters to 0.
The parts cost is around 10E per module so the hole project would be around 100E for the whole device without box, so I decided it's still too expensive and I will not make it, but I thought I share the work I've done so far.

[Sparc]

Interesting design.  Why not use a microcontroller?  When I was a kid and didn't have much, I wanted a resistor switch box.  What I did have though, was a few dip switches and bags of 1% resistors.  Knowing about computers and binary math, I built a real simple binary resistor box.   It had 15 bits.  Hehe.  Maybe the binary switch box is for computer programming nerds.   

I think this simple idea could be expanded to a more user friendly project.   A microcontroller takes user input from a keypad or rotary encoder, convert to binary and drive the relays, display current R value on a LCD.   The annoying part is working out the binary weighted resistor values.   How to get 65536 ohms, for example?   Some combinations of series/parallel resistors.

[sleemanj]

Get a small servo, mount on the bottom of a pcb with the shaft going up through the pcb, attach a "wiper" to the shaft that contacts 10 pads on the pcb around it, et voila you have yourself a micro-controllable 10 way switch.  Add your resistors, duplicate N times, and your servo driven digital decade box is done :-)

Otherwise, it's quite a few relays to get to a megohm - although small relays are pretty cheap, you should be able to get them for a buck a piece at the most - the coil currents would be fairly substantial (lets say 60mA per coil, 20 relays, 1.2A to drive your decade box :-))

What's the typical holding current of a servo?

[kxenos]

Nothing against uCs. I just didn't want it in this design. Part of the reason being I could imagine some oldfarts winning "You f*cking kiddos, use uCs everywhere. Back in the day we..." in which I somewhat agree.
Servos are interesting if the angle accuracy is decent. I also don't know about the holding current, I guess there must be small servos with 100-200mA or less on 5V. But then again, you could use small stepper motors to rotate potentiometers I guess.

[Lunasix]

I'm also thinking about a resistor box. Your choice of 1-2-2-4 arrangement is very good since choice of resistors is easy, one value for each decade, 10 factor between 2 decades. Base value of 2 (and power of 10) is better, but any other value will also be valid. So, 6 same value resistors for each decade are needed.
I will use a uP because I want an usb interface to automatically calibrate PT1000 digital converters, and will be much easier as other solutions.
One or two relays can be added to directly short circuit high value decades, and so reduce power consumption of relays and parasitic resistor value of these relays (0.1 ohm) when low values are needed. I will use 0.1% resistors, easy to find with such arrangement, except for 20R or lower values, but absolute error is lower than contact resistor of relay. And uP allows to compute and predict this error, to display or correct it if possible (adding 1 or 2 relay in special arrangement for example).
I before tried other arrangements (1-2-4-8... or other not obvious) but it was hawful to obtain required values with 0.1% error (1 to 3 resistors in serial).
I change for 1-2-2-4 !
I also think that the box must be as small as possible to reduce parasitic values (L and C). A processor and one board would be better in your case.