Potentiometer instead of resistance decade box

[thanasisk]

I am in the process of setting up a small lab for developing audio effects. One of my requirements is to have the ability to replace several resistors within the circuit (both discrete and op amp-based preamps) with variable resistors;

I was looking at resistance decade boxes and I was of course amazed at the prices of ready-made and of diy ones (a lot of switches, a lot of resistors, a lot of soldering!).

I then came up with this idea of using 1MOhm high quality Conductive Plastic pots (such as the BI/TT P260 series) where e.g. the left and middle pin would be used for resistor replacement and the middle and right pin for measuring the resistance using a uC/adc. The uC would measure the complementary resistance, subtract it from the total resistance (manually measured) and present it in an LCD screen. And all of this without the metering being in-circuit.

And for the cost of a single  decade box  I could implement e.g. a 4 pots variable resistor with very accurate readings.   

Do you think this is a technically viable idea?

- One strange thing that I found when measuring some alpha pots (100K and 500K) was that the sum of the left and right resistances was more than the total resistance as given by my multimeter     Is this a known potentiometer thing or perhaps it is an effect of my uncalibrated auto-ranging multimeter? 

- I realize that I  would have to implement a reference current source and stable supply voltage source for achieving accurate ADC and uC readings/calculations, perhaps as accurate as a 4-digit dmm? 


Cheers!

[thm_w]

I think its a good idea for a project. The downsides of a potentiometer:

• High temperature coefficient
• Low maximum power dissipation

For P260 0.5W, lowest R = 2 ohms, gang error = +/- 3dB
http://www.mouser.com/ds/2/414/p260-12592.pdf

One way to get around the accuracy/tempco issues is if you have a stable reference and can do a calibration routine.
So even if the main and secondary wipers are not well matched, calibration would take care of it.

What readings did you get from the alpha pots? The wiper will have some resistance that adds in both directions, although it should be quite low.

[ScottK]

I once used your method for a production test fixture that required tuning a device via a variable resistance. I used a motorized 10 turn wire wound pot, so that setting the required value could be done hands-free or even remotely. The fixture was small (smaller than a typical decade box), and battery operated, so noise injection was not an issue. I did not need tremendous accuracy, and the fixture didn't give it. Time to settle on the correct value was under one second.

In my lab, when I want to adjust a resistor to achieve some desired goal, I just use a trim pot in the vicinity of the value I think I'll need. I adjust it in-circuit to get what I want, then remove it and measure it. If I know I want a certain value that I haven't got in the parts cabinet, I'll set the trimmer to the value I want and insert it. I've never needed a resistance decade box or equivalent.

The odd behavior you see when measuring from wiper-to-end vs end-to-end is caused by the nature of making contact with the surface of a conductor. Current from end-to-end goes through the entire bulk of the conductor (once you get a little way from the end contacts). Current through the wiper must exit through the conductor's surface where they mate. That forces the current through a cross section that's smaller than the conductor's cross section, causing "current crowding". The result is that the resistance from the wiper through the conductor surface to either end is higher than the from the entire cross section of the conductor under the wiper to either end. The degree and nature of this effect depends on the geometry and materials used.

Want to experience another pot oddity? Every time you get your hands on a log taper pot, connect your DVM from end-to-end and turn the shaft. You will eventually find a pot where the resistance changes slightly as you turn. And this is caused by a wiper that's not quite perpendicular to the travel path, shorting out a changing percentage of the track width as it moves from end to end. Non linear pots often have a conductive path that's tapered to yield the desired transfer function, hence the use of the term "taper". If you have a meter with sufficient precision, you can see this on linear pots as well, and the change will be random with shaft position. The wiper usually has multiple contact points to minimize noise from surface contamination. As you move the wiper, you short across varying amounts of the conductor width (and length because no wiper is perfectly orthogonal to the current path). Using a pot with a long travel path, such as a multi-turn spiral type, minimizes this effect. The current crowding region becomes a smaller part of the total path length.

Add this to your technical library... https://www.bourns.com/pdfs/OnlinePotentiometerHandbook.pdf

Page 165 discusses current crowding.

[jeroen79]

Another downside of a potentiometer is that you can't set it precisely without measuring it with an ohmmeter.
A decade resistance box has fixed positions that can be calibrated.

[slurry]

I actually built a box with two good quality multiturn pots for lab use, a 10kohm in series with a 200 ohm for fine adjustments, i think i used Bourns 3500.

The tempco i not that great but good for most uses anyway.
Actually, i think that WW have better tempco than cermet? i'm not sure here,
but anyway WW is better when humid conditions changes so i guess WW would be a better overall choice.

[Gyro]

This idea will only work with any success if you use wirewound pots (preferably multi-turn). A ScottK points out, the reason for your end-to-end resistance being lower than the sum of the two end to wiper resistances is that the wiper only touches the surface and not the bulk of the resistive element. A potentiometer is just that, the wiper is intended to tap off the 'potential' at that point into a Hi-Z load rather than passing current. Current causes 'wiper noise' when you turn the pot. As I say, only wirewound resistors give a chance of passing reasonable wiper current (as in traditional rheostats).

Why not build your own compact resistance boxes. The cheapest and most compact way is to use cheap Decimal coded thumbwheel switches from ebay (Note, Decimal, not BCD*)...

http://www.ebay.co.uk/sch/i.html?_odkw=thumbwheel+switch&_sop=15&_osacat=0&_from=R40&_trksid=p2045573.m570.l1313.TR0.TRC0.H0.Xthumbwheel+switch+decimal.TRS0&_nkw=thumbwheel+switch+decimal&_sacat=0

Then buy yourself some bags of 100R, 1k, 10k etc. resistors of whatever tolerance you wish and solder them between the taps. You could quite easilly build yourself a bunch of repeatable and very small resistance boxes for very little money. You wouldn't even need to build enclosures, just use some large diameter heatshrink (as used for battery packs etc).


Edit: * You can use BCD coded switches to make capacitance boxes too.

[Siwastaja]

I have never found any need for a resistance box, or any alternative.

I simply solder in different resistors, or a variable resistor into the actual circuit. If I expect the need of more than a few iterations, I solder in a simple pin header socket so that I can swap leaded resistors. This is simple, reliable and low inductance.

[SingedFingers]

I disagree with the wirewound requirements. I use cheap carbon potentiometers. I only smoked the 100R one once

I designed and built this that works fine for most use cases, which are primarily impedance measurement for me:



SW1 is a DPDT toggle. All pots are reasonable quality Omeg 0.5W ones.

Instructions:

1. Connect it in circuit between A, B.
2. Connect DVM between C, D and set to ohms autorange.
3. Set pots accordingly for example to knock off half your amplitude on the scope to measure impedance.
4. Flick the switch (power down the circuit if it will explode open-circuit before!).
5. Read the value.

Edit: lives in a little die cast alu box with 4x 4mm jacks. No knobs, labels or anything. Not worth it.

[Gyro]

Surely the slightest tap of the 1M knob will completely nullify the usefulness of the 100R pot. I would have though a better solution would be just to have three pots and use whichever is applicable to the needed resistance range.

With regard to pot type - if you are just handling signal level then of course a carbon would be fine. If you want to pass any significant current (say 1/4W dissipation)  then the wiper resistance and point contact dissipation and noise rapidly becomes very significant. I'm not suggesting that the OP actually uses a wirewound pot by the way - that would be uneconomical for high values. That's why I suggested the Chinese thumbwheel switches for an economical general purpose solution.

Edit: It's pretty hard to beat £4 for 10 digits (splitable) plus some resistors.

[SingedFingers]

TBH with respect to impedances, you only need to find within the range of one pot usually with possibly the one below it to trim the result slightly if you fancy a bit of accuracy. You usually only use two of them and the other one is zeroed out at the edge of the taper. If you turn the 1M to say 470k, then the 10k gives you 465-475k which is good enough when your entire circuit is 5% tolerances anyway. That gives you a good 50R, 10k, 1M impedance match which everything is usually in the range of.

I tend to go for impedance matching and power measurements instead of using a resistance box for more than 1/2w dissipation.

Sure you can build a better one for £4 and I might actually do that in the future. Tayda sell some rotary switches and resistors which are suitable for a reasonable fee. I don't like the thumbwheel ones as I had the misfortune of dealing with those resistance boxes at university and they were all FUBAR. Resistors were fine; the switches however...

[mariush]

Here's another idea, and potentially cheaper.

DIGITAL POTENTIOMETERS

They're tiny and available in SMT versions.
The resistance can be configured through SPI or i2c on most (some have simpler up/down mechanisms) .. with i2c and spi you can select the chip by address or by CS pin to program the new resistance valu 
i2c and spi potentiometes can be connected in series or in parallel on same circuit board to allow for finer control over the resistance value (most chips offer 128 or 256 steps, put two in series or parallel and you have 512 values)
you could combine a bigger value with a smaller value, for example a 256 tap 100k with a 127-256 tap 10k one, and then you'd have fine control within those 100k (about 40 ohm steps on the 10k pot) 
So you'd have only two pins going into the board but above the "resistor" you'll have a header with the i2c/spi pins and optionally the CS pins (to select individual chips if you have multiple on the board) and when you want to change the value, you just plug a microcontroller driven device which quickly determines the number of chips on the board, reads the values from each one and gives user ability to adjust values.

Downside :

They need power to work ... most can work from around 2.7v and up and require little power, picking a random microchip datasheet I saw 45uA at 2.7v , 15uA at 5v ... on a TI part (TPL0102) I see 0.2uA at 5v ...  so you could basically put a battery on the tiny pcb. A CR2032 with typical 225 mAh capacity will last about 1000 hours at 200uA load, much more at only 50uA that's supposed to be the typical consumption. So basically these will work for more than 1 month and then you'd need to replace the battery which is less than 1$ (cr2032 is so mass produced it's actually more like 20 cents if you buy 50-100, here's an example http://www.digikey.com/product-detail/en/panasonic-bsg/CR2032/P189-ND/31939 )

There's also non-volatile potentiometers, but you'd only need those if you want to retain the value when the chips are not powered, so maybe for long development periods where you'd have to replace the battey. but even then you could add a couple of pins to the "programmer" device and send 3v to your configurable resistor through the header (you'd have to use a couple of diodes to prevent voltage going into the battery or from battery to programmer, it's not hard) , pull out the battery and replace it, then remove the programmer header. 

the other potential downside is that the maximum current on the resistance pins is usually  very small, at around 2.. 2.5mA , with maybe bursts of up to 20mA .. for some applications this may be too small.

[ScottK]

You can use BCD coded switches to make capacitance boxes too.

There's a better case to be made for constructing one of these. Capacitors outside the E12 bins are hard to find and large value variable capacitors would need their own buildings. I have a couple small proto boards with DIP switches and arrays of caps somewhere here in the lab. They're easy to make, requiring 15 capacitors of equal value (parallel sets of 8/4/2/1) and one four position DIP switch per decade. Wiring is so easy you don't need a schematic.

As I recall (I'm not going to dig through the junk to find them) one PCB covers five decades from 10pF to 0.9999µF using 10pF, 100pF, 0.1nF, 1nF and 10nF caps. The other covers four decades from 10nF to 99.99µF using 10nF, 100nF, 1µF and 10µF caps. In all cases the caps are ceramic. I don't worry about tempco (nor accuracy) as I'm using the things in a benign environment (my lab, where my personal tempco is more important) and I can measure values after dialing them in. If there is a frustration with using low tolerance components in this situation, it's that the system isn't monotonic, so I have found myself spending a lot of time hunting for the switch positions that yield best performance/desired value.

[Zero999]

Here's another idea, and potentially cheaper.

DIGITAL POTENTIOMETERS

They're tiny and available in SMT versions.
The resistance can be configured through SPI or i2c on most (some have simpler up/down mechanisms) .. with i2c and spi you can select the chip by address or by CS pin to program the new resistance valu 
i2c and spi potentiometes can be connected in series or in parallel on same circuit board to allow for finer control over the resistance value (most chips offer 128 or 256 steps, put two in series or parallel and you have 512 values)
you could combine a bigger value with a smaller value, for example a 256 tap 100k with a 127-256 tap 10k one, and then you'd have fine control within those 100k (about 40 ohm steps on the 10k pot) 
So you'd have only two pins going into the board but above the "resistor" you'll have a header with the i2c/spi pins and optionally the CS pins (to select individual chips if you have multiple on the board) and when you want to change the value, you just plug a microcontroller driven device which quickly determines the number of chips on the board, reads the values from each one and gives user ability to adjust values.

Downside :

They need power to work ... most can work from around 2.7v and up and require little power, picking a random microchip datasheet I saw 45uA at 2.7v , 15uA at 5v ... on a TI part (TPL0102) I see 0.2uA at 5v ...  so you could basically put a battery on the tiny pcb. A CR2032 with typical 225 mAh capacity will last about 1000 hours at 200uA load, much more at only 50uA that's supposed to be the typical consumption. So basically these will work for more than 1 month and then you'd need to replace the battery which is less than 1$ (cr2032 is so mass produced it's actually more like 20 cents if you buy 50-100, here's an example http://www.digikey.com/product-detail/en/panasonic-bsg/CR2032/P189-ND/31939 )

There's also non-volatile potentiometers, but you'd only need those if you want to retain the value when the chips are not powered, so maybe for long development periods where you'd have to replace the battey. but even then you could add a couple of pins to the "programmer" device and send 3v to your configurable resistor through the header (you'd have to use a couple of diodes to prevent voltage going into the battery or from battery to programmer, it's not hard) , pull out the battery and replace it, then remove the programmer header. 

the other potential downside is that the maximum current on the resistance pins is usually  very small, at around 2.. 2.5mA , with maybe bursts of up to 20mA .. for some applications this may be too small.

You've forgotten to say, that the maximum voltage for a digital potentiometer is no more than the pot's supply rails, which will be only 3V if you're running it off a CR2032 cell.

It's even more restrictive than that. None of the digitial potentiometer's pins can float outside its power supply rails. Both pins must be connected to a voltage within the supply voltage. If you want to support both positive and negative voltages, then one side of the pot will need to be biased at half the supply voltage, so with a 3V supply, that limits you to +/-1.5V.

[Conrad Hoffman]

Sometimes a simple pot is very handy. In the bad old days, bridges and such used large diameter wire wound pots, sometimes with wiper position compensation, and big dials. They could do very well in the accuracy department. Take a hint from those old boat anchors and just  use a mechanically decent pot with a big printed dial or pointer. Measure points on the pot and mark the dial. Do a nice one in CAD or just live with the pencil marks, but it will be good enough for most needs.

[thanasisk]

Here are the alpha pot readings:

A500K :
end to end 460K /
measurements of left  + right section = total  on same pot:
320K    + 164,8K = 484,8
320,8K + 166,6K = 487,4
172,5K + 312,4K = 484,9
303,8K + 184,5K = 488,3
approx 24-28K extra resistance  (~5% -6% of end-end value)


B500K :
end to end 519K/
measurements of left  + right section = total  on same pot:
135,9K + 392K = 528K
63,3K + 463K = 526K
approx 7-9K extra resistance (~1- 2% of end-end value)


A100K:
end to end 89,4K/
measurements of left  + right section = total  on same pot:
71,5K + 19,42K = 90,92K
33,77K + 59,3K = 93,07K
7,80K + 86,4K = 94,2K
approx 1.5..4.8K extra resistance (2 - 5% of end-to-end value)


B50K :
end to end 48,5K/
measurements of left  + right section = total  on same pot:
34K + 15,89K ~ 49,89K
24,53K + 25,45K ~ 49,98K
47,8K + 1,85K ~ 49,6K
48,1K + 1,44K ~ 49,5K
48,2K + 1,164K ~ 49,4K
48,3K + 50 Ohm ~ 48,3K (! smaller !  )
approx 1,5K extra resistance (~3% of end to end value)


B10K:
end to end 10,51K/
3,666K + 7,09K = 10,75K
4,56K + 6,18K = 10,74K
5,83K + 4,91K = 10,74K
approx 0,23K extra resistance (~2% of end-end value)

[thanasisk]

Thank you all for the nice feedback, material and ideas, there was a lot of material to process. In my testing of small signal audio the power would be 1/4 W tops.

Regarding ergonomics, I really think that a potentiometer solution would fit my testing workflow better. I want “real-time” indication of the actual value and do not want to plug/unplug all the time since I intend to jointly optimize  several resistances at the same time (well, occasionally)! Plus I am worried about the reliability of thumbwheel switches (really reliable ones cost a lot (e.g. mouser 653-A7PH-203-1) and i do not prefer ebay ones without any specs!). On the other hand digipots are nice but as Hero999 observed they have their major shortcomings.

To solve the tempco and the “paradox”  of total wiper section resistance not equaling the end-to-end resistance, I am considering a calibration routine as follows:

A switch such as the one used in SingedFingers’ resistance box  takes the pot out of the circuit and connects it to the ADC/uC (both wiper sections!). I then manually turn the pot end-to-end and the uC stores the wiper section curves. Or even utilizing a mechanical pot for full calibration automation would be cool


However, regarding the High tempco of pots: 
- Is the temperature dependent on the current dissipation? So if half of the pot carries current, is the resistance of both left and right wiper sections equally affected by the tempco (I guess not, in which case the above calibration routine would not be accurate!).


Wirewound pots - what are their pros compared to the conductive plastic or cermet ones  for this project?

[SingedFingers]

Wirewound = higher power dissipation, longer life. However they are not 100% linear. As the wiper moves along the resistances are in steps.

[alm]

One thing to keep in mind is that the maximum dissipation of a pot is across the entire resistance of the pot. So a 100 Ohm 1W pot can have 10 V across it without exceeding the ratings. But if you wire it as a variable resistor and set it to 1 Ohm, then it should only be exposed to 10 mW max, or 100 mV. This gets even worse for multi-turn pots that you can set very close to 0 Ohm (with close to 0 W power dissipation).

No problem if you wire it in series with a current source, but it is easy to kill a pot with a voltage source (like a 5 V power supply feeding a LED). Decade bridges will usually have resistors with similar power ratings for different decades, so if the 100 Ohm decade is rated for 1 W, then the 1 Ohm decade is likely also 1 W. It is also harder to move a decade resistor from 100 Ohm to 1 Ohm by accident.

[ScottK]

Wirewound = higher power dissipation, longer life. However they are not 100% linear. As the wiper moves along the resistances are in steps.

This is not true for multi turn wire wound pots. The wire is wound in a helix around the shaft and the wiper slides along that wire, traversing the shaft as it rotates.

Here's an example, complete with cutaway view...
http://www.precisionsales.com/potentiometers/multiturn/MW22B-wirewound.htm

You can get pots with 1% accuracy, 0.1% linearity and almost no end dead band. If you can guess the correct decade for the resistor value you need in a circuit, a decade collection of such pot/knob combinations could prove useful.

[BravoV]

In my testing of small signal audio the power would be 1/4 W tops.

When you talk about potentiometer and precision, you need at least high wattage , high turn counts, and oh yeah, a very good linearity as well.

Such as these, the top one with dial attached is "only" 5 watt acompanied by those puny 1/2 and 1/4 watters, and the one I'm talking about is at the bottom.   




Still can't imagine on it's size ? Here you go ... 



This is 15 watt rated, 15 turns and with 0.02% linearity.

[SingedFingers]

And about a million dollars a pop.