Designing a "discrete" pot

[ADWSystems]

For one of the newest prototypes, I need a "discrete" pot. Basically a rotary switch changing the selected resistors (8 values total). The idea is to increase current draw in known steps (8 total) using 100 ohm resistors. The simpliest thing would be to connect an increasing number of resistors in parallel at the contact on the switch (ie., 1 on the first, two on the second, three on the third, etc.). That would add to 32 resistors, some could be reduced if the resistors in parallel equate to a typically available value.

One thought was to use a string of 8 resistors and select where in the ladder I enter the circuit would determine the resistance, but the values are too close together and not readily available or produced. The resistor values for this would be 11.1, 1.4. 1.8, 2.4, 3.3, 5, 8.3, 16.7, and 50; none of which are readily available and all are very close together.

I also thought of using diodes on the switch. So the 8 100 ohm resistors would be in parallel connected by diodes at the top to each position on the switch. When position 2 was selected, current would flow directly through R2 and through a diode to contact 1 and down through R1, but that ends up incorporating a successive number of diodes in series as the dial is increased (position 3 would go through two diodes to get to R1, etc.).

If you can follow my word scramble, I would love to hear your brainstorms.

 

[crutschow]

Unfortunately I didn't follow it very well. It's not clear what output you need from this "pot". You talk about 100 ohm resistors and then a sequence of odd values (11.1, 1.4, 1.8, etc.). Do you need a linear change such as a standard pot would do, or some type of non-linear (logarithmic?) change? Do you need a pot type output or a rheostat (variable series resistance) output? I'm confused.

 

[ADWSystems]

I need a linear change in current in discrete steps. So a pot won't do, that would give me the current draws in between the steps as well. It turns out the change in resistance is by parallelling successive 100 ohm resistors where each additional 100ohm resistor adds the next amount of current draw.

I've seen designs (the fan speed control for the cabin blower in my car for one) where the have a sequence of resistors in series, in the case 3. In that design the output is at one end and the input is between the resistors. If you enter the sequence at the top, you get max resistance and low speed going through all 3 resistors (R1-R2-R3), if you enter between R1 and R2 then you get speed 2 by only going through 2 resistors (R2-R3), speed 3 is between R2 and R3 going through 1 resistor (R3), and full speed is a short directly to the output below R3. In my case the resistance values would be strange numbers (as listed above) and very close together, so not practical.

I will try to draw out each of the designs I tried to describe and post them soon.

 

[Grossel]

Hi.

Is there a demand that both ends must have free ends? Eg. not like a constant current sink that needs ground level to work.

If you tell what you're using it for it would be possible to come up with alternative ideas.

 

[crutschow]

As usual, it's much better if you tell us the problem you are trying to solve, not just ask us to implement the solution that you've already selected (which may not be the best).

 

[carbonzit]

Not exactly what you asked for, I know, but this would work:

**broken link removed**

Since you said this is for a prototype, I ASS-U-ME that means this feature might not be on the production version. In any case, it would give you the functionality you ask for, being able to increase the load (by decreasing the load resistance) in discrete steps. Easy to implement.

 

[ADWSystems]

Not exactly what you asked for, I know, but this would work:

**broken link removed**

Since you said this is for a prototype, I ASS-U-ME that means this feature might not be on the production version. In any case, it would give you the functionality you ask for, being able to increase the load (by decreasing the load resistance) in discrete steps. Easy to implement.

True that would work, but it needs to be controlled via a single input.

The whopper idea I had today while driving to/from my other job, was to swap the DIP switch with a series of relays driven by a 3914. What do you think? How about low resistance Nchannel FETs instead of the relays?

 

[ADWSystems]

I can't divulge much due to the NDA. I need to select a load that increases the current based on additional 100 ohm increments when placed in parallel like the picture below. It does not have to be parallel, but the only series implementation I have come up with isn't feasible.

 

[ADWSystems]

Not following your comment. I think one end of the resistor pack will be connected to ground, the other to customer appartus. (I will need to confirm this, I was ass-u-me-ing the ground connection)

 

[carbonzit]

True that would work, but it needs to be controlled via a single input.

The whopper idea I had today while driving to/from my other job, was to swap the DIP switch with a series of relays driven by a 3914. What do you think? How about low resistance Nchannel FETs instead of the relays?

If you use FETs, you could actually implement the rotary switch control, by stringing the FET gates together with diodes and connecting an increasingly-long string of them to gating voltage with the switch. (If you don't get how this would work I could draw it out.)

 

[ADWSystems]

I'm picturing a string of 8 diodes with each switch position. There would be a switch position, a diode cathode, a diode anode, and FET gate all connected together at each but the first and last (they would be missing a diode cathode or anode connection). Here's my question, when selecting minimum current, the last voltage of the last FET gate will be 7 diode drops less than the first diode. When selecting maximum current the FET will see the most voltage, will the FET gate handle that or work under both extremes. I'll sketch something shortly.

P.S. I thought of a similiar design using relays but was fairly sure there wouldn't be enough voltage to trigger the last relay after 7 diode drops.

 

[ADWSystems]

 

[carbonzit]

Yes, exactamente.

If the last few diode drops are a problem (don't think they will be, just set things so that the gate with the most diode drops still gets sufficient gate voltage), you could use Schottky diodes with smaller drops (or germanium: 7 x 0.3 = 2.1V).

 

[crutschow]

You would want to use P-MOSFET or PNP bipolar transistors for you switches if you are driving a grounded load.

What is the supply voltage that these switches will be connected to?

 

[ADWSystems]

I definately don't want to use PNP transistors. Why P MOSFET for grounded rather than N-MOSFETs?

 

[ronv]

I'm still a little confused.
If the input is a rotary switch like you have shown why not just switch in the resistor values like 100, 50, 33, 25 etc.

 

[ADWSystems]

Because some of the values required aren't (readily) available such as 16.3 and 14.7 ohms.

 

[Jaguarjoe]

Because some of the values required aren't (readily) available such as 16.3 and 14.7 ohms.

People like Caddock make custom value low ohm resistors that can be very accurate:

https://www.electro-tech-online.com/custompdfs/2011/07/TypeMV.pdf

Seeing as they are low ohms, hand winding these wouldn't be too big of a task if the wire was small.

Or just use a handful of multiturn trimpots. Pots would allow you to dial out errors from switch contact resistance or leadwire resistance or jfet resistance which will be significant at low ohms and high precision.

 

[ronv]

How about 10 + 6.8 and 10 +4.3.

 

[crutschow]

I definately don't want to use PNP transistors. Why P MOSFET for grounded rather than N-MOSFETs?

Because to drive a grounded load an N-MOSFET has to operate in a source-follower mode, thus the gate of the N-MOSFET must be 10V above the drain (supply voltage) for a typical FET or 5V above the supply for a logic-level type MOSFET.

So what's your beef against PNPs?