Non-Linear Resistance Control

[AnalogKid]

alec beat me to it (stoopid day job). A LUT lets you map any input value to any output value. And depending on the assembly language, compiler, and the equations involved, it doesn't automatically take up more memory than the calculations it is replacing.

ak

 

[vne147]

I could use a lookup table to solve my linearity issues, but...

Perhaps I should have been more specific with my initial post. My desire to have a linear relationship between digi-pot setting and output voltage really stems from two main reasons. The first being that it will simplify my code, but the second and arguably more important reason is to maximize resolution.

If I were to find a way to make the relationship linear using a 256 position digi-pot, then the resolution of output voltage I could expect would be somewhere around:

[LATEX]\frac{10V - 1V}{256} \cong \pm 35 mV[/LATEX]

If I don't find a way to make the relationship truly linear, but use a lookup table to make it approximately linear by essentially skipping over some positions within the digi-pot, then the maximum resolution I could hope for would be equivalent to the largest change in Vout between any two digi-pot positions.

In my case using a 30 kΩ digi-pot, the largest ΔVout occurs between positions 0 and 1 and is about 750 mV. That's not enough resolution for me unfortunately.

I'm still trying to crack this nut.

 

[vne147]

I think I may have found a solution.

If I use a digi-pot in a voltage divider to make a nonlinear variable gain op amp as described here:

https://www.maximintegrated.com/en/app-notes/index.mvp/id/864

And then use the output of that as the Vin here:

https://www.maximintegrated.com/en/app-notes/index.mvp/id/3284

I think the end result would be a digi-pot that changes its resistance non-linearly when the digi-pot in the variable gain op amp configuration has its setting changed linearly.

Does anybody see any red flags with this idea? Am I making life difficult for myself?

 

[ChrisP58]

I'm not sure that that is going to give you what you need. It may give you a somewhat linear correlation between your control and the output, but I don't think that it'll give you the resolution that you need, as the number of steps in the pot used as RFB2 are the same.

 

[vne147]

I'm not sure that that is going to give you what you need. It may give you a somewhat linear correlation between your control and the output, but I don't think that it'll give you the resolution that you need, as the number of steps in the pot used as RFB2 are the same.

After thinking about it a little more, I think you're right. This would essentialy by the same as using a look up table. I just got excited when I thought I found something new. Back to the drawing board...