Difference in resistance of in dividual "steps" of digital pot

[Flyback]

Hello,
Do you know what is the differencve in resistance of the individual "step" resistances of this digital pot (MCP4014)?
https://ww1.microchip.com/downloads/en/DeviceDoc/21978c.pdf

...The datasheet tells 20% resistance accuracy, but not how similar are the individual "step" resistances.

 

[be80be]

About 33 ohms a step

 

[languer]

Probably some trick question, but here it goes:
According to page three on the datasheet, the Step Resistance is RAB/ 63 (AB being either 2.1kΩ, 5kΩ, 10kΩ, or 50kΩ). Table 6-1 gives you the values depending on which dash number you have. Since the 4014 is a rheostat, you want to pay careful attention to Table 6-2 since the wiper resistance will play a role on the overall resistance. Those parts are laser trimmed I think; so the different between steps (i.e. Table 6-2) is very well controlled. But INL will give you an indication of what amount of deviation you can expect from step-to-step.

 

[be80be]

There 64 steps so it 33 ohms about if your using a 2.1k digital pot. But there is lot's of things that can change it temp voltage so the op would more or less have to test the one hes using.

 

[Flyback]

thanks, we will now use it in "potentiometer" mode, (mcp4013), but still , the datasheet does not allude to the differences in resistances of each individual step?

 

[Gobbledok]

Hi Flyback I think you need to look at the non-linearity table.

At 5.5v the non-linearity is +/-0.25 LSB for the 2k1Ω IC.

 

[be80be]

Theres 64 steps any way you look at it you 2100 5000 0r what ever it is is / 64 = but from what the data sheet tells you to really find out you need to step it and read its ohm value. There's a lot of variables voltage temp non-linearity. I'd test drive one today

 

[Flyback]

the thing is the voltage will be the same for each step..so too the temperature, and presumably the non-linearity...so surely each step will be very similar? Does 0.25LSB means 0.4%?, ie 0.25/64?

 

[Gobbledok]

+/-0.25 LSB means that each step can vary from 0.75 - 1.25 the actual step value.

EG:
0.75 x (2100/63) = 25
1.25 x (2100/63) = 41.66663

Therefore each step can vary from 25 ohms to 41.663 ohms.

 

[Flyback]

thanks, though does it mean each step will vary by the same amount , so that all the steps tend to be the same value as each other, whatever that value is?

 

[Gobbledok]

From Wiki:
"Differential nonlinearity (acronym DNL) is a term describing the deviation between two analog values corresponding to adjacent input digital values."

So no, each individual step can be off by that amount. EG: step one may take you to 41.663 ohms and step 2 may take you to 66.663 ohms.

It's actually a bit worse than that; the datasheet says these are typical values. The actual deviation may be up to +/-0.5 LSB.

 

[be80be]

If you look at the charts you'd see that these are not prefect there two types only way to really see how its going to be is to test it.
They work like there counter part you turn it and it may go full scale way before you hit the last step.
Like at the start it may have a nice 33 ohm step then it jump to a 60 ohm step middle way test it.

 

[Flyback]

thanks, this is awful, why bother making such an inaccurate digital pot.?..are they all like this?...this is terrible....surely you agree the inaccuracy is dreadful?

 

[Gobbledok]

For sure it's terrible, I was quite surprised myself. I haven't had a look at a lot, but it seems they are all pretty much the same.

What is the application, maybe we can suggest something better?

 

[kubeek]

thanks, this is awful, why bother making such an inaccurate digital pot.?..are they all like this?...this is terrible....surely you agree the inaccuracy is dreadful?

Most likely that beacuse for a lot of purposes it doesnt really matter.

 

[ronsimpson]

It's actually a bit worse than that; the datasheet says these are typical values. The actual deviation may be up to +/-0.5 LSB.

thanks, this is awful, why bother making such an inaccurate digital pot.?..are they all like this?...this is terrible....surely you agree the inaccuracy is dreadful?

Some DACs are made with a string of resistors just like the digital POT. +/-0.25LSB or +/-0.5LSB is common. I have seen larger than 1 LSB.
Mechanical POTs are not real linear. (the wire wound seem good) I find dead spots after usage and noise. A mechanical pot has 'backlash' so the same point will measure differently each time. At least the digital pot will measure the same after usage and has to backlash.

Using ADC or DAC a +/- 0.5 LSB is fine. Why is it a problem in a POT?

 

[Flyback]

for the 2k1 version of mcp4013 digital pot.......every step is 32.8 ohms.
And 0.25 * 32.8 = 8.2 ohms.
So any step could be anywhere between 32.8+/-8.2 ohms.....that's terrible....that's 25% error...most resistors you buy are 1% ....

I mean, in theMCP4013, you could have the first 10 steps being 41 ohms each , instead of 32.8 ohms, and that gives you a very inaccurate output voltage from the pot.

 

[Flyback]

Hello
I don't think we have interpreted the MCP4013 datasheet correctly.
page 41 of the mcp4013 datasheet shows that in fact,10 steps would have an accuracy of 328 ohms +/- 8 ohms, which is very accurate, do you agree with "328 ohms +/- 8 ohms"?

 

[be80be]

It's like I said dang test it. Where you think i got that 33 ohm steps at just made that up LOL.
And no most resistors don't have 1% tolerance and most are at 20% but then agin that depends on what your doing.
You turn up the volume on the radio it don't matter much would it?
But if you needed a dead on 1.1 volt from a 5 volt divider then you buy a 1% resistor.

 

[Flyback]

theres no pont in testing it until we understand the batch to batch tolerance...if we test, we can only test one, or a few parts, that's not enough...we need to know, say, if we are on the 30th step, then in the ideal case, that would give us a voltage of 30/64 *1V2......but there is a tolerance in this, and we have no idea what it is.......yes we can see it in one part...but that's no good, since as you know, the next batch may be totally different.