# Digital Potentiometer

#### Anna_123

Joined Sep 20, 2023
6
Hello All,

I want to sense a temperature inside a controlling Unit ( which is a ultra low freezer), I am trying to design a testing board, I wanted an Digital potentiometer which can give a 0.1 degree precision when compared with the PT1000 temperature sensor Table. Since we have to achieve an 0.1 degree precision,I need a IC with the higher resolution, Can you recommend how to overcome this issue?

#### MrChips

Joined Oct 2, 2009
29,802
Welcome to AAC!

You need to look at the resistance vs temperature chart of your temperature sensor over the applicable temperature range.
For example, if your range is 0-50°C, then ±0.1°C is 1 part in 500. The resistance deviation could be lower than this.

Look for a digital potentiometer with 1024 steps.

Another solution is to use two potentiometers in series. For example, one pot could have a span of 100kΩ (course setting) while another pot could be 1kΩ (fine setting).

You can also replace the course setting pot with high precision fixed resistors to span different test temperatures.
Put the fixed resistors in a box with a rotary switch or a collection of toggle switches to select the desired resistor.

#### crutschow

Joined Mar 14, 2008
33,325
If you are trying to emulate the resistance of the PT1000 temp sensor, that's not easy, since digital potentiometers have poor accuracy of their absolute resistance.
Don't know offhand how to generate an accurate variable resistance electronically for many steps.
How many steps do you need?

#### Anna_123

Joined Sep 20, 2023
6
Welcome to AAC!

You need to look at the resistance vs temperature chart of your temperature sensor over the applicable temperature range.
For example, if your range is 0-50°C, then ±0.1°C is 1 part in 500. The resistance deviation could be lower than this.

Look for a digital potentiometer with 1024 steps.

Another solution is to use two potentiometers in series. For example, one pot could have a span of 100kΩ (course setting) while another pot could be 1kΩ (fine setting).

You can also replace the course setting pot with high precision fixed resistors to span different test temperatures.
Put the fixed resistors in a box with a rotary switch or a collection of toggle switches to select the desired resistor.

Another solution is to use two potentiometers in series. For example, one pot could have a span of 100kΩ (course setting) while another pot could be 1kΩ (fine setting).

How can this be achieved?

#### Anna_123

Joined Sep 20, 2023
6
If you are trying to emulate the resistance of the PT1000 temp sensor, that's not easy, since digital potentiometers have poor accuracy of their absolute resistance.
Don't know offhand how to generate an accurate variable resistance electronically for many steps.
How many steps do you need?

0.1 degrees per one step with the resistance of 0.39 ohms

#### crutschow

Joined Mar 14, 2008
33,325
0.1 degrees per one step with the resistance of 0.39 ohms
That doesn't tell me how many steps of resistance you will test.

#### Alec_t

Joined Sep 17, 2013
14,005
Don't know offhand how to generate an accurate variable resistance electronically for many steps.
In some situations it can be generated using PWM to switch a resistor in and out of circuit, but that approach might not be feasible for an accuracy of 1 part in 500.

#### Anna_123

Joined Sep 20, 2023
6
That doesn't tell me how many steps of resistance you will test.
I didn't get your point could be please explain a bit more precise?

#### ericgibbs

Joined Jan 29, 2010
18,212
Hi @Anna_123

Moderation.

#### Jerry-Hat-Trick

Joined Aug 31, 2022
444
Maybe I've got the wrong end of the stick, but the first question I'd ask is what range of temperature are you trying to simulate? My understanding of PT1000 temperature sensors is that the resistance changes by about 4 ohms per degree Celsius - or 0.39 for 0.1 degrees you mention. Bear in mind, it's fairly linear, but not completely. If you are trying to measure over a fairly narrow range, which is probably the case for a freezer application, it should make the solution simpler.

Apart from the fact that digital pots typically have an end to end resistance tolerance of +/- 20% I think the smallest end to end resitances available are 1K. Take a look at the Analog Devices AD5254 quad digital pot with 257 steps. Coincidentally, for 1K end to end resistance with 257 steps the nominal step voltage is 3.9 ohms, so how about using two and a half of these ICs for 10 resistors in parallel to get 0.39 ohms per step? To compensate for non linearity you could tweak the value of one or more of the resistors to compensate if you have an accurate way of calibrating. These parallel resistors would need to go in series with a fixed 680R resistor if you are looking to measure from say -80C to around -56C for example

#### crutschow

Joined Mar 14, 2008
33,325
I didn't get your point could be please explain a bit more precise?
Seems clear to me.
How many different values of resistance will you actually need to generate when this variable resistance is used to test the PT1000 circuit?
Is it 5 values, 10 values, 100 values...?

#### Anna_123

Joined Sep 20, 2023
6
Maybe I've got the wrong end of the stick, but the first question I'd ask is what range of temperature are you trying to simulate? My understanding of PT1000 temperature sensors is that the resistance changes by about 4 ohms per degree Celsius - or 0.39 for 0.1 degrees you mention. Bear in mind, it's fairly linear, but not completely. If you are trying to measure over a fairly narrow range, which is probably the case for a freezer application, it should make the solution simpler.

Apart from the fact that digital pots typically have an end to end resistance tolerance of +/- 20% I think the smallest end to end resitances available are 1K. Take a look at the Analog Devices AD5254 quad digital pot with 257 steps. Coincidentally, for 1K end to end resistance with 257 steps the nominal step voltage is 3.9 ohms, so how about using two and a half of these ICs for 10 resistors in parallel to get 0.39 ohms per step? To compensate for non linearity you could tweak the value of one or more of the resistors to compensate if you have an accurate way of calibrating. These parallel resistors would need to go in series with a fixed 680R resistor if you are looking to measure from say -80C to around -56C for example
Yes, I want to simulate for -100 degree to 100 degree temperature range with the resolution of 0.1 degree.

#### Anna_123

Joined Sep 20, 2023
6
Seems clear to me.
How many different values of resistance will you actually need to generate when this variable resistance is used to test the PT1000 circuit?
Is it 5 values, 10 values, 100 values...?
The temperature is from -100 degree to +100 degree, multiple resistance values but it should be within this range.

#### crutschow

Joined Mar 14, 2008
33,325
multiple resistance values but it should be within this range.
Okay, we are narrowing it down.
"Multiple" is not a specification, which is what I'm trying to establish.
Accurately defining the specifications for a project is often the most difficult part of the job.
So how may resistance values is "multiple"?

You do understand that the more values you need to measure, the more difficult the design is likely to be.

Last edited:

#### Jerry-Hat-Trick

Joined Aug 31, 2022
444
Forgive me, but I’m struggling to think of a real world application which needs to be able to measure temperature to 0.1 degree accuracy over a range of 200 degrees Celsius. If you are happy to do this in bands of 25 degrees with a manual, or even motor driven rotary switch with resistors in series with the array of parallel digital pots I previously suggested, to switch between bands, it’s maybe possible. Or you could use the two spare digital pots in series with the ten in parallel to give you the necessary 800 ohm range.

For each of the ten pots in parallel, put each one in series with small resistors in stepped values from zero to 3.9 ohms which will give you the possibility to fine tune the total resistance by small amounts. The calibration process would be complicated but I really wonder if this is a real world problem. Unless you are needing to automate this for volume manufacturing a simple manual resistance box with rotary switches would be a better engineering solution

#### MrChips

Joined Oct 2, 2009
29,802
1 calibration point would do if it is a straight line fit and you can establish another reference point.
2 calibration points would do if it is a straight line fit.
3 calibration points would do if there is a slight bow in the curve.
4 calibration points are better than 3.
5 calibration points would cover most practical temperature ranges.

#### schmitt trigger

Joined Jul 12, 2010
763
I believe that a R-2R resistor ladder made with precision resistors would provide a vastly more accurate reading.
A 9 segment ladder would provide a resolution of 512 steps.

As others have already mentioned, digipots have wide tolerances and a large Temperature Coefficient.