How to Voltage Divide Precisely & Temp Compensated?

I mean when i adjust a ratio, whatever if it wasn't the exact ratio i need, i should stay as it is. Let's say 1% drift in ratio. Is multiturn trimpots enough for that? Or should i find 1% resistors? metal film? os sth else?

Resistor tolerance specifications are not directly related to drift (either temperature dependant or long term). You can however expect the best stability from a matched voltage divider, two resistors in a single package. You can search major manufacturers (e.g. Vishay) or catalog distributors for respective parts. Variable resistors will be additionally affected by mechanically induced instabilities and surely aren't first choice for precise voltage dividers. If an adjustment capabilty is required, use a fixed + variable resistor combination with the variable ratio part as small as possible.

Thank you for your reply.
In my region it's not possible to find specific parts. I read that in a voltage divider that made of fixed resistors, the bigger resistance dissipate more power and get warmer. And if the temp drift is proportional to temperature, its value change more. Lets say if the cooler one have a resistance 1.1Ra then the warmer resistor will be 1.2Rb. Is that right? And the ratio will change. So why do you advise me to use fixed and variable resistors in series? Could you say what is the approximate resistance change of a fixed and a variable resistor(i have multiturn bourns) per *C ?
Thank you again

---------- Post added at 17:15 ---------- Previous post was at 15:57 ----------

OK!
I made a little test. I connected 1k trimpot to 5,04V supply and adjusted the divider to 31,7mV (one big, one small part of resistance) Then waited for about 10minutes but it remained constant. Then i put hot glue onto it and it rised to 33,2mV. The hot glue is approximately 150*C and i think potentiometers temperature raised 50*C at least. This means 0,1% change per *C.
Then i connected one 100R and one 1k2 standart 10% resistors in series to same supply and measure midpoint as 4,66V. Then waited 10minutes, no drift. Then put hot glue and it was still constant at 4,66V.
What about that

I read that in a voltage divider that made of fixed resistors, the bigger resistance dissipate more power and get warmer. And if the temp drift is proportional to temperature, its value change more. Lets say if the cooler one have a resistance 1.1Ra then the warmer resistor will be 1.2Rb. Is that right?

Click to expand...

It's a possible behaviour. But the assumption, that both resistors have equal actual TC values is just a speculation. A standard 1% metal film resistor may be specified with a maximum TC of +/- ppm/K. But how about the actual TC of individual resistors? There is a certain chance, that smae value, same production lot resistors have a similar actual TC, but it's not guaranteed. This is only different for the said voltage divider parts, or resistor array with a specified TC tracking.

The TCs of both partial resistances of a trimpot can be expected to track, however. So if you observe a ratio variation of a trimpot, it's either caused by an additional terminal resistance, or mechanical moves. Does the initial voltage restore in your test after cool down?

Then i connected one 100R and one 1k2 standart 10% resistors in series to same supply and measure midpoint as 4,66V. Then waited 10minutes, no drift. Then put hot glue and it was still constant at 4,66V.

Click to expand...

I guess, you won't find any TC or even TC tracking specification for 10% (most likely carbon film) resistors. But that doesn't mean, that they must be particular bad.

Yes the initial value restored after the cool down.

---------- Post added at 20:23 ---------- Previous post was at 20:22 ----------

I'll use as you said, two fixed and one small variable in the middle. Thank you for help

It is better to implemented in Integrated Circuit. It can be achieved with small effort.

Could you give me a code?

In integrated circuit, matching principle can be employed for your requirement.
Usually, common centriod technique is used.
The temperature difference in small area will be small too. So resistor could be matched well.

In my view, the reference to integrated circuits is mistakable in this place. It's a really hard job, to compete with the performance of good industry standard metal film resistors in IC design. In present technology of leading manufacturers, low matching T.C. and voltage independent resistors are exclusively of the thin film type. Obviously, discrete thin film resistors and even more matched pairs have comparable parameters.

Furthermore I think, utilizing IC voltage dividers isn't an option for the orignal poster.

It depends on requirement.
I think 1% for ratio accuracy can be realized for matched resistors in integrated circuits.
However, the absolute value will vary much/

i think it depends on
1> tolerance of register. it should be minimum as possible component as.
2> see the datasheet of standard resistor. it also depend upon temperature co-efficient.
3> given input voltage supply.
4> output of load. if load varies then output of voltage will very.

Thank you everyone. Could you give me such an ic name/code? Do you mean digital ic's like ds1868?

I expect you would do better with thermally matched resistors than a digital pot IC. For a high volume product try a resistor array like the Vishay ACASA1001E1002P1AT. Or for a home or school project just divide the high resistance into two BigRs in series. Put the LittleR between the BigRs, and thermally bond them all together with heatsink grease or thermal epoxy. For the best match use through hole parts of the same type and case size for all resistors, spaced off the board so they are in their own little thermal world. Preferably use the same brand resistors too.