voltage dividers

[Flump]

I am making my own voltage reference
using a precision 10v regulator but I would like multiple taps from it,
for instance 10 5 2.5 1.25  , earlier i made a basic series resistor divider
using two resistors the same value but it wasnt very accurate so i then
replaced the resistors with a multi turn VR and was able to dial in exactly
the resistance i wanted, but i dont want to have 3 VR's on the board.

So I have been looking around for voltage devider IC's but not been able to find any
Does such an IC exist ?

[retrolefty]

I am making my own voltage reference
using a precision 10v regulator but I would like multiple taps from it,
for instance 10 5 2.5 1.25  , earlier i made a basic series resistor divider
using two resistors the same value but it wasnt very accurate so i then
replaced the resistors with a multi turn VR and was able to dial in exactly
the resistance i wanted, but i dont want to have 3 VR's on the board.

So I have been looking around for voltage devider IC's but not been able to find any
Does such an IC exist ?

 What kind of accuracy do you actually require? Would not using 0.1% resistors (or better) be a solution?

[Flump]

yeah fixed resistors would be great, i did not try low tollerance ones before but i do have some
0.1% 100K ones so ill try them.

i wanted to get the same accuracy as the 10v regulator which is %0.05

[Stonent]

yeah fixed resistors would be great, i did not try low tollerance ones before but i do have some
0.1% 100K ones so ill try them.

i wanted to get the same accuracy as the 10v regulator which is %0.05

You can put the .1% resistors in parallel to get 50K @ %.05

[Andreas]

Does such an IC exist ?

Hello,

Try a LTC1043 capacitive divider.
Practical reachable absolute accuracy is some ppm (around 5) for each 2:1 stage.
With a stability below 1 ppm.
You will need a precision voltage buffer like LTC1050 or LTC2057 at the output of each stage.

With best regards

Andreas

[Dave]

You can put the .1% resistors in parallel to get 50K @ %.05

According to statistics, that is incorrect.
Two 0.1% resistors would give you a combined tolerance of 0.07%. You would need to use 4 to get 0.05% tolerance.

Combined tolerance = tolerance of resistors / sqrt(N)

You can read more about this here.

[Andreas]

You would need to use 4 to get 0.05% tolerance.

Hello,

You shurely wanted to say "You would need to use 4 with a gaussian error distribution of 0.1% to get 0.05% tolerance"
Since the resistors are selected or trimmed the tolerance is usually not with a gaussian spread.

With best regards

Andreas

[Wh1sper]

Take a look to this:
Hamon Resistor http://conradhoffman.com/HamonResistor.html
How to create a precision Resistor decade
http://www.gellerlabs.com/752AJunior.htm

[f5r5e5d]

So I have been looking around for voltage devider IC's but not been able to find any
Does such an IC exist ?

http://www.analog.com/library/analogdialogue/archives/44-04/ad5791.html

[babysitter]

So you want to have standard Voltage and Resistance in your private Cal lab besides Frequency, huh ?

Consider using a resistor network consisting of several resistors in a single package. Their temperature tracks closer and the ratio will be more stable due to this compared to single resistors. Also I want to point you to ebay-available settable ratio voltage dividers, look for ESI Dekapot or dekavider. There you have several dials, say three or four decade switches and a poti for the last decade, it all depends on the model. That would be a good fit to a 10 V Source.

Also consider a impedance follower on the output to prevent turning your voltage divider into a loaded voltage divider !

[Flump]

Hey babysitter

yeah I am making frequency , voltage , capacitance and resistance reference
I know you can buy them from voltagestandard.com but I like making things
and I am trying to make this cheaply so dekapot or dekavider would break my bank

yes I have been using high value resistors in the divider to make sure I do not
draw too many mA from my regulator

so far I have 5v and 10v regulators at %0.05
100r and 5k resistors at %0.05
15nF cap at %1
and some other caps at %0.5

once it all worked out I will use a precision resister as a load on
one of the regulators to draw about 1mA and that will be my current reference

[babysitter]

OK, so the ESI stuff is out of range. For a resistive divider consider the LT(C?)5400 or some Bourns network.

Fine to have the high value resistors in the divider to reduce its current consumption, but this also increases sensitivity to load impedance. OpAmps are your friends to act as buffer / voltage follower so you have a quasi-constant load of almost nothing.

[Stonent]

You would need to use 4 to get 0.05% tolerance.

Hello,

You shurely wanted to say "You would need to use 4 with a gaussian error distribution of 0.1% to get 0.05% tolerance"
Since the resistors are selected or trimmed the tolerance is usually not with a gaussian spread.

With best regards

Andreas

You're probably right, I remember hearing that a long time ago, I don't remember the source.

[Dave]

You would need to use 4 to get 0.05% tolerance.

You shurely wanted to say "You would need to use 4 with a gaussian error distribution of 0.1% to get 0.05% tolerance"
Since the resistors are selected or trimmed the tolerance is usually not with a gaussian spread.

Yes, you are right. Assuming a perfect gaussian distribution, that is what you would get.
The number of resistors would also have to be significantly higher, if you really wanted to see statistics kicking in, 2 or 4 is much too small.

If you were to measure each resistor individually and match them, you would naturally get much better results. Tenfold improvement on tolerance wouldn't be a problem with careful matching.