Accuracy of digital multi-meters?

[messerschmitt]

I recently built a digital to analogue converter using a resistor ladder matching them to .05% but measuring the output shows that some are out by a fair few mV but then it catches up towards the top, not as linear as I had expected. Could it be the multi-meter thats not linear? It's only cost £20. Has anyone got any experience with how linear cheaper meters are?

 

[kellys_eye]

What value resistors are you using in the ladder? What are you using to switch the resistors? Is the switching device 'contact' resistance linear?

 

[messerschmitt]

What value resistors are you using in the ladder? What are you using to switch the resistors? Is the switching device 'contact' resistance linear?

100k and 200k the switching is 4042 for the first 4 bits and a 4013 for the last 2. its a 6bit D to A. its from the mfos voltage quantiser I included the pdf.

 

[shrtrnd]

How did you 'match' them to .05%?
Did you use your inexpensive multi-meter to do that, or select manufacturer-marked tolerance resistors?
Somebody here will probably study your schematic, there's a lot of circuitry involved there, come of which could account for the discrepancy you see.

 

[BobK]

With resistors that high, the impedance of your voltmeter comes into play.

Bob

 

[Alec_t]

As Bob says. DMMs commonly have a ~10 Meg impedance. 10 Meg in parallel with 200k brings the resistance down by ~2%.

 

[messerschmitt]

I used a component tester that measures them, not sure how accurate it is but it reads to a few decimal places and is consistent so I can get ones that are very close in value so even if its a few ohms out doesn't matter as long as they are similar values correct?

 

[kellys_eye]

Not really. Close tolerance resistors are not just matched by 'resistance' they are also matched to their temperature profile so simply using 'selected' resistors from ( say ) a bunch of 1%ers doesn't guarantee 0.05% accuracy.

 

[messerschmitt]

Oh I see, I'm always amazed at peoples knowledge and help, just when I think I understand something theres always something more lol. its probably the resistors that are causing the non linear results then, once again thanks for everyone's input and help.

 

[Terry01]

Ha try being a total newby! I see resistors,read a little and think...cool,they're a thick pipe or a thin pipe for electricity. I understand that kinda as pipes was a big part of my job....but then....
Not really. Close tolerance resistors are not just matched by 'resistance' they are also matched to their temperature profile so simply using 'selected' resistors from ( say ) a bunch of 1%ers doesn't guarantee 0.05% accuracy.

EH??? WTF!!!...ok Terry...its confirmed..just as you think your starting to "get it". Your not,not even nearly!

 

[Terry01]

Lucky I like it and find it very interesting
The things you learn here are awesome!

 

[(*steve*)]

I've only briefly looked at your correct, but you seem to have a 6 bit DÀC made from discrete resistors.

R2R networks are very dependent on tolerance, but also that the ratios are accurate. If you matched the 200k resistors, did you use 2 of these in parallel for the 100k resistors?

Also, you don't specify what a few millivolts is compared to an expected difference in the LSB of the DAC. Two or three mV difference in 78mV send fine to me.

Perhaps you should check out the effect of resistor tolerance in R2R networks. Using 1% resistors (for example) wont give you voltages accurate to 1%.

I note your circuit uses an op amp to buffer the output voltage, so loading should not be an issue if you're reading the voltage at the output of the op amp.

 

[(*steve*)]

The Wikipedia entry on resistor ladders is pretty good. The penultimate section talks about the resistor accuracy required to retain monotonicity in the output (i.e. that the output always goes up as the binary value increases). The next constraint would be that the step is within 1/2 the LSB of the previous value, and the next that it be within 1/2 LSB of the desired value. For your ladder, the transition from 011111 to 100000 is the most demanding. If it's between 39 and 117mV higher, then your ladder passes the basic test. From the sound of it, yours does somewhat better.