0
\$\begingroup\$

I am using ADS1231 to get values from a load cell,i am using the ADC in following configuration.enter image description here i am powering the micocontroller and loadcell by 5Volt supply as shown in above image. i am getting the results as expected and everything works fine.

Questions :

load cell that i am using is 2mV/volt type, and it says in the datasheet that the ADC input range is form -19.5 milivolts to +19.5 milivolts. In my case i am supplying the ADC with only 10 mV.

is there any way to increase the range ?
am i using the full potential of ADC capabilities?

Thankyou.

\$\endgroup\$
  • \$\begingroup\$ Your load cell range should be specified as 2 mV per volt per "mass". We don't know the mass and we don't know full scale mass so your question can't be answered yet. \$\endgroup\$ – Andy aka Jun 18 '18 at 11:19
  • \$\begingroup\$ Are you sure you need it? Load cells are often around 1% accurate. Your 10mV over a 40mV range gives you ~22bits of the 24 bits which is a factor 40.000 more then what you need. \$\endgroup\$ – Oldfart Jun 18 '18 at 11:22
  • \$\begingroup\$ @Andyaka the load cell range is 2mV per volt i.e. for full load the output is 10 mV .i have a 20 kg load cell and it is fitted in a jig. so when i place 20 kg weight on load cell jig the output is 10 mV. effective range form no load to full load is 0 -10 mV. \$\endgroup\$ – Mr.Sky Jun 18 '18 at 11:24
  • \$\begingroup\$ @Oldfart Technically yes , but with the noise in circuit i am able to get only 40,000 effective counts also as i am using only positive (0-5v supply) i am totally losing out on the negative range . \$\endgroup\$ – Mr.Sky Jun 18 '18 at 11:27
  • \$\begingroup\$ @Oldfart It's actually surprising how much precision you need for load cells, and it's not just the naive approach. Have a read up on weigh scale design, it was quite eye opening when I designed a commercial unit... \$\endgroup\$ – awjlogan Jun 18 '18 at 11:31
3
\$\begingroup\$

In the comments you stated that the problem is noise in the circuit. That is a crucial bit of information that you left out of your original question. If the problem is circuit noise then changing the ADC range isn't going to help.

First, you need to understand the noise. Take an FFT of the noise and see if it has a large component at your local power grid frequency. If so, then you clearly need to clean up your power supplies. You should also try running the whole thing from batteries using only linear regulators to see if that helps.

Make sure that the noise isn't coming from the digital side of your design. The AVDD and DVDD lines should be well isolated. If these two supplies are tied together at some point you may need to add passive LC filtering between them. You haven't told us how you physically constructed the circuit, but the proper routing of power and ground on a PCB is also important.

If the noise is truly random with a Gaussian distribution and its amplitude is greater than 1/2 LSB then you can use oversampling to improve the effective resolution of the measurements. In this case, averaging four measurements will improve the resolution by 1 bit.

\$\endgroup\$
1
\$\begingroup\$

Eliot Alderson's answer is correct, but let me expand a bit. At 5 volts power supply, your ADC range is (roughly) 40 mV (2 x 19.5, close enough), with a resolution of 24 bits. A 0 to 10 mV input will, then, convert to 1/4 of full scale, or 22 bits. That's about one part in 4 million. This is far, far beyond other sources of error, such as non-linearity of the load cells. Just stop worrying about that part of it. You have resolution and to spare. In rough numbers, 1 lsb will give you 5 milligram resolution (not grams, milligrams) and that is not a realistic goal.

What you also have is noise. From your comment, about 1000 counts of noise. That should be your concern. At a start, put 0.1 uF caps at each ADC input, connected to ground. Of course, you should also get a good capacitance meter and match the two caps as well, since if they are not matched well their different impedances at any frequency will cause problems. But even with not-well-matched caps they will likely help, and the experiment is simple enough that you should give it a try.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.