I seem to be getting confused by the resolution/accuracy quoted. The Industrial Dual Analog In Bricklet quotes a resolution of 24bit, which on the 70VDC full scale would give (70/2^24) roughly 4 micro volts resolution. The accuracy is quoted as 4mV. Is the resolution really 1000 times better than the accuracy?

Basically, yes. To achieve real 24bit accuracy you would need a crazy expensive analog frontend in front of the adc. This is out of the reach of a Bricklet.

 

The performance of the Bricklet is comparable to a modern highish end 30000 count multimeter (e.g. http://www.amazon.com/Agilent-U1272A-Handheld-Counts-True/dp/B004URG9EU).

 

The big difference is of course that the multimeter not only has a +-30V range but also a +-30mV range.

 

 

Edit: Do you need a really high resolution in the full +-35V range? Or would a +35mV version of the Bricklet help in your application?

Thanks for the speedy reply!

I am measuring a 0-5V signal, where ideally I would like to achieve an accuracy of +-1mV.

If I could get the range down to a 6th of current, that would give +-5.83V range which assuming the accuracy would be also be 6 times improved would work for our application.

Would it be as simple as exchanging R8 and R10 to 4k7/6 resistor? Going for something readily available, like a 787 ohm (e.g. http://uk.farnell.com/1752470)?

 

Edit: Resistor labels changed to read R8 and R10

The circuit uses a differential amplifier as "frontend" (link to schematic). This amplifier multiplies the input voltage by R8=R10/R13=R14. Please note the formula in the lower right corner of the schematic. By changing the resistors R8,R10,R13,R14 you can adapt the measurement range according to your needs. 

Ah yes, sorry, I meant to say R8 and R10 in my last post and should have multiplied the resistor by 6 rather than divide.

 

If I change both R8 and R10 to be 33k (e.g. http://uk.farnell.com/1717610) and leave R13 and R14 at 280k, I should be able to get a range of +-5.09V (at the input to the U1B) to provide the required 0.6V the ADC0 net.

 

Will I then need to recalibrate U4? to convert the reading would it be as simple as dividing the result from the API by 7.021 (based on 35.745/5.091)?

@magnaparva

Have you try the modification?