MCP3301 A/D converter SPI reading

[Ioannis]

Your ADC is differential one. So, that means it can convert positive and also negative (in relation to the two inputs + and -) voltages.

If the + input is more positive than the - input of the converter then the output will be a plain 12 bit binary number.

But if the - input is more positive than the + input, then as you can understand the result of the ADC should be negative. This is shown in binary with the most significant bit (MSB) of the value being 1. That is the sign bit.

But the rest of the 12 bits follow a specific rule then. All bits are reversed and also -1.

For example, 1 1111 1111 1111 means negative number and this number is the reverse of 1111 1111 1111 plus one, or just 1. So in total -1.

Or, 1 0000 1111 0000 is also negative and the value is -3856

Hope this clear a bit of the mud!

Ioannis

More here https://www.cs.cornell.edu/~tomf/not.../twoscomp.html

[flotulopex]

Thanks Ioannis and sorry for the late reply,

So, if I would connect GND to the "-" input and measure i.e. a + signal from 0 to 5 Volts, I should have as 12 bits result, right?

Or, 1 0000 1111 0000 is also negative and the value is -3856

This is confusing to me. I get the positive and negative thing but I can't figure out how you come to this 3856 number?

[richard]

to expand a 13 bit negative two's complement number to a 16 bit number bits 13,14 an 15 need to be set accordingly
in general the "sign bit" needs to be extended to the new width
so 1 0000 1111 0000 becomes 1111 0000 1111 0000 qed -3856

[Ioannis]

Yes, Richard is correct. I forgot to mention you have to adjust the value to the byte, word or whatever width you have.

So the sign is the left most bit but the rest should also be 1. Remember we reverse the bits to get the absolute value disregarding the left most, sign bit.

Seems confusing at first, but has its logic.

Ioannis

[flotulopex]

Got it now!

Thanks a lot

[tumbleweed]

Remember we reverse the bits to get the absolute value disregarding the left most, sign bit

To convert a 2's compliment negative value to a positive number, invert all the bits (NOT) and add 1.

For example 2's compliment -1 = FFFF
invert = 0000, add 1 = 0001

[Ioannis]

I was reffering to the left most bits but yes. Absolutely correct, to convert a positive to negative after reversing all bits you add one.

Ioannis