Archived Forum

My understanding of it is that there are 1024 'steps', so  0 to 1023 is 1024 steps.  In this case 1023 would be correct.

Of course, I've been wrong before....

Dennis

I think it should be 1024. The Atmel document http://www.atmel.com/Images/doc2559.pdf throws a little more light on the subject, and the diagram Fig. 26-15 on Page 288 of Atmels doc2549.pdf pretty much settles the matter, albeit for a +/- 9-bit input, by showing that the last step is 1½ LSB wide.

Pedantic arguments aside, in practice I think it makes no difference because it is going to take a seriously expensive meter to tell the difference with any degree of certainty, even on a measurement of direct voltage. Especially when the gain error is +/- 2 LSB.

Ahh, i muddled myself and meant 1024 would be correct, with the 0 being 1st step, 1023 being 1024.

"Ahh, i muddled myself and meant 1024 would be correct, with the 0 being 1st step, 1023 being 1024."

And I thought you'd got it wrong for a totally obscure reason (the 1½ LSB last step).  When you look at that diagram, and think about "nearest value", it all makes sense and the last step being different is the unavoidable consequence.

Hi.

I think both the diagram and the formula (ADC = Vin * 1024 / Vref) state clearly that 1024 is the correct value.

This said, I still don't get the 1023 in

double I_RATIO = ICAL *((SUPPLYVOLTAGE/1000.0) / 1023.0);

First, in readVcc, the ADC is used to read the (sharp) 1,1 V voltage, to calibrate the Vcc reference :

ADC = Vin * 1024 / Vref, where Vin = 1,1 V and Vref = Vcc

in other words

Vref = Vin * 1024 / ADC 
Vcc(mV) = 1.1 * 1024 * 1000 / ADC = 1126400 / ADC

Then, the precise Vcc measure is used to rectify the measurement from analogRead :

ADC = Vin * 1024 / Vref, where Vin is what we want to measure and Vref = Vcc
Vin = ADC * Vcc / 1024

which yields
Vin = ADC * (Vcc(mV) / 1000) / 1024

By the way, I totally agree with Robert : I don't need this kind of precision at all. I'm just trying to understand the code.

My guess would be the '1023' comes from the same misunderstanding that was argued in your link. I think we ought to ask Glyn.

I could initiate a pull request with a patch for this, to raise attention. But it's a bit of an overkill considering the triviality of the (potential) change. Besides, it is not that important.

I changed the topic title to be more explicit.

I don't get where you got the Vin = 1.1 mV.

It is not 1.1 mV. It is the internal reference 1.1 V. See the Atmel data sheet.