12 bit A2D on 18F46K80, need a bit of help for figuring TAD and getting correct value

[ecoli-557]

Henrik-
Sorry for the late reply, been out with a respiratory illness - geeesh.
In your example :
ADValue = 3272
Temp = ADValue ** 35809 ' Multiply by ~0.5464
Result = ADValue + Temp ' Result now 5059

Why are you using 'Top 16 bits' for multiplying? and how did you arrive at '35809' as the multiplier?

I really want to learn from this, Thanks.
-Steve

[HenrikOlsson]

Hi,
We want a "raw" display value of 5060 when the ADC returns 3272. That's a factor of 1.5464, right?
The ** operator does a 16*16bit multiplication and then returns the 16 top bits of the intermediate 32bit result which is the same as dividing that intermediate result by 65536.

So, on a calculator, 3272 * 35809 / 65536 = 1787 (truncated). Add in the "original" 3272 and we get 5059 which is pretty close to 5060.

Just think of the ** operators as multiplying by units of 1/65536.
65536*0.5464 = 35809...
35809 * (1/65536) = 0.5464...
And so on.

/Henrik.

[ecoli-557]

Henrik-
So, I kinda get the top 16 bits thing, but I still don't know where you got the number 35809?
65536 * 0.5464 ----> shouldn't this be 65535 * 1.5464?
-Steve

[HenrikOlsson]

Hi,

65536 * 0.5464 ----> shouldn't this be 65535 * 1.5464?

Idealy, yes. But that would result in a value larger than 65535 which won't work without resorting to LONGs. So I treat what's on the right hand side of the decimal point separately from what's on the left hand side which is why I first multiply ADValue by 0.5464 and then ADD the ADValue to that result to get the 1.5464. See if this makes more sense:

Result = ADValue + (ADValue ** 35809)   ' ~Same as ADValue * 1.5464

Since you're using an 18F part you atleast have the option to use LONGs so if you ARE then by all means go ahead and try Result = ADValue ** 101345, it should work.

Hope I don't make it more confusing....

/Henrik.

[ecoli-557]

Is a bit more confusing........ BUT I will try what you had said. I just can't quite wrap my head about using sections of numbers....... I'll draw it out on paper and see if that helps....
I'll get it though, may take me a bit.
This forum is an amazing central clearing for knowledge and the fact that others are not only willing but patient is refreshing...
Thanks,
Steve

[HenrikOlsson]

Hi Steve,

On paper, Result = X * 3 is the same as Result = X+X+X, right?
If the above is correct then Result = X * 1.5 is the same as Result = X + (X/2), right?
And if we agree that the above is correct then I think we'll agree that Result = X + (X*0.5) is the same as Result = X * 1.5 too?

Same thing with the formula in the previous message but the number is 1.5464 instead of 1.5.

Again, ideally, we would want to do Result = ADValue ** 101345 (where 101345 is 65536*1.5464) but that will result in a compile time error if LONGs aren't enabled (because 101345 is larger than 65536 so there's a risk of overflowing the 32bit intermediate result thus returning the wrong value) so we resort to first calculating what 0.5464 is and then add that to the original value, exactly as we did in the example above.

Don't worry, I'm sure you'll get it!

/Henrik.

[ecoli-557]

Henrik-
Its probably a good thing you are not in Texas cuz your ears would be ringing over the loud 'Aha!' that I just said - Now I need to check my bathwater -(bad reference to Archimedes <grin>).
I think I can now see the light, just needed help with the switch......
Thanks Bud,
Steve