probably more of a math question...

Bobb Fwed

I am running an RC Time to measure voltage. I have a few data points, but I do not have the math know how to convert it to a readable voltage (I have a similar problem with a rctime temperature measurement).
Here is the info (collected off my own RC Time object). This is the voltage range I need to read, but reading beyond 10.41 and 15.85V would be great too. It doesn't have to be perfectly accurate (but within 0.2V would be great across the range below), and for ease of programming I can use integers with a 10x multiplier (e.g. an output of 123 is representative of 12.3V).
@0.1µF, >2ms low
26000 = 10.41V
24249 = 11.00V
21673 = 12.00V
19585 = 13.00V
17785 = 14.00V
16241 = 15.00V
15073 = 15.85V

Obviously having something written in spin, or some instruction on how I would do it would be great.

Bobb Fwed

Here is the best I can do... 2546050/rctime. This gives me exactly 130 (13.0V), but is 0.3V and 0.7V high at 14V and 15V respectively and the same difference going to a lower voltage. I know it is going to have something to do with exponents and whatnot. I'm just running off trial and error to get these answers. Any help would be nice.

lonesock

Well, I just used Excel's "solver" to optimize the equation:

Voltage = A/counts + B

A = 195598.0736
B = 2.9486494

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
lonesock
Piranha are people too.

Bobb Fwed

thanks! is that part of excel? or is that some type of extension? I have excel 2002 (10) does it have that feature, if so, where can I find it?

lonesock

Bobb Fwed said...
thanks! is that part of excel? or is that some type of extension? I have excel 2002 (10) does it have that feature, if so, where can I find it?

You are welcome. The solver is an Add-In. It comes with Excel, but is not available by default. Go to the Tools->Add-Ins menu, then check the "Solver Add-In" box and hit OK. It will install for you, then "Solver..." will be available under the Tools menu.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
lonesock
Piranha are people too.

Tracy Allen

This just came up a day ago in another thread:
http://forums.parallax.com/showthread.php?p=753026

I plugged your values into the formulas:
V1: HIGH voltage applied to circuit = 15.85
RCT1: corresponding raw RCtime value = 15073
V2: LOW voltage applied to circuit = 10.42
RCT2: corresponding raw RCtime value = 26000

K1 = 10 * (V1 - V2) * ( RCT1 * RCT2 ) / ( RCT2 - RCT1 ) = 1951062
K2 = 10 * V1 - ( K1 / RCT1) = 29

The spin code should be then,

Voltage := 1951062 / RCT + 29 ' corrected typo, 1951062 not 21951062

This is the explicit solution of two equations in two unknowns:
V1 = K1 / RCT1 + K2
V2 = K1 / RTC2 + K2
and applies whenever V1 and V2 are sufficiently greater than the Prop threshold (~1.65 volts).

edit:
These values are close to the ones that lonesock found from the EXCEL solver, which I think uses a least squares fit. The above formula fits the curve exactly at two points.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Tracy Allen
www.emesystems.com

Post Edited (Tracy Allen) : 9/23/2008 5:33:45 PM GMT

Bobb Fwed

What worked much more accurately: 1978085 / RCT + 29
run the numbers...this is much better (within 0.1 -- it looks like -- until you get above 15.5V).
(It only give 1 decimal value, but that is good enough for me)

Bobb Fwed

actually... 21951062 / RCT + 29 isn't very good at all....
... 24249 == 11V ... 21951062 / RCT + 29 gives me 934 (9.34V?)
... 17785 == 14V ... 1263 (12.63?)

Tracy Allen

Sorry, a typo, an extra 2 in front of the number. Corrected now.
Voltage := 1951062 / RCT + 29 ' not 21951062

1951062 / 26000 + 29 = 104 (10.41 volt ' reference point)
1951062 / 24249 + 29 = 109 (11 volts)
1951062 / 17785+29 = 138 (14 volts)
1951062 / 15073 + 29 = 158 (15.85 volt ' reference point)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Tracy Allen
www.emesystems.com

Phil Pilgrim (PhiPi)

With enough data points, you could also use linear interpolation, without needing floating point.

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Still some PropSTICK Kit bare PCBs left!