RCTime result "wrong"?

jmspaggi

Hi there.

I'm doing something wrong, but what?

I'm trying to use RCTime with a resistor sensor but I'm not able to translate the result.

I'm trying the RCTime schema 9.6a page 169 with a 4.7uF capacitor, the 220 resistor, and R=1200. I have replaced the sensor by a 1.2KO resistor to simplify the research.

RCTime is returning 316.

Based on the fomula time = τ(ln(Vbegin/Vend)).
With t = R

jmspaggi

I just tried with 9,4uF instead of 4,7uF and got 647.

(1200 * 0.0000094) * (ln (5/2.5)) / 0.0000086 = 909.

So both values barely double, which is correct, but they are still not equivalent.

JM

Peter Verkaik

Have you measured the actual component values?

If I do the calculations I calculate rccount = (R*C*ln(2))/(8.68usec) = 450

316 is therefor (316-450)/450*100% too low = -30%

If you use a electrolyt cap, these usually have a tolerance of +/- 20-50%.

Doubling the cap value almost doubles the rccount value,
so the circuit appears to work.

regards peter

jmspaggi

Hi Peter,

Thanks a lot for your reply.

I measure the R values and they are 215 and 1182. Which should give me 895 instead of 909.

Regarding the cap, it's a tantal one. Should I replace it by and electrolyt one?

Can it only come from the cap efficiency?

Thanks,

JM

Peter Verkaik

A lower crystal freq also leads to lower count values,
but you probably suffer from a not fully discharged cap,
which means the cap is charged to threshold voltage
in a shorter time, hence the low count.

Make the rcTime pin a low output and delay for 10msec (CPU.delay(105))
Then call rcTime(rcPin)

regards peter

jmspaggi

Hi Peter,

Here is what I have so far:

CPU.writePin(CPU.pin14, true);
CPU.delay(1000);
System.out.println(CPU.rcTime(32000, CPU.pin14, false));

Do you mean I should "rever" my schema, and try:

CPU.writePin(CPU.pin14, false);
CPU.delay(1000);
System.out.println(CPU.rcTime(32000, CPU.pin14, true));

?

Thanks,

JM

jmspaggi

Just did what you proposed, and I now get 361.

How is the 220 resistor modifying the results? Should I put something closer to the 1.2K I have on the other side?

JM

Peter Verkaik

Here is the schematic I have in mind

······················· +5V
························ |
····················· [noparse][[/noparse]1200R]
························ |
JavPin14 o----[noparse][[/noparse]220R]

---

+
························ |
······················ =====··4.7u
························ |
·························V· GND

The 220R is used to discharge.
You can use·5*RC to fully discharge = 5*220*4.7u = 5.17·msec --> use 10msec

CPU.writePin(CPU.pin14,false); //discharge
CPU.delay(105); //10msec
int value = CPU.rcTime(32767,CPU.pin14,true); //changes pin14 to input

regards peter

Peter Verkaik

Remember, the lowest voltage accross the cap will be
(220/(220+1200))*5 = 0.775V

Charging·from 0.775·to 2.5

U =·0.775 +·5*(1 -·e^(-t/RC)) = 2.5
5*(1 - e^(-t/RC)) = 2.5·-·0.775 = 1.725
1 -·e^(-t/RC) = 1.725/5 = ·0.345
e^(-t/RC) = 1 - 0.345 = 0.665
-t/RC = ln(0.665) = -0.42312
t = RC*0.42312 = 1200*4.7u*0.42312 = 2.386 msec
count = 2.386 / 8.68 * 1000 = 275

(316-275)/275 * 100% = ·15%

That is very good,·considering not using the exact component values
You get better results if the 220R is lowered or the 1200R increased.

regards peter

jmspaggi

Hi Peter,

Just tried that too.

My delay is bigger to ensure cap is fully charged. Should I reduce it? I'm only processing data every 10 seconds. So big delay is not an issue for my application.

I also tried to chance the cap by other 4.7 caps, but result's still the same.

I will update the delay with 105 and see the result, but I don't think this will change it.

Thanks,

JM

jmspaggi

Probleme is it's for a temperature sensor for hot water.

R is 1000 at 0C, and should be 1232 at 60C. If I have a 15% error, than mean it can be between 54C and 56C. Range is way to big for my application.

How can I have a more accurate value?

Increasing 1000R is easy. I just have to add another 10K in serie. So sensor will go from 11000 to 11232.

Can I "simply" replace the 220 by 110, and 1000 by 10000 and redo the same math as what you did?

Also, thanks a lot for all those formulas! Very helpful!

JM

Peter Verkaik

That should work and result in a count roughly ten times the current count.

regards peter

jmspaggi

Thanks for your help Peter.

I did some tries with different R values, and results are pretty linear, but always at about 20% from the expected result.

What I noticed is that they are constant. If it gives me 485, it will never vary more than 2% from this value.

So I have put my temperature sensor on the water, and in the fridge, and I will take the value for 0C. Then I will boil it, and take for 100C. I will use those 2 values do extrapolate the others.

JM