ADC maths... exercise in futility - SOLVED

[Gabriel]

There is no problem with my code, it works fine.
I'm just looking for suggestions, its a long post.

I'm sure something good can come out of listening to all your opinions.

[Ttelmah]

Several comments:

First having significantly larger capacitors on the output of a regulator, than it's input, can endanger the regulator, resulting in it becoming reverse biased when you turn it off. Texas recommend a reverse biased diode across the regulator if you do this.

Then the best regulators are still an order of magnitude worse than a good voltage reference, but a zener is _not_ a good voltage reference. Look at bandgap references.

Then double check your maths. If you sample 68 times, and remove the high and low values, you still have 66 samples, not 64....

The easiest way to do this is to sum as you go:

[Gabriel]

EDIT : Hi Ttelmah.......

(my bad... sorry about that... i glanced at your name but read my version number on my footer at the same time and it say PCM... and ... you see where im going)...



i meant 66 times... sorry for that.. typo.

Thanks for the information on the reverse bias...ill implement that immediatly...

[Ttelmah]

The ADC functions internally as a capacitor fed by a resistor.
The resistor is a combination of internal resistances in the multiplexor, and the other circuitry. The capacitor is what is used to hold the voltage to be read, while the ADC is disconnected from the external source. The ADC is disconnected while the reading is made, and takes time to recharge. The time needed to re-acquire, is Tacq. Now this depends on the chip you are using, but is typically 10uSec on the commonest chips, dropping to just a couple of uSec on some of the newer chips. Assuming a reasonably fast clock rate, the comparisons, loop and sum, will only take a couple of uSec, which then means the capacitor _won't_ have properly recharged.

Several questions apply:
Chip?.
What clock rate?.
What clock rate are you using for the ADC?.

You talk about taking 2 to 3 seconds, but 66 readings done (say) 15uSec apart, will only take 990uSec, so what are you doing that takes so long?....

Best Wishes

[Gabriel]

I got it.
I'll add the delay.

I'm using a 16F88 @ 8MHz, using div_16 adc clock as per the datasheet on the TAD table.


EDIT remember i put everything through a lowpass filter (see OP)....
it take 2-3 secs for the value to stabilize in the terminal... 2-3 seconds is the time i think it takes... i see values changing immediately on the screen when i touch the sensor but because of the filter it takes a little bit to reach the final value.


Thanks for your help i really appreciate it.

G.
_________________
CCS PCM 5.078 & CCS PCH 5.093

[Ttelmah]

Probably 80% of the delay is the time it takes the chip to actually heat/cool. Remember the sensor is the IC right inside the chip...

Best Wishes

[asmboy]

I did a work for hire a while back - incubator stage heater system for approx human body temp center point with an LM34 - not the 35 -
even though the output is presented in cent degrees -
the servo code does not care.

I used an LM385-1.25 ref diode with 5uf on the ref input of the pic
and an 18F2423 - and it was very successful, stable and accurate.

I did use TWO LM34s and combined the data, but that is another story.

Now, does anybody care to tell me why i used:

** 1.25v ref,
** the pic that i did and the
** LM34 instead of the LM35 ??

[Ttelmah]

You missed one thing that would be degrading your result. Parameter A20 in the data sheet.

Best Wishes

[ckielstra]

[Ttelmah]

Yes, that is one I normally 'PIC up'.....

Best Wishes

[miro]

[Gabriel]

Hey guys thanks for all the good info.

The small caps are onboard however, i have them closer to the PIC/Chips so i didn't include them in the description of the power source.

Ttelmah, you referenced parameter A20, which by my understanding requires a minimum +Vref of 2V.... I'm using 2.5V so i "should" be Ok... feel free to correct me.

anyways, last night i tested some code and sadly i found that although I'm getting "acceptable" results (mean values)... my LM35s have become fabulous oscilators....in the kHz range...
They are connected to the board via a ~1meter cable... its 4 wire+1 bare conductor, aluminum foil shielded cable...unused wires are tied to GND on both ends.

I checked my 5V line on board and on the LM35 with my O-scope and it was smooth as a baby's bottom... so its definitely the LM35 oscilating... probably due to capacitance on the cable...sigh.

I ended up deleting all my code and using:

[temtronic]

Yes the LM34/35 are good oscillators if not wired up according to the datasheets! Used them for 20+ years for remote energy control systems.
I used 1/2 an LM358(IIRC) as a buffer at the sensor end along with some power supply decoupling(22r/10mfd).Ran simple 'quad' about 150' without any problems.
A one 'chip' solution may be the DS18B20 if you don't mind the 1 second response time for max resolution.
Yes it costs more,but weighed against an LM34/opamp/passives/Vref/pcb/etc. it probably comes out cheaper and/or easier to use.

hth
jay

[Gabriel]

@temptronic:

Jejejeje i just finished drawing up the art work for a small pcb to place at the sensor end with a LM358 for x2 amplification and buffering!

hahaha... almost exactly what you posted..


i think im going to take out the x2 amplification to save on the Resistors and extra soldering work... what do you think?

Edit: and also to avoid having MORE error introduced by resistor inacuracies... IE not exactly x2 amplification but x2.17...

G.
_________________
CCS PCM 5.078 & CCS PCH 5.093

[ckielstra]