Failure (aparent) on a test board, from a temperature sensor

[Poty]

Good morning (at least here). As my first post, I want to introduce me (briefly). My nick is Poty, I`m on GMT-3, and I work exclusively with PICs. English isn´t my mother language, but I try to do my best. Corrections on all this (PICs and language) will be acepted.

Anyway, to fulfill the subject, I have a test board with a ZIP socket, which allows me to connect up to 40 pins PICs. It is a nice piece of work, but I found a little failure. The temperature sensor is a TC1047A, SMD type mounted on the main board, which is not very useful, but to practical purposes, works. Its temperature range, is -40 to +125ºC, and its output, goes from 100 to 1750mV. Of course, I use ADC, and all the music to read it,(considering a lineal response, with a formula that considers offsets, etc, etc, etc).
Well, the problem is that I can´t get even an approximate value to the real temperature, and repeating lectures (ADC), gave me always different values.
It has not OP AMPS in the middle (connection is DIRECT to an ADC pin of the pic) so, there is no circuit in the middle that can modify the signal or so.
Repetitive lectures are realized every 3 seconds, time more than enough to correctly finish ADC each time.
As contrast, I used 2 different tools: a dedicated termometer with termocuples (I don´t remember the brand now) and a Brymen tester with a termocuple. Both instrumens gave me the same value, testing the TC1047A, while this one sent odd values to the processor.
Any suggestion?
Cheers.

[Sideshow Bob]

Could it be it be that  the high on low byte in the conversion result are merged in the wrong order. This can give quite errattic readings that appear very wrong. Done something like this my self You need to in some way trace the raw reading from the ADC(two bytes) and compare them to the input voltage

[solutions]

Agree - pull the sensor out and inject a low noise DC level and read it with your code as a temperature. Do same for another voltage. And another.

[Poty]

Good evening. I´m sorry for the delay, I got muted because someone thought I was another muted user getting a duplicate account.
Unfortunatelly, even on ADSL, in this place, connection cuts are more than common, and that means a change of IPs... probably I got one that other user used to annoy someone.

About the invitation, you´re very kind, Sideshow Bob.

Going back to the topic, I`ll use both suggestions. Fortunatelly, I don´t need to disconnect the sensor (it is a SMD one), just disconnect the PIC from the ZIF socket (no zip.. ) and this will allow me to test it directly. Then, I`ll verify the bytes´ order, and let you know what I find.
Regards,

[pickit2]

the TC1047A is an easy device to test, you just put a meter on the output and read 10mv per degree.
Used it a lot 10 years afo, if memory is correct.

[metal]

seriously, how do you read ADC, and how do you check ADC readings are being correctly done? put your ADC code here so we can help further, what you are trying to do is extremely simple..

[Poty]

Good morning, mates. I´m a little late with those suggested tests, but as soon as I finish with some rubbish things that I have surrounding me, I´ll show you some advance about this topic.

Regards,

[pickit2]

Microchip has a pigtail board, I have a few in my loft, but they can stay there, till forever.
if memory serves it only has a strip of connector pins, to fit dev-board.
on the pcb it has one TC1047A and one cap. thats it.

if you apply power and connect a meter you can read the temp as 10mV per degree.
they were very accurate but we found you could fine tune a programed pic&TC1047A by the suply voltage.

our boards always used var-regulators by design, and saves cost.

[Sideshow Bob]

Also you will never get dead stable readings from a PIC ADC. The least significant bit(s) will always fluctuate. How many bits that will fluctuate will depend on many factors like how good your power supply is. You design and layout will have its say. Anyway what may help you is to have your readouts be an average of several samples not just one sample alone. You can easy divide a binary number (at least a posetive) by shifting it to the right. If you shift it once you divide it by two, once more 4. Next time 8 and so one

[LabVIEWguru]

I'm not familiar with your device, but do you have it wired with proper polarity? I hear frequently "Device "X" doesn't work - it shows -40 degrees C" and I tell them to reverse the connections to the data logger and it works like a charm.

[Poty]

Good morning. Any idea is a good idea to make someone think. No issues with polarity. Besides it comes mounted on a test board, or "educational" board, I checked how it was connected. It is very interesting to verify the stability of the power source. I`m using a commuted one, to get 9V, and the test board has a 7085 with some capacitors to get 5v nice and clean, but may be useful to check with the oscilloscope, and see what happens. Good suggestions.
About average readings, Sideshow Bob, it is not a bad idea, but I have first to check the status of the sensor.
Let me arrive to tests, and I´ll let you lnow.
Regards,