Temperature Indicator, Math, Hair Loss

stinky · Joined: 05 Mar 2012 Posts: 99 Location: Central Illinois

CCS PCM C Compiler, Version 4.132
PIC 16LF1828

I've been inside AN1333 for 48 hours now and I'm questioning my sanity. Given an ambient temperature measured at my work bench I've calculated what I "expect" the ADC to read from the temperature indicator. I understand it isn't accurate and in fact this is part of the process to perform the single point calibration detailed inside of AN1333.

The following formulas, from AN1333, were used to calculate the "ideal adc result":

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

See if this helps:
http://www.ccsinfo.com/forum/viewtopic.php?t=48263&start=3

And please don't write function names in ALL CAPS. The C standard
is for #define'd constants to be that way, but not function names.
It makes the code harder to read.

stinky · Joined: 05 Mar 2012 Posts: 99 Location: Central Illinois

forgive me for being dense, but I am not sure what I'm supposed to be seeing in the post you linked to.

I put all function names in lower case to improve readability. I also added an extra while loop to ensure that value displayed by my leds is only the most recent adc conversion. I also tried running at 4MHZ.

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Mike Walne · Joined: 19 Feb 2004 Posts: 1785 Location: Boston Spa UK

Can't argue with your logic, but two things strike me:-

1) AN1333 recommends you only use HIGH mode for Vref above 3V6.
2) Your method for reading the ADC looks rather convoluted, why not use an LCD or RS232?

Mike

stinky · Joined: 05 Mar 2012 Posts: 99 Location: Central Illinois

PCM:
-I understand now. Do you see something at fault with the way my code is trying to obtain that data? Aside of the method of displaying the data, and that I added code to poll the go_done bit, I haven't spotted a difference in operations. Please advise. I do not mean to be obtuse so I will try your code when I get back to the bench.

Mike:
-I'm hoping your first point leads me to my problem. I also spotted a table in the 16LF1828 datasheet that said those same constraints on the range setting. According to my calculations, with a Vdd of 3.292V at -40C the voltage produced by the temp indicator is 0.656V. At 85C it is 1.316V. These are both well with in the range imposed by my ADC voltage reference.
The formulas I used to calculate those numbers are:

Mike Walne · Joined: 19 Feb 2004 Posts: 1785 Location: Boston Spa UK

CCS provides sample code for both RS232 and LCD, so with your new board you should be OK.

Like I said before I can't fault you logic.
Without knowing more about what goes on inside the PIC we can only guess at why they need the voltage overhead to drive the temperature measuring system.
The app note and data sheet are both saying the same.
My intuition tells me you should be getting an error the other way (i.e. reading too hot rather than cold)

Mike

Ttelmah · Joined: 11 Mar 2010 Posts: 19512

Remember the pull down in the temperature sensor circuit, is a current sink. This has a minimum operating voltage, which is dependant on the supply. Hence the minimum working voltages. Like you, I'd probably have expected the voltage to go to high when this stops working, but your behaviour suggests this isn't the case.

Best Wishes

stinky · Joined: 05 Mar 2012 Posts: 99 Location: Central Illinois

Trying this at a higher voltage:

Vdd set at 4.56 Volts, regulated benchtop supply.

Decimal result from ADC read of Temperature Indicator: 409.
Expected result from application note: 503.

Still having this consistent offset.

stinky · Joined: 05 Mar 2012 Posts: 99 Location: Central Illinois

Something else to chew on here.

I just changed a line of code to:

Mike Walne · Joined: 19 Feb 2004 Posts: 1785 Location: Boston Spa UK

I'd be looking at the register settings.

Mike

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

I think the reason for the temperature offset between AN1333A and
your test results is due to the variation between PIC families, production
batches, and individual PICs.

Look at Figure 2 on the first page of AN1333A:
http://ww1.microchip.com/downloads/en/AppNotes/01333A.pdf
This is a graph of the forward voltage drop vs. temperature, of an
individual diode in a single 16F1937. The lowest temperature that
they measured was -40 degC, and they measured 0.659v as the drop.
That's where they got the magic number of 0.659 for their equation.

But this number is really only good for the one PIC that they tested.
For example, I tested a 16F1824 which is in the same family as your
16F1828. With a Vdd of 5v and at 27 degC, I got an ADC output of 476.

The ADC measures the voltage drop across the current source, not
the diodes. So you have to calculate the drop across the diodes:
(476/1023) * 5v = 2.33v, which is the drop across the current source.
So the drop across the four diodes is then 5.0v - 2.33v = 2.67v.
The voltage drop across one diode is 2.67v / 4 = 0.668v and that's
at 27 degC, which is close enough to their test condition of 25 degC.
Note to Ttelmah: I am using 1023 because that's what they use in
AN1333A. I'm doing it for reasons of consistency.

Look at the graph in Figure 2 in AN1333A. They measured 0.5735v
at 25 degC. But I measured 0.668v at 27 degC. That's a huge
difference in the center voltage between the two PICs. In fact, my
center voltage value of 0.668v is about what they measured at -40 degC.
This explains the huge temperature offset that you're seeing. And not
only is there an offset, but the slope could be different as well.

So basically, each PIC needs to be calibrated, and the calibration value
needs to be stored in eeprom. They explain this in the rest of AN1333A.
Then whenever you read the ADC and calculate the temperature, you
use the calibration value to adjust the result.

Here's the test program that I used to study this:

temtronic · Posted: Sat Feb 02, 2013 6:16 am

Ok, I have to ask...aside from the 'mind challenge' of getting it to work what's the real purpose of using the internal temperature sensor?
Seems like a HUGE amount of time (and $$) has, is and will be spent on this when a simple external sensor ie: MCP9701a or DS18S20 could be bought, installed and code for far less time and money.
I'm all for tweaking out a design but even I have a limit!
Maybe the OP can explain the reason to me ....

jay

stinky · Joined: 05 Mar 2012 Posts: 99 Location: Central Illinois

Jay, My real purpose for using the internal temperature sensor is coarse temperature indications. My application does not need super accurate temp info but it does need to know arbitrary hot and cold points. Your comment is quite valid and I wish I were simply using an external sensor but as of yet my time invested has not warranted a BOM change. Like I said, I don't need anything more than coarse hot and cold points.

PCM:
I believe you are quite correct. I thought there would be variance in those threshold voltages but I incorrectly assumed that they would be at or around the given forward voltage on the AN1333 test pic. Thanks for catching my mistake.

Everyone:
Thanks for your input. Invaluable members in this forum.

harlequin2 · Joined: 11 Jun 2011 Posts: 21

I'm very pleased to have read through this forum, it makes me feel a little less stupid! I, also, have been reading through AN1333 and wondering why I/they bothered.
I happen to have a small board with a 16F1826 on board driving an LCD and I thought I would turn it into a digital thermometer, just out of interest. Its not easy to make any hardware changes, but the software, well, that can do anything!
My plan also is to power it with 3 x 1.5 AA cells (the lcd works down to 3 v) and make it work forever by putting it to sleep and waking up every so many minutes, doing a temp measurement and displaying it.
I wrestled with AN1333 and I think I see how to do it all, except for measuring Vdd. I guess one would set the adc to measure ref-lo = FVR, ref-hi = Vdd and select FVR as the input channel? I'm not sure!