Temperature is one of the most often overlooked
specifications in electronic design. We spend most of our time in controlled
climates and it’s easy to forget that the products we create might not. The
problem is that most component suppliers know this and either don’t include any
data about their product over temperature or only give you their ratings at the
peak conditions, i.e. the ideal temperature, voltage, light, etc. The casual
electronics person/lifetime tinkerer might not even consider designing for
environmental conditions because in their case they just need something that
works in one specific instance. In fact, the reason I chose to write briefly on
this subject is because I have been caught by temperature restrictions in the
past, and I was almost caught again with my pumpkin animation project.
Last Halloween, I attempted a lamer, less thought out
version of this project that involved some kludge of LEDs, resistors,
batteries, and switches. It didn’t go well in part because I used Lithium coin
cells as the power source. To this day I have not confirmed my suspicions, but
Lithium-ion batteries are notoriously sensitive to temperature variations, and I
think because of the low temperatures outside the battery voltage dropped and
the LEDs did not stay lit for very long. There are of course many other reasons
to explain why my first attempts were failures – one being that I have little
skill designing stuff – but I am inclined to think it was the temperature that
did me in. My main reason for thinking that is because once I brought the
circuit back inside and it returned to room temperature, the LEDs started to become
bright again.
To avoid being screwed over by lame batteries again this
year I decided to pick up some 9V Alkaline Enercell batteries from our local RadioShack.
The idea here is that I will step down the voltage from 9V to 5V to power the
PICs that control the lighting sequence using a linear regulator (currently on
its way from Digi-Key). To make sure that I wouldn’t lose any battery voltage
when the temperature drops at night, I tested one of the Enercells at varying
environmental conditions. The first test (pictured below) was conducted at
around room temperature. The second was done after I left the battery out in
the night air for a few hours. The last was after I left the battery in the
freezer overnight.
Figure 1. Battery voltage across various temperatures (From Left to Right: room, fall night, below freezing)
As you can see, there shouldn’t be any voltage drop on the terminals
at all based on any realistic temperature conditions on Halloween. I expected
this given the battery chemistry, but RadioShack is infamous for not providing
datasheets for any of their parts so I had to be sure before I relied on them.
As long as Halloween night stays above absolute zero everything should work
fine as far as the batteries are concerned. Current projections have the nighttime
low set a 42° F.
Unfortunately, 42° F could be problematic for another aspect
of my design. In trying to find a workaround for my oscillator timing problem I
came across Figure 2 in the datasheet for the PIC 16F690 (lazily colorized for
your enjoyment).
Figure 2. HFINTOSC frequency accuracy over supply voltage and temperature
What this graph is showing is the high frequency internal
oscillator (HFINTOSC) accuracy over temperature and voltage. Remember how I stated
in my last post that the accuracy of the internal oscillator is only quoted at
1%? And remember how I said earlier that datasheets often quote the best
metrics up front without qualifying the conditions for their execution? Well according
to their own chart I can only get 1% accuracy if I run the PIC on 3.5V at room
temperature, plus or minus about 10° C. Running the PIC on 5V alone is enough
to push the accuracy to 2%, and when you consider that temperatures could get
as low as 40° F (4° C) I am trending dangerously close to pushing that accuracy
to a horrendous 5%. While all this sounds really bad I am not in panic mode
just yet. Based on some reading I have been doing I think I can work around
this by shortening up the animation times, which will limit the amount of time
the four oscillators in the chips have to drift away from each other. I am also
working with putting the PICs to sleep when they don’t need to turn on LEDs and
then waking them back up using the Watchdog timer (WDT). From what I can gather
the watchdog timer triggers a processor reset when it overflows so I think I can
use that to re-sync the clocks after each light sequence. With only four days
left to work, getting this project up and running could come down to the wire.
2 comments:
Down to a hare's breath, even
Isn't the watchdog timer stepping based on an oscillator also? Or is it some independent hardware counter with timing based on physical characteristic of the circuit? I really don't know.
Post a Comment