Tests reveal something interesting
As a follow up to my previous post, I had a few minutes to mess with the erratically behaving voltmeter today and I believe a few quick tests might have found the problem – two problems anyway. The solution is not yet clear, but I have some idea on how to proceed now at least.
In reading the very helpful Tips for using Single-Chip 3½ Digit A/D Converters – thank you Intersil – I performed a couple of tests using my multimeter and scope
By the way, to those other hobbyists out there – you really can’t live without a scope, even a basic analog one like mine. Helps to see what the hell you are doing.
So my first thought was to probe around and collect some voltage readings to see how effective (or not) my input voltage divider is. As it turns out, mostly, it works! Within a certain amount of error of course, but I can trim that out easily. I measured the voltage coming in to the IN HI pin of the IC and it returned values as expect in all ranges except for 2V. So there’s problem number one: I need to put a proper resistor value between IN HI and IN LOW for the 2V range. I’ll solve that later.
What was most illuminating was the 20V range. As before, a 5V input revealed close to expected voltages at that decade setting on the IC pin. It did also using a 12V input BUT at 12V input, on the 20V setting – the number was off by half it seemed. Curious.The correct voltage was reaching the IC IN HI pin, but why was it displaying something different?
In looking at the Tips datasheet (linked above) one of the questions in the accuracy section asked about a non-linear voltage at higher input levels. Well damn, that sounds like what’s happening. From the data sheet:
Problem – Above a certain input voltage level, the displayed reading does not linearly track the input.
Action – Observe the waveform at the output of the integrator stage (pin 27) of the A/D converter. There should be no clipping at the positive and negative peaks of the ramped waveform. The value of RINT or CINT may be too small, or the oscillator frequency may be too low, allowing the integrator to saturate. See previous section on component value selection.
So the integrator could be the culprit. I tried to check the waveforms with my scope and I have a hard time interpreting the results. My scope is old and hack calibrated so I’m not quite sure what I should be looking at. I probed the integrator pin and bounced up and down on DC coupling like a low frequency square wave. I probably should have it on AC coupling which revealed a very short ramp up long sustain and quick drop. The datasheet showed an example waveform that looks just like a triangle wave to me (ramp up and down) so this is not quite what I was expecting. Either I get a low frequency square wave (dc coupling) or something completely wrong (AC coupling). It said to watch for clipping, so this could be what’s happening. A flat sustain I guess could be interpreted as extreme clipping. Anyone want to chime in?
So now I have a possible plan of action. I may breadboard up another unit to make easy changes and experiments without having to solder and de-solder the built one, then apply the changes to the built one. It also suggested the oscillator frequency might be too low, but as far as I can see on the scope, it is correct and producing nice waveforms. Here’s what I have to do:
- Re-check voltages to make sure the voltage divider is not at fault
- Check value and replace the caps on the integrator pins of the IC, possibly with a bit higher value
- Probe the integrator to see if that produces the nice ramp-up/ramp-down waveform
Hopefully fixing the integrator will solve the problem and I hope I didn’t fry three ICs doing this.
Later that day…
OK well this is encouraging. I googled around (what I do when I’m obsessed with a nagging problem) and came across something I didn’t see before. Two other gents, using the same schematic are having very similar issues. Best part is they solved it (at least for them, who knows about me yet). Read here
So the coles notes version is that not only do I have to trim up my divider scheme on the input but I am probably missing a connection! I did think of this fix last night but I didn’t try it for some reason or other. It makes sense to because this ties REF LO, IN LOW, and ANALOG COMMON together giving us a fixed reference point. Both claimed that it fixed the issue and it started behaving properly which is good news!
What I find confusing is that it isn’t clear if I am supposed to connect IN LO to ANALOG COMMON and to ground, or if I’m only allowed to connect IN LO to ANALOG COMMON or ground (not both). The official datasheet seems to suggest I can’t do both but the forum comment claims success with connecting both. For safety reasons (the ICs are $3.80 a pop) I will try just connecting IN LOW to ANALOG COMMON/REF LO and see if I have success. If not, i’ll try connecting the lot to ground also.
As I am at my girlfriend’s place for a few days I will be unable to test it (oh why can’t I have everything in the same place?) but it will also give me a much needed break from swearing at the thing.
Here is my updated fix-it list:
- Connect pin 32 (REF LO) to IN LOW
- Smoke test power on
- Check that the REF HI pin reads exactly 1V from and adjust if not
- Conduct test measurements and rejoice if fixed, howl if not
If this turns out to be the culprit i’ll be so pleased as I don’t have to mess with swapping out caps, playing with the integrator, or checking every bloody joint on the board. Wish me luck!