If it helps, the code I used is copy-able. Not sure if it fits into your code easily, but it’s available if you want it. The files are tk-voltage.h, tk-calibration.h, and bistro.c; look for “BATTCHECK_VpT” to find the relevant sections.
The method I used is a table listing the raw ADC value followed by a packed number of blinks. So, two bytes per entry — 8 bits of ADC value, 3 bits of 1.0V blinks, 5 bits of 0.1V blinks. It iterates over that table to find the correct entry, unpacks the blink values, then, um, blinks. This was the most compact way I could think of to do it without sacrificing accuracy.
A smaller but less accurate method is to assume that 0.1V equals exactly 5.0 ADC units. This eliminates the table. But in my testing, it’s usually in the range of 4.4 to 4.8 ADC units per 0.1V, so it will skew as it gets farther from its origin value. This effect could be reduced by using the full 10 bits of ADC measurement and doing everything in 16-bit math, but that takes extra room too.
Funny, because "I think" my brownout support isn't actually turning ON brown out detection. I'm using fuses: E2/DF/FF and it still works fine. So guess I should have called it simply NOINIT or something equivalent? I believe I tried it disabled and enabled at 1.8V and notice no difference.
Didn't realize (or forgot) bout the extra parasitic drain. Is there anything more to it? Any other initialization I'm unaware of? I'm thinking the extra parasitic drain is coming from enabling it via the fuses?
I use this var in my code:
// OFF Time Detection volatile byte noinit_decay __attribute__ ((section (".noinit")));
and it seems to work fine. I don't think there's any other initialization needed. This is in my main start-up code:
if (OffTimeEnable) { if (!noinit_decay) { // Indicates they did a short press, go to the next mode next_mode(); // Will handle wrap around's SaveConfig(); } } else modeIdx = 0; if (modeIdx == 0) { TCCR0A = PHASE; blink(2); } // set noinit data for next boot noinit_decay = 0; // will decay to non-zero after being off for a while
Well, actually any volt-nut worth his salt needs three of them… man with two clocks never knows what time it is…
With a little patience, you can probably find a decent used one for less than $3000. I paid around $2400 for each of mine.
Proper calibration will probably set you back $500. I have the equipment to calibrate them (the SR104 10K resistor retails for $10K, you don’t want to know what the Fluke 732A/732B voltage references cost). There is a 3458A on Ebay right now that is up to $2000… calibrated… closes in 3 days.
That device looks cool! Still $66 USD, plus shipping I'm thinking but in a reasonable range. I gotta talk to my EE friends at work - I think they've done this without high end instruments. We got scopes, bench PS's, signal generators, flukes, etc., but don't think we have anything like that bench DMM -- not sure though.
Actually, any decent quality multimeter should be able to read down to micro-amps. I use Tektronix DMM912/914/916 meters. No longer made (?) but usually available on Ebay. With some patience you can probably snag one for less than $100.
Ok - think I got the new "n.n" battery status blinking working pretty well, thanks from T_K .
I've partially integrated in her new headers, and tweaked the "n.n" blinking. With the default values for blink timing/rates she had, I really couldn't figure it out. The values I'm using now are more like the blink timing/rate a NiteCore MH20 does on it's switch LED. I took out the period blinking and made the blinks longer/slower. I find it really nice this way. I used the PWM values for voltages T_K had defined, measured by Dale for the wight FET+1 driver. There were only 5 values defined for the original 1-5 blinks, so I filled in all the values in-between for the full 2.0v to 4.4v levels. I tested it on a few cells at various levels and seems to always be within 0.1v. Probably could be tweaked a little but seems very close and reliable.
So far I'm pretty happy with how it's working in the UI - from OFF, do 1 click followed quickly by a click&hold - then the first digit of the volts blinks out (2-4 times), then pauses, then the 2nd digit blinks (0-9 times). This repeats over and over with a 2 sec delay in-between. One click aborts it.
Oops, did I make it do the readout too fast? I tried to make it quick but not too quick, but maybe I went too far?
BTW, if you want to get all the necessary voltage values, simply measure a full cell and an empty cell with battcheck.c, then feed the values into battcheck.py and it’ll spit out ADC values for each 0.1V value. Example input files are in battcheck/readings/*.volts .
It may then require minor tweaking to get things just right, but the values should at least be pretty close.
It’s weird having to offload all floating-point math since the attiny is terrible at it. I’m spoiled by full-featured CPUs where floating point is quick and easy.
Really dunno. Did you test it? The msec values looked really low. Was wondering then if they were actually intended to be in msecs?
I tried it with your values and could not distinguish the blinks from each other. The ratio of 1:4 of ON:OFF doesn't seem appropriate for this type of thing. Dunno if it's just my ol eyes/reaction or what. But I found the MH20 NiteCore really clean - very easy to follow, so sort of modeled the timing after it.
Edit: Nevermind. Think I didn't have the same BLINK_SPEED as you had - I did something wrong. Look'n again at the bistro.c code, you have 500 msecs pauses around the period - that's not how I ran it... Still, I like the way it came out.
I zipped up the whole solution/project. There's 3 required header files now - they are required to be in the same folder as the main source code file (eswBrOutCfg.c). It requires Atmel Studio 7.0 now to use it as-is. That google drive folder also has the batch files for doing the AVRDude stuff: fuses, dnld, etc. The fuses haven't changed.
Next: expand simple strobe to a bunch of blinky modes.
Mike - this sounds awesome btw! Sounds like you got about everything covered. I haven't looked into this much because of not having a dual switch light to use it on. I recall seeing a thread or post w/details on this mod, but not sure where now. I think
Got more done this eve on adding blinky modes - think the UI is good. Didn't test it yet, but should have a 12.5 Hz strobe, random strobe, police strobe, and 10 sec beacon mode. Can easily add more or pick and choose at compile time now - planning on a config UI for these "special" modes (what I call them).
Also fixed a couple of bugs I noticed in strobe and batt check. You can also now abort from batt check quickly.
Hhhhhmmmmmm......
.
.