I bought one of those Real Time Clock modules:
I noticed it has a diode and resistor to charge the battery on the back. But I plan on using a non-rechargable CR2032 cell. So these had to be removed:
I also removed the power LED, since I plan on using it in an alarm clock with a partly translucent enclosure, and I don’t want the light to disturb me at night.
I think these modules mostly get used with primairy (non-recharable) lithium cells, so why was that diode/resistor there in the first place? Charging those cells might lead to leakage (at best), or more spectaculair failure (explosion?) at worst.
If you had battery leakage or other issues with these modules, please let me know in a comment.
If you have one of these DS3231 modules and plan on using it with a standard CR2032 coin cell (non-rechargable), I’d strongly recommend removing the diode and/or resistor, as shown in the pictures above.
I’v measured the 32 kHz output on my 2 modules with a Agilent 34401A. According to the datasheet, this should be 32.768 kHz, typ@3v3. One of my modules outputs 32.763 kHz, the other 32.761 kHz at 3v3. This is 5 resp 7 Hz slow: 153 resp. 213 ppm, way over the specified 3.5ppm over temperature… It would cause me to be late at a rate of about 2 minutes a week. For comparison: my wristwatch does 30 Seconds a Month. So, probably I got out-of-spec or fake modules… Darn!
EDIT2: Those modules get used/sold with rechargeable/secondary cells as well. Makes little sense considering how long a primary cell would last in this application, but at least explains why there is a charger circuit on board.
EDIT3: More about the 32-Khz-out-of-spec problem, in Dutch, here: https://www.circuitsonline.net/forum/view/147354. Also:
I’ve bought a known-good chip from Farnell and plan on comparing performance – more later. The chip I bought is a DS3231MZ+ , which has a MEMS-oscillator, not a crystal like the …SN has. So cannot be compared 1-to-1, because its 32 kHz output is not compensated (only the internal 1 Hz is, within 5 ppm.)