AVR Freaks

Why use a 1337 and a kludged CR2032 battery when you could be using a 1307 with its proper backup battery design that doesn't power other parts of your circuit?

...but there are pullup resistors on the SDA and SCL lines somewhere as well?

What are those diodes for after the regulator?

Yes.

So you have a bit-banged I2C, do you do illegal push-pull operations on the line too or does it magically work without pull-up resistors 99% of the time by using leakage currents?

Within what spec? Two of them just drops more voltage while single diode would suffice. It's a 5V chip after all.

Which datasheet would that be? I have here the Maxim DS1337 datasheet which says it is good for 5.5V operation, absolute max being 6V. Do I have the wrong device type or is it a some subtype of DS1337 or from different manufacturer?

If it really is a max 4V device, then it has already went kaboom, as it won't consume anywhere near 1mA, and two shottky diodes won't drop even near 800mV at 1mA.

You may be on track here, you read datasheet version 012003 which does have 4V maximum, while I read a datasheet version 7/09 directly from manufacturer website which has 5V maximum.

So based on that, maybe that is why the diodes are there, trying to drop the voltage from 5V to 4V, but as far as I can tell they fail at the job (because they are Schottky diodes so about 0.3V drop per diode), and look suspicious. Thus the question what they do there, or trying to do, was a bit rhetorical and we can't know for sure, so there is no right or wrong answer in my opinion.

What the real problem (Again!) is that a hardware designer does sloppy job and software engineers have to live with that, compensating quirks in software.

If the I2C bus specification says there must be resistors on the bus, then there must be, and that is final. Another thing is if it can be compensated with software tricks, and to some extent it may be possible. And my answer to that is, not reliably on an AVR, as you can't have reliable transitions on bus that has to go through lo-hiz-pullup or pullup-hiz-lo sequence. Stray capacitance may help a bit. Some people also use push-pull SCL output, but it requires that all other chips never try to drive SCL.

Edit: Actually that battery-resistor-diode is somewhat valid, as seen from some Dallas/Maxim datasheets. Schottky diodes have quite a big leakage current, and the resistor limits charging of battery from regulated 5V when available. It may even be some kind of UL requirement to have both.

You are right, the drop is max 650mV per diode at 20mA current, but the RTC chip does not consume even 1mA. Diode has minimum drop of 260mV at 2mA, so even if the RTC chip consumed 2mA, the voltage could still be 4.48V.

It is unfortunate that us software engineers need to be the ones that tell hardware engineers what works and what does not. So it is a good thing to get involved in hardware design reviews so that you have a chance to point out hardware guys how to connect microcontroller pins the optimally and what is possible with software and what must be done in hardware. Friends! :)

Why do you need to change bit-bang programming if only pull-ups are added? So you have used the AVR IO pins to output logic high, haven't you?

By illegal bus use I mean you must never ever force a logic high on the bus by setting AVR pin as output high, as it is an open-collector bus.

Devices are only allowed to pull bus low, or let it float high with the resistors. Otherwise there is a chance your AVR pulls high while slave chip pulls low, so there is a shortcut of two drivers fighting and extraneous current flowing through IO pins.

Chips do not need to operate reliably under those conditions.

I already said the battery resistor *may* be because of UL safety requirements. Safety requirements say you must have either resistor or another diode there as a backup solution if single diode fails.

And no the Schottky diode is not enough to prevent charging, as safety requirements need less than 1uA of charging current, and the diode itself cannot guarantee that if the device is used in ambient temperature is larger than about 60C.

Same with the two VCC diodes - either they are required by safety regulations or it is a bad way to try to drop voltage.

We don't know why, we can only speculate.

Fortunately, the schematic shows you're not using /INTA, so you can change over to a DS1307 and cure this problem once and for all at the next board spin. Meantime you can drop a DS1307 in the same footprint to test it. Pull out the diode and resistor in series with the battery and connect it (the battery) directly to pin 3, and short out the other two diodes to supply full 5V to pin 8. With a couple of lines of firmware change you're done.

There are other similar clock chips that fit this same footprint, eg M41T82 (ST Microelectronics) and FM3130 (Ramtron), but they're more complicated to drive.

(the question should have been "can I use the internal pullup resistor on the input pin rather than the external pullup?" - to which the answer is more debatable (but I wouldn't do it personally))

Since you're using the square wave as a stable timing interrupt, you should use a pullup of 10k or so. No doubt it will work using just the internal pullup, but depending on the stray capacitance, you may get a shark-fin rising edge with a jittery switching point.