BQ32000 Reading and Writing Registers

Dean,

I believe that all this information is in the datasheet.  Please take a look and let me know if anything is still missing.

http://www.ti.com/lit/ds/symlink/bq32000.pdf

Hi Ryan,

Unless I can't see the forest through the trees, the only mention that I've found in the data sheet is, "The time registers are normally updated once per second, and all the registers are updated at the same time to prevent a timekeeping glitch." (p.1 in Description section)

I'm currently running a continuous test where I set the clock by writing registers 0-6 in one I2C transfer and I read them back in one transfer.  So far, I haven't seen any issues with the read-back being different than expected  (some tests wait 1.5 sec for the clock to rollover before reading back the time ... all tests assume that the timer could increment once between the write and the read and take that into consideration when determining whether the value read is correct).

I'm assuming that the BQ32000 has an intermediate buffer, or similar, that all the registers are copied into when an I2C address match with Read is detected and that this intermediate buffer is what feeds the SDA line so that there's no chance of one register changing while another one is being shifted out.

I also assume that if I set up my test read method to read registers 0-6 one register at a time, that I would see occasional roll-over errors if SECONDS was read at 59 on one I2C read, then it rolled over to 0, then MINUTES was read on the next I2C read.  (even if the two reads were very close together)

The data sheet doesn't comment on this.  It says that the registers are all updated at the same time.  Meaning, I assume, that at any particular instant there is no error internally in the presented data.  Unfortunately, since I can't read all registers in the same instant, because they have to be shifted out serially, I'm not guaranteed that the values read will be glitch free. (unless there's an intermediate buffer of sorts)

So I was just wondering if there was a preferred method for reading the part to guarantee that the values read were correct.

Thanks for taking a look at my request.

-dean

Dean,

The BQ32000 has an internal buffer which is updated with the RTC counter data every second. The user only has access to this internal buffer and not the real time keeping counters. If the user is in the process of reading the time from the internal buffer, the 1 second update is stopped until the read is complete. The internal buffer is then updated immediately after the read is complete (instead of waiting for the next 1 second update).

As a result the user will never see a rollover condition while doing a read of the registers. I hope this answers your question.

Thanks Ryan.  I believe I have the answer to my question, which is that I need to read registers 0-6 in one operation to prevent rollovers in *my* accumulated result.  i.e. your description tells me that the scenario I painted earlier would cause a rollover in the data if I assembled it by reading the RTC one register at a time.

e.g.

RTC has 10:01:59 (hh:mm:ss) on the clock

Read SECONDS (get the value 59)

RTC "one second" update occurs during the previous read, but is held off until read is complete

RTC now has the value "10:02:00" on the clock

Read MINUTES (get value 02)

RTC still has the value "10:02:00" on the clock

Read HOURS (get value 10)

Assemble time inside microprocessor from values read produces the time 10:02:59 -- clearly this is not correct.  It should be either 10:01:59 or 10:02:00 but because the registers were read one at a time, instead of six at a time, the rollover in the RTC found its way into the microprocessor.

Reading the registers the other direction (from HOURS to MINUTES to SECONDS) wouldn't help as a rollover between minutes and seconds would produce a time of 10:01:00.

So, I have to use one TWI bus transaction to read registers 0-6 in a single operation.

Please forgive me for being so pedantic and thank you for giving me a better idea of what's happening inside the RTC.

-dean