Hi Didier,

True this could be done in SW, and I should have mentioned that I
considered that in my post. However, these chips offer several other
functions that would add complexity to the code, and I've been looking for
a reason to do a simple HW project and open source it through OSH Park
anyway. I'm not looking to do this for just myself.

Looking across the TI and Microchip lines a bit, it strikes me as odd that
more micros supporting integrated RTCs actually use two I/Os for a 32 kHz
crystal option. Why they support an RTC is not mysterious at all, but why
not the option at to drive the 1 Hz clock directly rather than dividing 32
kHz down to 1 Hz and using an extra I/O is odd when these I/Os are usually
configurable anyway.

Russ

You want to drive the RTC with an external PPS to get time/date into an

Hi

In volume a 32KHz crystal is a sub 5 cent item. In high volume very sub 5 cents. The RTC chips are way more expensive.

Bob

On Nov 11, 2013, at 8:06 PM, Russ Ramirez russ.ramirez@gmail.com wrote:

Because the 32Khz xtals are profoundly cheap and accurate to some extent.
Long divide by 2 chain is very simple. The xtal could be easily adjusted
close to the correct frequency. It all sort of came from watch technology.
Regards
Paul
WB8TSL

On Mon, Nov 11, 2013 at 8:06 PM, Russ Ramirez russ.ramirez@gmail.comwrote:

Hi Russ,

I will venture that the vast majority of applications are served with 2 pins and a $0.10 crystal rather than the external silicon implied by a 1Hz input.

The advantage of off-chip timekeeping is the low power consumption of dedicated RTC chips that makes them able to run from a coin cell forever. If you have a GPS receiver or worse, an OCXO or a Rubidium, low power at that level cannot be a serious concern.

As a matter of fact, aside from time-nuttery, in 45 years of designing hardware and almost as much writing software, I have never come across the need for precision timing in a microcontroller project.

Complexity in code is something you only have to pay for once. You pay for hardware with every product you build. In a small run and if you are pressed for time, it often does make sense to buy hardware rather than write software, but in the vast majority of cases, software is the solution.

Didier KO4BB

Russ Ramirez russ.ramirez@gmail.com wrote:

--
Sent from my Motorola Droid Razr 4G LTE wireless tracker while I do other things.

Typically you need more fine time resolution them just a seconds counter.
Especially if you are using an Arduino, you are maybe building a robot or
real-time controller and want to measure something like "Milliseconds per
revolution"  so you need a faster running counter and then you want to
calibrate that counter

A typical method is to have the PPS trap the faster counter, than you can
see how many periods per second your counter is moving.  You can mount for
a 1,000 seconds and get a pretty good idea.  Then you use the fast counter,
now that you know it's rate for you timing.

So you don't need an external chip if you are willing to track the rate of
the free running clock.

--

Chris Albertson
Redondo Beach, California

Many times the 32KHz oscillator is just used for ultra-low power rudimentary CPU timing, not so much for a "time of day" clock. The crystal gives more reliable timing than an RC delay, and the low power works fine for a "press to wake up" type application.

Bob LaJeunesse

Le 12 nov. 2013 à 03:13, Chris Albertson a écrit :

True, if you don't need to know what your epoch is.  Although the DS indicates GPS PPS is a choice for the reference, why bother as you get your date and time to a better resolution on first fix than any RTC is going to give you. The advantage of this chip I think, is that if you have a reference, you get to keep time with much better resolution than a normal RTC without GPS. If I am interpreting the DS correctly, the 1Hz out is held within 7,8ms of the reference, which is pretty good over very long Tau. Maybe the key is that one of the accepted references is 50/60Hz  power line frequency. That nets billions of end users.