'Change crystal freq on the fly'
I need a 1% accuracy 500KHz ceramic resonator clock for serial data
transmission, and desire 50ppm stable (crystal) oscillator for time keeping.
I would like to conserve power by switching a low speed when not recieving
serial data & processing other tasks. And I want to conserve hardware, real
estate & cost too!
It would be nice if I could clock the RTCC with a 32KHz clock, prescale it
to generate an interrupt, waking up the 16C84 a few time/second to update
the clock and poll inputs. From what I gather from the data book, sync
circuit between the prescaler & RTCC will prevent the RTCC from clocking or
an interrupt from occurng. The WDT is no where near accurate enough.
Can I count on the prescaler to toggle through the TMR0/RA4 input when the
C84 is in sleep? Even if the RTCC won't toggle, maybe I could check the level?
Switching resistors with an RC oscillator is great. Too bad I need crystal
precision. As for switching crystals, I found I could switch the 500KHz
low-z ceramic across a 32 KHz high-z quartz. When switching up to 500KHz,
the clock amplitude at 32KHz decreases slowly, around 2 seconds. I also
tried jamming a 500KHz seperate bjt clock, enabled by a port, into the
osc-in pin, and once again obtained a slowly changing, composite signal. It
would be nice to sleep &
wakeup with the 500KHz clock, that would cost more chip's & time. Although
both methods worked, I worry about software crashing and I will miss serial
data, which is what causes the increased clock frequency.
I tried a 4011, using two nand-gates as gated oscillators, a gate to
complement the enables and one to combine the clocks. Worked great. Then I
measured the current. Turned out the 32KHz sucked 40uA, and the 500KHz
around 160uA, as much as the PIC! My bjt osc only draws 15uA. I tested
4011's from a couple different manufacterer's, & found they had different
characteristics. The best result I had was using a BJT 500KHz oscillator and
a NAND gate 32KHz oscillator & for switching, and got the current down to
60uA, not including the PIC. I still
worry about asynchronous signals causing glitches & crashes.
Has anyone compared current draws vs. clock speed by CMOS glue chips from
What I'll try next is a 4060 14-stage divider/oscillator, & interrupt the
pic to test ports & update the time 8 or 16 times per second. My calc. show
at 32KHz It should draw around 30uA at 3v, and the 4060 is cheap :-)
Maybe someone has a great 20uA, single crystal, single chip solution under $1?
Maybe Microchip can put a programable clock prescaler. This would help
applications needing to shift gears to handle changing demands, & keep the
PIC away from the steep side of the Idd vs freq curve :-)
More... (looser matching)
- Last day of these posts
- In 1995
, 1996 only
- New search...