----- Ursprüngliche Nachricht -----
Von: Mark Rages
Gesendet am: 11.Juni.2009 18:54:13
>> My own experiences with the UART programming is that the internal RC clock
>> is way too unstable and exact to produce a stable communication.
>> With 4MHz RC clock I was unable to establish a reliable link with 9600 Baud,
>> while with the same code and an external 8 MHz quarts even a baudrate of
>> 115KBaud was no problem.
>> At least you'll need an external 32KHz clock-quartz and some timer-driven
>> code to build something like a software-PLL to keep the RC oscillator in the
>> proper range.
>Were you using the calibrated clock?
>"Internal Frequencies up to 16 MHz With Four Calibrated Frequencies to<1% " <-
>from msp430f22x2 datasheet.
>Later on in the datasheet it claims <2.5% over frequency, still close
>enough for reliable async comms.
>If the frequency is off, it's off -- the baudrate makes no difference.
Won't help on the 1611, as it has no calibrated clock source :)
Also, even the calibrated source won't help if the device has a large operating
temperature range.
If it has to operate in industrial environment with -25 to 85 degree celsius
(or even just 0..70) the temperature drift is likely way too much for a stable
communication. So you still have to measure temp and adjust the
divisor constantly.
The 1% is for one of the four given frequencies, the 2.5% over the whole
frequency range, but all just for a given temperature.
I don't have one of the 2xxx datasheets as we don't use it, but what does it
say about temperature drift?
----- Ursprüngliche Nachricht -----
Von: Hardy Griech
Gesendet am: 11.Juni.2009 21:39:44
>> So we decided to put an 8MHz quartz on it and all problems were gone.
>> Together with the increased speed as bonus.
>I hope, that either your CPU clock is not directly connected to the 8MHz
>or your board uses 3.6V as supply voltage. See the datasheet of the
>msp430f1611, figure 1.
Sure we have 3.6V. It usually works with 3.3V too (the 8MHz dropout limit is
somewhere around 3.1V), but we didn't take a risk.
Anyway, the quartz can be connected to the CPU clock even with lower voltage,
but then MCLK (and all others) need a divisor of at least 2. (I guess this is
what you meant)
JMGross
