[LinuxPPS] linuxpps 2.6.32-rc8 question

[Jeff Angielski]

clemens at dwf.com wrote:
>> I have my PowerPC 85xx target connected to a GPS receiver that is 
>> generating a PPS which is in turn generating an interrupt on my host.
>>
>> I am trying to figure out why my system clock is showing such a large 
>> discrepancy in time (10ms/s) and was wondering if anybody had any tips 
>> on tracking this down.
>>
>> Does everybody runs this in a calibration mode to determine the drift of 
>> the local clock and then use adjtimex() to fix the offset?  The error is 
>>   so large that NTPD won't use it as a reference source.
>>
> 
> Well, I know nothing about this specific setup, but motherboard manufacturers
> seem to feel that spending more than about 10c for a quartz crystal is a waste
> of money.  Still, yours seems to be particularly bad.  Almost reminds me of
> owning a Sun.
> 
> But seriously, you MIGHT want to try using 'tickadj', which is in the ntp package
> to bring the clock on your computer closer to the 'real world' .  There is a 
> description of the program and its arguments in the HTML section of the
> ntp sources.

Problem solved.  Our target boards were stuffed with 33.000000 MHz 
oscillators instead of the designed 33.333333 MHz.  So the system clock 
[and all the derived clocks] were being programmed incorrectly in u-boot.

I fixed u-boot and there was much rejoicing.

-- 
Jeff Angielski
The PTR Group
www.theptrgroup.com