mhx / gpsdo-config Goto Github PK

Re the following clause near the top of solver.cpp::find_solutions():

  if (!is_div_one_or_even(f1, lcm_freq) or !is_div_one_or_even(f2, lcm_freq))
  {
          // NCn_LS must be 1 or even
          lcm_freq *= 2;
  }

It seems to me that the arguments being passed to is_div_one_or_even() in the above clause are backwards.

Reasoning:

My interpretation of what you're trying to do there is to enforce a value of lcm_freq for which NCn_LS is either unity or an even integer. (Let's ignore the unity constraint since it doesn't apply to the CMOS part.)

So, using f1 as an example, then it seems to me that what you want from the above clause is to conditionally scale lcm_freq by a factor of 2 such that

    f1 == lcm_freq / NCn_LS

for some potential even value of NCn_LS. This in turn requires that lcm_freq/f1 be even. (And similarly for lcm_freq/f2.)

But the second portion of the test performed by is_div_one_or_even() divides the first arg by the second, and then tests whether that quotient is divisible by 2. So the way that the args are currently passed in the above clause, it is the ratios f1/lcm_freq and f2/lcm_freq that are being tested for even-ness, rather that their reciprocals.

If you disagree with the above reasoning, can you please explain patiently why the test is being done the way it is?

Thank you.

Hi Marcus,

Per your earlier request for such examples (final comment in Issue #1), please see attached writeup.

Glenn

note.txt