Noise fitting frequency limits from hwp freq

[mmccrackan]

Branch to address #1071. If the config file upper-frequency limits (either the white noise limit or the global upper limit of allowed frequencies) are greater than the hwp frequency (if hwp is present), they will be restricted to 0.95 x hwp frequency (matching demodulation code) and a warning is printed. It will fail if the lower frequency limit ends up being >= upper frequency limit.

Also a small change to the limits on the noise fitting for fknee. Sometimes, zeros can occur and if alpha < 0, this gives NaN warnings.

[msilvafe]

Looks good, one comment to address and just want to confirm you tested:

1. when high_f/fwhite[1] are > hwp spin rate you got the expected correction in the values
2. the appropriate errors are raised when low f > high f
3. The divide by zero errors are eliminated by the addition of sys.float_info.min in the bounds

[msilvafe]

low_f > high_f is a check that should happen regardless of the check on hwp spin rate so I think this check/error raising should happen outside of the hwp_angle field check block.

[msilvafe]

Same comment as line 371

[mmccrackan]

Looks good, one comment to address and just want to confirm you tested:

1. when high_f/fwhite[1] are > hwp spin rate you got the expected correction in the values
2. the appropriate errors are raised when low f > high f
3. The divide by zero errors are eliminated by the addition of sys.float_info.min in the bounds

Just realized this won't work for the raw signal PSD after testing on the ISO configs. Will fix.

[iparask]

'hwp_angle' in aman should return true if it exists in its fields. So there is no need to specifically check in _fields

[iparask]

same

[msilvafe]

Looks good, one comment to address and just want to confirm you tested:

1. when high_f/fwhite[1] are > hwp spin rate you got the expected correction in the values
2. the appropriate errors are raised when low f > high f
3. The divide by zero errors are eliminated by the addition of sys.float_info.min in the bounds

Just realized this won't work for the raw signal PSD after testing on the ISO configs. Will fix.

We chatted about this but putting it here to not forget. I think the way we want to go forward with this is as follows:

1. When proc_aman is created in the run function add to the empty axismanager a field called frequency_cutoffs
2. Initially add a single field called signal which has a float matched to the cutoff of the IIR filter (should be 63 Hz but can compute from the iir filter params)
3. As additional signals are added and/or lowpass filtered (i.e. demodQ, demodU, dsT) their associated frequency_cutoffs should be added
4. In the noise calculation code check for the frequency_cutoffs.<signal_name> field for the signal_name you're computing the noise off of. If it exists check that the upper bound is < and if not set it equal to the cutoff if it doesn't exist do nothing.