Error: channel code does not exist for this value of dt

[dorogyn22]

Seisflows version: 2.3.1
pyatoa version: 0.2.2
pysep-adjtomo version: 0.4.1

I am running a simulation in SpecFEM2D with a high sampling rate (4,975 Hz). I have no issues modeling the observed data, but when I try to create synthetic traces, I receive the following error:

py3.11.egg/pysep/utils/fmt.py", line 34, in channel_code
raise KeyError("Channel code does not exist for this value of 'dt'")
KeyError: "Channel code does not exist for this value of 'dt'"

Based on the channel code appendix in the SpecFEM2D documentation, this should fall into the F band code and thus proceed to find this as the appropriate channel code. I noticed that the observed data deposited in scratch/solver/001/traces/obs uses the B channel instead of F as it should based on the time step defined in the Par_file. What causes this? Is there a workaround to force the channel code so pysep does not throw an error?

After initially configuring the parameter.yml, I update with the following parameters where I try set client to null as a workaround to this issue:

!seisflows par ntask 1
!seisflows par nproc 1  
!seisflows par materials acoustic  
!seisflows par start 1  
!seisflows par end 3    
!seisflows par data_case synthetic               # need both INIT and TRUE models
!seisflows par components Z                        # Z-component seismograms
!seisflows par adj_src_type cc_traveltime   # options: 'cc': cross-correlation, 'mt': multitaper, 'wav': waveform
!seisflows par step_count_max 10               # limit the number of steps in the line search
!seisflows par smooth_h 5000                     # smoothing distance 
!seisflows par smooth_v 5000                     # smoothing distance
!seisflows par min_period 0.025                  # minimum period of the filter
!seisflows par max_period 0.200                 # maximum period of the filter
!seisflows par client None                             # no client
!seisflows par filter_corners 4                      # limit the number of steps in the line search
!seisflows par plot True                                # plot waveform figures and source receiver maps during the preprocessing stage

[dorogyn22]

A manual workaround is to edit seisflows/lib/python3.11/site-packages/pysep_adjtomo-0.4.1-py3.11.egg/pysep/utils/fmt.py such that too high of a sampling rate does not throw an error.

`def channel_code(dt):
"""
Specfem outputs seismograms with channel band codes set by IRIS. Instrument
codes are always X for synthetics, but band code will vary with the sampling
rate of the data, return the correct code given a sampling rate.
Taken from Appenix B of the Specfem3D cartesian manual (June 15, 2018)

:type dt: float
:param dt: sampling rate of the data in seconds
:rtype: str
:return: band code as specified by Iris
:raises KeyError: when dt is specified incorrectly
"""
if dt >= 1:
    return "L"  # long period
elif 0.1 < dt < 1:
    return "M"  # mid period
elif 0.0125 < dt <= 0.1:
    return "B"  # broad band
elif 0.001 <= dt <= 0.0125:
    return "H"  # high broad band
elif 0.004 <= dt < 0.001:
    return "C"
elif 0.001 <= dt < 0.0002:
    return "F"
else:
    return "B" ## for example, to avoid any naming mismatches at later steps
    # REMOVED: raise KeyError("Channel code does not exist for this value of 'dt'")`

[bch0w]

Hi @dorogyn22, that is a high sampling rate! Glad you found a workaround but it is intriguing that this didn't work since the F channel code should still be applicable.

Is your observed data actual data or is coming from a target model simulation? I think you might run into issues later in the processing workflow when channel codes mismatch, but if they're both B then maybe not.

An example of one of your waveforms might also help to diagnose this. Thanks!

[dorogyn22]

@bch0w Indeed, I manipulated the sample rate to ensure it should still fall in the F code range but that obviously did not work. The data I am using is purely synthetic from a target model simulation. I did encounter issues later during the line search (after the first gradient) that resulted in a breaking error, although I do not know if the channel code and this issue are related. That is a problem for another thread, but I would like to make sure this is not breaking things down the line. I've attached here the waveform of station 1, XZ component after 5-40Hz bandpass. Let me know how I can help debug.