Need some clarification regarding the getting of commanded states

[jzuhone]

Hi @taldcroft and @jeanconn,

@Gregg140 and I had some questions about what is going on in this code snippet:

dpa_check/dpa_check.py

Lines 267 to 277 in ca60736

	# If cmd lines options were not fully specified then get state0 as last
	# cmd_state that starts within available telemetry. Update with the
	# mean temperatures at the start of state0.
	if None in state0.values():
	state0 = cmd_states.get_state0(tlm['date'][-5], db,
	datepar='datestart')
	ok = ((tlm['date'] >= state0['tstart'] - 700) &
	(tlm['date'] <= state0['tstart'] + 700))
	state0.update({'T_dpa': np.mean(tlm['1dpamzt'][ok])})

1. Why do we go back 5 values from the end of telemetry in line 272?
2. Is it necessary to take an average in lines 274-276?
3. Is there any particular reason why we can't just pick some arbitrary time in the past long enough ago to pick a starting temperature, instead of the last commanded state with available telemetry?

We're just looking for ways to simplify this particular part of the code. We will be implementing these changes primarily in acis_thermal_check, but since this is the code that is running at the moment I thought we'd comment here.

Thanks!

[jzuhone]

cc: @richardjedgar if you want to follow along

[taldcroft]

Why do we go back 5 values from the end of telemetry in line 272?

That's somewhat arbitrary, but I just wanted to back up a little to be sure the value is not right on the edge.

Is it necessary to take an average in lines 274-276?

Not strictly necessary, but it seemed like a reasonable thing because IIRC the DPA temperature telemetry is fairly quantized so getting a sampling over 1.4ks might give a better starting estimate. But it's true that it is not critical.

Is there any particular reason why we can't just pick some arbitrary time in the past long enough ago to pick a starting temperature, instead of the last commanded state with available telemetry?

You don't want to go back any further than necessary because then propagation errors are larger. That probably wasn't in your plan but just want to be clear.

Looking at this code again, the point is that it uses these discrete command states for thermal propagation and so you need to join the starting temperature (from available telemetry) to the available command states. The existing code does that, but I realize now that one could modify the start time of state0 to match the starting temperature time. This might avoid a situation where your propagation is longer than necessary because state0 happens to be a very long one and then current code backs up to the beginning of that state.

Something along the lines of:

    if None in state0.values():
        state0 = cmd_states.get_state0(tlm['date'][-5], db,
                                       datepar='datestart')
        state0.update({'T_dpa': np.mean(tlm['1dpamzt'][-10:])})  # last 10 samples
        state0['datestart'] = tlm['date'][-5]  # middle of the last 10 samples
        state0['tstart'] = ...