Multiple overbiasing methods and cfg file entries

[dpdutcher]

There are different overbiasing methods that use different configurations:

• bias_to_rfrac uses util. overbias_dets

  sodetlib/sodetlib/operations/bias_dets.py

  Line 139 in 04e2791

  sdl.overbias_dets(S, cfg, bias_groups=bias_groups)

  • This uses per-bias group settings from cfg.dev.bias_groups, namely overbias_voltage and cool_voltage, as well as cfg.dev.exp['overbias_wait']. As applied, the voltage values must be specified in actual Volts.
  • This uses the pysmurf function S.set_tes_bias_bipolar
  • Sidenote: cool_voltage is used in a confusing way here, where it gets applied in high current mode, but then (if high_current_mode=False, which it is by default) it switches to low current mode with const_current = False. This means either (a) cool_voltage is applying more power than we want it to, or (b) there's a sudden drop in power that could potentially latch detectors.
• take_iv uses the pysmurf function S.overbias_tes_all , wrapped in a helper function overbias_and_sweep

  sodetlib/sodetlib/operations/iv.py

  Line 639 in 04e2791

  S.overbias_tes_all(

  • This uses settings from cfg.dev.exp['iv_defaults'] for overbias_voltage, overbias_wait, cool_wait, and cool_voltage.
  • cool_voltage is only applied if the IV is run serially over bias_groups. I have no idea why. But at least it gets applied in low current mode.
  • S.overbias_tes_all does not take per-bias line values; it only accepts floats that get applied to all specified bias lines.

This is confusing. The overbiasing that happens for an IV curve and the overbiasing that happens for biasing the detectors in transition is the same process and should be identical between steps, and therefore should only be specified in one place (the need for IV curve to have a large bias_high, in order to have a long enough normal branch to fit Rn, is a distinct step separate from driving the TESs normal). Also, more granular control over per-bias lines is better than being forced to apply the same values to every bias line.

So, I propose that:

1. take_iv be moved to use util.overbias_dets
2. The current mode switch in util.overbias_dets be moved to after setting the bias to cool_voltage.
3. The overbias parameters be moved out of cfg.dev.exp['iv_defaults'] and into cfg.dev.bias_groups, where they are specified per-bias group. Also move overbias_wait here, so that it's not off on its own.
   a. This will have a side-effect of making scheduler calls to take_iv less flexible, as these parameters won't be accessible. I can think about that more though.

[tristpinsm]

I agree this is quite confusing.

If we are re-thinking how these functions are structured, could we include the pysmurf one as well? The pysmurf overbias_tes and overbias_tes_all are clearly performing very similar operations, and if we could modify them such that they are configurable enough for sodetlib to do what it needs without re-implementing that would remove another layer of confusing redundancy. Maybe this could be achieved by breaking up those pysmurf functions into more narrow sub-functions that could be invoked by sodetlib in the specific order it requires.

One particular area of concern that could be addressed at the same time is that set_current_mode accesses pysmurf methods that are meant to be private, which could be fragile (issue #482). I had to work around this for the move to rogue 6 (#485).

[dpdutcher]

I think the intention of sodetlib.util.overbias_dets is a convenience function to use the parameters stored in the SO device config, so in principle it could be a very thin wrapper for S.overbias_tes_all, but the latter needs to be expanded to accept per-bias line settings.

As to set_current_mode, which is also used in overbias_dets, that was added in #171, for the reason quoted below

Alright, I tested this quite a bit more, resolved inline comments and made the following changes:

1. I wrote a custom function to handle switching between low/high current mode, which has a few main benefits:

• When switching to high-current-mode, the cryocard relays and DC bias voltages are set with a single EPICS call, minimizing the amount of time we're in high-current-mode with the high DC voltages, and minimizing the amount of heating. When going HCM->LCM I've found that it actually takes longer for the DC voltages to settle than it takes for the smurf to switch back to low-current-mode, so if you try to set these at the same time there will be a current dip and the detectors will latch, so I've added a short sleep statement to account for this.
• This will prevent the DC voltages from changing if you are already in HCM / LCM and are trying to run it (at least if the relay software registers are set). This means you can safely run this to toggle HCM with constant current without the assumption that you're currently in LCM
• Switches all bias groups with a single epics call which I think should fix the problem where all bias groups do not consistently switch.