Recording setups¶

Many timing results in the atlas hold per recording setup: a delay, a drift rate or a correction mode that is the same for every recording made on one setup and differs between setups. This page says what a setup is, how the atlas recognises one, what depends on it, and where the uncertainty models should use it.

What a setup is¶

A setup is the equipment that made a recording: the recorder PC in one operating theatre, its serial cables, and the particular units plugged into it (one SNUADC converter, one BIS monitor, one patient monitor and its oximeter module, one ventilator). The release never names a theatre, a PC or a serial number. It does record, in every .vital file, each device's Windows serial port (COM port), and those ports recognise setups well.

How the atlas recognises a setup¶

1. The PC key (the working definition): the serial ports of the four core devices, SNUADC · Solar8000 · Primus · BIS, for example COM4 · COM9 · COM10 · COM20. It is complete for 90% of recordings. Port numbers are assigned by Windows on each PC, so one port number alone is a poor key (the SNUADC was on COM4 in 2,824 recordings, on several PCs), but the combination of four rarely repeats across PCs: Two heart-rate delay groups, one poll apart found 44 PCs this way. Knowing the PC key puts the SNUADC's drift within 3 ppm of its setup's median in 97% of cases, against 34% for the release's median (in the correction scan's cases; on the whole release's correction intervals 71%, an open question on Identifying the recording setup of every case).
2. Coarser keys, where the PC key is incomplete: one device's port (75% within 3 ppm for the SNUADC), or the SNUADC and BIS ports together (96%).
3. Clock fingerprints, which need no port: a device's crystal runs at its own rate to about 1 ppm, so drift identifies physical units (Physical units, shared drift and time order), and the oximeter's frame clock identifies its module (The oximeter's frame clock is a fingerprint of the setup). They can also split one PC key into the different units it carried over the years.

Some results were grouped differently before this page existed: the PPG delays by the SNUADC port plus its correction mode (tools/setup_offsets.py), the drift units by port pairs. They are coarser versions of the PC key. A setup is not the same as a correction mode: one setup can record in either mode, and the mode changes some constants (The EEG–ECG offset is fixed by the SNUADC correction mode). None of these keys identifies a theatre, a date or a patient.

What varies by setup¶

The setup register¶

Ports are recorded for almost every device: the four core devices' ports, and so the PC key, are complete for 5,755 of 6,388 recordings (90%); the rest lack a BIS or have no port for one device. They form 65 distinct PC keys, 37 of them with at least 20 recordings (5,631 recordings).

How much a grouping tells you about a device's drift. Share of cases whose drift is within 3 ppm of their group's median, against within 3 ppm of the median of all cases (groups with at least 5 cases):

Setups with at least 20 recordings, by PC key (SNUADC · Solar8000 · Primus · BIS port). Drifts are medians over the cases where they were measured.

Where the uncertainty models should use the setup¶

Several parameters are quoted over all recordings although they are known per setup. For a recording whose setup is known, the per-setup value has a much smaller uncertainty; the all-recordings range applies only when the setup is unknown. The SNUADC step drift (SNUADC step-mode drift), the BIS drift (BIS drift against the recorder PC) and the oximeter frame clock (Oximeter frame clock against SNUADC) now list their values per setup; the rest are on the work plan.

• PLETH delay and sawtooth rate. The PAT correction works far better with the setup's creep rate as a prior: without it, few teeth reach a confident tier (Pulse arrival time in two correction steps).
• SNUADC and BIS drift. A track without a whole tooth already takes its port's or unit's typical rate in the stamp model (The stamp sawtooth model for SNUADC and BIS); the uncertainty of that rate should be the port's spread, not the release's.
• Numeric delays. The heart-rate delay is two-valued and set by the PC; its uncertainty for a known PC is far smaller than the 1.9–4.2 s range.
• Planner and clock domains. The synchronisation planner uses one drift per clock domain; per-setup drift would tighten its windows for recordings whose setup is known.

Open questions¶

They are kept, with what would settle each, on Identifying the recording setup of every case, and appear with every other study's on the Open questions page.