The blast's clock

Every ViB unit keeps its own clock, so two records of the same blast do not line up by their timestamps. The platform places each ViB record linked to a blast on one clock, the blast's clock, and says how sure it is of each placement. The line-up, the Visualiser's replay and the ViB Analyser's comparison with other monitors all read the same placement, so no two of them put a record at a different time.

t = 0

t = 0 is the blast's first planned fire: the earliest firing time in its plan. Every hole's fire time is counted from there. When a det log has been imported, the holes take its actual times, so a hole that fired late is late on the clock, but t = 0 stays at the first planned time.

A record is placed by its trigger, the moment the unit started recording. Its pre-trigger lies before that, so its trace begins a little before its placement.

How a record is placed

Each record gets the best placement its data allows, and a chip says which. The line-up and the analyser word it a little differently:

Records with GNSS timing

Records whose trigger a GNSS receiver marked are pinned to each other on GPS time, to about one sample. With timing pulses the true sample rate is measured; without them the record is read at its nominal rate, and the error grows the further into the record you read.

GPS time is not the blast's clock, so the GNSS records as a group are placed on it by the same physics as every other record (see From the first arrival). The first motions of all of them are fitted to the arrivals the plan predicts at once, which places the group more tightly than any one record could place itself. With three or more GNSS records whose distances from the blast differ by at least 50 m, the fit also measures the ground's wave velocity, and every other record on the blast is placed at that velocity.

• A GNSS record whose first motion stands far from the rest of the group, more than three times its own uncertainty, leaves the group and is placed from its own first arrival.
• A group with nothing to tie it to the plan, such as no holes with firing times or no monitor placed on the ground, is placed against itself: its earliest trigger at t = 0, its records as close to each other as their receivers say, and the group as a whole only as sure on the blast's clock as its longest record.

The end-to-end accuracy of GNSS timing has not yet been measured on the hardware, so a GNSS record's ± is what its timing data says, not a certified figure.

From the first arrival

A record without GNSS timing is placed by its first motion: the first point where the ground clearly moves above the record's own noise before the blast. The plan says when the first wave should reach the unit: the earliest, over every hole, of the hole's fire time plus its distance to the unit divided by the wave velocity. The record is placed so that its first motion lands there.

This needs the monitor placed on the ground in the blast's grid, holes with positions and firing times, and a first motion the analysis could find.

The wave velocity is the one the blast's GNSS records measured, when they could. Otherwise it is the P-wave velocity of the blast's rock domain where the project's site model sets one, else 3,000 m/s, and it is taken to be good to ± 25 %. So a placement from the first arrival is good to about a quarter of the wave's travel time to the unit, plus 2 ms for the pick and the detonators' own scatter:

An estimate this wide is still useful: it puts every monitor on one time axis, close enough to see which arrived when. It is not good enough to compare arrival times to the millisecond; only GNSS records are.

Matched to a neighbour

A record placed from its first arrival is then matched to its nearest neighbour that is placed better: a GNSS record, or one placed by hand on the clock, whose ± is smaller than its own. The two records' envelopes (the combined motion of all three axes) are laid against each other over the stretch from 50 ms before each record's first motion to 300 ms after it, allowing for the later arrival of the wave at the farther unit, and the record moves to where they agree best. Its chip then names the neighbour, From arrival, matched to G4 (0.91) · ±…, where the figure is how well the envelopes agree (1 is a perfect match). When the neighbour's lane is not on the page it reads matched by envelope instead. The analyser says From the arrival, matched by envelope for a pair whose arrival placements were all matched.

The match only refines the estimate; it never places a record on its own. So it is made only when:

• the neighbour stands within a quarter of the record's own distance from the blast, so both units see the holes fire in the same order and their records keep one shape;
• the best match lies inside the record's own ±: a better one beyond it, or none, leaves the record where its arrival put it;
• the envelopes agree to at least 0.8;
• the match is sharper than the record's own ±: a long smooth burst can agree almost as well over tens of milliseconds, and then the match says nothing the arrival did not.

A matched record is as sure as its neighbour, plus a millisecond for the match, the width of the match (the lags that agree within 0.05 of the best), how far the arrival gap between the two units may be off with the wave velocity, and how far each hole's own gap between the two units departs from the first wave's (a long pattern reaches two units at a spread of gaps). It is never less sure than it was. Matched to a GNSS record, it is also as close to the GNSS records as that one is, plus the same terms: its chip gives both figures, From arrival, matched to G4 (0.91) · ±4 ms between GNSS units · ±25 ms to the blast, and the analyser compares it with another GNSS record by the tighter one. Matched to a record placed by hand, it gives only the one.

Placed by hand

A placement made by a person is kept: no rerun replaces it, whatever else changes on the blast. It comes in two forms.

• Against another record. On a record's event page, line another monitor's record up with it and press Save on the alignment chip (see Nudge the offset). What is kept is the pair: the other record triggered so many milliseconds after this one. Every time the blast is placed again, the other record is put that far from wherever this one now sits, so the two stay as you lined them up even when this record moves. Its ± is this record's on the blast's clock and one sample of the coarser of the two records, together. Against GNSS records it is as close as this record is to them, plus that sample.
• On the clock, at a time a person gives.

Either replaces the other for that record. A placement against another record is dropped when either record leaves the blast, or when someone who can edit the project's blasts presses Clear saved placement on its chip, on the event page of the record it was placed against; the record is then placed by the usual rules again. Wherever its place is shown, the record it was placed against is named: Set by hand against G4 in the line-up, Placed by hand against G4 on the alignment chip. If the record it was placed against cannot be placed, neither can it: it is laid on its own trigger until that record is placed.

A placed-by-hand record is never used to place others: it measures no velocity and does not join the GNSS records.

Records that cannot be placed

A unit's own clock counts whole seconds and drifts between syncs, so on its own it can be out by a second or more: many times the gap between two monitors' arrivals. A record with nothing better to go on is not placed:

• Device clock only when the record has the unit's clock and nothing finer;
• On its own trigger when it does not have even that.

The usual reasons are a monitor with no place on the ground, a blast with no plan or no firing times, or a record with no clear first motion. The line-up hatches an unplaced record on the blast's clock and draws it under First motion and Peak, which line records up by their own shape. The analyser lays it trigger on trigger and gives no arrival difference.

What the ± means

A ± is about one standard deviation: the placement is probably within that of the truth, not certainly.

• In the line-up, each lane's ± is against the blast's clock, and is drawn as a soft band at its trigger.
• Between two records, as the analyser compares them, the ± is on the gap between them: each record's own error, combined. Two GNSS records share the error of the group's place on the blast's clock, which cancels between them, so their ± is about a sample, far smaller than either lane's in the line-up.
• The analyser names the weaker of the two records' methods for a pair: an offset is only as good as the placement it rests on.

What moves a placement

The records of a blast are placed together, and placed again shortly after any of these:

• a record is analysed or re-analysed, or its GNSS timing is read;
• a record is linked to the blast or unlinked, or its suggested link is confirmed;
• a monitor is placed or moved on the ground;
• a det log or a blast plan is imported or deleted;
• the blast's fired time changes;
• a blast is made from events.

Because the records are placed together, one change can move others: a newly analysed GNSS record can measure a velocity that moves every first-arrival placement on the blast. A placement made by hand on the clock never moves; one made against another record moves with that record, and keeps its distance from it.