The time associated with each data point in a SEG-2 data file generated by a Geode is related to the time of the “trigger” event which was instrumental in the production of the file and its content.
The Trigger Master and Trigger Distribution
The trigger event occurs at the Geode designated within the Controller software as the Trigger Master. Although all Geodes are capable of being Trigger Masters, there must be one and only one Trigger Master in any properly functioning Geode system. The Controller automatically takes care of this requirement when the designation is made by a user, and when the system is established at the time of Controller start-up based on a previous designation (or a default setting in the case of a “new survey”). All other Geodes in the system will have their Trigger Master circuit disabled. A trigger event can be initiated by an external electrical pulse provided to the trigger input connector of the Trigger Master Geode, or by a command sent via Ethernet from the Controller to the Trigger Master (usually for test purposes), but only when all conditions are satisfied to allow data recording. There is also a special trigger initiation situation, called “self-triggering” which will not be discussed further here.
Upon acceptance of a trigger event, the Trigger Master will distribute the trigger signal to all Geodes in the system, itself included, via an RS-485 network that resides within the digital interconnect cabling. (Proper termination of this RS-485 network is automatically taken care of by the Controller.) The trigger signal is propagated through the cabling and Geodes at the nominal speed of 70% of the speed of light, or approximately 2.1x10^8 m/sec. The maximum distance of successful propagation depends on a number of factors such as the number of Geodes involved, the noise environment, the quality of the cables, and the acceptable amount of timing uncertainty for the particular application. Distances approaching or exceeding 1km should be given careful attention in this regard. In a 3-D Geode system involving LTUs, each LTU, unlike a Geode, will reconstruct the trigger signal before sending it on, effectively confining the maximum distance issue to each sub-network separated by LTUs. The penalty is an additional delay of about 100nS for each LTU in the route.
The External Trigger Circuit
The external trigger input is capacitively coupled, with a 2mS time constant, to the midpoint of a resistive voltage divider. The voltage difference between the two ends of the divider constitute a voltage "window", which size is set by the trigger sensitivity parameter and can range from essentially zero at the highest sensitivity, to about +/- 2.5V at the lowest sensitivity. The Geode will trigger (if enabled) if and when the coupled signal exceeds the window, in either direction (i.e., positive or negative going). The signal, after the capacitor, is clamped by diodes to the range between the trigger signal ground and +5VDC. The trigger detector output is disabled when the system is disarmed, during a parameter change, and during a shot, up to the trigger hold-off time after the end of the shot. The trigger hold-off time is a parameter set by the user.
Preceding the coupling capacitor (i.e., essentially the node accessible at pin A of the external connector), there is a 3.3K-Ohm pull-up resistor to +5VDC (relative to pin B). Also a fast transient suppressor clamps the input at about +/-14VDC. It is advised that the DC + AC level of any voltage applied to pin A relative to pin B be kept within the range of +/-7V, giving some margin of safety.
If a DC voltage somewhat less than +5VDC is applied when the connector is first mated, the instrument may trigger at that moment. But, subsequently, because of the capacitive coupling, it will trigger on the next positive or negative going pulse that exceeds the window level. If the duration of the applied voltage pulse is less than the record length + delay time + hold-off time, then the Geode will effectively be ready to trigger on the same edge of another similar pulse.
The Geode supports a sub-sample timing synchronization feature used for synchronizing the data acquisition after a trigger event to the distributed trigger signal, so that subsequent time points will be known to within 1/32 (~1/20 at the fastest two sampling rates) sample interval. It does this by increasing the sample interval at the trigger time by 0 to 31/32 of a sample interval in increments of 1/32, so that the first sample after the trigger would represent a time of one sample interval after the trigger event, with a tolerance within 1/32 of a sample interval. The following samples continue from there at the expected intervals. For example, with a selected sampling interval of ¼ mS and a recording delay of 0mS, the first sample in the recorded file for each channel would represent data at 250 to 258uS after the trigger event.
This of course potentially introduces a small discontinuity at the time of the trigger, observable depending on the nature of the channel waveform(s). (The zero-phase anti-alias filter will smear the discontinuity into the nearby samples both before and after, consistent with the bandwidth of the filter.) Sub-sample synchronization can be disabled if it is deemed to be detrimental for the particular application, at the expense of losing the 1/32 interval timing accuracy.
The principal errors in Geode timing are of two types: those associated with the trigger mechanism and which are static over the duration of the record, and those associated with the time base and which change over the duration of the record. Excluding the trigger propagation delay mentioned above, the trigger timing uncertainty is about 1uS. The known fixed errors have been lumped together and are reported in the SEG-2 file trace headers as channel SKEW. (The actual channel skew is zero, since all channels are effectively sampled simultaneously, but the SKEW value in the header is used as the only place permitting small timing corrections. Note that the SKEW value for every channel is identical.) If the size of this correction is important to the application, the SKEW value should be added to the calculated time points when the data is being processed.
The Geode time base has a +/-15ppm stability over temperature (-20C to +70C) and component variations. Thus time drift relative to absolute time and relative to other Geodes is possible. (However, all channels within any Geode enclosure use the same time base, so there is no relative drift between channels in the same enclosure.) Therefore timing uncertainty increases from that existing at the time of the trigger until the time of the next trigger (or end of record).
Special Timing Issues Involved with “Continuous” Recording
“Continuous” recording is a method that allows unending 100% time coverage with recorded Geode data. It produces a series of time-overlapped records created by the use of a negative time delay set equal to the record length such that each record consists of completed history at the time of the trigger event. This technique circumvents the problem of data transmission overrunning data acquisition. The principle constraint is that the cycle time from trigger to trigger must always be less than the chosen record length. Otherwise, gaps rather than overlap would result. Commonly it is used with GPSderived triggering in order to provide time-stamping of each trigger event.
Upon consideration of the above, it will become clear that the time-stamp associated with a particular trigger event will pertain to the data in the following record, not to the data in the record in which the time-stamp is written. This comes about because the trigger event ends the record.
Because there is data overlap between records, the precise trigger point in the following record at which the time-stamp applies can be found by comparison of the data values at the end of the former record with those near the beginning of the subsequent record. The overlapping data will be exactly identical in both records (since they are read from the same memory location, twice). The earliest data in the subsequent record that goes beyond the data of the previous record is the data that is one sample interval (assuming sub-sample synchronization is enabled) past the time-stamp.
Note well that this comparison must be made independently for at least one channel of each 8-channel Geode board set, because the discrete time at which data values are written to the memory buffer, relative to the trigger event, is a function of each individual board set in the Geode system.
Correct GPS Time-Stamping
There are differences between various GPS models that can affect accurate time stamping. The 1PPS signal from a GPS has a “timing edge” and return edge, of which only the former is the true whole-second edge. Some models use a rising edge as the timing edge, some the falling edge, and some have it selectable. Consult the GPS manual to determine the definition of its timing edge. As indicated earlier, the Geode can be triggered on either a rising or falling edge. It is important to insure that the Geode is being triggered on the proper edge in order to avoid timing that may be a fraction of a second off. This is expanded upon below.
Some GPS units provide a very narrow timing pulse, others one that has a nearly 50/50 duty cycle. For the narrow pulse units, almost certainly it is the leading edge (rising or falling) that is the “timing edge”. This case can be easily handled by using the Geode Trigger Hold-off feature. If a 10-second cycle time is desired, set the Trigger Hold-off time to about 9.5 seconds. In this case, there is a very small chance that the very first trigger could occur on the wrong (trailing) edge, but from then on the leading edge will be used as the triggering edge.
If the GPS provides a 50/50 duty cycle edge, and it is not alterable, then the Geode by itself could as easily start on the wrong edge as on the correct timing edge, and continue thusly until restarted. For this case, Geometrics can provide a Trigger Timing Interface Box (TTIB) that will correct the situation. The TTIB can be programmed to respond only to the correct edge (rising or falling), change the polarity if needed, and gate through only one of every N 1PPS pulses, where N is programmable. (The TTIB also incorporates an alarm system that can provide a remote alert if a record is missed.)
Another potential issue comes from the variations between GPS models of the time that the serial time string (containing the time value of the associated 1PPS) is issued relative to the 1PPS itself. The Geode Controller attempts to pick the correct serial string based on a calculation involving the known record length, the PC times, and the trigger notification message from the Geodes. But if the GPS issues the serial string at an unusual time (and the time has been seen to vary somewhat with a given GPS unit) then it could pick up the incorrect time, off by 1 second. If rare, it can be subsequently detected and corrected during data processing, but if consistent it may not be easily detected. Again, the TTIB can accommodate the situation by only gating through to the Controller PC the string belonging to the gated-through 1PPS pulse. The Controller Serial Input Time Window can then safely be widened to 2 seconds (assuming the cycle time is more than 2 seconds) if need be, to expand the Controller’s search for the string around the calculated trigger time.