Cosmic Origins Spectrograph Instrument Handbook for Cycle 17

5.4 TIME-TAG or ACCUM?

---

COS exposures may be obtained in either a time-tagged photon address (TIME-TAG) mode, in which the position and time of each detected photon are saved in an event stream, or in accumulation (ACCUM) mode in which the positions, but not the times, of the photon events are recorded. The TIME-TAG mode of recording events allows the post-observation pipeline processing system to screen the data as a function of time, if desired, and to make other corrections. The COS TIME-TAG mode has a time resolution of 32 ms.

Some pulse-height information is available for all COS FUV science exposures. (No pulse height information is available for COS NUV science exposures.) The pulse height distribution (PHD) is an important diagnostic of the quality of any spectrum obtained with micro-channel plate detectors:

1. In FUV ACCUM mode, the global PHD is accumulated on-board as a separate data product along with the photon events.
2. In FUV TIME-TAG mode, the individual pulse height amplitudes are recorded along with the position and time information of the photon events, so the PHD can be screened by time or position on the detector if desired during the calibration process.
3. Post-observation pulse height screening is useful for reducing unwanted background events, and can often improve the signal-to-noise ratio in the extracted science spectrum.

Recommendations:

Simply put, TIME-TAG should be used for COS observations whenever possible as it provides distinct post-pipeline advantages for temporal sampling, exclusion of poor quality data, and - for the FUV - better background removal. TIME-TAG should always be used for exposures that will generate count-rates of 21,000 counts sec^-1 or less from the entire detector (including both detector segments for the FUV). In the 21,000-30,000 counts sec^-1 range, TIME-TAG may be used to obtained properly flux-calibrated data, but loss of some continuous time-periods within the exposure will occur (see the discussion under BUFFER-TIME below). At present, TIME-TAG should not be used for count-rates greater than 30,000 counts-sec^-1. ACCUM mode should be used only when absolutely necessary, such as for high count-rate targets.

5.4.1 TIME-TAG mode

The preferred mode of data acquisition for both the COS FUV and NUV channels is TIME-TAG mode. The flight software does not make any adjustments for the doppler shift of the spectrum when observing in TIME-TAG mode. However, the doppler correction will be applied during ground processing of the data in the calcos pipeline.

An important distinguishing characteristic of COS/FUV TIME-TAG observations is the inclusion of pulse-height information for each individual photon event (the pulse-height distribution, or PHD). Note that no pulse-height information is available for COS/NUV TIME-TAG observations.

Important considerations for BUFFER-TIME

All TIME-TAG observations must have a specified BUFFER-TIME (equal to 80 or more integer seconds), which specifies the estimated minimum time in which 2.35 × 10⁶ photon events (half of the COS data buffer capacity) will be accumulated during the exposure. BUFFER-TIME is a required parameter if the target is not WAVE. If the target is WAVE, then BUFFER-TIME may not be specified.

It is important for you to actually calculate an accurate value of BUFFER-TIME using the COS ETC. Do not simply specify the minimum BUFFER-TIME in your proposal! Observations that fail because of observer error of this kind will not be repeated.

If the predicted total number of events from a TIME-TAG exposure exceeds the total COS data buffer capacity of 4.7 × 10⁶ photon events, data must be transferred to the HST on-board science recorder during the exposure. Transfers of data from the COS buffer during an exposure will be made in 9-MByte blocks (half the buffer capacity). The value of BUFFER-TIME should be the half-buffer capacity (2.35 × 10⁶ counts) divided by the anticipated maximum sustained count rate in photons per second. We recommend that you give yourself a margin of error of about 50% if at all possible; i.e., to take the time just estimated and multiply by 2/3.

Note that BUFFER-TIME should include expected counts from the detector dark current and stim pulses as well as the detected photon events, factoring in the instrument quantum efficiency. On-board commanding utilizes the predicted buffer-time to establish the pattern and timing of memory dumps during the exposure.

During the first BUFFER-TIME of an exposure, counts are recorded in one of the two 9-Mbyte buffers of memory. After that first BUFFER-TIME is completed, data recording switches to the second of the two memory buffers, and the first buffer is read out. No data will be recorded in a buffer until it has been read out completely. Therefore, if the second buffer fills before the first has read out, all subsequently arriving counts will be lost until the first buffer is read out completely and again available for data-taking.

If BUFFER-TIME is incorrectly overestimated, the on-board data buffer may fill before the scheduled memory dump. Subsequently arriving photons will not be counted; they will not overwrite earlier recorded events. Therefore, a gap in recorded data will occur. NOTE: the pipeline will correct actual exposure times for any such gaps, so flux calibrations will be correct.

A conservative value of BUFFER-TIME is recommended (err on the low side) to avoid data loss. You should not merely specify the minimum allowed BUFFER-TIME for all exposures, as this may lead to operational inefficiencies.

The absolute minimum BUFFER-TIME of 80 seconds corresponds to a maximum average count rate of ~30,000 counts sec^-1 over the entire detector, which is the maximum rate at which the flight software is capable of processing counts. Note that the first buffer readout of an exposure requires 110 seconds to complete; this means that the maximum average count rate that will always produce no gaps in the recorded data is ~21,000 counts sec^-1.

If BUFFER-TIME < 110 seconds, Time_Per_Exposure should be less than or equal to 2 × BUFFER-TIME so that the exposure can complete before data transfer is necessary.

Note that TIME-TAG exposures of high data-rate targets have the potential to rapidly use up the HST on-board storage capacity. Caution is advised on any exposure with an exposure time greater than 25 × BUFFER-TIME, which corresponds to ~6 × 10⁷ counts, or about 2 GBits (close to 20% of the solid-state recorder capacity).

Doppler correction for TIME-TAG mode

No corrections are made for shifts in the spectrum from orbital motion while in TIME-TAG mode; this is done later in pipeline processing.

Pulse-height distribution data for TIME-TAG

The FUV detector provides five bits of pulse-height information with every photon event. These data are down-linked with the science data and are used to create a PHD later during data processing. See also Pulse-height distributions.

5.4.2 ACCUM mode

ACCUM mode should be used primarily for brighter targets, where the high count rate would fill the on-board buffer memory if the data were taken in TIME-TAG mode.

Doppler- or other corrections for ACCUM mode observations cannot be performed in the post-observation pipeline as the identity of individual photons was lost in the ACCUM process. The on-board flight software will adjust for the doppler shift of the spectrum due to the orbital motion of HST when observing in ACCUM mode. The doppler correction is updated whenever the HST orbital motion shifts the spectrum across a pixel boundary.

Note that ACCUM Mode exposures longer than 900 seconds that use the G130M or G160M gratings may blur the FUV spectra by 1 to 2 pixels (about 1/6 to 1/3 of a resolution element) due to wavelength-dependent deviations from the mean doppler correction.

Observing efficiencies with ACCUM

In certain cases on-board readout overheads can be minimized with ACCUM mode. This will typically be of interest for very bright targets that must be observed with ACCUM anyway.

Two ACCUM FUV images may be placed into on-board memory as ACCUM exposures read out only that portion actually illuminated by the target (about 1/4 of the full detector area). FUV ACCUM image readouts require one-half of the total COS memory so it is possible to acquire two FUV images before dumping the on-board buffer. Similarly, for the NUV detector, up to nine ACCUM images can be placed in memory.

If multiple exposures with the same setup configuration are required in ACCUM mode, (e.g., a time-series of observations on a bright target), then utilization of the Number_Of_Iterations Optional Parameter can be useful (the "repeatobs" option). Unlike the TIME-TAG case, no data may be acquired during an ACCUM readout, so the NUV detector is more efficient for repeatobs observing as more images can be placed in memory prior to readout.

If FP-POS=AUTO is specified with Number_Of_Iterations > 1, the exposures will be obtained in the order Number_Of_Iterations of exposures at each FP-POS position between moves of the grating.

Doppler correction for ACCUM mode

The COS flight software adjusts detected events for the orbital motion of HST. The doppler correction is updated whenever HST's motion changes enough to cause the spectrum to cross a pixel boundary. This is done via a small table of values computed at the start of each exposure based on the orbital motion and the dispersion of the grating in use.

Pulse-Height Distribution data for ACCUM mode observations

Some limited pulse-height information is also available for FUV ACCUM observations. A PHD histogram is dumped for every ACCUM mode image with the FUV detector, consisting of 256 bins (128 bins for each segment) of 32 bits each.