Generalities

During VSR1, two types of hardware injections took place:

• regular injections, taking place with the ITF kept in science mode, at moderate SNR values
• loud injections, performed with the ITF in adjustment mode, at relatively high SNR values
  

In the following we provide details and tables for both kind of injections

Regular injections

During VSR1 science mode state, regular HI sessions took place at randomly selected epochs, with an average separation of 8 hours, except at the very end of the run; see Note at end of this section.

The list of the injection epochs was planned beforehand, to avoid introducing any human bias in its selection, and is available here. Please note that at each epoch injections took place only if the detector was actually locked and in science mode.

Sessions of burst injections and inspiral injections alternated, each one lasting for 5 minutes.

Burst sessions

Each burst session comprised the injection of several cosine-gaussian waveforms, spaced in average by 5 seconds. The waveforms had all Q=9 and central frequencies

    f0 = 70, 105, 153, 235, 393, 554, 850, 914, 1304 (2000, 3068) Hz.

Frequencies above 1304Hz could not be injected, because they were found to require too high driving voltages, thus leading to DAC saturation and wideband noise.

The explicit formula for each cosine-gaussian waveforms is

    sigma = Q / ( 2*M_PI*f0 );
    h[i] = cos(2*M_PI*f0*t)*exp(-0.5*(t*t)/(sigma*sigma));

and it was generated over a window

    window = 10*sigma;

around the central time t=0.

A table of the burst injections performed is available here in ASCII format.

The columns include
GPS: timing of the center of the event, as resulting from the Sc_NI_LoopIn injection channel (thanks Benoît!)
tBefore, tAfter: duration of the event before and after GPS
hrss: nominal value of the HRSS of the event. May be inaccurate because of uncertainties in the calibration of the actuators.
SNR: nominal value of the SNR of the event. Depends upon the actual sensitivity during the run.
Q: quality factor
f0: central frequency

The nominal SNR has been 7.5 for the first half of each session, 15.0 for the second half.
Preliminary information about SNR re-estimated for Hrec v2 frames is available below.

Note that the GPS injection time is not calibrated for delays introduced by the analog actuation chain; a crude estimate of the expected delay and reference to a refined one are available below.

CB sessions

Each CB session comprised the injection of a single inspiral event, generated using a Taylor approximation at 2PN order.
The library used for creating the waveforms is the Virgo "inspiral", version v0r12.

The starting frequency for the generation was set at 50Hz.

Inspiral events were rotated among a list of four, differing for the mass values and the starting phase.

m1:  1.39, 1.41, 1.43, 1.45
m2:  1.47, 1.46, 1.42, 1.40
phi0: 0.34, 1.50, 0.33, 1.20

A table of the CB injections performed is available here in ASCII format.
The columns include
GPS: timing of the nominal end of inspiral phase, as resulting from the Sc_NI_LoopIn injection channel (thanks Benoît!)
tBefore, tAfter: duration of the event before and after GPS
hrss: nominal value of the HRSS of the event. May be inaccurate because of uncertainties in the calibration of the actuators.
SNR: nominal value of the SNR of the event. Depends upon the actual sensitivity during the run.
m1: mass of the first star
m2: mass of the second star
distance: nominal distance of the event (optimally oriented) in Mpc. May be inaccurate for the same reason as for hrss.

The nominal SNR has been set at 12.0.

Information in Frame format about the injections (Burst and CB) is also available here as a TAR+GZIP archive containing frame files.

SNR values

The nominal SNR value has been set using a sensitivity measured on HrecOnline upon July 25th, 2007 at 7h27 AM.

Using a sensitivity measured on Hrec V2 data from May 23, 2007, the actual SNR displays some differences

• For inspiral signals, the SNR is about 1.4% larger than nominal
  
• For burst signals, which sample different frequency regions, the differences can be more marked as in the following table:

    f0        SNR (Hrec V2)
    1304    16.2
    914     15.3
    850     15.0
    554     15.3
    393     15.6
    235     16.5
    153     14.6
    105     16.0
    70       17.5

where the values on the second column are to be compared with the nominal SNR=15 used in the second half of each burst injection.
This SNR information should be taken as preliminary.

Waveform shaping and timing uncertainties

The waveform generation has been performed translating the h(t) signal into a Riemann force, then calibrating the Riemann force as a voltage to be applied to the driving coils.

The steps are as follows

• h(t) signal has been generated at 20kHz
• A filter is applied to compute the double derivative, and to scale by M*L, where M=21.33kg, L=3000m
• A low pass filter (butterworth, sixth order) is applied to limit the signal below 5kHz
• Force is translated in voltage assuming that V = 240.78 F. This results from the assumption that the mass driven is suspended by a pendulum having a resonance at 0.6Hz, and by the measured value of the characteristic dx/dV = 13.7µm/Volt
  
• Data are resampled at 10kHz to be sent to the DSP and applied as control signals to the North Input mirror.

The shaping filters can introduce delays, however these are immaterial for VSR1 hardware injections, which were not coherent with LIGO ones: the timing of the HI events is recorded by the HrecOnline process, reading the digital injection channel Sc_NI_LoopIn, with any delay upstream included.

However, in between this channel reading and the actual force application, there are other stages which are believed to introduce delay:

1. an analog signal is produced by the DAC with one DSP cycle of delay (1 DSP cycle = 0.1 ms) wrt to the Sc_NI_LoopIn sample.
2. The analog signal is filtered by an analog low-pass filter of 7th order with a band of 3.7 kHz. This causes an equivalent delay of 0.2 ms.
3. The DAC holds for a whole cycle the same value, introducing a delay of 1/2 cycle (DAC hold time: 0.05 ms).

Overall, the timing of actuation on the mirror should be delayed by about 0.35ms wrt the time recorded in the tables.

This is a crude estimation: a more refined analysis, supported by measurements, has been made by the calibration team and is available in the Virgo note Timing calibration during VSR1 - A.Masserot, B.Mours, L.Rolland, VIR-028B-08.
The results in the LN (low-noise) mode used during hardware injections appear consistent with the crude estimate within 0.025ms.

Note

Starting at GPS 875191432 (set 30 2007 12:43:38 UTC), the injection code run into a bug and started to repeat the injections sessions every 5 minutes instead of 8 hours. The resulting last 18 injection session, ending at GPS 875202716, are therefore potentially suspicious because they did not occur after a quiet period. The injection lists include also these extra events.

Loud injections

During VSR1, each week an 8 hours shift has been dedicated to detector maintenance. Some of these shifts could accommodate a session of loud burst injections, intended to provide data useful for testing veto safety against channel cross-talk.

These loud burst sessions took place once per week, for periods of 20 minutes, while keeping Virgo in adjusting mode, starting at the epochs:

UTC                                 GPS
08-Aug-07 07:15:09     870592523
16-Aug-07 06:06:24     871279598
22-Aug-07 07:15:31     871802145
05-Sep-07 13:26:28     873034002
12-Sep-07 06:08:18     873612512
19-Sep-07 06:08:13     874217307
26-Sep-07 07:08:02     874825696

The same burst waveforms as for regular injections have been used, except that the nominal SNR was 75 during the first 10 minutes, and 150 during the second 10 minutes of each injection session.

An ASCII table of the loud burst injections is available here, using the same format adopted for "normal" injections.

The information is also available in frame format here, stored in a TAR + GZIP archive of frame files.

Technical details, utilities

The timing of the events has been recorded by the HrecOnline process in FrSimEvent structures, stored in the HrecOnline frame files.
The nature of the events, and their parameters, was saved in log files.
A set of simple VEGA scripts has been used to parse HrecOnline frames and injection logs to produce the information provided in the previous sessions; the scripts are available here in a  TAR+GZIP archive.