The DAGA2 Data Acquisition System for Gravitational Wave Antennas

This work briefly describes the DAGA2 data acquisition system now running for all the antennas of our group. It was presented by S. Frasca at the conference ``Time Series Analysis in Astronomy and Meteorology'' held in Padova from September 6th to 10th.


The Rome group of gravitational wave research has now in operation three cryogenic bar antennas: EXPLORER at CERN, ALTAIR at IFSI and NAUTILUS at INFN in Frascati. The data acquisition is done by microvaxes, operating under VMS and equipped with an ADC card (DEC ADV11-D) and an IEEE-488 card (DEC IEQ11), controlled by the RT-Integrator software (derived from VaxLab).

DAGA2 is the software system that manages all the aspects of data acquisition and analysis. It derives from the DAGA system, initially developed for the room temperature gravitational antennas GEOGRAV and AGATA. It is installed not only on the acquisition computers, but also on some of our workstations because it has three operative modes:

- acquisition mode, the normal one, with real physical data

- simulation mode, with the real data substituted with simulated ones; in this mode the user has the same facilities of the normal mode, but time can run faster than in reality (more than 40 times with a 110 specmarks alphastation). This mode is very useful for testing filters or new procedures and to develop the system itself.

- analysis mode, that uses the acquired data to reanalyze them.

The programs are all written in Fortran and DCL. In the following, in square brackets is reported the case of the Explorer antenna.

The Base Software

The data acquisition is performed by a group of independent programs that communicate each other by means of a big shared memory area and a few event flags. The shared memory area is divided in three COMMON areas, named SHARED_IN, SHARED_STAT and SHARED_OUT, containing respectively the data pertaining the input, the status and the output. There are some batch programs operatING at high priority. The most interesting are

- DAGA2_ACQ that reads the data acquired by the ADC card and puts them in a big ring (circular vector) in the SHARED_IN, together with the time and other informations. In simulation mode this program is substituted by a simulation program. In analysis mode it is absent.

- DAGA2_FORM that reads the data in SHARED_IN, analyses them, archives some of the raw data, produces spectral estimates, statistics and lists of events. It reads the real data, named input or physical channels, [in the case of Explorer antenna they are now 12 channels], all at the same sampling time (220 Hz on Explorer) and then produces a certain number of output or software channels, that are particular types of processing of the input channels, typically filtering [Explorer has now 32 output channels]. The sampling time is different for different channels: there are three different output sampling rates plus a "super slow" [220 Hz, 55 Hz, 3.4 Hz and .03 Hz]. Some of this output channels are archived; for almost all of them it is computed a real time short-time mean [6 minutes] and 4 histograms, besides of hourly means and adaptive event search. Here is a list of the Explorer output channels

1: 1 <--> x nu -

2: 2 <--> y nu -

3: 3 <--> x nu +

4: 4 <--> y nu +

5: 5 <--> Calibration

6: 6 <--> Direct Acquisition

7: 7 <--> x out of resonance

8: 8 <--> Seismic l.f.

9: 9 <--> Seismic resonance

10: 10 <--> Electro-magn. monit.

11: 11 <--> Top second

12: 12 <--> Low freq. antenna

13: 13 <--> Press. He bath[mbar]

14: 14 <--> He Evapor flux l/min

15: 15 <--> Top second 2

16: 16 <--> Top second

17: 17 <--> R^2 nu -

18: 18 <--> ZOP nu -

19: 19 <--> W.K. nu -

20: 20 <--> R^2 nu +

21: 21 <--> ZOP nu +

22: 22 <--> W.K. nu +

23: 23 <--> Wide Band phy_0

24: 24 <--> W.K. average

25: 25 <--> W.K. minimum

26: 26 <--> AAP

27: 27 <--> Dir. Acquis. Temp.

28: 34 <--> Hilbert Mat. Filt.

29: 35 <--> H.M.F. nu -

30: 36 <--> H.M.F. nu +

31: 41 <--> Post Matched Filter

32: 45 <--> Rephas. Match. Filt.

- DAGA2_SIHQ that estimates some varying parameters (like the resonance frequencies) and computes the adaptive matched filters that is used by DAGA2_FORM.

- DAGA2_CLOCK that corrects [every half an hour] the VAX time; the precision obtained is normally of the order of 30 ms.

There are two fundamental interactive programs:

- DAGA2_CONT, a control program pertaining the management and control of the acquisition. By it one can start and stop the acquisition, insert and read experimenter's comments, set operation flags (that say if the antenna is in helium or nitrogen refilling or if there is someone working around it and for this reason we can expect noisy data), consult some log files (like the time corrections) and check the correct behaviour of the acquisition programs (many variables are displayed in real time).

- DAGA2_MON, a monitor program pertaining the acquired data analysis. One can analyse

a) data directly as they are acquired (physical and software channels)

b) histograms and statistics of recent data [last 30 hours]

c) two-hour spectra, events, variable parameters and hourly means of the present run and of all the past runs

d) archived sampled data

For all the above data it is possible to do many types of processing, analyses, statistics, graphs, as the program is linked with our signal analysis program SNAG. There is also a "Custom Menu", with a particular choice of the possible statistics. An important feature of this monitor program is the possibility, for the simulation mode, to simulate and analyse in real time events in input to the antenna. It is possible to program this event generator in many different ways (varying automatically the parameters of each event).

There is also a batch program that archives on tape cassette the closed data files and one that "supervises" the correct behaviour of the acquisition programs, in some cases looks for solutions and alerts the experimenters with automatic electronic mails.

One important feature of the DAGA2 system is that there are many possibility to "taylor" it. The acquisition can be adapted to local necessities by means of a file that acts like a program. Also the programs that actually run can be easily changed, changing the definition of logical symbols (there is a local file containing all the logical definitions). Among the symbols, there is also the shared common that can be changed in any part and in particular in the dimension of some vectors. Furthermore, the programs DAGA2_CONT, DAGA2_FORM and DAGA2_MON have external local subroutines that generate custom menus and particular filters of local interest.

Other utility programs are present in the DAGA system. Every interactive program is menu-driven and has an internal help (also callable at DCL level). Now the DAGA2 system occupies (in shrunk form), about 15 Mb, 3 Mb of which are source code, 4 Mb are executable code. A larger area is needed for the archived data [500 Mb].

Two other programs are worth of mentioning for the use of DAGA2. They are external to the system (normally are used on bigger Vaxes) and give easy access to the two data-bases. They are

- LEGGI that "reads" the archived rough data; it can access one or two data streams (from two different antennas) at the same time, residing on tapes, magnetic discs or optical discs, and provides graphics, statistics, spectrum estimation and correlations both for "hardware" channels (the actual data) and for "software" programmable channels (derived from the actual data by filtering and/or composition). This program can also build the second order data-base, with events and hourly statistics. An event is defined when the data from a channel (hardware or software) goes over a preassigned adaptive threshold; we archive the beginning time, the real amplitude, the statistical amplitude, the length, two parameters describing the shape, which channels were excited and which amplifiers were saturated during the event.

- ADES that "reads" the second order data-base. It can access conditionally the events and hourly statistics and provides many types of analysis.

Data Bases and Log Files

DAGA2 acquisition (also in simulation or analysis mode) produces many types of data, that are organized in data bases, accessible by some programs. The most important are

- sampled data archive, that contains files [one file every 36 hours] composed of records [a record every 29 seconds] with the data of some channels, with one, two or three different sampling rates and an header with the time of the first datum of the record and other informations. There are other types of records containing the physical parameters of the run, the DAGA2 configuration and the experimenters comments. These files are transferred on magnetic tapes, but remain on disc until there is lack of space. These data are also transferred during the night on a big disc in Rome and will remain here for about one month, in order to do other types of analyses and network low level processing (e.g. correlations between different antenna data>); this centralized archive will be now enhanced to 10 Gbytes; this will permit about two months of raw and filtered data for three antennas.

- spectra archive that contains all two hours spectra of the principal acquisition antenna channel.

- events; we say that an event has begun in an output channel when the data from this channel goes above a certain threshold; the threshold is defined adaptively, depending on the recent statistics (computed in an autoregressive way); an event lasts until it remain under the threshold for a certain time [3 seconds]. For each event the channel, the time of beginning, the time of the maximum, the maximum amplitude, the statistical significance, the length, the energy or something resembling it (depending on the channel), the number of relative maxima and, in some cases, some other variables like the "chrominance variables" of the chromomatched filter are recorded.

- hourly statistics, i.e. mean, standard deviation and number of event for the output channels, and varying parameters computed from the spectral estimation. The files contain the informations on the physical parameters and on the DAGA2 configuration as well.

There are a certain number of log files, containing e.g. the time corrections [every half an hour], the experimenter comments, the start and stop of the runs, the logs of the batch acquisition programs.


The problem of the timing is a very important problem in the detection of gravitational waves. It has two faces: absolute timing and uniform sampling. It is solved in the DAGA2 system in the following way. The sampling time is given dividing digitally the frequency 5 MHz of a rubidium clock or almost equivalent precision oscillator; this precise oscillator also gives a narrow top of the second pulse. There is a radio controlled clock (until now using the swiss time signals HBG, in the near future the GPS) that gives two informations: an ASCII string with the UT and a wide top of the second pulse. With the ASCII string the VAX internal clock is corrected every half an hour [the precision is better that 30 ms]. The two top of the second pulses are combined as a single signal and acquired as an input channel; so it is possible to know continuously the time shift of the three independent clocks: the low precision, but high reliability Vax clock, the high precision but sometimes lacking radio time and the high precision, but without absolute reference sampling trigger. The precision that can be obtained with this method is, in principle, that of the sampling time [now 4.5 ms, but in the near future 0.2 ms]. The DAGA2_FORM real time analysis program provides output channels containing all these informations.

DAGA2_HF technical documents (in development):