2008 IceCube Update - Section 5
V. DOM HARDWARE
Each DOM contains a downward-facing 10" (25 cm) Hamamatsu R7081-02 photomultiplier tube and associated electronics in a 35 cm diameter pressure sphere. The PMT has a standard bialkali photocathode (Sb-Rb-Cs, Sb-K-Cs), with a peak quantum efficiency of about 25% at 390 nm. The minimum useful wavelength of about 350 nm is set by absorption in the pressure sphere. The electronics includes a Cockroft-Walton high voltage power supply, electronic timing calibration systems, light emitting diodes for photonic calibrations, and a complete data acquisition (DAQ) system. The PMTs are currently run at a gain of 107, wwith typical high voltages of 1300-1500 volts. An average single photoelectron produces a pulse about 10 mV in amplitude and 5 nsec width into the 43 Ω impedance of the DAQ system. The charge resolution for single photoelectrons is about 30%. The DAQ system is designed to record the arrival time of all detected photoelectrons, with a relative precision of better than 5 nsec, across the entire array.
A block diagram of the DAQ system is shown in Fig. 2. The central elements of the DAQ hardware are two waveform digitization systems, the Analog Transient Waveform Digitizer (ATWD) and the fADC ('fast' ADC). A digitization cycle is initiated by a discriminator trigger; the threshold is set at a voltage corresponding to about 1/4 photoelectron. When this happens, the FPGA will start ATWD and fADC digitization on the next clock edge. To make up for delay in the trigger circuit, the signal goes through a 75 nsec delay line before the digitizers. This delay line limits the system bandwidth to about 100 MHz.
The ATWD digitizer uses a custom switched-capacitor array chip. Each ATWD chip includes four parallel inputs, each with 128 capacitors. When launched, the system acquires data at 200 to 900 megasamples per second (MSPS); IceCube runs the ATWDs at 300 MSPS, providing 400 nsec of recording capacity. Three ATWD channels are run in parallel, with input gains in the ratio of 16:2:1/4, providing more than 14 bits of dynamic range. After acquisition, the voltages on the capacitors are digitized with 128 10-bit Wilkinson ADCs, each multiplexed to the four capacitors which acquire a single time sample. A fourth ATWD input (not shown) is used for electronics calibrations. Each DOM contains two ATWD chips. They are operated in a ping-pong fashion – while one is digitizing, the other is live; this greatly reduces the dead time. The fADC digitizer uses a 10-bit, 40 MSPS commercial ADC chip. When triggered, the system records 256 samples, covering 6.4 μs.
Each DOM also contains a 'flasher' board, which has 12 blue (405 nm) LEDs mounted around its edges. These LEDs are used for a variety of calibrations, measuring light transmission and timing between different DOMs, to check the DOM-to-DOM relative timing and study the optical properties of the ice.
The entire system is controlled by a 400k gate Altera Excalibur FPGA, which incorporates an ARM9 hard-core CPU. The FPGA controls the data acquisition and digitization cycle, compresses (losslessly) and formats data for transmission to the surface, and oversees calibrations.
Data is transmitted to the surface via a single twisted pair, which also provides ±48 VDC (96 volts total) power. Each DOM consumes about 3.5 W. The cable also includes local coincidence circuitry, whereby DOMs communicate with their nearest neighbors; they can also pass messages onward. A more robust connector is used than in AMANDA, and a higher fraction of IceCube OMs survive 'freeze-in.' On the surface, the cables are connected to a custom PCI card in a PC; the remainder of the system is off-the-shelf.
IceCube DOMs have several operating modes. They are currently operating in "Hard Local Coincidence" mode: data is only saved when a neighbor (either nearest or next-to-nearest) DOM also sees a signal within 1 μs. In "Soft Local Coincidence" mode, an abbreviated 'coarse charge stamp' is saved even for isolated hits. It consists of the largest 3 fADC samples out of the first 16 samples. Saved data is formed into packets for transmission to the surface.
The system uses a 40 MHz system clock. Since this clock is used to record the hit times, a precision oscillator is used. The oscillator has frequency stability (Allen variance) of better than δf/f < 10-10. Despite this accuracy, maintaining the required 5 nsec precision requires frequent synchronization.
Timing calibrations are performed automatically every few seconds (currently once every 0.5 s). During each calibration, the surface electronics sends a timing signal down to each DOM, which waits a few μs until cable reflections die out, and then sends an identical signal to the surface. To maintain the symmetry, both the surface and DOM electronics use identical DACs and ADCs to send and receive signals. With the symmetric setup, transmission times in the two directions are identical. Even though the 3.5 km cable transmission widens the signals to ∪1 μs, the transmission time is determined to better than 3 nsec . This accuracy is maintained across the entire array; it has been verified using muons and artificial light sources. The software tracks the timing difference between the in-DOM oscillators and a surface based master clock, and appropriate corrections are applied to the data.
Amplitude calibrations are done using an ultraviolet (peaked around 374 nm) LED that is mounted on the main electronics board. It is flashed repeatedly at low intensity (<< 1 photoelectron in the PMT). A charge histogram is accumulated in the FPGA and sent to the surface, where it is fit to find the single photoelectron peak. This is done for a range of high voltages, and the system is then set to the correct HV to give 107 PMT gain. These calibrations are extremely stable.
Amplitude linearity calibrations take advantage of the 12 LEDs on the calibration board. The LEDs are flashed individually, and then together, providing a ladder of light amplitudes that can be used to map out the saturation curve.
One other critical requirement for the IceCube hardware is high reliability without maintenance. Once deployed, it is impossible to repair a DOM, so the system was designed for very high reliability. About 98% of the DOMs survive deployment and freeze-in completely; another 1% are impaired, but usable (usually, they have lost their local coincidence connections). Post-freeze-in, reliability has been excellent, and the estimated 15-year survival probability is 94%.
(fig 5) IceCube event displays for (top) a muon or muon bundle (multiple muons) in IceCube 40 (the 40 string configuration running in 2008), a simulated νe (middle) and a simulated ντ (bottom). The latter shows the classical 'double-bang' topology. Each dot is from a single struck DOM. The size of the circles indicates the number of detected photons, while the color gives the time, from red (earliest) to blue (latest).