Delay Required by Photomultiplier Signal Path

The IceCube digital optical module depends upon the Analog Transient Waveform Digitizer chip to capture and digitize photomultiplier pulses. Since the ATWD only acquires data when 'launched' or 'triggered', and since the triggering is done synchronous to the system main clock, the photomultiplier signal must be delayed on its way to the ATWD analog inputs long enough so that it arrives after the trigger signal.

The delay duration must be at least one clock cycle, plus the logic delay in the FPGA trigger circuit, plus the trigger comparator propagation delay, plus the propagation delay of signals from point to point around the PC board, plus a three sample start-up delay of the ATWD sampling circuit itself. In addition to those strictly hardware requirements, waveform analysis may be improved or simplified by capturing several baseline samples before the abrupt edge of the waveform. For a 40 MHz fast system clock speed (25 ns clock period), the necessary delay is estimated to be 60 ns.

This delay products tested so far are:
RG-62B/U, the 95 Omega coax cable used for the AMANDA String 18 DOMs,
RG-180B/U, another 95 Omega coax cable, but smaller diameter,
PCB Strip-line done in FR4,
Lumped Delay Line similar to one used in KAMLAND
PCB Strip-line done in low-loss laminate.

For PMT waveforms through RG-62B/U, through RG-180B/U, through FR-4 strip-line, through lumped delay line, and through low-loss strip-line

There is still time to think of novel solutions. Ideas here.

The AMANDA string 18 digital optical modules used coax cable of an impedance that best matched other requirements of the signal path (100 ohms).

Test 1: 49 foot length of RG-62B/U (polyethylene) coax cable like that used for delay for AMANDA string 18 DOMs.
Comment: Performance Acceptable for IceCube Setup: 50 ohm source impedance. Parallel 100 ohm resistor at inpjut,
and series 50 ohm resistor at output to scope internal 50 ohm termination. There are other cables in the 93-100 ohm range.

Test 2: 25.5 foot length of RG-180/U (teflon) coax cable like that used for AMANDA string 18 DOM PMT anode signal.
Comment: Performance Acceptable for IceCube

Test 3: Printed Circuit Bord Delay line produced at the direction of Dan Wahl, of U of Wisconsin SSEC. The delay line is actually a strip-line embedded in printed circuit baord. The impedance measured 105 ohms by TDR method using fast pulser and fast oscilloscope. The delay (cursors) includes 4ns delay in the test setup. The oscilloscope rise-time measurements are based on the standard 10%-to-90% reference points.
Comment:

Impedance nearly the target value.
Delay about 10% shy of the target value.
Output waveform rise-time slower than desired for this application. (consistent with 50 MHz bandwidth)
Delay characteristics well behaved and free of the bumps, wiggles and oscillations characterizing the lumped delay.
~7:1 delay to rise-time ratio is very marginal.

Test 4: Data Delay Devices brand lumped delay line (62 ohm impedance). Setup: The source impedance was 50 ohms. The input impedance of the delay line was matched to the source with a parallel 260 ohm resistor. The 62 ohm output impedance of the delay line was matched to the oscilloscope input with a 12 ohm series resistor.
Comment:

The behavior of the input sections of the delay line appear to introduce a distortion of the input waveform.
The pre-undershoot at the leading edge of the delayed waveform could introduce difficulties to waveform analysis or feature extraction.
The jaggedness of the waveform near full amplitude could, likewise, introduce difficulties for waveform analysis or feature extraction.
The transfer function of the delay line, when convolved with the PMT waveform, may produce a waveform at the ATWD input that may not meet IceCube requirements for waveform fidelity. [The waveform fidelity requirement for IceCube digital optical module electronics has not been established and ratified as of Oct. 10, 2002]

Low Loss PCB Delay Line driven by square pulse

Test 5: Printed Circuit Bord Delay line produced at the direction of Dan Wahl, of U of Wisconsin SSEC. This delay line is also a strip-line embedded in printed circuit baord. The laminate stack-up includes thick, low absorbtion layers above and below the strip-line layers in the middle. The impedance measured 96 ohms by TDR method using fast pulser and fast oscilloscope.
The input waveform data file in binary form, and in spreadsheet form.
The output waveform data file in binary form., and in spreadsheet form.
The 1.3 meter active oscilloscope probe cable effectively subtracts 5.0 ns of propagation delay from the measurement
of the delay PCB. So, the total, corrected, delay is 62.4 ns.
Comment:

Better rise time than Test 3 above
Waveform free of strange distortions or ringing.
Delay within 4% of target.
10:1 delay to risetime ratio may be adequate. (PMT waveforms are only slightly distorted)

Test Ref: The input matching network used for the coax cable tests was connected directly to the output matching network to demonstrate that the system was well behaved.

For all the above measurements, the instrumentation was as follows:

A Tektronix Type 111 pulse generator provides the input pulse (blue trace)
The input pulse waveform measurement is made with an active scope probe.
The output pulse is impedance matched into the terminated input of the oscilloscope
The oscilloscope bandwidth is 500 MHz.
Waveforms were averaged over 25 'events'

Questions and comments should be directed to the author: Gerald Przybylski,