Local Coincidence in AMANDA string 18, and IceTop.
Background:
The idea of using local coincidence, a trigger level filter in
the DOMs, arose in 1999 for the purpose of reducing the amount of data
that needed to be transmitted from the DOMs to the surface over a cable
with limited bandwidth. The need for better bandwidth
utilization arose as the estimates of trigger rates, and the set of
data that constituted an event continued to expand.
The gist of the idea is this: If two adjacent DOMs are 'hit'
within a time window of a few hundred ns of each other, then they most
likely both saw light from the same event. The chance that two
noise hits overlap in that time window is finite, but very small.
Circuits were added to the DOMs that allowed them to send short signals
to each other, bidirectionally, over one pair of wires. The
circuit depends on the port to port isolation characteristic of a power splitter.
The IceCube and IceTop DOMs extend the flexibility of the LC
configuration, and increase its immunity to deployment
limitiations, like LC cable impedance mismatches. IceCube
DOMs include an independant LC circuit for each communications
direction, and bipolar (window) comparators for signaling
detection.
Power splitters/combiners are transmission line products, which produce
best results when, and in some sense, depend upon impedance matching at
every port. Power splitter modules consist of coupled
inductors, a transformer, and a resistor that defines the system
impedance. An external shunt resistance can, of
course, be used to 'pull' the impedance of the module in order to match
the impedance of lines that connect modules together.
The basic configuration of the DOM, which is based on 75
Ω power splitters, and a 2:1 transformer, matches a nominal
150 Ω cable from DOM to DOM. Shunt
resistors, and several other resistor substitutions permit matching
impedances in the range of [120Ω to 150 Ω]. The available LC
cable (choice made by the IceCube cable engineers) will, in the
end, place constraints on the LC circuit in the DOM.
Component values in the DOM will be adjusted to provide the best match
and noise margin.
The AMANDA circuit
The IceCube circuit
2003/2004 First Deployment of the IceCube circuit
A 2003/2004 Summer Season deployment of IceCube DOMs,
colaboration between U of Wisconsin Madison Physics (IceCube project
office) and U of Delaware Bertol Institute (IceTop project office)
brings together all the components... and brings to light all the
surprises one might expect from a three, or more way
colaboration. Late in the integration process it was
discovered that the custom manufactured cable with a target impedance
in the range of 140 Ω actually measured in the range of 120 Ω to 130
Ω. The available quad cable left over from AMANDA deployments
measured approximately 160 Ω. The significance of these
cable impedance mismatches to a system designed for a target impedance
of 150 Ω had not been considered, much less appreciated.
Cable mismatches are usually characterized as a ratio of impedances
where the larger impedance is divided by the smaller impedance (by
convention) and the quantity is known as the Standing Wave Ratio.
The SWR predicts the amount of power reflected back to the source by an
impedance mismatch. The SWR of 160/120= 1.33
seemed disturbingly high. Therefore, simulations of various
possible cable combinations and LC circuit design impedances were
undertaken to see whether special attention was needed (or how
forgiving the circuit was of cable mismatches).
Simulation components:
The simulation exploits some inside knowledge of power splitter
construction in order to simplify the circuit, and reduce the number of
components involved. The simulation also
utilizes a lossy transmission line model, though for such short line
lengths, a lossless model would suffice. Real transmission
line components are necessary to observe the reflected signals that are
a consequence of impedance mismatches.
The characteristic impedance of the LC transceiver depends on the
values of three resistors. The characteristic impedance of
the transmission lines in the spice model depend on adjustable
parameters of the model... expressed in L, the inductance per meter,
and C, the capacitance per meter.
The impedance and delay are given by the following equations:
| Zo
= |
|
Td =
|
|
The above is a system of two equations in two unknowns, and easily
solved L and C for given Zo and Td, .
The delay is adjusted by a factor of 0.66 to account for the dielectric
constant of polyethylene (a delay of 5.05 ns / meter).
Summary:
- Even the worst
case mismatch simulation below should produce a received level at the
comparator input which is robust against the side effects of
reflections.
- The
configuration that produces the greatest margin against noise adjusts
the DOM LC circuit to match the 130 Ω, and uses 140 Ω cable between DOMs.
- Somewhat worse is the
configuration which adjusts the DOM LC circuit to match the 140
Ω, and
uses 140 Ω cable between DOMs.
- Somewhat worse is the
configuration which adjusts the DOM LC circuit to match the 130
Ω, and
uses 160 Ω cable between DOMs.
- With severe cable
mismatches, the worst reflections in a real DOM are predicted to be
about 40 mV, measured at the comparator input port. The
optimum threshold appears to be around 62mV.
Seven simulations are posted below.
To renormalize for IceCube DOM levels, divide voltage
values by 2.
The IceCube DOM transmitted LC level, measured at the combiner input
port is 0.25V>
The IceCube DOM received LC Thresholds are 75mV, symmetric about
the
LC system offset level, for an expected 125mV received signal level.
Comparison of the plot files yields a good feel for how mismatches
produce waveform distortions.
Simulations1(plot) (sch) (sim) (zip):
DOM LC impedance = 130 Ω
Pair Cable = 130 Ω
Black cable = 130 Ω (as a check)
This simulation is a check of the basic simulation which should, and
does, perform perfectly. i.e. no reflections.
Simulations2:(plot) (sch)
(sim) (zip):
DOM LC impedance = 130 Ω
Pair Cable = 130 Ω
Black cable = 140 Ω (New, Solid
Polyethylene dielectric Ericsson quad cable)
This simulation for the case where DOM MB components are adjusted to
match the 130 Ω cable, and 50 feet of the new
Ericsson quad cable connect the two DOM penetrator pigtails..
Simulations3:(plot) (sch)
(sim) (zip):
DOM LC impedance = 130 Ω
Pair Cable = 130 Ω
Black cable = 160 Ω (Old, Foamed
Polyethylene dielectric Ericsson quad cable)
This simulation for the case where DOM MB
components are adjusted to match the 130 Ω
cable, and 50 feet of the old Ericsson quad cable connect the two
DOM penetrator pigtails..
Simulationsr4:(plot) (sch)
(sim) (zip):
DOM LC impedance = 150 Ω
Pair Cable = 130 Ω
Black cable = 140 Ω (New, Solid
Polyethylene dielectric Ericsson quad cable)
This simulation for the case where DOM MB
components are not adjusted, and 50 feet of the
new Ericsson quad cable connect the two DOM penetrator pigtails..
Simulationsr5:(plot) (sch)
(sim) (zip):
DOM LC impedance = 150 Ω
Pair Cable = 130 Ω
Black cable = 160 Ω (Old, Foamed
Polyethylene dielectric Ericsson quad cable)
This simulation for the case where DOM MB
components are not adjusted, and 50 feet of the
old Ericsson quad cable connect the two DOM penetrator pigtails..
Simulationsr6:(plot) (sch)
(sim) (zip):
DOM LC impedance = 140 Ω
Pair Cable = 130 Ω
Black cable = 140 Ω (New, Solid
Polyethylene dielectric Ericsson quad cable)
This simulation for the case where DOM MB
components are adjusted to match the 140 Ω
cable, and 50 feet of the new Ericsson quad cable connect the two
DOM penetrator pigtails..
Simulationsr7:(plot) (sch)
(sim) (zip):
DOM LC impedance = 150 Ω
Pair Cable = 130 Ω
Black cable = 100 Ω (This simulation
tests the robustness of the face of gross mismatches)
This simulation for an arbitrarily chosen worst
case case mismatch condition where DOM MB components are not adjusted.
Ten foot lengths of the colored pairs are used, and fifty feet of 100
ohm couples the sections of colored pair together.
(sim) Pspice (demo) simulation file containing
data points for all nodes in the model. Suitable for examining
results in detail or creating new plots.
(plot) pdf file of plot of indicated voltage test points.
(sch) pdf schematic of the circuit simulated
(zip) Zip file containing all the pspice files for the simulation.
Sorry... but Pspice makes some pretty crummy color choices.
However, the pdf files are editable with Adobe Illustrator.
Comments? Questions? Corrections? Contact Gerald Przybylski,
