Noise & Grounding Andrei Nomerotski (U.Oxford) 17 July 2007
Outline This talk limits the discussion of the noise issues to those important for mechanics General principles Grounding of conductive pieces Breaking the ground loops Thanks to Marvin Johnson for many insights
General Remarks Current follows the path of least impedance Inductance and capacitance become important at high frequencies –Two 1 cm diameter 1 m long conductors 1 cm apart : Z_capacitive = Z_inductive at 30 MHz If one of these conductors is noisy GND and the other is silicon detector readout – half of the noise will be transferred to the readout General Recipe Provide low impedance path for currents, including the return currents Minimize capacitive coupling
Capacitive Couplings Generally would like to have a single point connection for GND to avoid ground loops But : tight packaging leads to large capacitive couplings which at MHz frequencies can dominate impedance and make single point connection not practical DZero Layer0 example Capacitor is formed by (sensor+analogue cable+hybrid) and (support structure) Referenced to common GND by long (inductive) lines large impedance at 10 MHz Susceptible to noise! SensorSVX4 Analog Signal in Long Cable Ground in Long Cable Ground in Support Dig.Signal Out Ground Out c
Capacitive Coupling Cnt’d Get noise if plates of capacitor are at different potential due to some source of noise – relevant for Layer0 where the capacitor itself was feeding the SVX4 preamp Also: Get noise if preamps are between the plates of capacitor – this is relevant to any layout Solution : Short the capacitor out using low inductance connections to GND Layer0 is an excellent antenna to catch any kind of noise so it is an extreme example. Most likely sensors with integrated electronics will be more robust to these effects. However the noise sources in the pit can’t be fully understood until the detector is installed.
CF is a Conductor DC : CF conductivity is times less than copper At high frequencies CF is as conductive as copper Measurements used capacitors formed by CF and copper plates and found no difference between copper-copper and copper-CF capacitors
Mesh and Wrap-Arounds Copper mesh can be co- cured with CF, see photo Ground connection to sensor using Kapton wrap-arounds, see sketch Local HV bypass capacitor was very important to reference HV to GND locally This scheme provided virtually zero pickup noise for D0 Layer0 silicon Carbon Fiber Flex Circuits Vias Low-pass filter board for HV Sensor Copper Mesh
Low Inductance Connections Inductance is maximized for objects of cylindrical shape and is minimized for flat extended objects (so currents in different parts talk less to each other) Multiple, distributed in space connections are good Single, long wire connections are bad Confirmed by measurements for Layer0 prototypes
Spacers Control of proximity to minimize capacitance had a mesh Kapton spacer between analogue cables Spacer had very little material, fill factor a few % analog cable pitch adapter mesh spacer wrap-around bumpers
Possible Support Materials for ILC VD CF is very conductive, already discussed Foams –RVC is conductive and quite soft so possibly difficult to ground –SiC is non-conductive Self-supporting silicon is conductive Be is conductive and difficult (but possible) to ground
Isolation of Two Sides CF (or other conductive) support can short two sides of the detector since it is non practical to have a gap at z=0 This ground loop can enclose a lot of noise sources (ex. Magnet Coil) Electrical isolation of two sides is essential and possible by electronics means (opto, transformers, differential signals, floating PS, resistors in bias lines etc.) –Need to keep in mind that all channels are connected in parallel. Achieved 10 Ohm overall isolation for Layer0.
Isolation of Two Sides, cnt’d Single wire diagrams are essential to make sure all sources are accounted for
Other issues Capacitive coupling to beampipe –~1 mm distance gives enough isolation for 10MHz 5 T magnetic field? How conservative we want to be?