Stubs
in the high speed tracks
Stubs
(branches from the main routing path) cause unwanted reflections
if they exceed the maximum allowable stub length for the signal
carried by the bet. This generates additional EMI as well as causing
a signal distortion.
Pouring
copper for reducing characteristic impedance
A
surrounding shield provides a local return path for EMI. Track
shielding also reduces the characteristic impedance of high-speed
nets. A common way to reduce emissions is to fill spare areas
of each layer with copper, thus providing an effective screen
to all signal tracks on the layer. However, these areas of copper
can act secondary radiators of EMI and should always be connected
to a power plane
Loop
Antenna Elimination
A
tracking loop constitutes a loop antenna for EMI. Radiation is
generated in the plane of the loop. This rule considers fully
closed loops only. The severity of each loop is approximately
proportional to the square of the operating frequency times the
approximate loop area
Mitering
the track
Right-angled,
and more acute corners, cause field concentrations which increases
EMI, particularly where the 'edge rate' is high.
Impedance
setting and the design technology
Areas
of high characteristic impedance give higher emissions. In order
to reduce EMI and Susceptibility for a design, it is good practice
to reduce the inductance and increase the capacitance of the signal
tracks. The characteristic impedance of a transmission line is
approximately: sqrt(L/C). So, to reduce the impedance, you need
to increase the capacitance (C). You can increase the capacitance
of a track by shielding it, or by altering the dielectric between
the track and the Power plane.
Reducing
the impedance
In
order to reduce the EMI and Susceptibility you can attempt to
reduce inductance and/or increase capacitance in the signal tracks.
The
characteristic impedance of a line is approximately the square
root of L/C (the unit length inductance divided by the unit length
capacitance). So high impedance areas are an indication that additional
capacitance is required.
Ground
planes
Some
of your designs may make use of multiple Power planes at different
voltages.
It
is not good practice to overlap the planes of different voltages
(this can lead to undesirable noise current distribution throughout
the system).
It
is also good practice to overlap corresponding Power planes (this
gives additional capacitive decoupling).. For example, the 5 Volt
plane and its corresponding Ground plane).
Overlapping
Ground planes are capacitively coupled to each other - so noise
can transfer from one to another.
In
addition, large copper areas are excellent radiators of EMI, so
it is good practice to prevent high frequency signals from appearing
on Ground planes.
On
any design using high-speed components, at least one Ground plane
should exist to ensure reasonable control of emissions. The existence
of a Ground plane also ensures that the characteristic impedance
of routes is well defined, and is effective in suppression of
excessive inter-route cross talk. Partial plane s are also useful
in this situation.
A
Ground or Power plane may become excessively perforated by through-hole
pins, vias and thermal relief cut-outs in a particular area. This
has the effect of increasing the resistance of the plane in that
area, and the impedance of the total plane. This can result in
transient conditions on the board, leading to increased EMI.
Efficient
placement
Placing
the highest speed/power components in close proximity to the source
of power reduces transient problems. For boards without ground
planes, radiation is reduced due to the reduced power signal loop
areas.
Setting
the power rails
Certain
designs require several voltage power rails which are associated
with different Power planes. For example there may be 5V Logic
power rail, together with 12V rails for the Analogue devices.
These
power rails will have corresponding ground rails which may also
be Power planes. It is good practice to overlap ie. Keep separate
power planes one over another as different layers. This will increase
the capacitive de-coupling of the power supply.
De-coupling
capacitors
The
appropriate de-coupling of components is essential in order to
reduce transient conditions on supply connections. The de-coupling
capacitor appropriate to a device depends on a number of electrical
parameters associated with the device itself. These parameters
include voltage levels, rise times, and power supply transient
current values.
Appropriate
de-coupling of active components reduces component noise and power
plane transients. Reducing the voltage transients on power supply
connections reduces emissions from the signal loops formed by
the power connections. Although circuit conditions affect the
choice of de-coupling capacitor to a certain extent, the technology
of the device being de-coupled is the most important factor in
the choice of capacitor and method of de-coupling.
Signal
cross talk
The
cross talk levels expected on the design must be quantified in
order to reduce interference between lines.
Cross-talk
is caused by parallel, or near parallel, tracks on the same board
layer. or on adjacent board layers. A change of state on one line
causes interference on another due to mutual capacitance and inductance
between the lines.
The
net causing cross-talk is known as the active net. The net receiving
(picking up) the cross-talk is called the passive net. The active
signal is the signal on the active net.
There
are two types of cross-talk:
Backward Cross-talk
This
is usually the dominant form. The backward cross-talk signal travel
down the passive line in the opposite direction to the active
signal. It is always of the same polarity as the active signal
Forward Cross-talk
This
is a secondary effect which occurs predominately on the outer
board layers. The forward cross-talk signal travels in the same
direction as the active signal, and is usually of opposite polarity.
Route
segments on the adjacent layers of multi-layer boards should run
at right angles to each other. This helps to reduce electromagnetic
interference by reducing capacitive coupling between routes on
adjacent layers.
Track
resonance cannot be over ruled
As
the time-length of a route (given by the delay along the track)
approaches a multiple of a quarter of the signal wavelength (or
a harmonic of that signal), its efficiency as a radiating antenna
increases. This increase the electromagnetic radiation from the
route, and also enhances the route's susceptibility to externally
generated electromagnetic interference.
The
Track Resonance rule uses the frequency of operation, and rise
time of a signal on the net, to derive a set of harmonic frequencies
where quarter-wavelength tuning is significant.
High
frequency nets with long routes should be terminated to avoid
extra high frequency harmonics. Each track has a critical length
above which reflections are a maximum, and below which they are
reduced. This critical length occurs when the round-trip delay
time along the track is equal to the rise-time of the signal.
It is recommended practice to terminate lines whose length is
close to or above this critical value.
|
