Digital Signal Processing Reference
In-Depth Information
signal path of the receiver. The complexity of the windowing circuit should be
kept to an absolute minimum in order to avoid capacitive coupling between
the clock lines and the received rf-signal. The most critical factor here is
charge injection due to the overlap capacitance between the steering gate and
the drain/source terminals of the mos transistors. The repetitive nature of the
injected signal (caused by the opening and closing of the receive window) gen-
erates spurs in the signal path of the receiver and can cause clipping further on
in the baseband amplifier.
The dimensions of the transistor switches should be kept as small as possible,
with some obvious consequences for the isolation performance of the window.
For example, a circuit can use a shunt switch to divert unwanted interferer
power to a (virtual) ground node. The transistor that is used for this purpose
has a certain on-resistance. This means that a certain amount of power leaks
into the signal chain during the off-state of the receive window which results
in implementation losses (il).
However, a considerable part of interferer power leaks into the receiver any-
way, during the on-state of the receive window. On average, the leakage power
makes up only a small portion of the total interferer power at the output of the
windowing circuit. If the duty cycle of the receive window is 10%, the suppres-
sion factor of narrowband interferers is 10 dB. It makes no sense to demand an
isolation of the shunt transistors which is several orders of magnitude better
than the suppression ratio of the time-domain filter itself.
For example, suppose that an implementation loss of 0
.
4 dB can be tolerated.
It can be calculated that this boils down to the fact that leakage-induced power
makes up about 10% of the total interference which enters during the active
phase of a receive slot. It follows that the isolation of the switch must be
S
21
=−
20 dB (or better) with respect to the insertion loss in pass-through
mode. Then, considering the impedances at the surrounding circuit nodes, the
maximum value for the on-resistance of the shunt switch can be easily deter-
mined. Finally, based on the maximum on-resistance, the minimum dimensions
of the shunt window switch can be determined.
Important notice!
A second measure that significantly reduces the effects of charge in-
jection is to place the window switch in front of the downconversion
mixer. In the design of the chip that is described here (see figure be-
low), the window switches were incorrectly positioned at the outputs
of the downconversion mixer. As will be shown later on in the mea-
surement section, leakage of clock signals into the sensitive baseband
section of the receiver resulted in spurious components at the symbol
