https://www.mpdigest.com/2019/07/24/low-loss-high-isolation-mmwave...

by Graham Pearson and Liam Devlin, Plextek RFI

Many emerging applications at Ka-band and above, such as 5G and broadband satellite
communications, are adopting architectures that require low-loss, high-linearity switches. FET-based
switches on both GaAs and SOI (Silicon on Insulator) are used to realize high performance, low-loss
switches at RF frequencies. They have been used in high-volume products for many years, and
provide a low-cost practical solution. At higher frequencies, switches can still be realized using these
technologies, but as the operating frequency increases, the losses creep up. PIN diode MMIC
technology offers a means of realizing mmWave switches with very low insertion loss. The PIN diode
MMIC technology and the design approach for realizing mmWave switches is discussed, and a
specific Single Pole Double Throw (SPDT) switch example is presented that offers a measured on-
state loss of 0.5dB and an off-state isolation of 52dB at 28GHz.

Solid-state Switch Technologies

With a FET-based switch, the drain-source voltage is held at zero volts and the gate is used to control
drain-source resistance. With the FET in its on-state, the drain-source path behaves like a low value
resistance. With the FET pinched-off the resistance is high. This allows the convenient realization of
high-performance, voltage-controlled RF switches.

The difficulty with FET-based switches is that when pinched-off, there is a parasitic drain-source
capacitance that offers a low impedance path to RF signals at higher frequencies. This capacitance
can be reduced by using smaller transistors, but then the on-state loss increases and the power
handling reduces [1].

PIN diode switches have a much reduced parasitic off-state capacitance for a comparable on-state
resistance, and so offer a means of providing low insertion loss and high isolation at mmWave
frequencies. A PIN diode takes its name from its structure; it comprises a region of high resistivity
intrinsic material sandwiched between a region of P-type semiconductor and N-type semiconductor.
When the PIN diode is forward biased, charge carriers are injected into the I region, lowering its
resistance. Thus at RF and microwave frequencies, a PIN diode behaves as a current-controlled
resistor [2]. They are optimized to achieve wide resistance range, good linearity, low distortion, low
drive current, and high power handling capability. These properties mean that PIN diodes can be
configured to make excellent RF/microwave switches and also find applications in variable
attenuators and phase shifters.

1 of 5 21-08-2019, 12:02
 https://www.mpdigest.com/2019/07/24/low-loss-high-isolation-mmwave...

Figure 1: Electrical equivalent Figure 2: Typical simple PIN diode

model of a PIN diode RF SPDT schematic Figure 3: The design path

Figure 1 shows a simple equivalent electrical circuit model for a PIN diode at RF/microwave
frequencies. This model is for the intrinsic diode: the feed structure and contact metallization also
need to be modelled to accurately predict the performance at mmWave frequencies. The forward
current through the diode controls its resistance Rj. For zero or reverse bias, with no current flowing,
the resistance Rj is high—in the region of several kΩ. As the forward bias current through the diode is
increased, the charge carriers injected into the I region reduce the value of Rj to values as low as one
or two ohms, depending on the diode structure.

Circuit Design

The fact that a PIN diode is essentially a voltage-variable (current-controlled) resistor means it is
extremely well suited to switch realization. A simple series-mounted diode can be used in each arm of
a multi-pole switch. With the diode forward biased and conducting a DC current, the switch arm is on.
With the diode reverse biased, the switch arm is off. This arrangement is shown in Figure 2 for a
Single Pole Double Throw (SPDT) switch and can be easily extended to multi-throw implementations.

Even though the off-state capacitance of the diodes is lower than that of a switch FET, it still limits the
isolation that can be achieved at microwave and mmWave frequencies when used in series, as
depicted in Figure 2. Decreasing the size of the diode (which involves selecting a different diode in
the case of discrete designs) will reduce the off-state capacitance (so increasing the isolation) but will
also increase the on-state resistance (so increasing the on-state insertion loss). Switch
implementations using combinations of series and shunt diodes, switching complementarily, can be
used to provide additional isolation.

Although the off-state capacitance of PIN diodes is very low, at mmWave frequencies this can still
represent a significant reactance. An off-state capacitance of just 0.05pF would reduce the isolation
of an SPST to just 3.6dB at 28GHz. Achieving adequate isolation from a series-mounted diode would
require a diode so small that the design rules of the process would be infringed. Additionally, the on-
state resistance would result in unacceptably high insertion loss. This dilemma can be resolved by
using a switch approach that requires only shunt-mounted diodes.

A reverse-biased shunt-mounted PIN diode presents a low capacitance to ground. The key to
designing mmWave switches is to absorb the shunt capacitance into a low pass filter structure. This

2 of 5 21-08-2019, 12:02
 https://www.mpdigest.com/2019/07/24/low-loss-high-isolation-mmwave...

results in a very low loss through path when the diode is off (reverse biased). When a forward bias
current is allowed to flow in the diode, it presents a low resistance path to ground. Thus the same
low-pass filter structure will then provide a high insertion loss (or high isolation) when used in a
switch.

The design route to absorbing the diode capacitance into the low pass filter is illustrated in Figure 3.
The starting point is to design a conventional low pass filter with acceptable upper operating
frequency [3]. The diode or diodes are then selected to have a size that has the equivalent—or
slightly lower—capacitance. All of the shunt capacitors are replaced with the diodes and the filter’s
performance is re-optimized. Choosing the number of shunt capacitors in the filter (and so the
number of shunt diodes in the switch) is a compromise between high isolation and lower die
area/current consumption. In the SPDT design example reported here, two shunt diodes were
selected. In the case of mmWave switches, the series inductors are very low in value and are
replaced by short lengths of high-impedance microstrip transmission line. Again the performance of
the filter is re-optimized to give acceptable insertion loss and match.

In order to realize multi-throw switches, the input impedance of the filter structure with the diodes in
the on-state (low-impedance) must approximate an open circuit. If this is the case, a number of switch
branches can be connected to a common port to realize a single pole multi-throw switch. The series
transmission line at the input of the filter is used to transform the low impedance of the on-state diode
to a high impedance at the input of the switch arm.

In summary, the filter structure that has been designed is well matched, having a low insertion loss
when the diodes are “off” (reverse biased) and a high insertion loss with a near open-circuit input
impedance when the diodes are “on” (forward biased). The final step of the design process is to
incorporate DC bias tees to allow the injection of bias current to turn the diodes “on” and the
application of a reverse bias to turn the diodes “off.” This can be implemented by the addition of a
high impedance shunt transmission line, RF connected to ground, to each switch branch. The length
of the line is adjusted to present a high impedance in the band of interest. Series capacitors (DC
blocks) are added to the input and output of the filter structure to keep the RF ports DC isolated.

Realization and Measured Performance

The switch was realized on the PIN diode MMIC process of WIN Semiconductors. It has a 3µm thick
I-region offering low capacitance. A photograph of the MMIC is shown in Figure 4. All RF ports are
configured with Ground-Signal-Ground (GSG) pads to facilitate RF test. The common RF port is on
the center bottom edge, with the other two RF ports on the left- and right-hand edges. The two DC
bias/control pads are numbered 2 and 3 on the top edge. On-chip decoupling is included at each pad
with a series bias resistor to set the forward bias current to 10mA per diode from a +3V supply.
Choosing the forward bias current at which the diodes are operated is a compromise between
improved isolation (for higher bias currents where the on-state diode impedance is lower) and lower
current consumption but reduced isolation (with the on-state diode impedance of the diode being
higher). The off-state switch branch is reverse biased. Good small-signal performance is achieved
with 0V, but a higher reverse bias voltage is required for RF operation at power. The switch should
handle RF power levels of 1W with a reverse bias of 10V and 4W with a reverse bias of 20V.

3 of 5 21-08-2019, 12:02
 https://www.mpdigest.com/2019/07/24/low-loss-high-isolation-mmwave...

Figure 4: Photograph of the 20 to 30GHz SPDT switch Figure 5: Measured (solid traces) to simulated (dashed
MMIC traces) s-parameters of a typical switch

Figure 6: Measured on-state and off-state insertion loss of a typical switch

The RF performance of the switch MMICs was measured by RFOW probing. Figure 5 is a plot
showing the measured on-state insertion loss and input return loss compared to that simulated. The
measured performance is shown by the solid traces and the simulated performance by the dashed
traces. Across the 20 to 32GHz frequency band, the measured on-state loss is 0.55dB ± 0.1dB, with
the measured isolation greater than 47dB.

Figure 6 shows the measured on-state loss compared with the measured off-state isolation. Very low
insertion loss and high isolation is clearly evident across 20 to 32GHz, with useful performance
extending beyond this key band.

The switching speed has also been measured on-wafer using a crystal detector driving a high speed

4 of 5 21-08-2019, 12:02
 https://www.mpdigest.com/2019/07/24/low-loss-high-isolation-mmwave...

oscilloscope. The measured switching speed (10% RF to 90% RF) was around 20ns, but this was
limited by the driver circuitry used to control the switch, so the actual switching speed is faster than
this.

Conclusions

PIN diode MMIC processes are well suited to the realization of low loss, high isolation mmWave
switches. The design approach has been discussed and the design and measured performance of an
SPDT switch MMIC covering 20 to 32GHz has been presented. The measured on-state loss across
this band is 0.55dB ± 0.1dB and the measured isolation > 47dB.

References

[1] Liam Devlin, “The Design of Integrated Switches and Phase Shifters”, Proceedings of the IEE
Tutorial Colloquium on “Design of RFICs and MMICs,” Wednesday 24th November 1999, pp 2/1-14.

[2] Doherty, Bill, “PIN Diode Fundamentals,” MicroNote Series 701, Microsemi Corp., Watertown, MA.

[3] Matthaei, Young and Jones, “Microwave Filters, Impedance-Matching Networks and Coupling
Structures,” Artech House 1980, ISBN 0-89006-099-1.

(97)

Related posts:
1. Design and Evaluation of a 125W L-Band GaN Power Amplifier
2. Design of a Broadband L-band 160W GaN Power Amplifier Using SMT Packaged Transistors
3. The Design of Multi-Chip SMT Front-End Modules for mm-Wave 5G Applications

5 of 5 21-08-2019, 12:02