Product Application Note March 29, 2002
Design Guidelines for Microwave Cavities
Background: Placing a microwave circuit, active or passive, in a cavity below its
waveguide cut-off frequency will decrease the effects of radiating signals (such
as feedback, gain ripple or decreased isolation) from the microwave elements.
Method: The theoretical equation for the lowest waveguide cutoff frequency
(mode TE01) for a rectangular waveguide is
Fc = c / 2b
where Fc = cutoff frequency (Hz)
c = speed of light (3x10^8 m/sec)
b = longest dimension of the rectangular waveguide cavity (meters)
This equation assumes a dielectric constant of 1 (air). In reality, the waveguide
cavities with microwave circuits contained therein are partially filled (usually
<10%) with a substrate that has a dielectric constant greater than 1. There exist
more complicated design equations that can calculate the cutoff frequency for a
partially filled cavity, but this extra complexity is usually not worth the effort. The
many other factors that influence the characteristics of radiating signals, such as
bondwires, circuit patterns, or passive components, typically have a larger effect
on the final result.
A second equation is used to calculate the attenuation of the microwave signal
as it travels down the waveguide channel even though it is below the cutoff
frequency. This is called the evanescent mode; the signal strength decays rapidly
with distance and is proportional to the ratio of the cutoff frequency to the signal
frequency
Alpha = sqrt ((pi/b)2 – (2*pi*f/c)2)
where Alpha = attenuation constant (Nepers/m)
pi = 3.14159
b = dimension of waveguide (meters)
f = signal frequency (Hz)
c = speed of light
Then the attenuation in dB is,
Attenuation = 20*log(exp(Alpha*z)) dB
where z = length of waveguide channel (meters)
For microwave circuits with a lot of gain (30+ dB) or with a high isolation
requirement (such as the stopband of a filter) a good rule of thumb is to make the
cutoff frequency 3x to 4x the highest signal frequency of interest. As the gain or
isolation requirements moderate the cutoff frequency can be moved closer to the
signal frequency (the microwave channel can get wider).
Examples:
Design #1:
Design a stable amplifier chain consisting of four 20 dB amplifiers and 10 dB of
losses (attenuators and filters) at 7 GHz.
Objective: Reduce the non-propagating signal in the channel from output back to
the input by 40 dB greater than the forward gain (70 + 40 = 110 dB loss). With
the feedback signal 40dB lower than the input signal a gain ripple (or error) of
less than 0.1 dB is expected due to the worst case vectoral addition of these two
signals. The equations for this calculation are
Amplitude error (worst case, additive) = 20 * log(1 + 10^(fb/20))
Amplitude error (worst case, subtractive) = 20 * log(1 - 10^(fb/20))
Phase error (worst case) = arctan(10^(fb/20))
Where fb = level of feedback signal relative to the level of the main
signal in dB.
This value will be negative if the feedback signal is smaller
than the main signal at the input (as it needs to be).
The waveguide channel dimension was limited by the size of the filters to 0.25
inches (b dimension). The length (z dimension) of the channel was approximately
0.90 inches. So the cutoff frequency is
Fc = c / 2b = 3e8 / (2*0.25 / 39.3701) = 2.36e10 or 23.6 GHz
The calculated attenuation of the evanescent wave at 7GHz is
Alpha = sqrt((3.14159/(0.25/39.3701))^2 – (2*3.14159*7e9/3e8)^2)
= sqrt(2.45e5 - 2.15e4)
= 473 Np/m
Attenuation = 20*log(exp(473*0.9/39.3701)
= 93.9 dB
This value is less than the desired 110 dB of loss required for this design, but it
doesn’t take into account the probe losses – the gain of the “antenna” that
launches the signal from the microstrip environment to the waveguide mode. In
this design the most effective radiating element was the 5-pole bandpass filter
that was at the output of the amplifier chain. It is very difficult to calculate the loss
from imperfect antennas such as filters or bondwires. However, it was estimated
that this filter had an “antenna gain” of no more than –10dB. At the input to this
amplifier chain, the “receiving antenna” was considered to be the bondwires
attaching to the input of the first MMIC. Their gain had to be less than –10dB
(since they are so much smaller than the filter elements) so it was considered
that this design met its goal (93.9 + 10 + 10 > 110dB). The empty waveguide
cavity was measured with simple 50 ohm lines placed into each end and two
bondwires attached to their ends. The measured isolation was greater than 110
dB (the value exceeded the dynamic range on the instrument). The completed
design, with all amplifiers and filters in place in the narrow channel, exhibited a
feedback signal more than 110 dB below the output signal.
Design #2:
Microstrip edge-coupled bandpass filters are often used at millimeterwave
frequencies because of their compact size, lack of vias, reasonable line
width/spacings, and good performance (insertion loss, bandwidth, filter
selectivity).
Objective: Design a bandpass filter at 25 GHz with a 5% passband and 40dB
rejection at 0.95*center frequency. Achieve flat passband response and a 40dB
loss in the stopband while taking into consideration the variables affecting the
filter performance such as substrate dielectric constant, substrate height, line
width/spacing dimensional tolerances and temperature.
After estimating the influence of all of the variables a filter can be designed with
sufficient margin to meet the specifications. It is typically a 5 to 7 section design.
When tested on an open carrier plate however, the lower frequency skirt of the
filter misses the 40 dB rejection level at the specified frequency by 15 to 20 dB.
This is due to the resonators of the filter radiating a portion of the signal which is
“received” by the next resonator. A linear circuit simulator cannot predict this
response because it doesn’t include radiative coupling between the open ends of
the resonators.
The solution is to place the filter in a waveguide cavity. The chosen width of this
cavity was 0.12 inches and the filter length was approximately 0.3 inches. The
design values are
Fc = 3e8 / (2*0.12/39.3701) = 49.2 GHz
Alpha = sqrt((3.14159/(0.12/39.3701))^2 – (2*3.14159*25e9/3e8)^2)
= 886 Np/m
Attenuation = 20*log(exp(886*0.3/39.3701) = 58.6 dB
Even though this cavity produced a cutoff frequency-to-signal frequency ratio of
only 2x, it was sufficient to bring the response back to the predicted values. In
the absence of radiated energy the shape of the filter matched the simple linear
model. Also, the interaction with the lid (reflections) was low enough to avoid the
use of absorber.
These equations provide a good first order approximation to the problem and can
sometimes highlight serious radiation issues before the design is frozen. Due to
the many other variables which can add to, or subtract from, the radiating or
propagating signal (bondwires, substrates, microwave structures, filters, passive
components, etc), it is best to stay as conservative as the design will allow. At
higher frequencies such as 40 GHz it becomes difficult to build a channel below
the waveguide cutoff frequency (b=0.147 inches at 40 GHz) and still support the
circuit element sizes. To achieve a 3x ratio at 40 GHz would require a channel
width of 0.049 inches and a height from module floor to lid less than this value.
If the design dictates that active components, such as MMIC amplifiers, be
placed in a propagating waveguide channel, it is prudent to limit their gain to 20
or 30 dB maximum. The use of absorber on the lid in this case will almost always
be required and some gain ripple due to radiative feedback of the output signal
can be expected. The best course of action is to keep everything very close to
the ground plane. This reduces to a minimum the radiation of components such
as bondwires and other transitions. It is not uncommon for a MMIC amplifier with
15 or 20 dB of gain to lose about 1 dB when a lid with absorber is placed above
the MMIC. This is an indication that the radiative signal level is not negligible.
Recommended Approach:
1. Use cavities below waveguide cutoff as needed to eliminate radiative signals.
A ratio of 3x to 4x (cutoff frequency to signal frequency) is advisable.
2. Minimize gain at high frequencies whenever possible.
3. Break up gain into smaller blocks (place filters between gain stages)
especially if the amplifiers are in a propagating cavity or near the cutoff
frequency.
4. Use absorber to reduce reflections from the lid. This is often necessary even
in small channels well below the cutoff frequency.
For additional information, please contact TriQuint Texas Applications Engineering Department
at 972-994-3647.
