APPLICATION NOTE
One Technology Way • P.O. Box 9106 • Norwood, MA 02062-9106, U.S.A. • Tel: 781.329.4700 • Fax: 781.461.3113 • www.analog.com
Rev. 0 | Page 1 of 8
AN-851
ADF4153 ADF4360-8
~100Ω ~1kΩ
06147-001
NF = 0.9dB NF = 0.9dB NF = 7.5dB NF = 8dB IIP3 = 33dBm
IIP3 = 18dBm IIP3 = 18dBm IIP3 = 24dBm IIP3 = 12dBm NF = 8dB (AT MAX GAIN)
OIP3 = 35dBm
Figure 2 shows the spurious trajectories for a downconverting Next it is necessary to consider the final IF. The second down
mixer using low-side LO injection. In order to cover the converting mixer has the benefit of a narrower input band-
2.3 GHz to 2.4 GHz bands while minimizing the number of width, typically no more than 20 MHz. This allows several
spurious mixer components that can interfere with the desired possible second IFs to be considered, with ranges from 10 MHz
signal, inspection of the spurious trajectories indicates that the to 70 MHz and higher IF frequency bands centered at 107 MHz
first IF should be within a range of 210 MHz to 400 MHz. The and 140 MHz. A 70 MHz IF is selected in order to avoid higher
grey shaded area indicates the fractional bandwidth bounded by order LO harmonics that can leak back to the antenna receive
fRF/fLO and fIF/fLO for a 374 MHz IF using a low-side LO (LO1 = port and fall in the desired receive band. Additionally, the
1926 MHz to 2026 MHz). It is possible to substitute other IF 70 MHz IF allows for a wide selection of commonplace SAW
center frequencies, but caution must be taken to ensure that the filters to be applied, or alternatively, practically lumped element
spurious responses of the first mixer do not cause in-channel LC passive filters.
interference. For this design, a first IF of 374 MHz is selected.
There are several commercially available SAW filters of various After establishing the target IF frequencies it is possible to
bandwidths from Sawtek, as well as other manufacturers, that consider the cascaded performance. Table 2 depicts the
address this center frequency. anticipated dynamic performance of the double downconver-
sion receiver from the output of the band-select filter down to
1.0
the ADC interface.
L
-2L
4R-4
0.9
3L
2R
L
0.8 R-
Signal Chain depicted in Figure 1
5L-
-4L
-5L 3L
3R
2R
4L
3R
R
0.5
fIF
2L
3L
0.4
37 −7.7 4.6
0.3 1.0
23.5 −5.6 9.7
4L-
0.2
3R
0.1
0
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
fRF
06147-002
fLO
Rev. 0 | Page 2 of 8
AN-851
IMPLEMENTATION
25 5
The first component in the receiver chain is a band-select filter.
Several manufacturers, such as Anatech Electronics, Inc., K & L
Microwave Inc., and Digital Communications Inc., provide 20 4
GAIN (dB)
2.5 GHz WiMax deployments. The cavity-based filters are
NF (dB)
capable of low insertion loss of less than 2 dB with up to 60 dB
10 2
of stop-band rejection at only 25 MHz offset from the center of
the pass band. The lower cost ceramic-based filters offer closer
to ~50 dB of stop-band rejection at 150 MHz offsets. The choice 5 1
of band select filter strongly depends on the image rejection
performance of the receiver and the expected magnitude of the
0 0
06147-004
interfering signals present in the vicinity of the band of 2100 2200 2300 2400 2500 2600
FREQUENCY (MHz)
operation. For this demonstration, a Digital Communications
Inc. 10-section band-pass cavity filter is selected at a center Figure 4. ATF541M4 Gain and NF, Input Return Loss is Measured to be −8 dB
with an IIP3 of 18 dBm
frequency of 2350 MHz.
In order to help improve the image rejection of the front-end
The first LNA stage follows the front-end, band-select filter. The
design and to minimize the presented broadband noise to the
insertion loss of the filter and the noise figure of the first stage
first mixer, a simple high-pass filter was used between the first
dominate the overall cascaded sensitivity of the receiver. As a
and second LNA. The rejection requirement is not very critical
result, it is critical to achieve very low noise figure in the first
considering the first band-select filter provides >60 dB rejection
LNA stage. The ATF541M4 GaAs E-mode pHEMT was selected
of unwanted signals appearing at the image frequency. Using a
from Agilent Technologies for its low noise and high OIP3. The
374 MHz IF with low-side LO injection, the fundamental image
circuit implementation for the first and second stage LNAs is
band is from 1552 MHz to 1652 MHz for a 2.3 GHz to 2.4 GHz
presented in Figure 3, along with the measured gain and NF
input frequency range. A simple 3-pole lumped element filter
performance in Figure 4.
was designed between the first and second LNA stages. The
5V
filter provides better than 20 dB of image band rejection and
10nF less than 2.5 dB of insertion loss in the desired pass band. The
1nF
measured frequency response and circuit implementation is
32.4Ω depicted in Figure 5.
294Ω 1.21kΩ
8.2pF 0
50Ω 8.2pF
–5
5.6nH
1.5nH –10
1.5nH
–15
8.2pF
3.3pF
06147-003
0.6pF –20
S21 (dB)
–25
Figure 3. Basic LNA Implementation Schematic
–30 4.3nH
–35
–40 2.2pF
2.2nH 2.2nH
–45
–50
06147-005
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
FREQUENCY (MHz)
Rev. 0 | Page 3 of 8
AN-851
30 12
After the second LNA stage, the signal is downconverted to a
IIP3
fixed high IF of 374 MHz using the ADL5350 mixer. The 25 9
ADL5350 is a single-ended passive mixer with an onboard LO
20 6
buffer amplifier. The mixer relies on off-chip filtering networks to
06147-007
2200 2250 2300 2350 2400 2450 2500
constructed using lumped element components. Without the RF FREQUENCY (MHz)
filter the second harmonic of the LO would cause some addi-
Figure 6. Measured Performance of ADL5350 Mixer
tional high frequency noise power to be downconverted to the
desired IF, degrading the reference sensitivity of the receiver. During the downconversion, the phase noise of the PLL is
The measured dynamic performance of the ADL5350, includ- imposed onto each subcarrier of the OFDM modulated signal
ing the external LO buffer and filter networks, is presented in through convolution. To minimize the impact on the receiver
Figure 6. The mixer provides better than 24 dBm input IIP3 sensitivity level, the first agile PLL for the generation of LO1 is a
with ~8 dB conversion loss and single-sideband NF. fractional-N PLL. It is designed with the ADF4153 synthesizer.
The closed loop bandwidth is about 30 kHz and the estimated
phase jitter is 0.3 degrees.
2.7nH 3V 3V
VP VDD
100pF
0.47µF
2.2pF
100nF 3dB PAD VCO190-1960T
1nF 1nF 18Ω 18Ω 1.5kΩ 10MHz
CP REFIN XTAL
100pF 100pF REF
600Ω ADF4153
MUXOUT
18Ω CLK SPI
AD8353 100pF DATA CONTROL
820pF 40nF 2.7nF
EVALUATION BOARD RFINA LE BUS
51Ω
RFINB
RSET
100pF
06147-006
Rev. 0 | Page 4 of 8
AN-851
The first IF is then passed through a 374 MHz SAW filter for The first filtered IF signal is then downconverted to the second
channel selection. The excellent stop-band rejection of the IF of 70 MHz using the AD8344 active mixer. A low-side LO
Sawtek 855898 SAW filter allows the receiver to have injection scheme was used to ensure optimal spurious response
exceptional selectivity, improving immunity to adjacent rejection and to achieve greater gain in the mixer. In general,
interfering signals. The filter is matched to 50 Ω using the the AD8344 offers slightly higher conversion gain when using a
external LC component indicated in Figure 7. The filters low-side LO. The 304 MHz LO is provided by the ADF4360-8.
frequency response characteristic is presented in Figure 8. This is an integrated PLL + VCO providing good cost and
0
board space savings over discrete solutions. Two external
inductors that are the tank inductors of the on-chip VCO set
–10
the center frequency. The differential outputs of the ADF4360-8
–20
are combined in a balun. Using a 304 MHz LO the AD8344
–30 offers a conversion gain of ~11 dB, with 12 dBm IIP3 and 8 dB
RESPONSE (dB)
–40 SSB NF. The 70 MHz output is then passed through a fourth
–50 order low-pass filter to help reject LO feedthrough and higher
–60
frequency mixer spurii.
–70
–80
–90
–100
06147-008
AD8344 1nF
EVALUATION BOARD
5V
2.43kΩ
1nF
5V
12 11 10 9
240nH L2
1:4
RFOUTA 470Ω
RSET
100pF 4.7kΩ
RFOUTB
06147-009
Figure 9. The AD8344 Active Mixer Driven by the ADF4360-8 Through an External 1:4 Impedance Ratio Balun
Rev. 0 | Page 5 of 8
AN-851
The final 70 MHz IF signal is then passed through the AD8370 to raise the presented loading impedance at the desired IF
digitally controlled variable gain amplifier before being IF frequency. The filter network provides an impedance
sampled by the AD9246. The AD8370 provides a high output transformation from 100 Ω to 600 Ω and has the effect of
IIP3 and greater than 40 dB gain adjustment range. This allows stepping up the voltage by ~8 dB. The step-up transformation
the overall receiver conversion gain to adjust and tailor the needs to be accounted for when analyzing the receiver line-up.
cascaded input dynamic range to accommodate widely varying For more information regarding the design approach used to
input signal powers. yield the driver/ADC interface, please refer to AN-827, A
Resonant Approach to Interfacing Amplifiers to Switched-
The selected ADC provides excellent spurious free dynamic Capacitor ADCs. The key details of the AD8370 and AD9246
range out to greater than 200 MHz IF frequencies while only network interface are captured in Figure 11. The simulated filter
consuming ~250 mW. The high analog input bandwidth of response is plotted against the measured response through the
650 MHz allows the AD9246 to be applied to higher IF ADC in Figure 10.
frequencies. In this demonstration a 70 MHz IF was selected
0
using an 80 MSPS sampling clock, placing the IF in the second
Nyquist zone of the ADC. –5
–10
In order to prevent degradation of the ADC’s sensitivity level it MEASURED
was necessary to employ an anti-aliasing filter. The anti-aliasing –15
RESPONSE (dB)
filter helps to reject higher frequency spurious signals such as –20
LO leakage from degrading the perceived noise floor of the SIMULATED
–25
ADC. Additionally the anti-aliasing filter helps to reject
wideband noise generated by the driving amplifier stages that –30
would otherwise alias additional noise into the desired Nyquist –35
band. Elliptical low-pass architecture is selected to help provide
–40
better rejection of strong spurious components in the higher
Nyquist zones. A resonant parallel tank network is formed by –45
06147-011
20 40 60 80 100 120 140 160
the 72 nH bias inductors in combination with the total FREQUENCY (MHz)
presented input capacitance. The resonant tank network helps Figure 10. Simulated and Measured Response of the Anti-Aliasing Filter into
the AD9246 ADC
AD9246
VCOM
SERIAL CONTROL INTERFACE 39pF
1nF
1nF 1nF 47nH 56nH
16 15 14 13 12 11 10 9
1nF
1nF 1nF 47nH 56nH
06147-010
0.1µF 0.1µF
39pF
5V
Rev. 0 | Page 6 of 8
AN-851
–85 –5
–90 –10
06147-012
30 35 40 45 50 55 60 65
–15
SIGNAL GAIN (dB)
Figure 12. Second and Third Harmonic Distortion for the Full Receiver –20
EVM (dB)
0 7 –30
–35
–5 6 WITH ADF4153 AND
ADF4360-8 PLLs
–40
–10 5
–45
USING EXTERNAL GENERATORS FOR LO SIGNALS
SSB NF (dB)
–50
IIP3 (dBm)
–15 4
06147-014
–30 1
–35 0
06147-013
30 35 40 45 50 55 60 65 70
SIGNAL GAIN (dB)
Figure 13. Two-Tone IIP3 and Single-Sideband NF for the Full Receiver
Cascade Measured vs. Receiver Conversion Gain
Rev. 0 | Page 7 of 8
AN-851
In order to access the selectivity of the receiver, it is necessary to The typical spectral characteristics of the full receiver is
measure the relative response for single-tone inputs swept over depicted in Figure 16. By using the ADC Analyzer software, it
a reasonable input frequency range. Figure 15 presents the is easy to see the instantaneous spectrum and capture time
selectivity of the full receiver in the absence of the front-end domain data for later signal processing. Whereas the IF is at
band-select filter. The frequency selective nature of the double 70 MHz with an 80 MSPS sampling clock, the signal appears as
downconversion receiver provides very high immunity to a 10 MHz signal in the first Nyquist zone. The ADC Analyzer
nearby interferers. With the addition of the front-end band- software makes it possible to quickly analyze and debug
select filter, greater than 60 dB rejection can be achieved at spurious clock and LO signal components that can otherwise
adjacent RF frequency bands. compromise in-band receiver performance. The multiple IF
10
filter stages and LO filters helped to minimize spurious clutter
to negligible levels. Note the weak spur at 32 MHz. This is the
0
LO to IF leakage of the first mixer.
NORMALIZED RESPONSE (dB)
–10
A complete summary of the receiver’s performance is provided
–20 in Table 3. The design provides better than −25 dB EVM per-
formance for input signals greater than −74 dBm, with more
–30
than 60 dB image rejection and excellent adjacent and nonadja-
–40 cent channel rejection.
–50 Table 3. Performance Summary (measured with 64-QAM ¾
rate coding OFDM 10 MHz bandwidth input signal)
–60
Measured
–70 Parameter Performance
06147-015
–70
REFERENCES
–80 Newman, Eric and Reeder, Rob. 2006. “A Resonant Approach to
–90
Interfacing Amplifiers to Switched-Capacitor ADCs.”
–100
Application Note AN-827. Analog Devices, Inc. (January)
–110
IEEE Std 802.16-2004. IEEE Standard for Local and
–120 Metropolitan Area Networks - Part 16: Air Interface for Fixed
–130 Broadband Wireless Access Systems. Institute of Electrical and
06147-016
0 5 10 15 20 25 30 35 40
FREQUENCY (MHz) Electronics Engineers, Inc. (June).
Figure 16. FFT Plot of ADC Spectrum at Mid-Gain for a 64-QAM ¾ Rate
Coded, 10 MHz Bandwidth OFDM Input at 2.342 GHz (Note that the signal
appears as a 10 MHz IF but is actually a 70 MHz input in the second Nyquist
zone of the converter.)
Rev. 0 | Page 8 of 8