RF System Architecture and Budget Analysis Anurag Bhargava Application Consultant Agilent EEsof EDA Agilent Technologies Email: [email protected]

You Tube: www.youtube.com/user/BhargavaAnurag Blog: http://abhargava.wordpress.com

RF Architecture Design During RF architecture design engineers determine how many, what type, parameters, and the order of stages required to meet system specifications

3

1

?

2

Coupler IL=2 dB CPL=20 dB DIR=30 dB Z0=50 ohm

or

Spectral Domain Techniques Page 2

The Architecture is Critical • Architecture is the system foundation

• Poor architecture causes poor performance • Poor architecture causes more design turns

• Poor architecture causes increased cost • Poor architecture causes increased time to market

Spectral Domain Techniques Page 3

Traditional Method Limitations • SPICE, linear, and harmonic balance simulators are designed for components, not for complex system realization • Programming spreadsheets is extremely difficult when image noise, intermod filtering, mismatch effects and multiple paths are considered • Summary: traditional tools are ill-suited for analyzing RF architectures

Spectral Domain Techniques Page 4

What is Required for RF Architecture Analysis? Spur identification and resolution • Identifying root causes is critical in systems with hundreds of leakage, harmonic and intermod tones.

Noise analysis that considers the channel bandwidth

In-channel and out-of-channel intermodulation analysis Accuracy: mismatch, phase, images, measured vs. ideal models, etc.

Spectral Domain Techniques Page 5

Typical RF System LNA

Attenuator

Mixer

IF Filter

IF Amp

Antenna RX Filter

Demodulator Buffer Amp

Synthesizer or PLL

Switch or Duplexer

Buffer Amp

TX Filter

TX Power Amplifer

Mixer

IF Filter

Modulator

Copyright © 2012 Agilent Technologies 6

Typical RF System

How do you determine what types of stages to use, the order to place them, and the parameters for each block?

Copyright © 2012 Agilent Technologies 7

How would you respond to the following questions? 1. How do you currently determine what and where spurious and intermod products are being generated in the RF portions of you design? 2. How do you identify possible leakage paths in your design? 3. Does your current method of designing wideband systems include the effects of frequency dependencies of complex load impedance? 4. How do you access the impact of phase noise on noise figure?

5. How do you do budget analysis in the presence of modulated signals? 6. How do you compute noise figure of a path while other paths are actively connected to this path?

Copyright © 2012 Agilent Technologies 8

Why was Spectrasys Developed? • RF Architecture (or Systems design) are the titles used during the design stage where engineers determine what types of stages (filters, amplifiers, mixers, etc), the ordering of these stages in the design, and their parameters. • Cascaded equations are used during this phase. G1 nf1 OP1dB1 OIP31

G2 nf2 OP1dB2 OIP32

G3 nf3 OP1dB3 OIP33

𝐺 = 𝐺1 × 𝐺2 × 𝐺3 𝑛𝑓2 − 1 𝑛𝑓3 − 1 𝑛𝑓𝑡𝑜𝑡𝑎𝑙 = 𝑛𝑓1 + + 𝐺1 𝐺1 × 𝐺2 1 1 1 1 = + + 𝑂𝐼𝑃3𝑡𝑜𝑡𝑎𝑙 𝑂𝐼𝑃31 × 𝐺2 × 𝐺3 𝑂𝐼𝑃32 × 𝐺3 𝑂𝐼𝑃33 1 1 1 1 = + + 𝑂𝑃1𝑑𝐵𝑡𝑜𝑡𝑎𝑙 𝑂𝑃1𝑑𝐵1 × 𝐺2 × 𝐺3 𝑂𝑃1𝑑𝐵2 × 𝐺3 𝑂𝑃1𝑑𝐵3 Copyright © 2012 Agilent Technologies 9

Why was Spectrasys Developed?: Continued

How do you use cascaded equations on this?

Copyright © 2012 Agilent Technologies 10

Why was Spectrasys Developed?: Continued

It would be nice if every design was two port AND just contained behavioral models. But its not! Copyright © 2012 Agilent Technologies 11

SPARCA These needs led to the development of the SPARCA method, an acronym for Signal Propagation And Root Cause Analysis. • • • • •

New method developed in 2002 Based in the frequency domain Bilateral signal flow of all paths Magnitude and phase modeling (vector) Channel concepts integrated into the method

Spectral Domain Techniques Page 12

How SPARCA Works Signal generation and propagation: • All signal sources and noise sources propagate through all paths to all nodes • Harmonic and intermod products are generated by non-linear models and propagate to all nodes

Measurements: • Power and noise are each is integrated across the channel bandwidth • In-channel and adjacent channel measurements are calculated from the spectrums at each node

Imperfections modeled: • Devices are bilateral • The true filtered and non-flat transfer function including phase is modeled • Mismatch, reflections, and reverse propagation are accurately modeled

Spectral Domain Techniques Page 13

SPARCA Method SPARCA does not replace other analysis methods (except perhaps spreadsheets), it complements them. It is optimized for RF architecture design. Advantages

• Identifies root causes • Handles multiple paths • More accurate than ideal model and scalar methods • Handles more signals & tones than other methods • Handles channel concepts • Integrates well with other software and lab tools

Spectral Domain Techniques Page 14

Example: Aviation Band Downconverter RF BPF FLO=105 MHz FHI=139 MHz N=4 R=0.1 dB

(1)

1

5

RF Mixer CL=8 dB LO=7 dBm 6

R

L

I

IFAMP1 G=9 dB NF=3 dB OP1DB=25 dBm

16

IFAMP2 G=20 dB NF=3 dB OP1DB=30 dBm

9

10

13

RF: 108 to 136 MHz RFAMP G=?9 dB NF=2 dB OP1DB=15 dBm

Detector CL=8 dB LO=23 dBm 17

R

L

I

3

LO ATTN L=6 dB

IF BPF FLO=21.45 MHz FHI=22.45 MHz IL=2 dB

(3) BFO: 20.75 MHz

BASEBAND: 0.95 to 1.45 MHz 7

4

(4)

BASEBAND AMP G=46 dB NF=6 dB OP1DB=36 dBm

2

(2) LO: 86.3 to 114.3 MHz

The analyzer uses a tunable 1st LO to convert 500 KHz segments of the aviation band to 0.95 to 1.45 MHz for baseband processing.

Spectral Domain Techniques Page 15

Simple Cascade Noise Figure

In this equation block, noise figure cascade formula have been entered much like could be programmed into a spreadsheet. This is being done to illustrate the errors introduced by spreadsheet system analysis.

Spectral Domain Techniques Page 16

Results of the Cascade Calculation RF BPF FLO=105 MHz FHI=139 MHz N=4 R=0.1 dB

(1)

1

5

RF Mixer CL=8 dB LO=7 dBm 6

R

L

I

IFAMP1 G=9 dB NF=3 dB OP1DB=25 dBm

16

IFAMP2 G=20 dB NF=3 dB OP1DB=30 dBm

9

10

13

RF: 108 to 136 MHz RFAMP G=?9 dB NF=2 dB OP1DB=15 dBm

Detector CL=8 dB LO=23 dBm 17

R

L

I

3

LO ATTN L=6 dB

IF BPF FLO=21.45 MHz FHI=22.45 MHz IL=2 dB

(3) BFO: 20.75 MHz

BASEBAND: 0.95 to 1.45 MHz 7

4

(4)

BASEBAND AMP G=46 dB NF=6 dB OP1DB=36 dBm

2

(2) LO: 86.3 to 114.3 MHz

Here the block diagram of the downconverter is repeated along with the resulting cascade noise figure after each stage in the downconverter. The resulting noise figure at the baseband output is 5.28 dB.

Spectral Domain Techniques Page 17

Results by Spectral Propagation

Shown at the left is a level diagram in SPECTRASYS. This level diagram depicts the cascade noise figure in red, the stage noise figure in blue and the percentage noise figure in green. Numeric cascade noise figure data after each stage is also given in the final column in the table at the lower right. Notice that the noise figure after the final amplifier computed using SPARCA is 7.83 dB, rather than the 5.28 dB computed using classical cascade NF equations. The popular and classical formula is in error by 2.55 dB! Spectral Domain Techniques Page 18

Source of Noise Figure Error What is the source of this error? • Mixer noise figure assumes noise only exists in the desired signal channel. In reality, noise exists in both the desired and image channel. Therefore, the true noise figure of a mixer is 3.01 dB higher than the specified noise figure.

Why not just specify the mixer noise figure 3 dB higher than is normally done? • Because in a system with gain and an image filter preceding the mixer, the effective noise figure of the mixer approaches the specified value. In that case, using the higher noise figure would give the wrong result. The fact of the matter is, accurate assessment of system noise figure requires noise propagation analysis. If the position of the input filter and the preamp are changed and the loss of the input filter was set to zero. Simple cascade analysis would predict the same noise figure for both systems. In reality, the difference with a 30 dB preamp with 8 dB noise figure is 2.7 dB. Approaching 3 dB. But look what happens when the noise figure of the preamp is reduced to 2 dB. The difference is then 1.4 dB. Again, accurate noise analysis requires noise propagation techniques.

In a similar way, spreadsheet analysis of intermodulation and dynamic range are also flawed. Spectral Domain Techniques Page 19

Measurements Different applications require different forms of output data. In the next few slides, brief definitions and examples of SPECTRASYS output data are provided.

There are over 85 built-in measurements.

Spectral Domain Techniques Page 20

Measurements: Gain & Signals in the Channel Cascaded Gain

Cascaded Gain (All Signals) Channel Frequency

Channel Power Channel Power (Desired) Gain Gain (All Signals) Total Node Power

Spectral Domain Techniques Page 21

Example: Gain & Signal Level Diagram Gain & Signal 50

130

DB[GAIN], DB[CGAIN]

30

100

70

20

40

10

10

0

CF (MHz), DBM[CP], DBM[DCP]

DB[GAIN] DB[CGAIN] CF (MHz) DBM[CP] DBM[DCP]

40

-20

-10

-50 1

5

6

R

L

I

16

9

10

17

R

L

I

7

8

4

Node

Spectral Domain Techniques Page 22

Measurements: Noise Added Noise Carrier to Noise Ratio Cascaded NF Channel Noise Power Image Channel Noise Power

Image Noise Rejection Ratio Minimum Detectable Signal Percent Noise Figure Stage NF Spectral Domain Techniques Page 23

Example: Noise

Spectral Domain Techniques Page 24

Short Video

Removed due to size of the Video……

Intermodulation Cascaded intermod equations are NOT used by SPECTRASYS. Just as with noise, cascade intermod equations are flawed because they: • Assume interfering signals are not filtered and maintain the same gain as the desired signal through all cascaded stages • Assume all stages are perfectly matched • Assume two equal tones • Assume infinite reverse isolation

Signal propagation generates intermod signals as they appear in real systems. Signals are not limited to two tones. All intermods are propagated through the system using the system transfer function. Consequently, intermod measurements are accurate for filtered and non-ideal systems.

Spectral Domain Techniques Page 26

Measurements: Intermodulation & Compression Cascaded Gain (3rd IM)

Stage Output 2nd IM

Channel Frequency (Tone)

Stage Output 3rd IM

Channel Power (Desired 3rd IM)

Stage Output Saturation Power

Channel Power (Tone)

3rd Order Intercept (Input)

Gain (3rd IM)

3rd Order Intercept (Output)

Gain (All Signals)

3rd IM Power (Propagated)

Percent 3rd IM

3rd IM Power (Generated)

Stage Output 1 dB Compression

3rd IM Power (Total)

Page 27

Spectral Domain Techniques

Example: Intermodulation & Compression

30

200

24

160

18

120

12

80

6

40

0

0

-6

DBM[IIP3] DBM[GIM3P] DB[SDR] PRIM3

-12 -18

-40 -80

DBM[GIM3P], DB[SDR]

DBM[IIP3], PRIM3

Intermodulation & Compression

-120

-24

-160

-30

-200 1

5

6

R

L

I

16

9

10

17

R

L

I

7

8

4

Node

Spectral Domain Techniques Page 28

Measurements: Dynamic Range Spurious Free Dynamic Range

Stage Dynamic Range

Spectral Domain Techniques Page 29

Example: Dynamic Range 120

-120

90

-126 DB[SFDR] DBM[IIP3] DBM[MDS]

60

-132

30

-138

0

-144

-30

DBM[MDS]

DB[SFDR], DBM[IIP3]

Dynamic Range

-150 1

5

6

R

L

I

16

9

10

17

R

L

I

7

8

4

Node

Spectral Domain Techniques Page 30

Measurements: Out of Channel Adjacent Channel Power

Adjacent Channel Frequency Channel Frequency (Offset) Channel Power (Offset) Image Frequency Image Channel Noise Power Image Noise Rejection Ratio Mixer Image Channel Power

Mixer Image Rejection Ratio Spectral Domain Techniques Page 31

Phase & Complex Models The SPECTRASYS SPARCA simulator propagates signals using complex multi-port transfer functions for the system. • Phase data is retained and utilized • Circuit models and schematics may be mixed with system models in the schematic • The effects of mismatch are simulated • Bilateral signal flow is simulated COUPLER1_1 IL=1 dB CPL=10 dB [DIR=30 dB]

(1)

1

COUPLER1_2 IL=0.5 dB CPL=24 dB [DIR=30 dB]

RF_PHASE_1 A=?10° [Z0=50 Ω]

4

5

11

6

15

COUPLER1_3 IL=0.5 dB CPL=10 dB DIR=30 dB

RF_DELAY_2 T=0.53 ns [Z0=50 Ω]

8

8

10

10

7

ATTN_VAR_1 IL=1 dB L=?0.25 dB

RFAMP_1 G=17 dB NF=8 dB OPSAT=50 dBm

14

RF_DELAY_1 T=0.27 ns [Z0=50 Ω]

(2)

3

RF_PHASE_2 A=?10° [Z0=50 Ω] 7 9

15

2

13

12

3

14

SPLIT2_1 [IL=3.1 dB] ISO=100 dB PH3=0°

ATTN_VAR_2 IL=1 dB L=?0.5 dB

RFAMP_2 G=38 dB NF=8 dB OPSAT=43 dBm

Spectral Domain Techniques Page 32

X-Parameter Models are Supported

Spectral Domain Techniques Page 33

The Link to ADS

Schematics can be exported to ADS for further simulation. Spectral Domain Techniques Page 34

Digital / DSP & RF System Cosimulation

Modulation and Spectrasys (SystemVue)

The RF Link part is used in a data flow schematic to point to a Spectrasys schematic.

Copyright © 2012 Agilent Technologies 36

How RF Impairments affect Modulation Quality

Tunable Impairment List: Gain Imbalance Phase Imbalance LO to RF Isolation Noise Figure Phase Noise ADC (# of bits, clock, ref voltage)

Copyright © 2012 Agilent Technologies 37

Gain Imbalance effects on Constellation

Blue – Balanced Gain (Square) Red – Imbalanced Gain

Copyright © 2012 Agilent Technologies 38

Phase Imbalance effects on Constellation

Blue – Balanced Phase (Square) Red – Imbalanced Phase

Copyright © 2012 Agilent Technologies 39

LO to RF Isolation (DC Offset) effects on Constellation

Blue – High Isolation (Square) Red – Lower Isolation

The LO leaks into the RF and appears as another RF input signal

Copyright © 2012 Agilent Technologies 40

LO to RF Isolation (DC Offset) effects on Constellation

Blue – High Isolation (Square) Red – Lower Isolation

The RF can also leak into the LO and this is used as another LO signal

Copyright © 2012 Agilent Technologies 41

Noise effects on Constellation

Blue – No Noise (Square) Red – Thermal and Phase Noise

Copyright © 2012 Agilent Technologies 42

Modulation Analysis Software (works with Hardware)

Copyright © 2012 Agilent Technologies 43

Summary A new system simulation technique referred to as Spectral Propagation and Root Cause Analysis (SPARCA) offers a new set of tools for the optimization of system architectures • Improved accuracy over conventional cascade analysis

• Rapid root cause discovery using spectral component display at any node • Wide set of built-in as well as user specified measurements • Complex, bilateral and mismatch analysis

• Integration of other spectral domain simulation and synthesis tools

SPARCA is optimized for system architecture development and saves multiple architecture and component design turns. Spectral Domain Techniques Page 44

Looking Ahead..!! Kindly send your requests on what we can cover in our future webinars to: India: Mukul Pareek Marketing Engineer Agilent Technologies India Pvt. Ltd Email: [email protected]

Elsewhere: Please contact your local Agilent representative.

Resources to help you..…. • Subscribe to the Blog: http://abhargava.wordpress.com to receive latest news, tips & tricks on RF design and industry news.. • Subscribe to the You Tube Channel: www.youtube.com/user/BhargavaAnurag to see various tutorial videos and learn the concepts better • Agilent EEsof Home Page: www.eesof.com • For software evaluation license or for any other queries:

– E-mail: [email protected] – Toll Free Number: 1800-11-2929

Q&A