1. Power Amplifier Design using ADS
PA Workshop
Wilfredo Rivas-Torres Technical Support Application Engineer October 12, 2004

2. Outline Introduction DC and Loadline analysis Bias and Stability
LoadPull
Matching using Smith Chart Utility
SourcePull
PA Characterization – Did we meet the specification?
Optimize/Fine Tune the design
Test Design with real world modulated signals
Layout

3. Why do we need a Power Amplifier?
Power Amplifiers (PA) are in the transmitting chain of a wireless system. They are the final amplification stage before the signal is transmitted, and therefore must produce enough output power to overcome channel losses between the transmitter and the receiver.

4. PA requirements
The PA is typically the primary consumer of power in a transmitter. A major design requirement is how efficiently the PA can convert DC power to RF output power.
The design engineer has to often concern himself with the Efficiency of the Power Amplifier. Notice that efficiency translates into either lower operation cost (e.g. cellular basestation) or longer battery life (e.g. wireless handheld).
PA linearity is another important requirement, the input/ output relationship must be linear to preserve the signal integrity.
The design of PAs often involves the tradeoff of efficiency and linearity.

5. PA Design Requirement RF Output Power: 50 W PEP
Input Drive Level : 1 W
Output Load (RL): 50 Ω
Efficiency (η) > 50%
Bias Voltage: 28 V
Device: MRF9045M

6. DC Curves
MRF9045M_DC
This schematic can easily be created by using ADS built in templates as a starting point.
On a blank schematic select File>New Design.
Once you have named the design look for the Select Design Templates pull down menu field
Select “FET Curve Tracer”
Set drain and gate voltage sweep limits as needed.
The LDMOS transistor is from a Design Kit from the vendor. Notice the Tech_Include statement.

7. DC Curves
By examining the Load Line we can see that we will be operating in Class AB.

8. Bias and Stability
MRF9045M_LSSP_Stab – push into MRF9045M_AMP subcircuit
A quarter wavelength transmission line was added to the drain for RF isolation and an inductor to the gate as an RF choke.
Capacitors are either bypass or DC filtering.
The Lumped components are from the ADS built in library. We want to use realistic parts as much as possible.
Added series and shunt resistors for stability.
Note that our amplifier has been made into a subcircuit.

9. Stability Analysis
Notice that we added the Measurement blocks that are available straight from palettes for this analysis.

10. Stability Analysis This chart shows the Rollett Stability factor (K).
For conditional stability K>1 is required.

11. Impedance Matching
The need for matching circuits is because amplifiers, in order to perform in a certain way(e.g. maximize output power), must be presented with a certain impedance at both the load and the source ports.
For example in order to deliver maximum power to the load RL the transistor must have termination Zs and ZL.
The input matching network is designed to transform the generator impedance Rs to the optimum source impedance Zs.
The output matching network transform the load termination RL (50 Ω) to the optimum load impedance ZL.
A LoadPull measurement will help the designer determine the optimum load impedance ZL.

12. LoadPull Setup Examples Search MRF9045M_LoadPull
There is no need to create this schematic from scratch, we will copy it from a LoadPull example (or you can use the Amplifier DG).
We use the Examples Search tool by clicking on the icon on the main window.
In the query field type “load pull” and click on the Search Now button.
The schematic above was create using the “HB1Tone_LoadPull.dsn” schematic in the LoadPull_prj project.
Once you copy the design to your project add you design and setup the different parameters (e.g frequency, power, …).
You can also copy the data display file.
Examples
Search

13. LoadPull Contours
The data display was also copied over from the same example.
Note that from marker m2 that Pout optimum is 44.2 dBm.
This is not the required 47 dBm goal. Is it time for a redesign?
The answer is no. We still need to do at least one Source Pull analysis.

14. Matching using Smith Chart Utility
output_match_design.dsn
Before we do any Source Pull lets match the output with the optimum load.
We will use the Smith Chart Matching Utility.
The first step is to place a Smith Chart Matching Smart Component. This will hold the matching circuit we design.
You can find the Smith Chart Utility under the Filter, Amplifier or Mixer Design Guides. In the Amplifier and Mixer DG look under tools.

15. Matching using Smith Chart Utility
In the Circuit Data section you can look at all s-parameters. In the above chart we actually look at the magnitude of S21.
Once you are satisfied press the Build ADS Schematic button.

16. Output Match
ADS has created the matching circuit in a subcircuit inside the Smart Component that you placed.

17. SourcePull Contours MRF9045M_SourcePull_wOutputMatch
Again we used a schematic and data display from another ADS example.
The output power is 46 dBm which is still short by 1 dB off the requirement.
The PAE is 58.1%.

18. Matching Circuits Output Match Input Match
Despite not meeting the requirements we will continue to design the matching circuits using realistic parts, we will then use optimization to “fine tune” the circuit.
We have replaced the ideal capacitors for more realistic vendor part.
The Ideal TLines have been replaced with Microstrip Lines.

19. Complete Design Power Sweep
MRF9045M_HB2
Notice that we have added a Power Sweep (Pin) to the HB simulation.
The power sweep will allow us to see the Power Compression characteristic of the Amplifier using the Input/Output Match.

20. Power Compression Curve

21. Gain Compression Curve
Note that at the nominal input power the Gain Compression is –1.77 dB. We aim to improve this as much as we can.

22. Power Added Efficiency

23. Getting ready to Optimize the PA
The next step is to optimize the design to meet the requirements.
The Designer can take the opportunity to see if other requirements, such as layout will require any changes before proceeding to optimize.
Example: we notice that the transistor pads are rather wide and the Tlines leading up to it are not the same width.
Since the Tlines are much narrower, we could add a taper so we have a nice transition.
We included a MTAPER at the input and output side
However note that the output power has decreased. This is because the Taper was not included in out matching design.

24. Optimization Setup MRF9045M_HB4
You will need an Optimization controller. At a minimum you need to select an Optimizer type and the number of iteration. In this particular design the Gradient optimizer is the most efficient based on experience dealing with the design.
The fact that we start with some gain is a plus. If the output power were much lower to the point there is no gain it will make the problem much harder to solve. Bottom line, the closer (meaning a good initial starting point) to the requirements the better chance of a successful optimization.
In Goal1 we are requiring that the Amplifier output power be greater than 47 dBm (50W)
Goal2 was included so that the harmonic products do not get out of control.
We could add more requirements as needed but make sure these are realistic.

25. Optimization Results After Before Touché!
Remember to update the design with the optimization values. Select: Simulation>Update Optimization Values
Before
After

26. Optimization Values
If you observe the status window, it will show the results of the optimizable values during the simulation.
In ADS the length are all converted to meters during simulation.
You will need to convert to mils in the data display.

27. PA Results
The results speak for themselves.

28. Power Added Efficiency

29. Gain Compression Curve
The original ideal optimum match had produce less power but also had a dB gain compression for the nominal input power.
The final optimized design only has dB of compression at the nominal input power. We have a more linear amplifier.
Is this the results of forcing the optimizer to a maximum 2nd order harmonic (Goal2). YES!

30. Complex Modulated Signal
The next step is to simulate a Modulated Carrier QAM 16

31. 16 QAM Modulated Source
The signal on the prior slide was created using the DSP capabilities in ADS with Agilent Ptolemy.
ADS can co-simulate DSP with Analog Circuits (Mixed Signal Analysis).

32. DSP and Analog Circuits Setup
Create a subcircuit with your analog design.
You need to add either Circuit Envelope or Transient controller to the analog circuit.
We use Circuit Envelope specifically with our PA since we have a Modulated Carrier.
Circuit Envelope will also allow the use of Fast Cosim (Automatic Verification Modeling – AVM). This will dramatically increase the simulation speed.
In the DSP schematic we will create the Modulated Carrier, feed it to the PA and collect the signal samples and spectrum.
The DSP schematic contains a Envelope Output Selector component used for interfacing between circuit subnetwork output and the signal processing components.

33. Ptolemy Cosim SchematicPA
QAM Source
MRF9045M_CE3

34. Fast Cosim Improvements
Simulation Time Benchmark:
Total bits: 1024 bits
AVM disabled : 410 sec
AVM enabled: 13 sec
AVM data reuse: 5.5 sec
AVM data reuse (16 Kbits): 17.5 sec
In the PA design the Circuit Envelope controller needs to be set for Fast Cosim for improved simulation speed.
Once you double click on the controller go to the Cosim tab and select the Enable AVM button.

35. Cosimulation Results - Spectrum
Carrier Power
25 dBm
Carrier Power
30 dBm
Blue – Input
Red - Output

36. Cosimulation Results - Constellation
Carrier Power
25 dBm
Carrier Power
30 dBm

37. Cosimulation Results Carrier Power 30 dBm Carrier Power 25 dBm
Equations used:
peakP_out = peak_pwr(RF_out, , ,"dBm")
peak_avg_out = peak_to_avg_pwr(RF_out, , "dB")
avgPout = total_pwr(RF_out,,"dBm")

38. Cosimulation Results – CCDF
Carrier Power
25 dBm
Carrier Power
30 dBm
Blue – Output
Red - Input
CCDF - Complementary Cumulative Distribution Function
The CCDF curve shows the probability that the instantaneous signal power will be higher than the average signal power by a certain amount of dB.
The independent axis of the CCDF curve shows power levels in dB with respect to the signal average power level (0 dB corresponds to the signal average power level).
The dependent axis of the CCDF curve shows the probability that the instantaneous signal power will exceed the corresponding power level on the independent axis.

39. EVM vs. Power Measurement
MRF9045M_CE4
We can take advantage of the Parameter Sweeping capabilities in ADS to produce an EVM vs. Power analysis.

40. EVM vs. Power Results

41. ADS to VSA link
MRF9045M_CE5
Using Agilent Ptolemy we can also access the VSA series software (a.k.a. Glacier)
The VSA Sink models provide a stream interface that lets you input digitized waveforms directly from ADS to series VSA software without using the series hardware.
The full functionality of the analyzer is available to analyze and display the ADS signal.

42. VSA Spectrum from ADS Cosim

43. VSA Constellation from ADS Cosim
Carrier Power
30 dBm
This constellation does not match the results previously shown.
The reason is that the VSA applies a Root Raised Cosine filter to the demodulated data. In the ADS simulation we did not add this filter.
The VSA setup will need to match that of the ADS Cosimulation if one wishes the results to match.

44. VSA Constellation from ADS Cosim
Carrier Power
30 dBm
These are the Glacier results by turning off the measurement filter.
These now agree much closer with the simulation results.

45. VSA EVM from ADS Cosim
Carrier Power
30 dBm

46. PA Layout ADS can generate the layout from the schematic.
Prepare you schematic for it to auto-generate the layout.
Make sure all component either have a related artwork or set the artwork property to “null”.
MRF9045M_AMP_Lay
You could also delete those component without an artwork before auto-generating the layout.
We chose to delete the Voltage Sources and the Grounds.
There is also the possibility to place each component manually.

47. PA Layout – Generated from Schematic
To auto-generated the layout select Layout>Generate/Update from the schematic window menu.

48. PA Layout – Ground Fill MRF9045M_AMP_Lay2
In order to fill in the ground first draw the entire ground shape, we chose a rectangle.
Select Edit>Create Clearance and follow the dialogue.

49. Other Possibilities
Run an EM Cosimulation – include layout effects in the simulation. Optimize design if necessary.
Run Loadpull for IP3 or ACPR. The optimum load would then be a compromise between all the requirements.
Use Connection Manager and real PA to validate design and compare vs. simulated results.
Create a behavioral data model that can be used to protect you IP yet give access to your design results.
If there is any topic about PA Design and ADS you wish to
discuss me: