Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Deviation of dain current of PA between simulation and measurement

Status
Not open for further replies.

ueckid

Member level 1
Joined
Mar 22, 2017
Messages
39
Helped
2
Reputation
4
Reaction score
3
Trophy points
8
Location
Osaka, Japan
Activity points
411
Hi,
Could anyone tell me what is the cause of the significant difference in drain current between simulation and measuremet of a PA which I just designed. Does this difference mean a difference in output power between Sim. and Meas.?
Pls check the attached figure. Solid line is Sim. and dotted line is Meas.
Thanks.

Screenshot_20210120-222259_Box.jpg
 

We get accustomed to seeing the real circuit act 'less than ideal' compared to a simulation.

For one thing gain differences among your transistors can account for it. Or different bias thresholds. Etc.

Also real components have resistances we don't necessarily think about. Add a few parasitic resistances to the simulation. These can be located in the likely places.
 
  • Like
Reactions: ueckid

    ueckid

    Points: 2
    Helpful Answer Positive Rating
All I understand from those curves that the model is not sufficiently accurate.
Because the currents are not same even under Low Power Level (almost small signal) so I cannot blame Implemented/Practical Circuit.
 
  • Like
Reactions: ueckid

    ueckid

    Points: 2
    Helpful Answer Positive Rating
All I understand from those curves that the model is not sufficiently accurate.
Because the currents are not same even under Low Power Level (almost small signal) so I cannot blame Implemented/Practical Circuit.
I agree that the large-signal model of transistor may be less accurate. However, the small signal model looks good as shown in the attached figure.
What do you think about there any any other critical causes of this issue?
Screenshot_20210121-181835_Box.jpg
 

How are you measuring the drain current in simulation?

How are you measuring Pin in your lab measurements? Are you actually measuring the input power to your amplifier or is it the setting on your instrument? The reason I ask this is because your S11 is very poor.
 
With this poor S11, cannot trust any Ids vs Pin measurements.
However, designing a power amplifier working from 1GHz to 10GHz is a real challenge.
 
  • Like
Reactions: ueckid

    ueckid

    Points: 2
    Helpful Answer Positive Rating
The Small Signal Measurements tell me that this transistor can be used up to 3GHz and all Calibration and Measurements should be done up to 3GHz.
Going beyond this point will just increase the complexity.I see also some notches and unpredictable behavior in both Measurement and Simulation.
There might literally be some serious design errors.Or some other models are also inaccurate..
I cannot say something different beyond this point because I didn't see layout and circuit..
 
  • Like
Reactions: ueckid

    ueckid

    Points: 2
    Helpful Answer Positive Rating
How are you measuring the drain current in simulation?

How are you measuring Pin in your lab measurements? Are you actually measuring the input power to your amplifier or is it the setting on your instrument? The reason I ask this is because your S11 is very poor.
I measured Id with a current probe in the drain bias line in simulation.
Pin was measured with a power calibration using power meter not just reading from the the SG.
This PA has actually a poor design of S11, this is not usual in PA design where S11 should be good while S22 should be poor.
--- Updated ---

The Small Signal Measurements tell me that this transistor can be used up to 3GHz and all Calibration and Measurements should be done up to 3GHz.
Going beyond this point will just increase the complexity.I see also some notches and unpredictable behavior in both Measurement and Simulation.
There might literally be some serious design errors.Or some other models are also inaccurate..
I cannot say something different beyond this point because I didn't see layout and circuit..
Thanks for your comments.
The targe operation frequency of this circuit is 2.4 GHz.
Below are my layout of this PA as well as additional S12 and S22 results.
1.PNG
2.PNG

--- Updated ---

Hi all,
Please check my large-signal results of this PA as shown below. There is a signifficant difference in large-signal performance between Sim. and Meas.
3.PNG

--- Updated ---

Anyway, if the problem is due to the accuracy of the non-linear model of the transistor, how can we evaluate the non-linear model provided by a manufacturer? I used a Qorvo TGF2977-SM device for my PA design with a large-signal model from Modelithics.
 
Last edited:

I see that you are not using any SMD decoupling capacitors on the DC bias, and rely only on the decoupling provided by the radial stubs.
This might be an issue, and could be the reason you are not having good results not even on the simulation.
Guessing the dimensions of your DC bias radial stubs, I don't think they are effective below about 6GHz.
 
I see that you are not using any SMD decoupling capacitors on the DC bias, and rely only on the decoupling provided by the radial stubs.
This might be an issue, and could be the reason you are not having good results not even on the simulation.
Guessing the dimensions of your DC bias radial stubs, I don't think they are effective below about 6GHz.
Thanks, but as you can see the small-signal results look quite good. In fact, I already tried to connect a bypass capacitor at the point of radial stub but the results didn't get better. I doubt that the main cause is, as indicated by BigBoss, the accuracy of the large-signal model of the transistor.
 

1.PNG


-Biasing
λ/4λ/4​
Lines are too close to GND Plane.They will create stray capacitances.See RED circles on the picture
-Modelithics Models are quite accurate models, no need to get manufaturer's model.
-I believe you forgot place more Via underneath of the Transistor, there should be larger GND plane and more via.
-There is a weird jump around 1.75 GHz and it generally comes from poor decoupling capacitor or GND connection.
-Have you ever checked the Stability with a Spectrum Analyzer ?
-You can send me ADS project file with embedded Modelithics model to review in detail..
 
  • Like
Reactions: ueckid

    ueckid

    Points: 2
    Helpful Answer Positive Rating
-You can send me ADS project file with embedded Modelithics model to review in detail..
[/QUOTE]
Many thanks, please find the link for my zipped ADS project: **broken link removed**
Actually, I don't have much experiences in designing PA using QFN package devices like TGF2977 of Qorvo. I checked stability on a SA and my PA was stable w/ and w/o RF signal in a wide frequency range.
I appreciate if you could figure out my design issue.
--- Updated ---

Dear all,
One more thing I would like to mention in my PA design: there is also a difference in output DC characteristic between the transistor model and measurement as indicated in below figures. Right one is measurement.
dc.JPG
 
Last edited:

The schematic and layout that you followed from the datasheet of TGF2977 is for 9GHz-10GHz applications (attached picture 1).
You mentioned that want to use the PA for a 2.4GHz application, so have to follow somehow the schematic and layout recommendations for 2.6GHz-4.2GHz. (see attached picture 2).
As I said above, the radial stubs you have now on the layout, should not be used at 2.4GHz, because are pretty much useless.
 

Attachments

  • 1.jpg
    1.jpg
    54.5 KB · Views: 103
  • 2.jpg
    2.jpg
    53.4 KB · Views: 113

The schematic and layout that you followed from the datasheet of TGF2977 is for 9GHz-10GHz applications (attached picture 1).
You mentioned that want to use the PA for a 2.4GHz application, so have to follow somehow the schematic and layout recommendations for 2.6GHz-4.2GHz. (see attached picture 2).
In fact, I didn't design follow the application circuits in the datasheet. I made my own design based on my knowledge, indeed.
I think that they don't use a radial stub at 2.6-4.2GHz is due to its large size. In addition, at X band, it is relatively difficult to find an SMD commercial bypass capacitor.
I feel that the issue doesn't come from the radial stubs and the radial stub should work at S band. There should be some critical issue in my layout or fabricated PCB that I have not yet figured out so far.
 

I am confused. You want to design the PA for 2.4GHz or for X-band?
The issue is definitely related to bias decoupling, bias topology and its layout.
 
I am confused. You want to design the PA for 2.4GHz or for X-band?
The issue is definitely related to bias decoupling, bias topology and its layout.
Thank you.
My targer operation frequency is 2.4GHz. I just would like to say that they don't use a radial stub for application circuit at 9GHz-10GHz is because it is difficult to find an available bypass capacitor which has a self resonant freq at this freq range.
I wonder if my issue is the radial stub, how my circuit works well at small-signal performance!?
In simulation, I already checked input high impedance if the bias line which includes a quarter lamda line and a radial stub at one end.
 

they don't use a radial stub for application circuit at 9GHz-10GHz

Actually THEY USE radial stubs for 9GHz-10GHz application (see my pictures posted above from Qorvo datasheet)., and THEY DON'T USE radial stubs for 2.6GHz-4.2GHz application, because the stubs will be too big at lower frequencies.
At 2.6GHz they use inductors (chokes) and capacitors, instead of transmission lines and radial stubs. This is the path you have to follow if you want to design the PA for 2.4GHz.
 
vfone is right, Radial Stubs may be trouble at that frequency.You should replace them with appropriate Capacitors that can easily be found for this frequency.
I cannot use your model files because they are not compatible with my ADS version ( ADS 2021 Update1) however I will use manufacturer's model to check the design.At first sight, there are some grounding problem on the layout ( vias are not appropriate and very few) and possibly others..

I have checked the Transistor itself and this device is not very appropriate for 2.4GHz Power Amplifiers because it's indeed intended to use 8-10 GHz. band. That's why you have chosen wrong device.Instead, you have to find a "Optimized for 2.4 GHz Band Operation" device.There are many disrete devices for this band.
Read carefully "Model Validity" of this device in its Modelithics Datasheet.

Why don't you look TGF-2965-SM ? It looks quite promising for your application.Look at especially its PCB evaluation layout how it's been designed for 30MHz-3GHz.
 
Last edited:
Dear Vfone, Bigboss,

I appreciate your time and kind advice.
About radial stub: there are some IEEE papers use radial stub for PA at S band like showing in the attachment as an example.
In the datasheet of TGF2977SM it says this device can work from DC to 12 GHz. However, in the datasheet of its model provided by Modelithics as attached, there is no load-pull data below 6GHz.
In addition, as I indicated in the previous reply, there is a difference in DC characteristic between Sim and Meas of this model.
@BigBoss: Is the via issue critical to performance degradation?
I used TGF2977SM for multiple purposes due to its wide operation frequency, that's why I didn't notice TGF-2965-SM device.
I do doubt that the large-signal of this Modelithics model is not really accurate compared with the real device that I bought.
 

Attachments

  • HMT-QOR-TGF2977-SM-001_datasheet.pdf
    960.7 KB · Views: 112
Last edited by a moderator:

About radial stub: there are some IEEE papers use radial stub for PA at S band like showing in the attachment as an example.

I am not going to continue this discussion but I have to mention this: I read a statistic saying that, in the hystory of 110 years of IEEE, only one in a thousand (1‰) of the published articles found their way to be implemented in the industry.
The same statistic mentioned that only 10% of the published articles are read by other engineers than their author(s).
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top