[SOLVED] Load Pull technique of amplifiers - any exprience?

Load Pull Expriences

Is there anybody who has great exprience about load pull technique of the amplifiers in the area of either simulation or measurements..

I'm curious about the accuracy of the simulation engines such as ADS ,Ansoft or MW Office etc. Which one is better ? or this depends on some special conditions ( I mean under some special operating classes some simulators may give more accurate results..??)

If you share your exprience , you're great..

Good Luck

it depends on what information you want and what class of amplifier you want to load pull.

For efficiency, P1dB and gain of class-A, AB or even C amplifier, the nonlinear model in ADS should be quite accurate.

But if it is class-E amplifier, it is better to have experimental load pull.

Also, experimental load pull is a better way to determine linearity performance rather than simulation

Load Pull

Dear BB,

1- I am used to characterize all devices I use in PA designs using a pair of manual tuners manufactured by Focus.

2- I only had a reasonable agreement between measured and simulated load pull for devices (either bipolar or FETs) biased in Class A.

3- The conclusion I came in terms of device characterization for power amplifier design is to use a test fixture that mimics the attachment you are going to use in the real PA including the real bias circuit, with all bypass, etc...

4- Instead of to perform a full load pull measurement that is very demand, I adjust the tuners to get the highest output power with the best input return loss for the desired operating point. These are the optimum source and load impedances for power.
Around this point I take four additional points de-tuning the output tuner so that I get 0.2dB less power (respect to the maximum power) The Impedances corresponding to these 04 points belong to a -0.2dB isopower contour. This procedure is repeated for some points more in the passband of interest.

4- Based in the optimum impedance points I get the input and output networks to provide the right impedance transform.

5- So I plot on the Smith Chart the optimum impedances, the -0.2dB isopower contours and the response of the network. This way you have a perfect idea about how much the output power of the PA will change.
Note: 0.2dB is used because that is normally my target spec for power fluctuation over the passband.

6- The result I have been getting with this approach is 100%. The most of the times all considerations made at design time are there in the real circuit,so that I don't have to perform any kind of tuning.

Best regards,

NandoPG

In @DS, the sinulator will pull the load regrdaless the maximum current of the device, sometime the load line go beyond the maximum rating. I think the best way is to use the REAL Load pull.

Thanks a lot for your kind answers.

During my load pull simulations with ADS and Ansoft the results seem to be quite different. Under the same bias conditions , two simulators give
independent Gamma L zone. But in fact driving levels, source reflection coefficients , bias conditions and nonlinear models are same.

Can it be some convergence errors in the simulations? Because I have seen that Harmonic Balance Simulalation engine of Ansoft is much more rapid. :?: Personally I don't trust to it very much and does not seem to be very robust and consistent..

Otherwise, perhaps I do some misinterpretings in simulation setups.

Using Load Pull setup is not very easy because I design a PA with 0.15u SiGe technology giving 26dBm output power and it's very small to measure with microprobes.

Anway , I have understood weel that the best way to measure ..

Thanks a lot again..

I dont use load pull in my designs because i never find similar experiences in the devices i design. The amplifiers have in fact a P1dB of about 56dBm with two devices (each of which had 2 devices in one case) and according to me it is too difficult to do a load pull. I do tests scanning the amplifier even every 2 or 5MHz in the UHF band because it is so easy to find holes in the erogations. So i still use a very experimental method trying to match the output device at the maximum power (where i have the model of the LDMOS). But i would like to try the load pull.... Anyone of you had experiences about this???
thanks
stefano

It depends on your nonlinear model that should be supplied by the device manufacturer or similar.
Most nonlinear models till now fail to get you close what the actual device will do.

I prefer doing actual loadpull without simulation. I use simulators only for linear circuits and passive/transmissionline networks. Anything other than that, I never trust the simulation results, because the nonlinear model does not include all the effects happen physically.

As nandopg said above, use a test fixture that will look very close to what you gonna build, including decoupling networks, same dielectric constant board and thinkness...etc.

Never trust nonlinear models, and they also very rarely supplied by the device manufacturer.

Cheers...

Gemini1706 said:

It depends on your nonlinear model that should be supplied by the device manufacturer or similar.
Most nonlinear models till now fail to get you close what the actual device will do. ...

Click to expand...

I agree with Gemini1706. Getting halfway decent nonlinear models is the most difficult task in the simulations. You can compare the I-V-Curves of the active devices between the simulators. If you want to simulate LOAD-PULL you will have to find a good nonlinear model that suits your actual device. Also note that the actual device might also vary over production lot(Threshold-Voltage for example).

Also, how you setup the harmonic balance simulator (number of harmonics etc.) is important.