Bild your own high power VNA front endusing directional couplers in order not to kill the VNA those directional couplers would be between the high power load and the outpit of youramplifier.

Why do you want to measure S22? A high power transistor load is not matched to the transistor small signal parameter.

I want to make sure I am injecting as much power as possible to 50 ohm matched antenna. I understand that there might be load pull, but still I am expecting that better matching to load will allow better signal injection in antenna

Output S22 for a power matched amplifier is very far from 50 Ohm as the output match is designed to ensure maximum current and voltage swing (in other words maximize the output power). The amplifier S22 is not of any value. Rather use a directional coupler and monitor the forward traveling wave going to your load.

There are some vendors that provide hot S22 measurement procedures (search on this term will provide more information), but I have not seen a reference to this for some time, so I suspect that they have realized that this might not be that valuable.

I fully agree with Cornell.

It is always a good idea to follow the power amplifier with an isolator. Then you can have a good output match towards the load so that will reflections from the antenna back to the amplifier will be absorbed.

I was using load pull contours for large signal parameters

Thanks Boris, but this is only in simulation right? My LDMOS transistor doesn't provide models, in this case I prefer practical impedance matching

If you don't have your own load pull capability, and your budget allows it - you might want to use an isolator (as Pieter Willem Van der Walt suggested)

Thanks, but I still believe that I need to get as best as possible reflection, because partial power will be wasted by heating RF isolator

Hello Audrius. An amplifier with a low S22 reflection is a small signal amplifier as used in receivers. This type of amplifier will not provide you with the optimum output power that the device is capable of as either the output current or the voltage waves will saturate prematurely. S22 is per definition is small signal concept and the impedance is only valid for small signal excitations. I suggest the books on power amplifiers from Cripps for a good description.

A high-power amplifier has a high reflection if you look back at it, but it will produce the largest amplitude current and voltage waves before saturation (hence the highest possible power). A high power amplifier assumes a load (antenna) with a low reflection (good 50 Ohm input). If the reflection from the load is low, then the high reflection from the output of the amplifier is not an issue. However, if your load is not so good (wideband antenna), then you can get that the amplifier is being de-tuned (resulting in less power from the amplifier).

There is only 2 things that you can do about this. Use a well matched load, or use a circulator to intercept reflections from the antenna and dumps them in an isolation load. The circulator is in effect a buffer that makes both the amplifier and antenna look like angels (at the price of bandwidth, cost and some additional power loss).

And regardless of what you implement, S22 is still not going to help you (sorry for the bad news).

Is your problem the design of a power amplifier, or the use of an existing amplifier in a test setup?

If your application does not require a wideband amplifier, then the one other option is to construct a balanced amplifier. Two identical amplifiers are placed between two 90-hybrids that split the input signal equally between the amplifiers (but 90 degree out of phase) and on the output combines the two amplifier outputs again. The added benefit of the 90 degree hybrids is that the output match of the amplifier will be good (un-used/isolated ports of hybrids are 50Ohm terminated).

And these are easy to design on PCB and will cost a fraction of a circulator. And you get close to double the power you would get from a single amplifier.

Dear Cornel,

Thank you for detailed answer.

Yes, we are talking about new design, but I am just curious, because my main knowledges are with lower power wideband amplifiers

Yes, I am familiar with balanced amplifier using 90deg hybrids, but is there any possibilities to have 180deg phase shift for high power? In case for reducing second harmonics and improve OIP2? Balanced amplifier do not provide second order intermodulation reduction, which in low power baluns does.

And one more question. In my lab I have FLH-20B power RF amplifier that is capable to work 20MHz-1GHz. Could you advice any possible amplifier topologies for 20W output and such widebnd operation? 90deg hybrids often starts at 700MHz at least from Xynger.

If you are designing an amplifier and do not have load-pull data, then I suggest the approach outlined in Cripps book.

This requires a equivalent circuit model that you can extract from measured S-parameters at the amplifier Q-point (bias point) along with cold S-parameters (VDS = 0). Building a TRL or TRM (my preference as it allow low frequency measurements) test fixture is relative easy and using cheap connectors like SMA will still get you up to 10 GHz with good calibration quality (well beyond most LDMOS devices max frequency). Just take care with larger LDMOS devices as their port impedances gets low and when presented with 50 Ohm at their ports they can oscillate. That can be terminal for a VNA (most test equipment in fact). But this can be overcome by designing the test fixture for a lower impedance and in effect measuring the S-parameters in a non-50 Ohm environment. For devices with less than 10W levels you should be able to get away with 50 Ohm test fixtures. And there are lots of publications on how to extract circuit models for LDMOS type FET devices. A direct extraction approach is easy to implement and good enough for this type of design.

If you have suitable load-pull data, then use that. Its equivalent. Just measured impedance levels that when presented to the FET will give you the optimum power match.

The answer to your question is yes, 180 phase shifts are possible. This results in a push-pull amplifier. But I suggest you ask Prof PW van der Walt (he has responded to your original question) for more information on the passive circuit options as he is a grand-master in this regard. I know there are LC equivalents for when frequencies gets to low for transmission lines but have no hands on experience with these.

Hello, to get back to your original question on VNA measurements of a PA. What can be done with a lot of success is to use a one path S11/S21 calibration. Not enough to characterize a transistor for design, but useful when dealing with a PA module. The trick is to define your attenuator as your through standard, which can be done on most VNA's as they allow a Touchstone S-parameter file to be imported to describe the networks used as calibration standards. Yes, this decreases the VNA dynamic range, but for a PA you normally have enough gain so that this loss of range on S21 is not a problem. And you can still measure the S11 with the normal dynamic range.

If your VNA has a sync input at the back that allow you control the sweep with an external TTL signal, then you can synchronize the RF measurements with other measurements such as DC current consumption and get a lot of data in few minutes. Just requires some programming.

Thanks Cornell,

Do you have any advice on practical stability check. Standard power LDMOS transistors for frequencies above 1GHz has capacitors at the gate and drain. Increasing those capacitance can lead to positive feedback and decrease of stability. Could you recommend any techniques (without modelling) to evaluate power amplifier stability? Normally I am doing measurements with high power attenuators, but later any kind of antenna can be connected.

Isolators might help, but is there any techniques for evaluation?

Stability problems normally occur at low frequencies where the device has more gain. So having more issues when you increase the size of the bias decoupling capacitors is understandable. You need to have a good understanding of the port impedances presented to your device by the bias-T networks you use in your test setup at low frequencies.

Previously when I had to do this, I placed my device in my test fixture with bias-T's and terminated the RF ports of the bias-Ts with 50 Ohm loads. The device was then powered up to the bias points required and I checked for oscillation with a spectrum analyzer using a sniffer antenna (I suggest a simple loop antenna).

Also use a power supply with an analog read-out for current. As you are biasing up the device keep a careful watch on the behavior of the needle. Any sudden unnatural movements of the needle is normally an indication of the device being pulled into a limit cycle and oscillating. Not very scientific, but you will recognize it when you see it (this behavior is not easy to see with digital read-outs).

If you pass these checks, then at least it should be safe to connect to a VNA that will present the device with similar port terminations.

Many moons ago I know that Cornell built an adaptor to their network analyzer that performed measurement directly with 10 ohms port references.

If you have a transistor that is unstable with 50 ohms terminations, the 10 ohm terminations will usually allow you to measure S parameters without the risk of oscillations.

As Cornell says, you must be wary of low frequency oscillations. The cure is usually to use some form of diplexer (a filter that separates the input signal two a low and a high frequency channel, with the input port facing the transistor) that will connect low resistor values to the transistor ports at low frequencies while providing a direct path to the signal ports for signal frequencies

S11 and S22 are small signal parameters. They are level dependent, so meaningless for high power levels. For the proper design of the Power Amplifier you would have to have load pull contours (usually they come with data sheets for LDMOS, but it might pay-off running your own measurements). Somebody recommended a book by Cripps, which is very good, but you can also look at my paper "Concepts of RF Power

Amplification" published in November 2017 in High Frequency Electronics (highfrequencyelectronics.com).