How do you verify an impedance matching network?

[AJ528]

If I have an RF amplifier where I know it's source impedance, I can either use math or simulations to find a theoretical matching network to turn it into 50 ohms. However, reality is often different than theory, so once it is built, how can I verify this matching network to make sure the components are optimal? If I replace the RF amplifier with an SMA jack, is it as simple as hooking up a VNA, measuring the impedance, and verifying it is the complex conjugate of the RF amplifier source impedance?
If that is a good approach, how do you make sure your measurement plane using the SMA jack matches the RF amplifier?

Is it also possible to verify the matching network by having the amplifier output a continuous wave, measuring the 50 ohm side, and verifying you are seeing the max power being output?

Essentially, I am thinking of making a test PCB with 2 identical circuits, except replacing the RF amplifier with an SMA jack on one (see attached image). Will this work how I think it will? Is there a better way to verify impedance matching?

Edit: The frequency of interest is 915 MHz, and when I say "source impedance", I mean I have the load-pull data for the amplifier

[szoftveres]

How about building your amplifier and matching network, then measure from the 50ohm side? If the output impedance isn't what you're looking for then you can continue tweaking your matching network.

[AJ528]

Wouldn't measuring impedance from the 50 ohm side only tell me if that side is 50 ohms? How does that help me verify I have matched the impedance between the amplifier and matching network?

[szoftveres]

How does that help me verify I have matched the impedance between the amplifier and matching network?

Well, if you don't get exactly 50 ohms then obviously you've failed to match your amplifier. On the other hand if you get exactly 50 ohms then you succeeded. Isn't this what you're looking for? What do you mean under "verification"?

[AJ528]

By verification I mean "how do you prove the impedance matching network is actually doing its job?" What measurements can I perform that will tell me everything is working properly?

Also, how would you do an impedance measurement on the 50 ohm side if the RF amplifier is an active device? If you do the measurement with the RF amplifier off it wouldn't show its true source impedance, but if you try to do an impedance measurement with the amplifier on it will mess with your measurement, right? I assumed the best I could do for measuring the output is measuring the power being output on the 50 ohm side. If I know the power being output by the amplifier, I can easily calculate what the losses are from the matching network right?

[Geoff-AU]

You do have to measure it powered to get sensible results, but you need to make sure the amplifier isn't going to transmit a signal (best case, interferes with your measurement. Worst case, damages your VNA).

Be very, very careful and thoroughly think things through before connecting potential transmitters to VNA/SA.  Inline attenuators to make sure worst-case malfunction won't cause overload is pragmatic (make your calibration reference plane after the attenuation, of course).

[AJ528]

Right, I figured that much. The transmitter outputs a max of +14dBm, so I'm planning to put a 20 or 30 dB attenuator in line before hooking it up to a VNA.

But that's also why I'm wondering if I can measure the impedance of the matching network by replacing the amplifier with an SMA jack. I could do a measurement without anything powered.

[radiolistener]

Just load your output with two different load around expected impedance and measure amplitude, then calculate it's actual output impedance.

For example, if expected output impedance is 50 Ω, you can measure amplitude for 25 Ω and 75 Ω load.

Let's assume you get U1=20V for R1=25 Ω load. And U2=35V for R2=75 Ω load.

Then:

Z = R1*R2*(U1-U2)/(R1*U2-R2*U1) = 25*75*(20-35)/(25*35-75*20) = 45 Ω


Another way (but less reliable) is to measure EMF voltage of your output (with no load) and U voltage for known R load.

Let's assume you get EMF=40V. And U=22V for R=50 Ω load.

Then:

Z = R * (EMF / U - 1) = 50*(40/21-1) = 45.24 Ω

But the second method cannot be used for devices whose output does not allow being open or when the load impedance differs significantly from the specified value.


PS: Make sure that load resistors for both methods are not inductive and not capacitive.

Since both methods are scalar, you will not be able to measure the exact value of the reactance of your output.

[EggertEnjoyer123]

What frequency are you working at?

Do you have the S parameters of the transistor from the manufacturer?

The best way is to build the transistor with the matching network, and then measure S11 with a VNA. The S11 should be very low. Make sure to put an attenuator on the output of the amplifier (and maybe the input too to avoid saturating the amplifier) to avoid damaging your VNA. (If you use attenuators you need to calibrate the VNA with the attenuator in place to cancel out any effects it would have). Then you can reverse the connections and measure the S11 again (assuming you have a cheapo Nano/LiteVNA). By switching the connections you are effectively measuring the output return loss or S22. You should also do a sweep from DC to the fT of the transistor, and then calculate the stability factor using the S parameters to ensure that your design is stable.

Measuring just the matching network is a bad idea, especially in the microwave range. Your connector will add extra length which can be difficult to calibrate out. Also, the transistor return loss is also sensitive to many factors, like biasing, emitter inductance, parasitics on your board, etc. It would be bad if you optimized your matching network, then found out that the impedance of the transistor was wrong.

[selcuk]

Are these components on the PCB and SMD parts? If so, replacing SMD parts with a SMA jack alters the results too much. At least I don't know an easy way to compensate it within the VNA. I would replace the RF amplifier with a 0402 or similar 50 ohm high frequency resistor (soldered between amp output pin and ground) and make the measurements from the load side SMA jack. I think that one is already in the circuit and not for test purposes.

[AJ528]

What frequency are you working at?

915 MHz

Do you have the S parameters of the transistor from the manufacturer?

No, I only have the source impedance (approximately 11 - j5 Ω)

The best way is to build the transistor with the matching network, and then measure S11 with a VNA. The S11 should be very low.

Where in the circuit should I measure S11? From the 50 ohm terminated SMA jack? If so, wouldn't the amplifier need to be turned on to present realistic impedance? And wouldn't that mess with the S11 measurement?

Make sure to put an attenuator on the output of the amplifier (and maybe the input too to avoid saturating the amplifier) to avoid damaging your VNA. (If you use attenuators you need to calibrate the VNA with the attenuator in place to cancel out any effects it would have).

This is a PCB-mounted design, how would I insert an attenuator between the amplifier and the matching network? Also, the amplifier is an output from a microcontroller, so I can only change the amplifier input through microcontroller settings.

Measuring just the matching network is a bad idea, especially in the microwave range. Your connector will add extra length which can be difficult to calibrate out.

Yeah, this is something I'm concerned about. I know at 915 MHz each millimeter rotates the smith chart by a couple degrees, but I don't know of another way to verify the impedance of the matching network, except possibly measuring the output power with the amplifier transmitting.

Also, the transistor return loss is also sensitive to many factors, like biasing, emitter inductance, parasitics on your board, etc. It would be bad if you optimized your matching network, then found out that the impedance of the transistor was wrong.

How would I even be able to tell if the transistor impedance was wrong?

[AJ528]

Are these components on the PCB and SMD parts? If so, replacing SMD parts with a SMA jack alters the results too much. At least I don't know an easy way to compensate it within the VNA.

Yes, these are SMD parts on a PCB.

I would replace the RF amplifier with a 0402 or similar 50 ohm high frequency resistor (soldered between amp output pin and ground) and make the measurements from the load side SMA jack. I think that one is already in the circuit and not for test purposes.

Replacing the RF amplifier with a 50 ohm resistor doesn't represent the source impedance of the RF amplifier. I would need to replace the amplifier with a (11 - j5)Ω component, right? I don't know how to create that.

Also, the load side SMA jack also exists for test purposes. Eventually I'll be replacing that jack with an impedance-matched antenna, but I don't think that's relevant for this question.

[radiolistener]

Replacing the RF amplifier with a 50 ohm resistor doesn't represent the source impedance of the RF amplifier. I would need to replace the amplifier with a (11 - j5)Ω component, right? I don't know how to create that.

11 Ω resistor in series with capacitor which gives 5 Ω at specific frequency:

Xc = 1 / (2*pi*f*C)
C = 1 / (2*pi*f*Xc)

for f=915 MHz and Xc=5Ω C=34.79 pF

[EggertEnjoyer123]

The best advice I can give is to build the entire design (with biasing, matching network, etc) on a board. Basically build everything as it would be in real life. Measure the S parameters and adjust the output matching network until it is at 50 ohms (basically get S22 < -15dB or so). The input matching network is highly dependent on the output matching, but not the other way around, so it is very important to adjust the output match first. Then, once your output matching network is good, you can adjust the input matching until you get -15dB S11. You will probably need to spend some time switching around components though.

[AJ528]

The best advice I can give is to build the entire design (with biasing, matching network, etc) on a board. Basically build everything as it would be in real life. Measure the S parameters and adjust the output matching network until it is at 50 ohms (basically get S22 < -15dB or so). The input matching network is highly dependent on the output matching, but not the other way around, so it is very important to adjust the output match first. Then, once your output matching network is good, you can adjust the input matching until you get -15dB S11. You will probably need to spend some time switching around components though.

I can give that a try, but I think if I measure any S parameters from the 50 ohm side, it will give me wrong measurements. Either I measure with the device not transmitting, in which case it's not not the right source impedance, or I measure with the device transmitting, in which case it's injecting current into the VNA and messing up the measurement. I did some more research and I think this is maybe where someone might utilize a "hot S22" measurement.

Additionally, my research turned up this application note from Maxim https://pdfserv.maximintegrated.com/en/an/AN5417.pdf where their recommended way to verify the transmitter matching network is essentially one of my original ideas: assemble the transmitter and matching network, then measure the output and tweak stuff until you get the max output power.

So I've at least found some evidence to support one of my approaches.

[EggertEnjoyer123]

Small signal S parameters only work when the transistor is operating linearly. The transistor's output power is linear with respect to the input power until you approach the P1dB point. In the linear region, the transistor basically has the same gain and impedances, and it doesn't matter what the input power is. It's like if you had an op amp amplifier, and you put in 1mV and got 10mV out, you know that if you put in 2mV in, then you'll get 20mV out. However, if you put in 10V, then you most likely won't get 100V out because the amplifier can't output that much voltage. The gain of the amplifier in this case would be 10 for most input voltages, but once you reach a certain point it won't be 10 anymore.

Check the P1dB of the transistor. Then determine how much power you expect to get at the output. If the power output is lower than the P1dB, then you can expect the S parameters to be the same when you are transmitting, and when you are using the VNA. If the power output is the same or higher then you might need more advanced testing equipment (I'm pretty sure the cheap VNAs do not have enough dynamic range to do a hot S22 measurement).

[AJ528]

Something I forgot to put in the original post is that my RF amplifier is actually an output from a microcontroller. So I don't think I can do any transistor-level measurements. I can turn the transmitter on, set the approximate power level, and turn it off.

And I agree, I'm sure my nanoVNA can't handle a hot S22 measurement, which is why hooking it all up and measuring for max output power seems like my best bet here.

[EggertEnjoyer123]

Is this Lora?

The maximum output power should be listed in the datasheet for the microcontroller. I guess you wrote +14dBm. Find the gain of the amplifier at 915MHz, add 14 to it, and then compare it with the P1dB.

[AJ528]

I'm just using it to transmit generic FSK packets, but the micro can also do Lora if desired. It's the STM32WL5X series. Yes, it has a listed max output power.

[EggertEnjoyer123]

What's the transistor?

[AJ528]

The microcontroller is directly driving the RF output, there is no external transistor. I didn't see any information about the internal transistor in the microcontroller datasheet.

[EggertEnjoyer123]

If there's no transistor, how are you amplifying it?

[tszaboo]

Well, you have 6 problems that need to be measured.
S11 of the source (known)
S11,S12,S22,S21 of your amplifier
S11 of the antenna
These need to be matched. To keep it simple you probably want to match everything to 50 Ohm. So you need 4 matching network.
You need to figure out how much gain you wish, which is the S12. Also how much return, S21, if you want to receive with the same antenna.
Make the setup, so you can measure each of these scatter parameters. I would build them separately until I can measure everything, and have space at least two L matching pad for each device. As an extra point, take care about the DC coupling, and treat it separately from your matching for now.
Oh and probably you want to measure S12 with an attenuator depending on your VNA.

[AJ528]

If there's no transistor, how are you amplifying it?

The amplifier is built in to microcontroller. So technically there is a transistor, but it's not a discrete part, it's part of the microcontroller.

[AJ528]

Well, you have 6 problems that need to be measured.
S11 of the source (known)
S11,S12,S22,S21 of your amplifier
S11 of the antenna

The amplifier is built in to the microcontroller, and I know the complex source impedance of the amplifier, so eventually I just need to attach the output of the amplifier to my antenna. I'm planning to match to 50 ohms in between, so I'll need 2 matching networks. My question here is "how do I verify the amplifier->50 ohm matching network is actually matched well?" And the solution I'm coming up with is "solder the microcontroller and the matching network on a PCB, have the micro transmit a continuous wave at the desired frequency, then measure the output power and make sure it's close to the theoretical output of the microcontroller." So if I set the micro to output a +14dBm output signal, I should measure something close to 14dBm on the 50 ohm side.

At least I haven't heard anyone say "that's a bad idea" or "that approach won't work".... so I think that's what my approach is gonna be (unless someone knows of a better way).