Matching an antenna and a chip pin to 50ohm

[Koen]

Hello,

     up to now, I only had to match an antenna to a well-defined 50ohm port.

     For this occasion, the pin of a chip also requires to be matched to 50ohm. The datasheet reads : "The LNA input requires two to three discrete components for input matching. Adjust these components values to provide a 50ohm input impedance".

     Now, do I match the antenna on one side, match the pin on the other side then later join both sides with a 0 ohm resistor ? As in this schematic :

     

     If it is so and none more adventurous, is there a way to verify the good matching of the whole circuit by probing somewhere ? I'll explain myself :

when both sides are unconnected by the missing zero ohm resistor :

• probing "ANT test pin" yields a RL S11 at 1GHz of -20,
• probing "LNA test pin" yields a RL S11 at 1GHz of -20.
• great !
when both sides are then connected by a zero ohm resistor :

• probing "LNA RF_In pin" yields a moved center frequency around 700MHz. I would have expected it to show RL S11 at 1GHz of -20, maybe is it the influence of the LNA internals ?

Is it maybe normal, should I probe somewhere else, should I simply trust that both sides of the uniting 0ohm resistor where correctly matched and leave it to that ?

Thank you very much, Koen

[uncle_bob]

Hi

If you have the S parameter data for the LNA, you plug that into your math and out comes a matching network. As for checking it, the issue is that the LNA's best noise performance and it's best match performance are likely not the same thing. It's easy(r) to test for match. It's not quite so easy to test for noise figure via an antenna.

The "normal" approach is to do the math and put in a circuit. A lucky intern then gets to try some changes while observing signal to noise over a live circuit. The swap out / capacitor bump process may or may not improve things much.

Even more exciting: your antenna may not really be 50 ohms....

Bob

[Koen]

Hello,

    and thank you for your answer which I haven't really understood.

    To reduce my first post, I'd like to ask :

• is optimizing the value of the VNA S11 Return Loss on the LNA matching test pin the correct way to match the LNA pin ?
• is connecting two 50ohm matched blocks (matched antenna + matched pin) with a solder blob or 0ohm resistor correct ? Or is there a trap ?

I do not have access to the LNA output port so can't run an S21 transmission test, sorry.

Thank you,
Koen

[uncle_bob]

Hello,

    and thank you for your answer which I haven't really understood.

    To reduce my first post, I'd like to ask :

• is optimizing the value of the VNA S11 Return Loss on the LNA matching test pin the correct way to match the LNA pin ?
• is connecting two 50ohm matched blocks (matched antenna + matched pin) with a solder blob or 0ohm resistor correct ? Or is there a trap ?

I do not have access to the LNA output port so can't run an S21 transmission test, sorry.

Thank you,
Koen

Hi

No matching the VNA return loss S11 is *not* the way it is done.

No, connecting two "50 ohm" pads is *not* how it works.

====

The optimum noise figure operating point is where you match the LNA at. The S11 match guarantees that you put the maximum power into the LNA. That also guarantees that there is significant loss in the source impedance. That limits the achievable noise figure.

A simple example:

I have a 1V 50 ohm source. By the "rules" of 50 ohm sources, that is a 2V voltage followed by a 50 ohm resistor. If I match it, I get 1V into the amplifier. If I have a "high impedance" load, I get 2V into the amplifier. If the amplifier has 1uV of noise, the signal to noise is twice as good in the "high impedance" case.

At RF things are not quite as simple as "high Z". The process is a bit more complex. You still have a match point that is not the same as the S11 match point.

The 50 ohm to 50 ohm thing is a little more complex. The process is essentially building a custom filter that achieves the correct impedance transformation. Unfortunately, just tacking filter sections directly to each other does not work. If you calculate a third order filter and compare it to a second order filter and a first order filter .... there is no simple relation between them. You need to directly calculate the proper filter.

Sorry, but there are no trivial shortcuts. You need to do it correctly. The only simple shortcut is to just connect the antenna directly with no match. You then have a less than perfect system. It probably works.

Bob

[Koen]

Thank you, I found this by opening a commercial module :



I do not understand "The 50 ohm to 50 ohm thing is a little more complex." part. I assumed the RF industry standard was to connect a 50ohm antenna to a 50ohm receiver which would be an equivalent ?

Now, if this isn't how its done, would you please have pointers on how to do it ?

Thank you very much,
Koen

[Koen]

I just found this in an NXP GPS LNA datasheet and it makes no sense to me that the input pin S11 (Fig 4) center frequency isn't around GPS L1 frequency (~1.6G) while the output pin S11 (Fig 5) is. http://www.nxp.com/documents/application_note/AN11420.pdf

[uncle_bob]

Thank you, I found this by opening a commercial module :



I do not understand "The 50 ohm to 50 ohm thing is a little more complex." part. I assumed the RF industry standard was to connect a 50ohm antenna to a 50ohm receiver which would be an equivalent ?

Now, if this isn't how its done, would you please have pointers on how to do it ?

Thank you very much,
Koen

Hi

How much training in circuit theory have you had?

Put another way, if I gave you a polynomial, have you done the process of turning that into a bandpass filter?

If you already have that training, there is no need to go back through it. If not, then a starting point a bit further back will be needed.

The "how to do it" is to put the numbers into a proper synthesis and derive the needed components. If the starting point is far enough back, it would be better to point you to a course or two on synthesis rather than try to teach it all here.

Bob

[Koen]

Hello,

      putting it simply, I have no training whatsoever in electronics.

      Does combining both matching networks create a filter ?

      Should the antenna be tuned to the actual impedance of the LNA pin instead of the standard 50ohm ?

Thank you, Koen

[uncle_bob]

Hello,

      putting it simply, I have no training whatsoever in electronics.

      Does combining both matching networks create a filter ?

      Should the antenna be tuned to the actual impedance of the LNA pin instead of the standard 50ohm ?

Thank you, Koen

Hi

Ok, what you are asking to do is a task that a normal EE would not be able to do after getting a 4 year degree. It would take further training past that point. Simply going into the "why" means getting you through at least three of the first 4 years of training.

Here it is very simply put:

The LNA *must* run from a very specific source impedance that is *not* 50 ohms. That is dictated by the physics of the LNA.

The "matching network" does not match a load, it provides the source impedance that the LNA needs for best performance.

To tune the matching network to a 50 ohm source, you need a noise figure meter, not a network analyzer.

The antenna may or may not be a 50 ohm source, this makes things even more complex.

The system may or may not need the high pass filter shown in one of your schematics. That adds complexity as well.

If you are doing a bunch of these as part of a product, rent the gear you will need and hire a guy for a day to to the work (no I'm not shopping for a job). It's a cheap thing and you will get it done right.

If you are doing a one off, just ignore the matching and wire it up.

Bob

[Koen]

Thank you for your answer. To my surprise indeed, connecting the antenna matching network to the LNA input pin directly yields an average but working end result. Now, the manufacturer datasheet recommends two component values and I have a commercial module with its LNA matching network available to see how they've done it. If I understood correctly, there is absolutely nothing I can measure with a VNA to ensure things are properly done on the LNA side ?

[uncle_bob]

Thank you for your answer. To my surprise indeed, connecting the antenna matching network to the LNA input pin directly yields an average but working end result. Now, the manufacturer datasheet recommends two component values and I have a commercial module with its LNA matching network available to see how they've done it. If I understood correctly, there is absolutely nothing I can measure with a VNA to ensure things are properly done on the LNA side ?

Hi

Not if you want to set it up for optimum noise figure. The "best match" point ( = what you get with a network analyzer) may give you a 3.9 db noise figure. The "best noise" point might give you a 0.9 db noise figure. Yes those are random (but likely) numbers based on typical parts. Simply wiring it straight might put you at 2.3 db noise figure.

Bob