transeiver impedance tuning

[justtry114]

Hi all,
I am so confused with the impedance matching and tuning skill when use a transeiver in our design. A standalone amplifier or some discrete part is easy to measure the impedance and performance, so the tuning is not a problem.
But these days I am designing some wifi and bluetooth product. We use the atheros SOC for wifi design and TI CC2540 for bluetooth.
For the atheros WIFI SOC.
I don't have any clue on how to tune the balun for the rx chain and tx chanin. For rx, wifi SOC has differential input. Differential input impedance for best sensitivity is unknow. There is single end matching component and LC balun in the rx lineup. Usually I use VNA to measure the impedance at single end. Then I tune the single end matching component and LC balun to make inband S11 better. But after that I can't get better sensitivity. That's S11 can't help me to identify if matching and balun is good or not. It's the same for TX. And also I can't control the SOC into pure TX only or RX only state.
For the TI CC2540
2540 has buildin T/R switch and one chain differential intferface to enxternal antenna. I use VNA to measure the single end input imedance, suppose that 2540 stay at RX at most of the time, I can mesure the input matching for RX. By using VNA we tune the matching component to get a acceptable S11, then use spectrum analyzer to measure the output power. If both the inut S11 and output power meet requirement. Tuning is done. But some times this tuning method doesn't ledad to a good field test result. Besides, the oddest thing was that the measured input impedance keep the same no matter CC2540 is power off or power on.

In a word, the tuning for this kind of SOC is a little bit bindness. So I am here to seek advice from you guys on how to do the tuning of this kind transeiver？

Thank you for your help!
Derek

 

[E Kafeman]

I have tuned RF<->antenna for a lot of circuits based on CC2510-2540. It is a clearly measurable difference in impedance when power off, TX or RX. All three modes differ clearly.
When I get the DUT for tuning can continuous, fixed frequency, TX or RX, be manually selected by software, but have no idea how it is done, as I not do the programming. Guess it else would be very hard to do any tuning.
In general is it good enough to use by TI proposed balun circuit values in combination with short traces and stable ground, and then just tune the T network that feeds the antenna.
Fine tuning of balun circuit, and it is possible to increase TX level with 0.5-1 dB above data sheet typical values.
Other reasons I have had to adjust balun values have been, due to that a fully differential solution was used and a good balun balance also reduce spurious levels.
Do not either know optimum impedance setting for RX but as TX output matching tuned for max output power, also result in good RSSI level, do I focus on TX level.
If PCB length is short or if RF ground not is ideal, do VNA measurement coaxial cable braid cause a detuning, that not exist in field test. This effect is can be reduced by ferrite tubes.

 

[justtry114]

Thanks Kafeman for your kindly help.
Yeah, I found that 2540 can be confiured to continuous RX state with TI SmartRF tools. Or I can ask our software engineer to program the device into the desired mode.

For the cable braid detuning, I think you are right when for pcb with small size. But I usually solder the coxial cable braid to PCB's ground firstly, then do the auto port extension to compenstate the cable delay. After doing that, it doesn't show difference either add the ferrite tubes to the cable or not when do the impedance measurement. This is base on my daily circuit tuning observation. Maybe I have make some mistake? Or I misunderstanding you?

I still looking for advice on Atheros WIFI balun tuning. It would be appreciate if anyone could help me.~ Thanks.

 

[E Kafeman]

Doing port extension is not related to that the measurement cable becomes an extension of ground plane.

Say that you is doing impedance matching for a lamda/4 monpole placed on a lamda/8 length of PCB.
Adding a coaxial cable and actual ground is extended enough to give a good behaviour of the monopole.
What happen is that when cable is removed, is it like cutting away a part of a halfwave dipole.
This is well known in cell phone industry and isolating measurement cable is not an option, it is an requirement to be able to do a good job for that size of PCB relative frequency.
Ferrite tubes fitted for actual frequency around coaxial cable, and the cable should leave PCB in a straight angle at a spot on PCB that not is hot.
See also https://www.antune.net/calibration.html , halfway down on that page is a photo.

Sorry, missed your Atheros question, but can not see your problem. Some of Atheros WiFi chip do not have differential RF pins. Can you provide a Atheros PDF that explains a bit?
A balun is common either as discrete components or as a winded core, but that is seldom a tuning problem? If you want to measure RX impedance just do so.
Depending on situation do you maybe need to make a balanced measurement probe. This is one of my 2.4GHz probes:
It is simple to make, it is a lamda/2 transformer. This kind of probe should be calibrated typically at 200 Ohm and is a bit limited in frequency range but within 2.4-2.5 GHz is it a very minor problem.
If your VNA supports full 2-port measurements is no balun needed for balanced S11 measurement, but I prefer to use a single coax with a diy balun as it is a bit more handy.

 

[justtry114]

Thank you Kafeman.

Doing port extension is not related to that the measurement cable becomes an extension of ground plane.
Sorry, missed your Atheros question, but can not see your problem. Some of Atheros WiFi chip do not have differential RF pins. Can you provide a Atheros PDF that explains a bit?
A balun is common either as discrete components or as a winded core, but that is seldom a tuning problem? If you want to measure RX impedance just do so.
Depending on situation do you maybe need to make a balanced measurement probe. This is one of my 2.4GHz probes:View attachment 92566
It is simple to make, it is a lamda/2 transformer. This kind of probe should be calibrated typically at 200 Ohm and is a bit limited in frequency range but within 2.4-2.5 GHz is it a very minor problem.
If your VNA supports full 2-port measurements is no balun needed for balanced S11 measurement, but I prefer to use a single coax with a diy balun as it is a bit more handy.

I have never use VNA to do differential S parameter measurement. But I know I can use two port VNA to measure the single-end impedance then convert them to differential impedance. I have interesting to your diy probe. How about the accuracy use this probe? especially when this probe is hand-helded but not soldered on board?

For the Atheros problem, I use AR9342 to design a AP. This SOC has differential rx input and tx output. I have never use ARtool to configure the SOC into continuous rx, I measure the input impedance before the matching network when the SOC is operating in normal operation mode. I think when there isn't anyother AP connecting to the AP undertest, then the AP should be at rx mode at most of the time. Then that's why I try to measure the rx input impedance even when the SOC is not in continous rx mode. But I don't get good result by this way. After tuned the impedance to 50Ohm, the senstivity is not good yet~

 

[E Kafeman]

The probe is handheld and my hand is not too steady but have no problem to get correct readings.
Accuracy within a such narrow bandwith as 2.4-2.5 GHz depends on how well it is calibrated.
Perhaps biggest problem is that a special calibration standard must be defined in VNA, and VNA do probably not accept that you set 200 Ohm as characteristic impedance. Either do you need to take in account that 200 Ohm shows as 50 Ohm in Smith chart in VNA or can you use an external software that allows 200 Ohm setting, showing corrected values while reading data from VNA.

Even if the Atheros chip is in RX mode most of the time, can it be many reasons that RX impedance measurement fails.
Do the chip for example contain internal switches which only opens for shorter periods, is correct bias set, is VNA power setting to high...

 

[justtry114]

Thanks for the information Kafeman.
I would like to try your diy probe later. But seems the calibration procedure is not the same as 50Ohm single end calibration. Do you have any material for reference to help me understand this kind of probe calibration? Thank you very much for your kindly help in advance!

 

[E Kafeman]

Calibration procedure is simple: Short, Open and Load. As load is a 200 Ohm SMD resistor used.

 

[justtry114]

Short is to solid connect both end of the two branch of the probe together,Open is to let end of the two branch of the probe untouch and stay in the air. After calibration, the characteristics impedance of the smith should be set to 200 Ohm. Is that right?

 

[FvM]

the characteristics impedance of the smith should be set to 200 Ohm

The VNA is still operated with 50 ohm impedance matching. Does your instrument provide scaling to 200 ohms? I would rather expect that the results have to be scaled by a factor of four manually.

There has been a discussion about ground plane size and cable effects. I understand, that it mainly applies for impedance measurements of embedded antennas, which is apperently not involved in this case. Finite common mode impedance of the balun will still introduce errors in combination with circuit assymetry.

 

[E Kafeman]

Construct the probe is relative simple but start with a slightly to long coaxial cable, and cut down until optimal length is found. With only the first coaxial cable, calibrata so VNA shows correct open. Connect next cable add cut it down until VNA measurment again shows best open.
Calibration is a minor problem. What probably intoduces a much bigger measurement error is if not a correct calibration standard is defined/selected in VNA.
After calibration, to verify that results are reasonable reliable, I measure a single inductor at a few nH and compare with manufacture S11 data.
Same procedure is repeated with a cap. Component error can be relative high at these values but have since long a small collection of componets that are assumed ideal according to manufacture data sheet.
It is a tool with many limitations but I have tried to measure assymetric situation (single component, big ground plane), handheld relative not hand held, with and without ferrite tubes, rotate it a half turn comparing results as well as situations with high VSWR and so far has this simple probe been good enough for my measurements needs. I like it beacause it is a very handy size and distance between the both tips fit most of my needs, and if not, is it simple to adjust them or make a new better fitting probe.
There is a problem that I not know much about, this probe causes funny loads, especially for harmonic frequencies. In an nervous TX circuit can this probably cause problem. Also, at DC is the probe a shortcut, can easily kill componets if measurement is done at wrong place. Verified by me.
Have planned to try to make a probe with serial capacitors included in the design and still keep the probe tips small, but it is a plan with low priority.

It is a scaling problem for Smith chart that none VNA I know about is able to handle. In my case do I seldom use VNA display for data readings. AnTune software is presenting more informative live data directly from VNA over GPIB and AnTune allows setting of any characteristic impedance.

 

[justtry114]

Thank you Fvm and Kafeman, your answers help me a lot.
to Kafeman, as you mentioned, cut the cable to show corret open, I think you are try to identify the length of half of lumda. The calibration was done on single end with a extened coxial cable which length equals to the proble.