This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

CC1352R: CC1352R1 RF balun

Part Number: CC1352R

Hi,

Background and problem definition:
TI provides a calculated RF balun for the PCB stack-up where the width of a single-ended 50ohm trace is roughly equal to 12.5 mils. Once we want to integrate AM335x and CC13x2 on the single PCB, we have to use a stack-up with a width of a single-ended 50ohm trace roughly equal to 5-5.5mil. Unfortunately, it's not possible to use RF balun from the reference designs because the PCB stack-up change is too significant.

Question:
Would you help us to find (at least) theoretical values and component placement for RF balun (CC1352R1) for a given PCB stack-up?

Regards
Adam

  • Which stack-up are you planning to use?

    Note that the only trace that is a true 50 ohm trace is to the SMA connector. The traces in the balun is bot 50 ohm traces and are too short to be a transmission line anyway. I would follow the reference design. Bu if the stack-up is different the parasitics will be different meaning that the balun may need some minor change. You can simulate the layout + the balun with the ref design stack-up and with the new stackout and see if S11, S21 change.

  • Hi,

    Which stack-up are you planning to use?

    Let's define two stack-up versions:
    Stack-up A

    • Copper 35um
    • Prepreg 1xPR7628 - 200um
    • Copper 17.5um
    • Core (FR4) 665um
    • Copper 17.5um
    • Prepreg 1xPR7628 - 200um
    • Copper 35um

    Stack-up B

    • Copper - 35um
    • Prepreg 1xPR2313 - 100um
    • Copper - 17.5um
    • Core (FR4) - 365um
    • Copper - 17.5um
    • Prepreg 1xPR2116 - 127um
    • Copper - 17.5um
    • Core (FR4) - 365um
    • Copper - 17.5um
    • Prepreg 1xPR2313 - 100um
    • Copper - 35um

    I'm referring to the real situation, therefore I'm giving all the details. However, I'm aware that the number of layers is not an important factor here. Space and isolation material between the top layer and the nearest solid plane is the key factor.

    Note that the only trace that is a true 50 ohm trace is to the SMA connector. The traces in the balun is bot 50 ohm traces and are too short to be a transmission line anyway.

    Yes, I fully understand. I was referring to a 50ohm trace just to emphasize the difference between the two used stack-ups.

    Description of the current situation:
    We already did several prototypes using the above-mentioned stack-ups.
    The schematic attached.


    Let's focus on the 2.4GHz part.

    Stack-up A
    Maximum output power (2.44GHz, 3.3V) :
    UFL (before diplexer): 3.6 dBm
    SMA (after diplexer): 2.8 dBm

    Stack-up B
    Maximum output power (2.44GHz, 3.3V):
    UFL (before diplexer): 3.4 dBm
    SMA (after diplexer): 0.4 dBm

    Question:
    Would you help us to adjust RF balun to be used in stack-up B?
    What are the necessary steps to follow?

    Bu if the stack-up is different the parasitics will be different meaning that the balun may need some minor change.

    We are looking for guidance and assistance in these adjustments.


    Regards
    Adam

  • Since you get close to equal output power measured on the UFL connector for both stack-ups this indicate that the balun is ok. But you loose more than 2 dB from the input to the diplexer to the SMA.

    - If you bypass the diplexer and measure on the SMA, what do you then see?

    - It could be that the trace to the diplexer and/ or from it needs to be adjusted. Is it possible to post a screenshot of how the layou looks from the output of the balun and to the SMA?

  • Hi,

    - If you bypass the diplexer and measure on the SMA, what do you then see?
    Configuration Port Output Power (2.44GHz, Vcc-3.3V, ChBW-100MHz)
    C305 – DNM; C306 – DNM; C307 – 12pF UFL 3.1 dBm
    C305 – DNM; C306 – DNM; C307 – 0R UFL 3.5 dBm
    C305 – 12pF; C306 – DNM; C307 – DNM; C308 – 0R; C309 – DNM SMA 2.6 dBm
    C305 – 0R; C306 – DNM; C307 – DNM; C308 – 0R; C309 – DNM SMA 2.1 dBm
    C305 – DNM; C306 – 12pF; C307 – DNM; C308 – DNM; C309 – 0R SMA 0.1 dBm
    C305 – DNM; C306 – 0R; C307 – DNM; C308 – DNM; C309 – 0R SMA -8 dBm

    Reference measurement:
    LaunchPad CC1352R1
    SMA (after diplexer)
    Output Power (2.44GHz, Vcc-3.3V, ChBW-100MHz) - 3.3dBm

    It could be that the trace to the diplexer and/ or from it needs to be adjusted. Is it possible to post a screenshot of how the layou looks from the output of the balun and to the SMA?

    Take a look, it's a bottom side (mirror).

    Regards
    Adam

  • From these measurements it could look like the loss through the diplexer is higher than expected. Are you able to measure the level on the diplexer input pad without the diplexer mounted? 

  • Are you able to measure the level on the diplexer input pad without the diplexer mounted?

    I can definitely try tomorrow.
    Unfortunately, we lack practical experience with such ad-hoc RF connections.

    I could try to use a 10cm long semi-rigid 50ohm RF cable and solder it to this pad.
    Usually, it's a problem with the ground soldering, because the shield of the semi-rigid cable has a large thermal capacitance. Once I have this cable soldered, I can measure the level using a spectrum analyzer.

    Is that what you mean?
    Maybe you or anybody can give some practical guidance on how to handle this?

  • Off topic (do not reply on this...)

    For several weeks now, we can admire beautiful colors on the new forum. I don't know if you are collecting comments about it. A few below (maybe useful for forum administrators):

    • There are no email notifications if someone replies (On this whole thread, I received notification only once - suggested answer. Your first reply.)
    • Membership points are no longer incremented
    • Text area for writing reply is extremely small
  • Yes, I was thinking about a semi-rigid. It's not that easy in this case since the pads on the diplexer is small and it could be difficult to place the tip on the (correct) pad. Ground is not a large issue since you have a reasonable large ground area available around where you want to measure. I typically use a dremel (or similar) to expose the copper for ground. This makes it possible to use plenty of solder on the ground for mechanical stability. You would need a soldering iron that can deliver some watts to do this. 

  • It seems like my points are incrementing. I will try to notice if your points increment or not.

    The e-mail notifications seems to come and go a bit after the last update. I will check on this in a couple of days if the notifications don't come back.  

  • Hi,

    Are you able to measure the level on the diplexer input pad without the diplexer mounted?

    2.6 - 3.1 dBm depends on the way how the semi-rigid is connected (C305 – DNM; C306 – 12pF; C307 – DNM).

    I typically use a dremel (or similar) to expose the copper for ground. This makes it possible to use plenty of solder on the ground for mechanical stability.

    Thanks for the tips.

    Regards
    Adam

  • Based on this it looks like you have a loss through the diplexer that is larger than the datasheet indicate. I note that the layout recommendation in the diplexer datasheet is different than in your layout. It looks like the ground pins on the diplexer should be connected together on the lop layer. It could be that the different parts of the diplexer is connected to different ground pins and when these are not connected directly together the performance will change. 

  • Could be a good guess... however, I have the same layout using stack-up A, and signal levels are completely different.
    I guess that the traces from the balun to diplexer and from diplexer to SMA are not correctly matched.
    The trace from diplexer to the SMA is the longest one.

    Unfortunately, I have no idea how to check this in practice. Any hints?
    Tomorrow, I'm going to check the signal level just after the diplexer.

    Regards
    Adam

  • Btw, it actually looks like you points is not increasing. Strange.

    Could be a combination since the parasitics to ground change. Measuring the output of the diplexer (and cutting the trace after the pin) will show where exactly you get the loss. 

  • On the diplexer output (C309 pad while C309 - DNM), we have 3dBm.
    When we connect this trace to the SMA central PAD, the signal level drops to 0.5-1dBm.
    When the SMA is dismounted, the signal at the SMA pad is 0.5-1 dBm.

    We use a THT SMA, so I think the problem can be narrowed down to the central THT pad of the SMA connector.

    I think that the loss is due to the wrong clearance at the internal plane layers.
    Clearance to RF on top and bottom planes is 25 mil, while on the internal planes (Power/GND) is 5mil only. This creates additional parasitic capacitance and we lose the signal...

    Moreover, I found that on stack-up A we have the same clearance on both outer and internal planes (25 mil),which further confirms my assumption.

    Regards
    Adam

  • Btw, it actually looks like you points is not increasing. Strange.

    Still the same number.

  • I have sent a question about the points (and very slow e-mail notifications) to the e2e team), thanks for reporting.

    SMA: We discussed this topic internally yesterday. We have seen a 0.3 dB power loss with insufficient clearance on one of our designs before but it looks like the low clearance in combination with the stack-up create a fairly large parasitic cap that cause a large loss. If you cut the trace just before the central pad and solder on a semi-rigid, hopefully you will see close to 3 dBm.  

  • If you cut the trace just before the central pad and solder on a semi-rigid, hopefully you will see close to 3 dBm.

    Yes, I can confirm. When I have a semi-rigid soldered at the end of the track, I get 0.1-0.5 dBm just before the central SMA pad. If I cut the connection to the central pad, the signal level rises to 3-3.6 dBm.
    Thanks for your assistance. The next step is to make new PCBs and confirm this BUG.
    According to your remarks, we will also correct the footprint of the diplexer (GND connection).

    Regards
    Adam