>1GHz S11 minimisation

Discussion in 'The Projects Forum' started by NickL, Jun 20, 2011.

  1. NickL

    Thread Starter New Member

    Jun 20, 2011
    5
    0
    Dear all,

    I'm new to this forum, and would like to ask a question about >1GHz circuit engineering. I have a circuit which is very simple - it consists of a surface mount directional coupler from minicircuits, with the output being filtered by another couple of surface mount packages. This circuit lives in a little copper box, and my problem comes going from semi rigid coax to the microstrip on the circuit board, via an SMA bulkhead connector. I am trying to minimise the S11 of my box, but am struggling to get it below -20dB. I'm a physicist, not an engineer, and so am a little stuck.

    Are there any decent tricks I can use to improve this? Any help much appreciated.

    Best regards,

    Nick
     
  2. CDRIVE

    Senior Member

    Jul 1, 2008
    2,223
    99
    We're going to need more specifics. If this device is the end of the line then it must look like a 50Ω load. Does minicircuits spec the return loss?
     
  3. NickL

    Thread Starter New Member

    Jun 20, 2011
    5
    0
    Thanks for the reply.

    The circuit board is supposed to look like 50Ω. I think I've got that bit right. But the thing I don't know about is the launching of the signal from my SMA bulkhead onto the stripline, or into the smcs. I think that I am not managing the discontinuity in geometry well.

    The specified return loss of the directional coupler is 30dB at 1GHz. I get about 18dB.

    Incidently, this device is not the end of the line, but the rest of the network is also supposed to 50Ω matched.

    Cheers,

    Nick
     
  4. CDRIVE

    Senior Member

    Jul 1, 2008
    2,223
    99
    Please post some photos so we can get a clear picture of the transition from your hard line to the SMA.
     
  5. NickL

    Thread Starter New Member

    Jun 20, 2011
    5
    0
    OK, here's a photo.

    [​IMG]

    Looking at this, it seems to me that a major candidate for causing problems is the ground plane next to the input. Is this likely? It's just part of the recommended footprint for the directional coupler - I don't know what it's for. Shall I just take a scalpel to it?

    The traces look particularly wide because the substrate is a low loss material (from Rogers). The lid screws on through the hole that can be seen towards the middle of the board. The black material on the inside surface of the enclosure where the out and coupled ports leave is Eccosorb, to kill any box resonances. There would be another one on the bulkhead where the in port is.

    I really don't know what I'm doing with this, so any comments welcome!

    Nick
     
  6. CDRIVE

    Senior Member

    Jul 1, 2008
    2,223
    99
    Back in the early 70's I worked in an electronics lab that designed RF Log amplifiers for military ECM. Enclosures like yours had covers on both sides but a small 1.5" x 3.5" enclosure had cover screws spaced on ~ 3/8" centers, 360° around. The mating surfaces of the covers and the casing were EMR sealed with a braided gasket.

    Before you modify anything on your board I think you should close it up and take some good return loss measurements with all ports terminated into 50Ω terminations, then measure it again with the ports connected in operation.

    BTW, the photo doesn't show use the transition from hard line to SMA. Remember that any mismatch on the transmission line or your module will increase radiation of the line itself.
     
  7. NickL

    Thread Starter New Member

    Jun 20, 2011
    5
    0
    Hi,

    I've measured the return loss with 50Ω terminations - it's 18dB at 1GHz. I can't measure it in situ, but I can see the standing waves between this and the next component down if i measure S(coupled)(out). I want to get rid of these as far as possible.

    If by connection to the hard line you mean the semi rigid - sma soldering - I'm not worried about that because a) I'm a legend at soldering semirigid and b) the 18dB mentioned above was measured with this box stuck directly onto the front of a network analyser.

    "Remember that any mismatch on the transmission line or your module will increase radiation of the line itself. "

    I'm afraid I don't understand this sentence. Does an impedance mismatch really increase radiative losses? Is this effect large?

    Cheers,

    Nick
     
  8. CDRIVE

    Senior Member

    Jul 1, 2008
    2,223
    99
    In an RF circuit power transfer loss can be looked at by at least three elements. The first is the "Maximum Power Theorem" that states that maximum power will be transfered/dissipated by the load when the load impedance equals the source impedance. This applies to both DC and AC circuits.

    RF / Transmission line theory throws other factors into the mix. Transmission line "Loss Per Foot", as spec'd by the manufacturer, is valid only when source and the load exactly match the characteristic impedance of the line. When this is true, loss will be predominantly caused by dielectric loss and less so by resistive loss in the line. Theoretically, when everything is perfectly matched, the transmission line's radiation should be insignificant. On the other hand, when the source or the load does not present a purely resistive element equal to the line Z, it will radiate and radiation translates to loss. A poorly matched transmission line (Low Return Loss/High VSWR) looks more like an antenna than a transmission line.

    It's been many years since I've done this kind of work so I'm quoting from memory.
     
  9. NickL

    Thread Starter New Member

    Jun 20, 2011
    5
    0
    I see.

    So, back to my box... I know I need to match the impedance everywhere. Presumably I'm not doing. Do you think the bit of ground plane (arrowed in blue) may be having an effect?

    Cheers,

    Nick
     
  10. CDRIVE

    Senior Member

    Jul 1, 2008
    2,223
    99
    Sorry, but I can't make that determination at this end. I would think that the larger the ground plane, the better.
     
  11. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    It can be quite a challenge determining where the problems lie when you're working in the GHz range. Parasitics come into play so strongly that the primary component specification seems to go right out the window.

    It looks like you have an SMT capacitor (0805 ceramic perhaps?) connecting the two SMD's between OUT and IN. At a former employer, we tested a number of multilayer SMD ceramic caps on a material analizer; the higher the frequency, the higher the apparent capacitance became. That sort of thing will really throw you off if you're trying to build LC filters in the GHz range. We went to using custom-trimmed single layer caps for HF stuff.

    Are you using 6dB attenuators on both of your network analyzer ports, and calibrating using standards (open/short/thru) that are not expired on their calibration verification? (The 6dB attenuators help to minimize errors).

    Getting S11 to -18dB is really not too bad at all; it's getting down from there that's the real challenge.

    One "trick" that may help you is to try using "gimmicks" - small pieces of tinned flat copper stock - soldered to the ground plane, and try bending them closer to and further away from the signal paths to see what that does to your S11/S21. If your S11 increases when you get the gimmick closer to the signal path, then you don't have enough capacitive coupling to ground. You can also use gimmicks across things like what I mentioned that I think is a multilayer ceramic cap.

    I'm somewhat surprised that you obtained some of that Rodgers board; it was not available for a considerable period of time a few years ago.
     
  12. CDRIVE

    Senior Member

    Jul 1, 2008
    2,223
    99
    This is an excellent tip, especially with impedance mismatch. If you can tolerate the loss it helps considerably.
     
Loading...