8
\$\begingroup\$

I am in the progress of ordering parts for a LNA board (2.4GHz, based around the broadcom/avago MGA-635P8). I was following the manufacturer's component list in the datasheet for their evaluation board.

enter image description here

They use a 1000pF DC blocking capacitor. I was wondering why one would use such a large value in a 50Ω system, when the operating frequency is 2.3GHz up to 4 GHz. Wouldn't using a smaller value improve the noise power, since the bandwidth of the system is decreased? Is there some other reason why I would choose such a high value of capacitance?

\$\endgroup\$
8
  • 1
    \$\begingroup\$ I'm mystified too. I checked the datasheet and they specifically recommend Murata PN GRM155R71H102KA01. Which has a resonant frequency around 200 MHz. \$\endgroup\$
    – The Photon
    May 7, 2017 at 16:40
  • \$\begingroup\$ They claim all numbers in the datasheet are done with that board, so it seems to work, but I find it so strange. It seems like a very inappropriate choice... \$\endgroup\$
    – Joren Vaes
    May 7, 2017 at 16:47
  • 2
    \$\begingroup\$ To be fair, the Murata capacitor will still block DC, and it's still less than 10 ohms (inductive) at 4 GHz, so it might not be too bad a choice. But if you don't want the whole frequency range that demo board is designed for (down to 450 MHz), you can probably find a better choice. \$\endgroup\$
    – The Photon
    May 7, 2017 at 17:03
  • 2
    \$\begingroup\$ @ThePhoton Avago brag about low NF @ 2.5 GHz...could that input network be tricked-out to provide optimum noise match at that frequency? Seems a little suspicious to see a 10pf capacitor at the "DC" end of the demo circuit's bias inductor. \$\endgroup\$
    – glen_geek
    May 7, 2017 at 17:22
  • \$\begingroup\$ I orderd a handfull of values, I will let you guys know what happens when we measure it with out VNA. \$\endgroup\$
    – Joren Vaes
    May 8, 2017 at 5:33

2 Answers 2

3
\$\begingroup\$

Self-resonant frequency (SRF) of a DC coupling cap does not do what you might think it does. Think about it: SRF is the result of the cap's inductance, and its capacitance value.

In a decoupling application, of course you want low inductance. But the SRF of the cap alone means nothing, it is the SRF of the mounted cap that counts, including via inductance etc. Datasheet SRF is only a specification, which tells you the maximum HF performance you can expect from your cap if you mounted it perfectly (like, with magic non-inductive vias).

This would be the case of C3, C4, C5, C6 here.

Now, in a DC-coupling application, things are different. Notice the cap is in series with the transmission line. It is also the same width as your transmission line's copper trace, and it is very low profile (0.5mm height).

Since the cap is mounted right on the PCB surface, and its plates sit very low on the PCB, almost aligned with the trace, it acts just like it is part of the trace. The extra inductance it adds compared to the "no capacitor" case is much smaller than its actual inductance.

Capacitor SRF does not matter here. What matters is the difference between a straight bit of trace and the capacitor. This difference is very small. It does not depend on the cap's value, only its dimensions. For example if it is tall, it will have more parasitic capacitance with surrounding GND traces, introducing a slight impedance discontinuity.

The cap is in series with the transmission line, so the resonances you'd be worried about would be making a LC tank with L1/C3 or having it resonate with your transmission line inductance, that kind of stuff, but that has nothing to do with the naked cap's SRF.

Also, the current in your transmission line travels in the copper which is closest to the surrounding ground. Since there is a ground plane below, the current will concentrate on the lower surface of the trace, and at really high frequency, current will only run through the plates of the capacitor which are closest to the PCB. This will change the cap value a bit, also its ESR... another reason to use a tiny, and low-profile part.

\$\endgroup\$
2
\$\begingroup\$

You can easily use a smaller capacitor with resonance below or at 2.4GHz.

You can use Murata sim-surfing to find your preferred cap. The use of 1nF cap is to cover larger frequency band.

You will not improve your noise power since you generally don't rely on the gain-block/LNA bandwidth to filter the noise. You use a filter for it.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.