How do you use an RC filter on an LNA output to reduce unwanted noise?

Question

We designed two low noise amplifiers, one at 144 MHz and another at 440 MHz using this LNA; each design is tuned for the particular band. We noticed amplification at 440 MHz is clean and crisp, but amplification at 144 MHz is scratchy with FM audio signals. (The 144MHz LNA measures S21=+28dB and S11=-19dB with a VNA.)

Someone at a conference we went to recently told us that an RC filter is often used on an LNA output to reduce unwanted noise in the intended band. Later I wondered what type of filter to use and how it should be tuned, but the conversation was in passing I do not have access to the person to ask directly. He said that RC filters on LNA outputs have been used for years, so it sounds like this is a common thing to do.

Primary question:

• What is the best practice for RC filters on an LNA output? HP, LP, or BP?

• What dB drop across the filter is acceptable (or necessary) in the passband to operate as an appropriate filter? (We get 28dB gain at 144MHz so, if necessary, losing a few dB in the filter isn't that critical.)

More detail mixed with my speculation

Here they are, outputs are on the female SMA side (with the flat-side of the PCB down, output is on the right):

Below is the matching schematic, roughly oriented to the 2m picture above. Notice that S7 is disabled in the schematic (and bare pads/missing in the bottom-right of picture above), so we were thinking of adding a cap to pull down higher frequencies (in lieu of an RC filter?):

This is the output match by itself, that shows a highpass behavior on the 270pF output cap with S7 empty (bottom right of the 2m PCB picture):

This is a simulation with a 22pF shunt cap in S7 (hoping to avoid a board-redesign, but we can if necessary):

There is a big issue with this 22pF shunt: In S2P simulation (no EM), it kills the S11 input match, going from -27dB to -10dB. Smaller shunt caps affect the match less, but of course with less high-frequency attenuation. (Not sure how the real life match is affected without trying it, thought I'd ask the experts here, first.)

Suggestions?

Answer 1

very often, you'll want to do both – filtering before and after the LNA. Yes, when filtering before insertion loss is a problem (because it effectively reduces the gain of the first active stage, and Friis' noise formula doesn't like that), but in the presence of blocking interference, you won't get around it. Filtering after the LNA is still a good idea – a bit of power loss doesn't matter that much anymore for overall noise figure, and:

Someone at a conference we went to recently told us that an RC filter is often used on an LNA output to reduce unwanted noise in the intended band.

Exactly, because if you happen to have a strong signal at your amplifier, be it in-band or out-of band, that makes your amplifier a bit less linear.

Let's think about a second amplifier stage after your LNA:

Any non-linear amplifier (and amplifiers, even without a narrowband interferer saturating it, are always a bit nonlinear) is effectively a mixer (if you take the output/input function and develop it into a Taylor series, you see a quadratic term appearing: that's your mixer right there).

That means that a tone at 190 MHz and a signal at 430 MHz yield an intermodulation product at at \$430 \pm 290\$ MHz, i.e. at 720 MHz and at 140 MHz. And the latter you really can't use. (So, you would filter as much of the non-140 MHz frequencies to begin with.)

Problematically, the same applies to harmonics of 140 MHz (which you are interested in) and receiver noise:

non-linearity in your LNA makes 0 Hz and 240 MHz, 420 MHz …; and as you always have receiver noise, further down the line you'll mix your wide band noise back down to 140 MHz.

So: get rid of these harmonics! A low-pass filter is almost always a good idea after a mixer.

Best practices are really hard to give. Of course, a band-pass makes the most sense here, practically: you want one band, and not the noise above or below, and if there's any interferer that made it through frontend filtering, you'll have to get rid of it anyway.

How that's implemented is a pretty different question – as a mater of fact, so good a question that the whole principle of the superheterodyne receiver is based on making that band-pass filter sharp: Buy / build / tune a filter that is really good, and move the signal of interest so that it fits that filter. (Intuitively, I'm always tempted to design a filter to "fit" the signal; superhet is the realization that good analog filters are hard, and if you need superb selectivity, moving the signal to the filter's passband is what wins.)

But: compared to a complex thing like a multi-stage analog filter for 144 MHz, be it implemented directly at 144 MHz or by means of mixing and filtering at a different frequency, an RC low pass is "nearly free"! Note how far the first relevant harmonic is from 140 MHz: this is a really easy filter to "get right enough" to have a benefit. It would be a shame if you didn't add that filter.