# How does lowering the RF Gain help with SNR?

Let's say for sake of example we are in the 40m band.

I've read articles and posts about riding the RF gain (or even just simply lowering it). And I can see how it is useful in overload situations.

What I don't quite get is how lowering the RF gain can help dig a weak signal out of the noise. A "weak signal" just means that it's not much stronger than the noise floor. Lowering the RF gain will lower the noise floor. But it will also lower the signal as well.

How is it possible that lowering the RF gain will attenuate the background noise faster than the desired signal. If you lower the gain by 3dB. Everything goes down by 3dB, right? The SNR doesn't improve.

I've also heard of "riding the RF gain" where you crank up the AF and use the RF as your volume control. How does this help? You are lowering the noise floor and the desired signal by the same amount with the RF gain. And then making it loud again with the AF gain.

There must be something happening when going from the IF to audio then benefits from having an overall quieter signal I guess. I just don't see how any of this can somehow know to attenuate static and amplify desired signal more.

Is dynamic range an issue with weak signals? I can see it being a problem with strong signals (hence the overload and need to attenuate). But with a weak signal, unless your noise floor is S9, you have enough dynamic range right?

Articles (i.e. forum posts)

https://www.eham.net/reviews/review/148466

Next, with the RF Gain still all of the way Counter-clockwise, so that the radio is nearly silent (don't move the RF Gain yet), we bring the AF Gain (volume) up to a quieted -- not quite normal listening level, where we cut hiss, but still just hear the signal clearly.

Once that AF Gain volume is set (about half for Icom 7300), you just leave it where it is, roughly, but you will twiddle it a bit up or down, to keep HISS down to lowest possible levels, as your slowly raise your RF Gain. You are now using your RF Gain as your volume control, by turning it clockwise to raise the volume, or counter-clockwise to lower the volume. While you are adjusting your RF Gain, you are looking for a happy medium between where signals are loudest, but the noise around them is quietest.

… I also twiddle with the AF Gain a little bit, if that helps the signal -- it really depends. Experiment!

With AF gain on full, while using the RF gain to control volume, and aggressive EQ settings on upper-middle frequencies, signals barely readable often rise to an easy 56 to 57.

My KX2's RF Gain, when used with either ATT or Pre-amp, and EQ, will dramatically cut hiss noise while digging out signals.

• The RF receiving system is very complex. An approach that works with any given signal, using a particular receiver, under specific conditions of propagation, with unique combinations of atmospheric, man-made and cosmic noise, might not work well with any other combination. – Brian K1LI Mar 19 '19 at 15:46
• it might be very helpful to steer the answers by linking to one of these articles! – Marcus Müller Mar 19 '19 at 16:24
• added a few to the question – Paul Mar 19 '19 at 18:04
• @Paul I tried to add the quotations from the links; didn't find anything clearly related to your question in your third link; if you could copy and paste the appropriate section yourself, I think this would help my understanding. – Marcus Müller Mar 19 '19 at 18:33
• @MarcusMüller The ones you added are pretty much the ones I'm wondering about. They all seem to imply that it improves the SNR. One going so far as to say the S meter reading goes up a unit. I'm starting to think the qrz and eham forums are full of more misinformation than I thought. How apropos when talking about poor SNR. – Paul Mar 19 '19 at 18:55

For an ideal, linear receiver, reducing the RF gain by 1 dB lowers signal and noise by 1 dB, so there would be no change in SNR.

However, receivers are not perfectly linear. This means that besides just amplifying inputs, they also mix them to some extent. So any input to the receiver which contains more than just a single frequency will generate some intermodulation products, that is, new frequency components at frequencies based on the sums and differences of the input components, and harmonics thereof.

The reason for this argument to reduce RF gain is that these intermodulation products aren't linear. For a 1 dB increase in gain, the 3rd order intermodulation products increase by 3 dB. Thus if intermodulation is contributing significantly to noise, decreasing gain improves SNR.

You can see an example in the animation below. The two central markers M1 and M2 are the "signal" inputs, two tones from signal generators. D3 and D4 are the third-order distortion products, and outside of those we also see the fifth, seventh, and ninth order products rise above the noise floor. The animation shows in each frame a gain increase of 1 dB: note how the gain of the intermodulation products increases faster than 1 dB per frame. This is because as gain is increased, the amplifier becomes less linear, and thus generates disproportionately more distortion.

By Ice Ardor - I measured the intermodulation characteristics of an amplifier myself on a spectrum analyzer, CC BY-SA 4.0, Link

Of course in practice the input to the receiver isn't just two tones in an otherwise quiet spectrum, but rather a cacophony of many signals and external noise. As a result, the intermodulation distortion isn't tidy, regular spurs such as this, but rather a raising of the noise floor. The distortion can be generated in the receiver front-end, and also in mixer stages when very strong signals can have an effect even through the filter's stop-band attenuation. Digital modes like PSK31 present a particular challenge since many signals all fall in the receiver passband, and thus can intermodulate quite strongly.

The extent to which this effect is significant depends significantly on the quality and design of the radio. In recent years, many radios have adopted direct-sampling or direct-conversion SDR architectures, and this has the effect of greatly reducing intermodulation distortion since there are very few analog stages to introduce such distortion.

In contrast, superheterodyne receivers have several intermediate mixer stages, and each one can add some distortion. A very high quality design might have very effective filtering and very linear stages, approaching the performance of a direct-conversion SDR. But a cheaper design can be quite terrible.

As an example, I have a Softrock RXTX, and a Yaesu FT-897. The former is a direct-conversion SDR which costs under \$100, and the latter is a superheterodyne receiver that cost over \$900 when it was new. Decoding FT8 with WSJT-X on the Softrock I get easily twice the decodes compared to the FT-897. The FT-897 simply generates too much intermodulation distortion.

Reducing the RF gain on the FT-897 does help, though it's never as good as the Softrock. On the other hand, the Softrock is so linear, adjusting gain makes no relevant difference because the intermodulation products are always well below the noise floor.

• Thanks for adding a demonstration of measured behavior, Phil. Rob Sherwood (sherweng.com) has quantified audio distortion and demonstrated that numerous popular, "top end" radios suffer from poor performance in this area. – Brian K1LI Mar 22 '19 at 13:54
• Phil with another killer answer. I have a follow up if that's okay to add to the original question: On an IC-7300 there is a feature called "IP+" that is designed to reduce those distortions. Since this radio is a direct-sampling SDR. And it has a specific feature to reduce IMD. Would there still be much of a benefit to lowering the RF Gain with this radio? Perhaps I should make this a separate question. – Paul Mar 23 '19 at 14:20
• @Paul Probably better to ask a separate question, and maybe someone with more experience operating that particular radio will have an answer. – Phil Frost - W8II Mar 23 '19 at 17:30

I've seen this argument before and never been entirely convinced by it either, but I think the theory is that it comes down to nonlinearity. The silicon in the amplification and detection stages of your rig is never perfectly linear. Nonlinearity at the high end is what causes overload issues, but there's also nonlinearity at the low end, where a signal that is too small is greatly attenuated or causes no reaction at all. So if you can push the noise down far enough by reducing the input gain, the noise level in the output will suddenly fall off of a cliff, "revealing" your weak signal, or so the theory goes.

However, since the signal is (in this scenario) very close to the noise, we must be adding quite a bit of nonlinear distortion to it as well, and indeed the nonlinearity must be multiplying the signal with the noise (since the gain in the nonlinear region is dependent on the sum of signal and noise), so it's not clear that the result is any better than what we started with. Perhaps, since the mixing means that even most of the distortion products we hear are somehow correlated with the signal, copyability should improve even though the absolute fidelity doesn't?

• I hadn't considered non-linearity at the low end. But that does make sense. Like you said, however, if the signal is just above the noise then it would be hard to find a spot where the noise is attenuated more than the signal. I assume that linear to non-linear "cliff" is not a clean cut line. I suspect that most people who try this method are going in with a bias that it works. They try to "dig" out a signal and it sounds good. And confirmation bias dictates it was their method that made it sound good. – Paul Mar 19 '19 at 15:37
• the nonlinearity on the lower end is typically done away with by the magic of dithering – for every time instant, the received amplitude isn't just the low signal amplitude, but biased with the random noise amplitude (which tends to be larger). By low-pass filtering after the amplification, we use the fact that the "lifting" noise is temporally uncorrelated and hence falls prey to the low-pass filter, and the correlated desired signal remains amplified. But that only works if the filtering happens after the amplification! If you start with a narrowband filter, you'll get nothing. – Marcus Müller Mar 19 '19 at 16:23

There is some systemic noise created by the front end which is proportional to its gain. So, lowering the gain lowers that systemic noise, thus raising the SNR. Of course, the degree of improvement depends on the front end and the frequency. 'Riding the RF' simply keeps the front-end gain as low as possible. You're right, it isn't a dynamic range issue.

• How strong is the systemic noise in relation to the background noise on the antenna? I would expect the antenna noise to be the issue and not the internal noise of the radio. – Paul Mar 19 '19 at 15:28
• The higher the frequency, the more systemic noise becomes a problem. In fact, many cheaper/older UHF radios are swamped by their systemic noise. It's typically never a problem for HF bands, though. You'd have to look at your receiver specs to know for sure. Or you can do your own measurement. – Digiproc Mar 19 '19 at 15:35
• I'm talking mainly the HF band. And to be even more specific: 40 meters. I will update the questions. – Paul Mar 19 '19 at 15:38
• The DSP noise reduction circuitry in my transceiver is the only thing I have found that works on 40m QRN. – Cecil - W5DXP Mar 19 '19 at 20:28
• @Cecil-W5DXP There's a lot to be said for some DSP noise reduction circuits! Which transceiver do you have? – Mike Waters Mar 19 '19 at 22:22

My understanding of how RF gain works (or does in my own SDR design) is that RF/IF gain alters the gain BEFORE the AGC mechanism and the AF gain alters the gain AFTER the AGC. For a signal that is strong enough to activate the AGC, the AF gain sets the volume of the audio.

With the AF gain set at a suitable level, lowering the RF/IF gain should reduce the background noise by stopping the AGC from lifting the noise up to the full level. Sure, lowering the RF/IF gain will NOT improve the SNR of any received signal but it can allow the operator to choose the worst case SNR that is coming out of the loudspeaker.

I set my SDR Rx so that the band noise is about 10dB lower than a strong signal - this lowers listening fatigue. If a weak signal is then received (not strong enough to activate my AGC) then it just sounds quiet - just what I want as it still reduces listening fatigue.

Tim G0ETP

• Hello Tim, and welcome to ham.stackexchange.com! – rclocher3 Aug 10 at 14:23