A great read from the Elecraft list:
Receivers are always ranked by the "2 kHz third order dynamic range", such as at: http://www.remeeus.eu/hamradio/pa1hr/productreview.pdf but do we really grasp the meaning of these specs? For instance, the Elecraft K3's (after synthesizer upgrade) number is 103 dB, good enough to be in the top ten. In fact, this number is so strong that very few hams will ever be affected by it. To the best of my knowledge, I have *never* been close to running out of dynamic range. To understand why, let's put "103 dB" into English.
Let's say you're on 20 meter CW, operating at 14.050 MHz.
You're listening through your fine Elecraft 500 Hz crystal filter when suddenly, and by incredible coincidence, two equally strong 49 dB over S9 signals begin transmitting at the exact same time, one on 14.052 and the other at 14.054 MHz, exactly 2 kHz and 4 kHz up from where you're listening. With the preamp off (which is totally believeable on 20 meters with a decent antenna) you will just barely hear a "ghost signal" right at the noise level... if you notice it at all. That "ghost" signal is the two-tone, 3rd order intermod product generated in the K3 receiver by those two hugely strong and perfectly placed signals.
Not a very likely scenario, but that's what 103 dB of dynamic range buys you.
I have assumed a noise floor or MDS of -130 dBm because it's a nice round number. If your 20 meter noise floor is higher than this, then the two signals would have to be *even stronger* to hear the intermod come out of the noise.
Even if each of those interferers was *60* dB over S9 -- pegging the S-meter-- the intermod product on 14.050 would still be only S5. Amazing. This kind of performance begs the question, "How much more dynamic range is really needed?" and some (like Rob Sherwood) have said that once you're above 90 dB, you already have enough, at HF at least.
Perhaps it's time to rank receivers by a different measurement, something that affects more of us. Looking through the table at the link above we see another measurement called "2 kHz blocking gain compression" and for the same K3 it is 143 dB. This is a measurement not of two interfering signals, but a single interferer just 2 kHz away. Since there's only one signal, it won't generate a "ghost", but it will reduce the gain of the receiver. ARRL defines this as the signal level that reduces the gain by 1 dB. One dB is really small, something like changing your RF Gain knob from the 3:00 o'clock position to maybe the 2:45 o'clock position. Barely noticeable. Nonetheless, for our K3 the signal required to do this is about +13 dBm, or 20 milliwatts, which is probably near the damage level of the receiver! (I'm quite sure that Wayne has made intercontinental QSOs at 20 mW.) It's a theoretical value that very, very few hams would ever encounter... only the ones living next door to a guy running a kilowatt. So this measurement is even less relevant to us.
Finally, we notice a measurement called "2 kHz reciprocal mixing dynamic range" -- probably the limiting spec nowadays for top tier receivers. In our example of the single strong signal, way before reducing the gain of the receiver, that signal will have another effect: it will mix with the phase noise of the K3's own local oscillator and deposit that phase noise right onto your desired frequency of 14.050 MHz. As you're listening there, you suddenly notice that the noise floor seems to be rising for no apparent reason. You listen some more, and notice that the noise is following some kind of CW keying. You glance at your panadapter and notice an enormous signal just 2 kHz away on 14.052. So there are two culprits: that strong signal, and the K3 oscillator phase noise. The K3 with upgraded synths has a spec of "-115 dBc", again near the top of the list, which means that a signal 2 kHz away and 115 dB above the noise floor will cause the noise floor to rise by 3 dB. For a K3 noise floor of -130 dBm this is -15 dBm, or about 60 dB over S9. The reason I say this is the limiting factor is because the chance of just one 60 dB over S9 signal nearby is greater than *two* of them at the right spacing as in our discussion of 3rd order DR.
For these reasons, we could start ranking receivers by 2 kHz reciprocal mixing dynamic range because reciprocal mixing is far more likely to happen to a larger number of hams. It's not a catastrophic effect, but it's quite noticeable. There's a problem, however, because sampling receivers don't follow the classical reciprocal mixing model. We need a measurement that hasn't been invented yet to compare modern receivers. Maybe we could simulate the worst-case contest by applying thousands of signals and noise to the receiver and seeing how much junk is generated to cover up the signal you're trying to copy at 14.050, something kinda like the noise power ratio test. But in order to compare apples to apples, the exact same test conditions must be used on every receiver, regardless of architecture.
Finally, the general unlikelihood of any of these conditions occurring also convinces me that other ergonomic factors -- not necessarily measureable-- should be considered when choosing a receiver. 'Usability' (whatever that means to you) is underrated. So is 'listenability' -- again, a very subjective term. For instance, I have come to believe that the AGC system in a receiver has everything to do with how it "sounds" and explains why receivers with similar specs can sound so different from each other. This area needs further study.
I hope that this helped turn mere numbers into an actual operating reality.