1. The noise figure (NF) for the formulas is the *actual* displayed noise level given your RBW (& other) settings?
Yes, of course!
The spec sheet gives -141 dBm (typ.), but only with att = 0 dB, RBW = 1 Hz, sample detector, trace average > 50, TG off. In practice, very few people use a RBW of 1 Hz & >50 trace average as it's so slow and not always needed. Increasing the RBW and decreasing the trace average will speed the refresh up but increase the noise floor. I think turning the TG on also adds some noise, which is why Siglent turned it off to get the best DANL for their spec sheet.
Well, it's always a tradeoff. If you need to measure very low distortion levels, then you will have to choose settings that provide the highest sensitivity, i.e. the lowest noise floor. For the usual harmonics of a transmitter, in the range up to -60 dBc, it should not be much of a challenge and you can use wider RBW. Practice will tell you what works.
Never use the preamplifier for distortion measurements, as it will lower the noise floor, but at the same time the intermodulation intercept points will drop by an even bigger margin, i.e. der maximum SFDR (Spurious Free Dynamic Range) will be reduced quite a bit.
Trace average does not reduce the noise floor DANL, it is just an optical aid. The noise level as well as low level signals are much better readable and distinguishable if the trace is averaged. Even low numbers (like 4 times averaging) is sufficient to get a decent effect and 50 times doesn't bring a huge further improvement in my experience.
The tracking generator TG is of course an additional source of spurious signals, so it's better to leave it turned off.
2. Is 'Second harmonic distortion (SHI)' the same thing as 2nd Order Intercept (SOI)? Siglent have an entry for SHI but not for SOI in their spec sheet (see my post above for attachment). If not, can anyone explain what Siglent are talking about in the spec sheet when they quote -65 dBc / +45 dBm (nom.) for the SSA3000X-Plus?
I could write a Novel about this, but I'll try to keep it short and simple. Either way, everyone using an SA should be familiar with these fundamentals…
2nd harmonic distortion is the amplitude of the 2nd harmonic for a given level of the fundamental signal. This is measured in dBc (Decibel below carrier). -60 dBc means that the 2nd harmonic is just 1/1000 of the fundamental, or in other words: 2nd order harmonic distortion is 0.1%.
SOI is the 2nd order Intermodulation Intercept point. To be more precise, we have to distinguish between input and output. We're generally dealing with input intermodulation intercept points here. The specification of SOI is rather uncommon, in practice you'll mostly only see the TOI (3rd order intermodulation intercept point). Why do we have that at all and do not just specify a 3rd order harmonic distortion instead?
First of all, 2nd order intermodulation products are the mixer products of two (first order) fundamental waves. 3rd order intermodulation products are the mixer product of the 2nd order harmonic of one signal with the fundamental of the other signal. In intermodulation intercept point measurements, both signals are the same level, and the 3rd order TOI is of special interest because it is the main cause of phantom signals in receivers. Other than the 2nd order intermodulation products, which are at very different frequencies as the individual fundamental signals, the 3rd order intermodulation signals appear in close proximity to the wanted signals, hence cannot be filtered out.
So far I should have explained why specifying the third order intermodulation is at least as important as the harmonic distortion, even though both are correlated in some way. A frontend/mixer with low harmonic distortions will produce low intermodulation distortions as well. The beauty of numbers like TOI is that you have all the information in a single number, whereas for harmonic distortion, you always need to specify the conditions - at least the fundamental signal level in addition to the distortion figure. On the other hand, if I tell you some receiver (= SA frontend) has a TOI(i) of +30 dBm (a very good one!), then you immediately know that an input level of -10 dBm will result in 3rd order intermodulation products of -90 dBm = -80 dBc.
Regarding the Siglent specs, the most recent data sheet does not contain the +45 dBm figure.
It shoud be clear by now what -65 dBc means in terms of 2nd harmonic distortion.
Maybe the +45 dBm were meant as a hint on the SOI – +45 dBm sounds at least plausible for this.
3. 1dB gain compression is quoted at > -5 dBm (nom.) given fc ≥ 50 MHz, att = 0 dB, preamp off. Given this is so much higher than the -30dBm level discussed above, Can you think of a scenario where you'd want to drive the input this hard?
If you don't need to drive the input this hard then you wouldn't as a smaller signal would yield less amplitude compression. You'd just add attenuation before the SA, but in doing so you'd give up any very small signal detail as it'd be lost int he noise.
Of course there are a few scenarios where the distortion of the SA frontend isn't a problem. Think of phase noise measurements for example. Since the CP already has 1 dB amplitude compression, we generally don't want to drive the input that hard indeed. I recommend to stay at least 10 dB below the CP.
As already stated, it's always a tradeoff between linearity (low distortion) and high S/N ratio. Most of the times low distortion is paramount, but in some other cases, low noise might be more important. This is also when you use the PA (not power amp, in an SA this is the preamplifier) – only in situations when the input signal is so weak that it doesn't come close to the CP even with PA turned on.
In case of an SA with only -5 dBm CP, the TOI will be correspondingly lower as well. And indeed, the TOI of the SSA3000X is only +10 dBm according to the data sheet. Siglent is obviously favoring high sensitivity over high level handling. But it is always better and easier when you have to use (additional) attenuators instead of preamplifiers. And again, there's no need for external attenuators, as long as the maximum allowable input power is not exceeded and, quite obviously, the maximum attenuation of the internal attenuators is sufficient for the task.