Total newbie here...

I'm researching how to detect broadband noise, and wanted to see if I have the basics correct.

Let's say the noise spectrum is wide ranging, from DC to over 100MHz, and I was interested in looking at the noise power between 47.5 and 52.5MHz (for example).

For noise detection, I would use some RF noise detection IC, TBD.

But first I should probably downmix. If I use an LO at 53.5 MHz, then the input is downmixed to 53.5-50 = 3.5MHz, start =1MHz, stop = 6MHz.

But before I downmix I should probably LPF with cutoff above 52.5 MHz, so that higher frequencies don't pollute the window I'm inspecting?
And after I downmix I should BPF between 1 and 6MHz?

My ultimate goal is to have the frequency window of interest be adjustable, which complicates how the filters and LO are designed.

 

The first question is: what type of signal do you want to measure? Noise will exist in any band (kTB noise aka "thermal noise"), but are you trying to detect if there is a signal or interference above the thermal floor within some band?

You could use a mixer to bring things down to DC, but if you are trying to measure power in your example band of [47.5, 52.5] using an LO of 53.5 MHz, you will need to be careful about IQ mismatch that will cause power from [54.5, 59.5] to image onto your band of interest at baseband (i.e. [+1,+6] MHz will image onto [-6,-1] MHz). As an example, suppose there is a strong TV broadcast in the upper 50's that can image onto your baseband of the upper 40's and lower 50's when mixing down with an LO of 53 MHz. You will need to match gain and phase in your mixer sufficiently well to avoid clobbering the band of interest with an image of the strong TV signal.

Alternatively, if you use an LO that is at the center of the band of interest (50 MHz in this case), you only image from the band of interest onto itself. From a power measurement perspective, the impact is minimal. You would likely need to remove a zero-IF-related DC spur, but otherwise this gets around the IQMM problem.

Another approach - especially if you are "only" measuring up to 100 MHz - is to sample the whole spectrum and use digital mixing and filtering to get what you want, though such ADCs or o-scopes start to get a bit pricey.