I've been planning for a while to build a tracking generator for my spectrum analyzer, but because time seems to be a rare commodity these days, I had to look for some quick alternatives that could get the job done. I had a crack at building several noise sources, but was not too happy with the result and since I didn't want to spend too much time on it, I went ahead and purchased one of those cheap noise sources from eBay.
They currently sell for about 20 pounds and they're advertised to work well up to 1.5 GHz. I need a lot better than this, but for my current project it was enough.
When buying a noise source, keep in mind that they can't replace a tracking generator. The power of the signal generated by the noise source will be spread across the entire spectrum, so every time you reduce your RBW, the displayed signal level will also drop, since less power will make its way inside the filter. It's easy to get fooled by this, so if you do get a noise source, make sure you're aware at all times of the total power that's present at the spectrum analyzer input, so you don't exceed the maximum limits.
Working with it
In order to operate, it needs 12 V and seems to sink about 220 mA. That's 2.64 W and I believe most of it goes into the 3 gain stages and the resistors feeding them. Because of that it gets rather hot, so keep that in mind when chosing an enclosure.
The output seems to be fairly flat up until 1 GHz, after which it starts dropping constantly until close to 3 GHz and it seems to put out about 0 dBm of total power (your millage may vary).
I don't mind the limited bandwidth, but the unit I got has an issue that makes it unusable. At some point I noticed that the signal level suddenly dropped by 10-13 dB and after that it kept going back and forth randomly, which made it impossible to make any meaningful measurements. I believe that's because one of the gain stages stopped working properly (maybe because of the heat). After using it like this for a while, this variation in output power has almost stopped and now it sits at -12 to -14 dBm and only rarely jumps back to 0 dBm.
If the power drop issue is isolated to my unit, then I see nothing wrong with it, it looks like a perfectly usable noise source up until 1 GHz and it's cheap. Chances are however that it's not dissipating heat properly and that causes the behavior I observed.
Leaving all the guessing aside, if I were to make a recommendation based on the unit I have, I'd recommend that you look for a different design and try your odds with that one.