marcusl said:
How are these observatories able to measure the CBE without big spurious signals from stars and galaxies. Don't they emit microwaves?
For the most part, they're just too dim. In the frequencies where the CMB is brightest, we can only barely see the dust in our own galaxy (except when looking directly at the center). Andromeda is also hard to see in these frequencies, and that is the galaxy closest to us.
There are, however, radio quasars that are somewhat bright across all frequencies, and the brightest of these are simply masked out.
That said, the small contributions from billions of galaxies out there do add up, and do contaminate the power spectrum at small scales. These aren't subtracted in the maps, but their level of contamination is estimated from the data (basically, the statistics of these contaminants, and thus the shape they add to the power spectrum, is very different from that of the CMB).
Anyway, you can see more detail here:
http://arxiv.org/abs/1303.5075
You can see the masks they used on page 5, where the black regions are the places where they don't use data. You can see that they don't use the data from our own galaxy, or for a series of point sources away from the galaxy.
The clearest plot of the effect of the foregrounds is found on page 12, where you can see that at very small angular scales (high \ell = small scales), the signals at the different frequencies start to go above the CMB line. Note that different frequencies are impacted by different amounts. This extra boosting of the signal at small scales is then subtracted for the final power spectrum estimate.