Some sources of measurement error give errors which are correlated between one measurement and the next; some sources of measurement error are uncorrelated from measurement to measurement. This second type of measurement error is usually referred to as noise.
All measurements have some noise. If your measurement uncertainty is due to noise, then averaging will reduce the measurement uncertainty, typically making your measurement more accurate by the square root of the number of averages you take.
Measurements also have other sources of error - thermal drift in a measurement setup, impedance mismatches at connectors, unwanted spurious signals. It is usually possible to measure and mathematically calibrate out such errors; in practice, people calibrate out the largest sources of error, and treat the rest as measurement uncertainty. If your measurement uncertainty is dominated by non-noise error, then averaging won't help.
For a quick check whether something is working, you can just look at the output on a spectrum analyzer. In continuous sweep, adjust the averaging, Resolution Bandwidth, and Video Bandwidth to get the noise to an acceptable level. Use good quality instrumentation and a simple measurement setup to minimize the effect of other errors.
For a scenario where you are guaranteeing performance to a customer, you will need to be much more rigorous. Consider all of the potential sources of error in your measurement setup, and analyze how each of those errors will affect the end measurement result (this is referred to as building a measurement equation). Your measurement equation will allow you to estimate the measurement uncertainty, and to see the contribution of noise to your measurement uncertainty.
If your measurement uncertainty is too high, the measurement equation is also useful as you figure out where to invest to improve your measurement. Typically, non-noise errors can be reduced by simplifying the signal paths, by reducing nearby interference, by investing in equipment which will minimize measurement errors, and by developing algorithms to calibrate out errors.
All measurements have some noise. If your measurement uncertainty is due to noise, then averaging will reduce the measurement uncertainty, typically making your measurement more accurate by the square root of the number of averages you take.
Measurements also have other sources of error - thermal drift in a measurement setup, impedance mismatches at connectors, unwanted spurious signals. It is usually possible to measure and mathematically calibrate out such errors; in practice, people calibrate out the largest sources of error, and treat the rest as measurement uncertainty. If your measurement uncertainty is dominated by non-noise error, then averaging won't help.
For a quick check whether something is working, you can just look at the output on a spectrum analyzer. In continuous sweep, adjust the averaging, Resolution Bandwidth, and Video Bandwidth to get the noise to an acceptable level. Use good quality instrumentation and a simple measurement setup to minimize the effect of other errors.
For a scenario where you are guaranteeing performance to a customer, you will need to be much more rigorous. Consider all of the potential sources of error in your measurement setup, and analyze how each of those errors will affect the end measurement result (this is referred to as building a measurement equation). Your measurement equation will allow you to estimate the measurement uncertainty, and to see the contribution of noise to your measurement uncertainty.
If your measurement uncertainty is too high, the measurement equation is also useful as you figure out where to invest to improve your measurement. Typically, non-noise errors can be reduced by simplifying the signal paths, by reducing nearby interference, by investing in equipment which will minimize measurement errors, and by developing algorithms to calibrate out errors.