You measure an amplifier and the screen shows a single peak, even though you are sure there are two carriers. Or you know a harmonic is there, but the noise floor sits above it. Or you finish adjusting and the analyzer sweeps so slowly that you wait on every pass. These three situations usually share one cause: RBW.
What RBW is, in brief
A spectrum analyzer does not look at the whole frequency range at once. It sweeps a narrow filter (the intermediate-frequency, or IF, filter) across the range and plots the power at each point. The width of that filter is the RBW (Resolution Bandwidth).
Think of RBW as the width of the "window" the analyzer listens through at each step. A narrow window lets you hear each frequency separately, but you have to move slowly. A wide window moves fast, but everything inside it is merged into one.
Adjusting RBW changes three things at once
| Parameter | Small (narrow) RBW | Large (wide) RBW |
| Separating close signals | Good | Poor, peaks merge |
| Displayed average noise level (DANL) | Lower | Higher |
| Sweep speed | Slow | Fast |
| RBW | 100 kHz | 10 kHz | 1 kHz | 100 Hz | 10 Hz |
| Noise floor (relative to reference) | reference | -10 dB | -20 dB | -30 dB | -40 dB |
Five measurement situations and how to set RBW
Two carriers at 900.00 MHz and 900.01 MHz, 10 kHz apart. This is what you will see on the screen:
RBW 3 kHz ▁▃▇▃▁▁▃▇▃▁ → splits into 2 peaks
Setting: RBW less than or equal to the spacing between the two signals. To make the two peaks clearly separate, choose about one third of that spacing, which is 3 kHz in this example.
Note: if the two signals differ a lot in amplitude, the smaller one can hide under the "skirt" of the larger one, because the filter is not infinitely steep. In that case you need an even smaller RBW.
Setting: first narrow the Span around the suspected frequency, then reduce RBW step by step (for example 10 kHz, 1 kHz, 100 Hz, 10 Hz). Each step lowers the noise floor by 10 dB, so any spurious signal buried in the noise gradually emerges.
Note: if you reduce RBW but leave the Span wide, the sweep time becomes very long. Do it in order: narrow the Span first, then reduce RBW.
A 100 MHz carrier modulated by a 1 kHz signal has two sidebands at ±1 kHz from the carrier.
Setting: RBW below the modulating frequency, about one third of it or lower. For 1 kHz, choose 300 Hz or less, otherwise the sidebands merge into the carrier.
A digitally modulated signal has a spectrum like a flat block of noise, several MHz wide. You do not need an extremely narrow RBW here, but you should not let RBW get too large relative to the channel either.
Setting: RBW of about 1 to 3% of the channel bandwidth. For example, for a 20 MHz channel choose about 200 to 600 kHz. The easiest way is to use the Channel Power measurement function, which sets a suitable RBW and detector for you.
Note: an RBW that is too large relative to the channel rounds off the channel edges and affects adjacent channel power (ACPR) results.
Setting: wide Span and large RBW (about 100 kHz to 1 MHz) so the analyzer sweeps quickly. When you see a suspicious peak, place a marker on it, narrow the Span around it, and only then reduce RBW to look closely. This is the most time-efficient approach.
Read more:
RF measurement distorted or worried about burning the probe: when to switch to a signal analyzer
Quick reference table
| Measurement goal | Suggested RBW | Combine with |
| Hunting for signals across a wide range | 100 kHz to 1 MHz | Wide Span, fast sweep |
| Separating two carriers | About 1/3 of the spacing between them | Span just wide enough for both |
| Finding small harmonics and spurious signals | Step down from 10 kHz to 10 Hz | Narrow the Span first |
| Modulation sidebands | 1/3 of the modulating frequency or lower | Span of a few times the modulating frequency |
| Wideband channel power | About 1 to 3% of the channel bandwidth | Use the Channel Power function |
VBW: it smooths the trace, it does not lower the noise floor
Next to RBW there is always VBW (Video Bandwidth), a post-detector filter used to smooth the displayed trace. On many Keysight analyzers, VBW defaults to the same value as RBW (a 1:1 ratio).
When the noise floor looks jagged like a saw blade, you can lower VBW (VBW/RBW = 0.1 or 0.01) to make the trace smoother and the peaks easier to read. But keep these points clear:
Common mistakes when adjusting RBW
| 1 | Setting a very low RBW but leaving a wide Span. The analyzer seems to freeze. Narrow the Span first. |
| 2 | Ignoring the "Meas Uncal" warning. This message means the analyzer is sweeping too fast for the RBW, so the measured amplitude can no longer be trusted. Leave Sweep Time on Auto. |
| 3 | Signal skirts spreading wide at a very narrow RBW. This is usually phase noise from the source or from the analyzer itself, not necessarily a spurious signal. |
| 4 | Peaks jumping around at a very narrow RBW. This is usually caused by the source drifting in frequency, and the analyzer is not necessarily at fault. |
Choosing RBW in four steps
Step 1. Set the Span to the frequency region you need to see.
Step 2. Turn RBW to Auto. The analyzer typically sets RBW to about 1/100 of the Span, enough to balance resolution and sweep speed.
Step 3. Decide your goal (separating signals, lowering the noise floor, or measuring channel power), switch to Manual, and set RBW from the quick reference table.
Step 4. Check the result: no Meas Uncal warning and stable peaks. If the sweep is too slow, narrow the Span first, then relax RBW.
