TX works at 433 MHz and sends a burst of 64 bytes every second. The burst itself lasts for 102.4 ms. The width of TX signal is 400 kHz.

RX is an Spectrum Analyzer (R&S FSL6) with a receiving antenna. RBW is 100 kHz, VBW is 300 kHz, span is 6 MHz, sweep time is 2.5 ms. "Max hold" is used to obtain the peak value of the received signal.

It is observed that (1) background noise level would rise ~10 dB once "Max hold" is selected: normal -70 dBm, max hold -60 dBm (no additional changes to SA settings) (2) If not read immediately after "Max hold" is selected, peak value would keep increasing gradually.


  1. What is the "valid period" for readings in "Max hold" regarding the burst signal? [From the comment, "Max hold" is regarded as valid eternally.]

  2. Based on the settings, did the SA sweep too fast or too slow? Could this be the reason for non-eternal "Max hold" reading?

  3. Is it normal to have the spectrum/noise level risen upon selecting "Max hold"?

  • 3
    $\begingroup$ it's a bit unclear what you're asking. You don't tell us anything about your Spectrum analyzer and how it's configured. Generally, "hold" means "valid for eternity", because that's what a maximum is. Also, classical sweeping spectrum analyzers are not, without further ado (meaning that you think very much about your parameterization and build stochastic models!), suitable for analysis of bursty signals, because you can just miss them, or just observe half of them. $\endgroup$ Mar 10, 2018 at 19:35
  • $\begingroup$ The Spectrum Analyzer is an R&S FSL6. The settings: RBW 100 kHz, sweep time 2.5 ms, target frequency 433 MHz. The peak just kept rising as the received signal level does vary a bit but sometimes the value would "grow" too much that become unreasonable. $\endgroup$
    – luw
    Mar 10, 2018 at 20:16
  • $\begingroup$ how much bandwidth do you sweep? What's bandwidth of your signal? Again, you should definitely describe your measurement setup, in detail, in all detail in your question (rather than in the comments). $\endgroup$ Mar 10, 2018 at 22:19
  • 1
    $\begingroup$ To measure transmitter power, what you want is Zero Span. Set the RBW to several times the bandwidth of your signal, and the centre frequency exactly at your signal. Set the sweep time to 200 ms, and use the video trigger to hold the display. Use the peak detector. Now you have a graph of power over time, and you won't need to use peak hold. $\endgroup$
    – tomnexus
    Mar 18, 2018 at 2:35

1 Answer 1


It is important to understand that the SA takes some finite time to sweep (calculate the FFT buckets) for the frequency range in question. It is quite possible with a short duration signal that the sweep will catch none or only a small portion of a narrow band signal on a single sweep. On another pass it will potentially catch the peak of the narrow band signal spectrum. Depending on how often and at what value the sweep catches the signal will determine if it is visible and to what amplitude.

Placing the SA in a max hold mode will cause the display to persist all maximum values from each FFT (fast Fourier transform) bucket. So now instead of a bucket briefly rising in value and then falling back to potentially noise level, it remains at its highest peak as long as the max hold mode is engaged. Eventually the entire spectrum and amplitude of the periodic signal will be persisted on the display as long as the max hold mode is engaged. The effect can be that the amplitude and bandwidth of the signal appears to grow over time. It will eventually stabilize in shape and amplitude.

Noise is detected in the same manner using max hold since noise is just another signal. Noise is generally random in nature so a simple single sweep may not catch it. When the SA is in max hold mode, it eventually finds a random noise event in each FFT bucket. This raises the apparent overall noise floor. In the UHF spectrum this noise is generally the result of thermal component noise and external RFI (radio frequency interference). RFI, in contrast to thermal noise, is often repeatable and identifiable.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .