This website uses cookies to implement certain functions. If you use this website you agree to our Privacy Policy.
News and Information about the Test of Electronics in Research & Design, Production, Maintenance, and Installation.  

Newsletter

Register to our newsletter
Every two weeks -
all news at a glance
captcha 

Latest Test and Measurement News

Background: Optimize Spectrum Analyzer Settings to improve Sensitivity

29 January 2014 - In many cases, the primary use of a spectrum analyzer is to measure very low-level signals. This may possibly be known signals that need to be characterized or unknown signals to be discovered. In either case, an understanding of how best to improve the sensitivity of a spectrum analyzer can aid this process. In this article we will discuss the optimum settings for measuring low-level signals. We will also discuss how to use noise corrections and noise floor extensions to maximize the instrument’s sensitivity.

Displayed Average Noise Level and Noise Figure

The figure-of-merit for the sensitivity of a spectrum analyzer can be found in the specifications guide for the instrument. It will be listed as either Displayed Average Noise Level (DANL) or Noise Figure (NF). The DANL specification is the amplitude of the noise floor of the spectrum analyzer over a given frequency range with the input terminated in a 50 ohm load and 0 dB input attenuation. The specification is normally given in units of dBm/Hz. In most cases, the averaging that is done is on a log scale. This causes a reduction in the noise floor by 2.51 dB. This reduction in the noise floor is what differentiates DANL from NF as we will find out further into this discussion. As an example, for a -151 dBm/Hz DANL specification in a 1 Hz Resolution Bandwidth (RBW), you should be able to lower the spectrum analyzer noise floor to at least this level using the settings provided in the specification. Incidentally, a continuous wave (CW) signal that is at the same amplitude of the noise floor of the spectrum analyzer will measure 2.1 dB above the noise floor due to the summation of the two signals. Similarly, a noise-like signal will appear 3 dB above the noise floor.

The noise floor of the spectrum analyzer has two components.  The first is the Noise Figure (NFSA) of the spectrum analyzer and the second is thermal noise energy. The amplitude of the thermal noise energy is given by kTB, where:

k = Boltzmann’s constant (1.38 x 10–23 joule/°K)

T = temperature, in degrees Kelvin

B = bandwidth in which the noise is measured, in Hertz

This is the amount of thermal noise energy present at the input of the spectrum analyzer when you terminate the input into a 50 ohm load. In most cases, the bandwidth is normalized to 1 Hz and at room temperature 10LOG(kTB) is calculated to be -174 dBm/Hz. The DANL specification in a 1 Hz RBW is then given by:

     DANL = -174 dBm/Hz + NFSA - 2.51 dB                      Equation 1

Also

    NFSA = DANL + 174 dBm/Hz + 2.51 dB                       Equation 2

*Note: If power RMS averaging is used in the DANL specification, the 2.51 dB term can be omitted.

The -151 dBm/Hz DANL specification would equate to a NFSA of 25.5 dB.

 

Settings that Affect the Sensitivity of a Spectrum Analyzer

The gain of a spectrum analyzer is unity. This states that the display is calibrated to the input port of the spectrum analyzer. So if a 0 dBm signal is applied to the input, the measured signal should be 0 dBm plus or minus the accuracy of the instrument. This is important when we discuss adding input attenuation or preamp gain to the spectrum analyzer. Increasing input attenuation causes the spectrum analyzer to increase an equivalent gain in the IF stage of the instrument to maintain a calibrated level on the display. This in turn raises the noise floor by an equivalent amount on the display therefore maintaining the same signal-to-noise ratio. This applies to external attenuation as well. You will also need to compensate for a RBW greater than 1 Hz by adding a 10LOG(RBW/1 Hz). These two terms allow us to determine what the spectrum analyzer noise floor will be at different settings of attenuation and resolution bandwidths.

     Noise Floor = DANL + Atten + 10LOG(RBW)                 Equation 3

 

Adding a Preamp

An internal or external preamp can be used to improve the noise floor of the spectrum analyzer. Typically, there will be a second DANL specification that covers the internal preamp case and the previous equations will apply as well. If an external preamp is used, you can calculate the new DANL by using the cascading noise figure equations, where the gain of the spectrum analyzer is unity. If we consider the system as a preamp and a spectrum analyzer we now have:

      NFSYS = NFPREAMP + ((NFSA-1) / GPREAMP)                 Equation 4

Using our above example of a NFSA of  25.5 dB and a preamp with a gain of 20 dB and a preamp noise figure of 5 dB, we can determine the overall noise figure of the system. We need to first convert the values into power ratios and then log the result:

      NFSYS = 10 LOG (3.16 + (355/100)) = 8.27 dB            Equation 5

Equation 1 can then be used to determine the new DANL with an external preamp connected by simply replacing the NFSA with  the NFSYS calculated in Equation 5. The preamp reduced the DANL significantly in our example from -151 dBm/Hz to -168 dBm/Hz. However, there are tradeoffs. Preamps tend to have significantly poorer non-linear distortion characteristics and compression point that limit the ability to measure large signals. This is where an internal preamp is more useful because it can be simply switched in and out as the measurement needs change. This is especially true in an automated test environment.

So far we have discussed how the RBW, attenuation and preamp can be used to improve the sensitivity of a spectrum analyzer. In most modern spectrum analyzers, there are methods for measuring the noise floor of the spectrum analyzer and then correcting for this in the measurement results. This technique has existed for many years.

 

Noise Corrections

When measuring a Device Under Test (DUT) with a spectrum analyzer, the spectrum that is observed is a culmination of the kTB, NFSA and the DUT input signal. If we disconnected the DUT input signal and terminated the input in 50 ohms, the new spectrum would then be the sum of kTB and NFSA. This trace is the noise floor of the spectrum analyzer. In general, noise correction is a process in which you measure the noise floor of the spectrum analyzer with a considerable amount of averaging and store this to a Correction Trace. You then connect the DUT input signal to the spectrum analyzer and measure the spectrum and put these results into the Measurement Trace. The correction is applied when you subtract the Correction Trace from the Measurement Trace and then display the results in the Resultant Trace. This Resultant Trace is the DUT Input Signal with the excess noise removed.                                           

               Resultant Trance = Measured Trace - Correction Trace               Equation 6
                                
= [DUT Input Signal + kTB + NFSA] - [kTB + NFSA]
                                 = DUT Input Signal

 *Note: All values are converted from dBm to mW before subtracting. The resultant trace is displayed in dBm.

 This process allows you better observe low-level signals as well as making more accurate amplitude measurements as the error from the contribution of the spectrum analyzer noise floor has been removed.

 Agilent-Optimising-Bild1

       Figure 1. Noise Corrections performed with trace math

Figure 1 displays a relatively easy method for performing noise corrections that can be done with trace math. You first average the noise floor of the spectrum analyzer with the input terminated. Save this trace to trace 1. Connect the DUT and capture this signal and save this to trace 2. Trace math can then be used to perform a power subtraction of the two traces and place the results in trace 3.  As you can see, noise correction shows the most benefit when the input signal is close to the noise floor of the spectrum analyzer. Larger signals have a much lower contribution of noise and the correction has little to no effect.

The main issue with this approach is that you have to disconnect the DUT and connect a 50 ohm load whenever you change a setting. A method for measuring the Correction Trace without removing the DUT is to increase the input attenuation (let’s say 70 dB) to raise the spectrum analyzer noise floor far above the DUT input signal and saving this to the Correction Trace. The Correction Trace will now contain:

     Correction Trace = DUT Input Signal + kTB + NFSA + Atten  Equation 7

If kTB + NFSA + Atten >> DUT Input Signal  we can omit the  and we can state that

        Correction Trace = kTB + NFSA + Atten                            Equation 8

 By subtracting the known attenuation from equation 8, we can get back the original Correction Trace we used in the manual method.

        Correction Trace = kTB + NFSA                                       Equation 9

  The issue with this process is that the Correction Trace is valid only for the current settings of the instrument. Changing settings such as center frequency, span and RBW will invalidate the values stored in the Correction Trace. A better approach is to know the specific at all frequency points and then apply the Correction Trace for any setting.

 

Noise Floor Extensions

The Agilent N9030A PXA signal analyzer provides a unique capability called Noise Floor Extensions (NFE). The PXA Signal Analyzer’s noise figure across the entire frequency range of the instrument is measured at the time of manufacturing and calibration. This data is then stored within the instrument. When the user turns on NFE, the instrument calculates a Correction Trace based upon the current setting of the instrument and the stored noise figure values. This eliminates any need for measuring the noise floor of the PXA, as was done in the manual procedure. This greatly simplifies the use of noise corrections and eliminates the excess time needed for measuring the noise floor of the instrument whenever a setting has changed.

Agilent-Optimising-Bild2

     Figure 2. Agilent N9030A PXA signal analyzer with Noise Floor Extension

In any of these methods described, the thermal noise kTB is subtracted from the measurement trace as well as the NFSA so you can in fact get results below kTB. These results can be valid in many cases but not all. This occurs when the measured values are very close or equal to the noise floor of the instrument. In fact, if they are equal, the result would be –infinity dB. Practical implementations of noise corrections normally include a threshold or a graduated level of subtraction close to the noise floor of the instrument.

 

Summary

We have examined some of the techniques for measuring low-level signals with a spectrum analyzer. We have found that the sensitivity is affected by the RBW, attenuation and use of a preamplifier. Noise reduction methods such as noise corrections and noise floor extensions can be applied to further enhance the sensitivity of the instrument. In practice, ensuring that external path losses to the spectrum analyzer are reduced can aid tremendously.

 

Additional Information

More information about spectrum analyzer measurements and analysis can be found in the following application notes:

“Agilent Spectrum and Signal Analyzer Measurements and Noise, Application Note 1303”, http://cp.literature.agilent.com/litweb/pdf/5966-4008E.pdf

“Using Noise Floor Extension in the PXA Signal Analyzer”,
http://cp.literature.agilent.com/litweb/pdf/5990-5340EN.pdf

“Spectrum Analyzer Basics, Application Note 150”,
http://cp.literature.agilent.com/litweb/pdf/5952-0292.pdf

 

Author:  Bob Nelson, Agilent Technologies, Inc.

 



Related Articles:

Upcoming Events

Embedded World 2024
Nuremberg (Germany)
09 to 11 April
Control 2024
Stuttgart (Germany)
23 to 26 April
Automotive Testing Expo Europe 2024
Stuttgart (Germany)
04 to 06 June

  More events...
  See our Trade Show Calendar
  Click here

 

Advertising
Advertising