I have to make measurements of the RF power which a circuit is delivering, and I am using an RF spectrum analyzer for that purpose. I am a beginner in this topic, so I apologize if the question is too silly.

So, I am using a 30dB attenuator before the Spectrum Analyzer input in order to ensure that the power levels entering the instrument is well within the design limits of the device.

When I obtained the RF power spectrum, I noticed that there was a setting in the Spectrum analyzer called 'Attenuation', which I could change by using the various knobs provided in the control. When I made the measurement, the 'Attenuation' setting was on 10dB.

So, my question is, have I done a mistake by using the 10dB attenuation setting when I was actually using a 30dB attenuator externally? Or does the attenuation setting mean that an additional 10dB of attenuation is provided inside the spectrum analyzer?

To sum it up: So, using my current setup, with a 30dB external attenuator and a 10dB 'Attenuation' setting in the oscilloscope, if I read a power level of 'x' at a frequency of interest, what is the actual absolute power which is being measured? Is it x+20dB? Or x+40dB?

Also, can you please explain what the attenuation setting in a Spectrum Analyzer does?

  • 3
    \$\begingroup\$ The att. setting is the same as you adding an attenuator at the input but if you add an attenuator of 10 dB you will read 10 dB less power on the SA. If you set the SA to attenuate 10 dB more then it will compensate the reading. So you don't have to subtract the 10 dB, the SA does it for you. Only if you have a very large signal, larger than the SA can handle (like more than +30 dBm) then you need an external attenuator to bring your signal down to below +30 dBm. Otherwise (your signal is less than +30 dBm), you do not need an external attenuator. \$\endgroup\$ Dec 17, 2017 at 21:39
  • 5
    \$\begingroup\$ What you should do: connect a signal generator ouput directly to the SA input and set the generator power to 0 dBm. Now play with the SA att. setting. Observe the signal power: it remains 0 dBm whatever the att. setting. Now add a 10 dB attenuator at the SA input and play again. Now you get - 10 dBm. Observe and learn. It's not hard. \$\endgroup\$ Dec 17, 2017 at 21:41
  • \$\begingroup\$ Thank you for your replies. I don't have a reliable way to take the function generator output to the SA, hence I asked the question. Shall I close the question? \$\endgroup\$
    – Harsha
    Dec 18, 2017 at 3:22
  • 1
    \$\begingroup\$ @Bimpelrekkie: If you're going to write an answer -- especially one that spans multiple comments -- please do it in the "Your Answer" section below. \$\endgroup\$
    – Dave Tweed
    Dec 19, 2017 at 0:48

5 Answers 5


Feynman gave a great answer (reference level for external attenuation and the noise floor increasing by adding attenuation). However, the reason for the increased noise floor isn't the attenuator. Enough attenuation will attenuate your noise down to -174dBm/Hz (at 290K). The spectrum analzyer's attenuator causes an internal increase in gain at the IF stage after the mixer and amplifies the device's noise floor (regardless of the input). A good explanation is given on page 8 here: https://reeve.com/Documents/Noise/Reeve_Noise_6_NFMeasSpecAnalyz.pdf

Example: If you run -100 dBm/Hz of noise through a 10 dB attenuator, then you get -110 dBm/Hz. Attenuators at room temp always bring the RF power down to 290K (or up to 290K for measurements below 290K). Good explanation here: http://www.ittc.ku.edu/~jstiles/622/handouts/Noise%20Figure%20of%20Passive%20Devices.pdf

As long as you're above the noise floor of the spectrum analzyer (for a signal or noise measurement), the spectrum analyzer's attenuator setting shouldn't change the measured value. Try setting a spectrum analyzer to 0 span, RBW 100kHz, VBW 100Hz and noise marker on (so you get dBm/Hz). See what your noise floor is. You can adjust the attenuator setting to see what the noise floor does. Then hook up 3 cascaded 20 dB amplifiers (with a terminated input to get just noise) and inject that into the spectrum analyzer. If their -114 dBm/Hz noise input is above the noise floor, you shouldn't see that measurement change with the spectrum analyzer's attenuation setting until you get below the noise floor (you might need a wider RBW to compensate for the less amplification... you'll get 3 dB more noise every time you double the bandwidth).


This answer is for the question in the title.

It is based on Agilent's presentation that can be found here: https://ewh.ieee.org/r5/denver/sscs/Presentations/2012_10_Agilent1.pdf

Attenuation setting is a control for a step attenuator right after your analyzer's input. It simply reduces your signals amplitude that is fed into spectrum analyzers mixer. Use it when working with large signals to prevent mixer gain compression and distortion due to high-level and/or broadband signals.

However, when working with smaller signals, you generally want the least amount of attenuation possible. This is because your analyzer's internal noise is generated after its mixer, so lowering your useful signal level (but leaving the noise level the same) lowers your SNR (signal to noise ratio). When taken to the extreme, your signal may become smaller than noise, so that it can no longer be observed on your screen.


Yes, the "attenuation" is an attenuator internal to the spectrum analyzer. The main purpose of the attenuator is to protect the input mixer. But it will also increase the noise floor (as with any attenuator). The analyzer also adjusts the y-axis scaling accordingly. So the analyzer shows the signal level that is applied to the instrument's input.

For your analyzer to show the actual power being measured in front of your external attenuator, you need to tell it that you have 30 dB external attenuation connected. Usually this feature is called "reference level offset" or something similar. The reference level offset is an arithmetic level offset only. It shifts the measured levels by a certain amount and changes the scale of the y-axis accordingly.

So when your signal in front of your external attenuator is -10 dBm, the analyzer will show -10 dBm.

You are also right about starting with high external attenuation to protect the analyzer (internal attenuation is easily lost or forgotten). If you find out your noise floor is too big, you can always decrease the total attenuation or maybe even add a pre-amplifier.


It depends on the device a bit.

For R&S analyzers, the level before the mixer should be -10dBm, so with the default reference level after preset at 0 dBm, the automatically selected attenuator setting is 10 dB.

Also, impedance matching is a bit dependent on input path -- the termination in the mixer diode is optimized for frequencies below 50 kHz (where you'd think about using DC coupling), so for higher frequencies I'd use the internal attenuator to get better matching (although if you're interested in power only, that isn't necessarily as important).

You need to add the external attenuator to the measured level yourself, but the internal attenuator should already be compensated for.


Attenuator setting in SA is automatically calculated out of the display. External attenuator obviously cannot.

Hence if your SA display shows (say) -20dB, then with an external attenuator of (say) 10dB, the actual power is (-20dB + 10dB) = -10dB.

For your explicit example, the correct value is X+30 dB


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.