Noise Figure Measurement [Gain Method] 

Nice video. You really need to terminate the input of that amp with 50 Ohms, as it will make a noticeable difference in your measurement; notice it is -141 dBm/Hz in this video but -138 dBm/Hz in your other video with the noise source off. Also, if the MDO uses digital down conversion to compute the spectrum then you are OK since it is taking the log of the average; i.e. it's all DSP following linear down conversion. Although in a typical spectrum analyzer it is taking the average of the log (a logarithmic amplifier), which is why they use a special noise marker mode, since white noise is statistically different than a sine wave, and there is a 1.4 dB difference between the two after logarithmic amplification. Anyway, nice videos.

Just noticed you were doing trace averaging though, which is interesting, since that is equivalent to averaging the log instead of log of the average. This would give the 1.4 dB error even with a signal analyzer such as the MDO. Need to reduce the VBW instead, which would average prior to the log.

Hi Sebastian, nice video showing the basic "Gain Method". Many modern spectrum analyzers offer very low displayed average noise levels and very small resolution bandwidths which are very helpful when using the "Gain Mehod". My spectrum analyzer provides a resolution bandwidth of 1 Hz when the frequency span is small which really makes this method a breeze, and the displayed average noise level is very low when running a 1 Hz resolution bandwidth.

I thought this was great! Thank you very much for a clear explanation of this procedure.

The best NF video on RU-vid. Thank you!

very helpful information. the MDO also should have signal path compensation preformed before a measurement like this.

How can you compare gain at 9V with NF at 12V. Both of them should be calculated at same voltage!

If you use kTB as input noise, the input has to be terminated with 50 ohm

Hmmm, is it necessary to account for the noise figure of the spectrum analyzer itself?

thank you very much

how to calculate noise when my output from amplifier is not a straight line as in video but have crest and troughs over the spectrum. where should I measure the noise. you have chosen the marker at 10MHz since my response is not a straight line where should the marker be? why is the span arbitrary, shouldn't it be equal to bandwidth of the amplifier? what does resolution bandwidth do and how do you choose a value?

On Amp datasheet we look at 9V and 10MHz, on thermal worksheet we use the 10kHz number and then after all the math we compare it to the 12V noise figure. I like the concept here but it seems like we didn't use all the right info to analyze actual noise figure

In my day, spectrum analysers typically had an inherent noise figure of about 25 dB. Unity gain, but NF 25 dB. So, with the input of the SA input terminated and a RBW of 10kHz, the display of the SA would typically be -134 plus NF of 25 = -119dBm . The gain of your DUT is lost in this. I don’t think this methodology is appropriate.

Would I be able to see -174dBm on spectrum analyzer when nothing to connected at the input? What is the setting? Setting RBW to 1 Hz, what about Video BW?

Million likes for that video straight to the karma!

Is the measured NF lower because the MDO's NF was not accounted ?

The real gain of DUT should be reduced by cable loss, about 0.5dB

Nice explained #discreteoptical

But leaving the input port open the noise figure is not minimized. According to my university lectures reminds there is an optimum source impedance which minimize the noise source. It is correct? Thanks!

video about stereo multiplex from scratches still waiting aI

it's somewhat strange, i am curious what about the sa that cannot have ktb measurement results. for 10kHz, it should be -134 dBm, but sa shows -89dBm without any DUT connection.

good video

Did you measure the noise spectral density of your spectrum analyzer before connecting the amp?

is it possible to meassure Scattering-Parameters with spectrum analyzer ?

Measuring NF is not this easy....without feeding actual signal you are just measuring the noise floor with amplifier on and off. But there is lot of stuff happens when you actually feed in signal as biasing will effect noise performance, your non linearity will and most of all this is not actual platform which noise will follow when the amplifier is in operation. Feeding through 50 ohm makes a lot sense coz you need antenna while feeding in the signal and how its input matching could deteriorate the noise performance you can see in simulations. That is the reason we measure noise figure with SNRinput/SNRoutput ratios.