Can a Noise Figure Meter such as the old HP 8970B be used to measure the noise generated by a power amplifier in the same way as a LNA's Noise Figure is measured. I want to know how much spectral noise is generated by a PA with no input signal.


Thanks in advance,

John

Yes, you can measure any amplifier. (Converters also)

Check the output power of the power amp with a 50 ohm load on the input to be sure you won't damage the NF meter. Measuring the power amp with no input should work. The exception to this would be class C amplifiers which won't turn on without a strong enough signal at the input.

If the gain is high or just to protect the NF meter you can add an attenuator after the amplifier. On the 8970, use 34.4 Special Function, "Enter loss after DUT" to compensate for the attenuator.

Joe thanks for the reply and sorry for the delay. If I'm trying to calculate the kTB power in dBm then do I need to place a filter on either the input or the output of the PA to simulate the operating BW? I'm only operating in a narrow bandwidth compared to its total operating frequency range.

Thanks again,

John

You don't need to add any filters. The reading is a spot noise figure at individual frequency. I am not sure what the internal bandwidth is for the 8970B, but there are internal filters.

I would measure several frequency points, depending on your operational bandwidth.

You can calculate the noise power using the operating bandwidth. You want noise figure and gain measurements for use in the noise power equation. These should be bandwidth independent.

Joe thanks again for responding.

Doesn't Noise Power equal;

NP=10Log(k) + 10Log(T) + 10Log(B) + NF + Gain ?

Based on above I can calculate a NP in a 1Hz Bandwidth but I'd also like to determine the Noise Power in dBm in it's operating Bandwidth. If I don't have an operating bandwidth filter after my Unit Under Test (UUT) then the Bandwidth is either the bandwith of the instrument or the bandwidth of the UUT which ever is less.

Am I correct in this assumption?


John

John,

The equation you have shown will give you noise power in dBm over your operating bandwidth. The 10log(B) portion of the equation changes the units of the equation from dBm/Hz to dBm. You need to calculate your operating bandwidth and use it in the equation. If your system is filtered, then the bandwidth is reduced, otherwise it will be the full bandwidth of the amplifier. The noise produced and transmitted is not only limited by the bandwidth of the amplifier. Although you are only using a narrow portion of the amplifier bandwidth, it will still amplify noise over its full bandwidth. The only system improvement you may see is an equivalent bandwidth reduction from the noise contribution of earlier components if they operate at a reduced bandwidth. If you don't care about the noise outside your operating bandwidth calculate the noise power based on your operating bandwidth.

As for the measurements, if you measure the noise power with a power meter it will not be band limited. You would need a filter to measure the noise power within your bandwidth. Unless the noise power is high, this may not be a good measurement because you may be operating at or below the noise floor of the power sensor. A measurement with a spectrum analyzer will have the resolution bandwidth of the analyzer determining the bandwidth of the measurement. Again this measurement is limited by the noise floor of the analyzer. The noise figure measurement is not determined by bandwidth. It is a unitless number describing the increase in noise due to the effects of the amplifier.

In the equatiion:
NP=10Log(k) + 10Log(T) + 10Log(B) + NF + Gain
NP will be in dBm provided you convert to mW when using Boltzmann's constant.

The units are:

k in J/degree K or W/Hz/degree K
T in degree K
b in Hz

Converting from watts to mW and 10log() converts mW to dBm. The noise figure and gain are in dB and have no other units.

I hope this clears things up rather than add more confusion.

Joe