Re: Agilent 346A and N9069A reading high?

Mike Willis

John Makes a good point - NF meters can typically measure in bandwidths of several MHz and are assuming the noise figure, gain and match is constant across this bandwidth.

When you design a preamp for VHF it is normally a good idea to make it narrow band, especially in the UK where we have very strong pager signals either side of 144MHz. To a DXer, there is nothing of importance outside the 144-145MHz range and a preamp may have a 3dB bandwidth of 2MHz or less to cover this range plus a bit. What this means depends on the filter shape and the NF meter bandwidth, but its quite possible to be missing nearly half the noise due to the filter.

As John says this will mean the ENR calibration won't be right so the difference between noise on and off will be lower than expected resulting in a higher recorded noise figure. You can fix this with a filter which is difficult, or by doing the maths if you know the relative bandwidths, which may be easier.

The negative noise figure issues is probably just noise. If you expect to get better than 0.1dB accuracy after calibration you will need to be very careful, do a lot of integrating and keep everything at a constant temperature and do it all in a screened room. You should see a mean value of zero. The fact it is reading -0.1dB consistently could just be the instrument needs recalibrating, or maybe you have not got the settings right for temperature, or indicate external noise is getting into it from local RF sources and disturbing the calibration. If you are doing this at home I would go with the QRM. Since they installed a mobile phone mast at RAL near the RF lab, we haven't been able to measure VHF/UHF noise figures accurately in the RF lab.


--- In, G3XDY <g3xdy@...> wrote:

Hi Conrad

I have also seen abnormally high NF readings on an example of the HA8ET
preamp. This preamp includes narrow band filtering, and it could be that
the noise bandwidth of the preamp is narrower than the Agilent test set
bandwidth. I recollect that you might be able to configure the Agilent
box for a narrower measurement bandwidth, but I don't have the manual to
check. The 8970A will suffer from the same problem and is not
configurable, the only way to test such preamps with the 8970 series
boxes is to build a low loss narrow band 144MHz filter, then calibrate
the instrument with the filter in circuit and then test the preamp via
the filter (bear in mind that the filter won't look like a flat 50 ohm
source so there will be errors due to changes in the impedance seen by
the preamp input unless you include an isolator after the filter).


John G3XDY

On 18/07/2013 13:44, Conrad Farlow wrote:
I have access to an Agilent 346A head with an N9069A noise figure
measurement application running on an EXA series spectrum analyser. It
seems to have number of issues.

1. After calibration the noise figure and gain are never truly zero, the
noise figure normally reads -0.1 or similar and the gain will be -0.15
or thereabouts, these are just typical figures.

2.When I measure noise figure I see good results on the gain but the
noise figure seems high. For instance my HA8ET Extra 2 LNA reads about
1.4dB whereas I am pretty sure that it is much better than this.

I have Tcold set to the temperature on the surface of the device as
measured by a thermocouple. I have used some loss compensation before
the device, I have generously allowed 0.1 dB for a high quality
female-female N type.

Does anyone have any advice as to what is going wrong? I do have an 8970
that I could use for a reference check but it hasn't been switched on
for a while, shocking I know.......

Bit puzzled..

Join to automatically receive all group messages.