SPF5189Z LNA for pre-compliance use : measured performance

[dazz1]

Hi
I brought a cheap Aliexpress SPF5189Z Low Noise RF amplifier and put it into an Aliexpress extruded case usable as a handle.  The aim is to use this LNA with EMC probes to give the signal a low noise boost before the Siglent SSA3021 spec-an.  The irony of using a low noise amplifier to measure noise is not lost on me.  The SPF5189Z gain varies a lot over the frequency range. 

The purpose of measuring the performance was that the SSA includes a feature to add gain corrections across a frequency range.  By saving the measured gain of the SPF5189Z LNA, I could save that as a CSV file.  This would allow me to generate a corrections file to upload back to the spec-an.  Applying the corrections file across the entire bandwidth would allow me to normalise the gain to 0dBm.  I can add to the corrections file the gain of any connected antenna.  That is a very powerful feature because it will allow me to take direct readings from the spec-an.

The format of the COR  corrections file is dead simple.  It is
frequency (Hz), gain (dB).  Below is a sample COR file created on the Siglent spec-an.

Code: [Select]

1000000,-25.00
10000000,-24.00
100000000,-20.00
1000000000,-14.00
2000000000,-10.00
I plan to use Excel to smooth the corrections data and reduce the number of data points. 

The measurement test setup was as follows:
                                                                             -------------
                                                                             |                \
TG output >>>-----------------[10dB Attenuator]---|  DUT LNA  >----[10dB Attenuator]--------------->>>RF input
                                                                             |                 /
                                                                              -------------

The LNA performs much as expected.  One of the images shows what happens if the 10dB attenuator on the output is removed.  The plot shows waves due to mis-match between the LNA output and the spec-an input.

A study of the photos will show that the power cable enters the aluminium enclosure through a rubber grommet.  This is poor emc design because the sheild of the USB power cable is not grounded to the enclosure.  Any noise on the outside of the USB shield will go straight into the enclosure.    This risk is mitigated by the ferrite choke very near the power entry point. The prevents the whole USB power cable acting as an antenna.

One of the really nice features is that the LNA is a 5V device that can be supplied from the spec-an USB connector. 
The correction feature is also really useful.

The combination of a dirt cheap LNA and aluminium enclosure now looks and performs reasonably well.  Entirely adequate for the required purpose.  The correction feature built into the SSA3021 spec-an will be use to make the LNA frequency response look flat.

Dazz

[tefe]

Maybe windows add an extra symbol after your order. But why not use Normalization?

[dazz1]

Maybe windows add an extra symbol after your order. But why not use Normalization?

Yes.  Windoze added [CR][LF].  I used Notepad++ to convert those to just [LF].

[Wolfgang]

Hi,

why not calibrate or normalize it out, then you dont need a manual correction table. For up to 10-20dBs this is fully OK, EMC measurements are not soo precise anyway

[dazz1]

Hi
I am using the correction table to normalize the amplifier gain.

[Wolfgang]

The idea of normalization is that your instrument computes the correction table automatically from a measurement sweep. Why make it by hand ?

[dazz1]

So I can create a library of normalisation curves that I can use in MS Excel to do calculations.  Loading those curves into the SA is a good way of validating them.

This specifically include antenna factor curves which cannot be conveniently measured.   

[Wolfgang]

Well, you could download the curve from your spectrum analyzer and use *this* for further computation. Its at least a valid measurement result.
Its clear that the antenna characteristics are not so easy to measure. You probably need an anechoic chamber for that. What you could measure, howver, is impedance over frequency.

[dazz1]

Well, you could download the curve from your spectrum analyzer and use *this* for further computation. Its at least a valid measurement result.

That is exactly what I did to produce the correction curve.  I downloaded the measured  amp gain curve, calculated the negative of each data point with Excel, then uploaded that back to the SA as a correction curve. 

Its clear that the antenna characteristics are not so easy to measure. You probably need an anechoic chamber for that. What you could measure, howver, is impedance over frequency.

Few have access to an anechoic chamber.  One of the many ways to measure antenna performance is to use two identical antennas. One as a signal source transducer driven by a TG, the other as the receive sensor providing the input to the SA.   If the antennas are a measured distance apart, the characteristics of the antenna can be defined.    The performance curve would be produced with a combination of measured data and mathematical calculation performed in Excel.  The resulting correction curve can then be uploaded back to the SA.

The feature to allow externally calculated correction curves to uploaded to the SA is really useful. 

[Wolfgang]

Would it be less effort if you try it this way:

- Put in your amp and normalize within the analyzer
- Now open the connection and put an identical antenna on both ends
- Repeat measurement. Now you get a transmission loss that can be used to derive the signal strength in your geometry (distance, orientation)

I see the following difficulties:
- To make sense the distance between the antennas should larger than a few wavelengths (otherwise you get a lot of near-field effects)
- antenna cable losses should also be calibrated out (you could use the same cable setup with the amp).
- it is a problem if the antennas are badly 50Ohms matched, making their feedlines radiate. This must be meticulously blocked (baluns, ferrites, ...), otherwise impedance and direction info will be wrong.

[dazz1]

Would it be less effort if you try it this way:

- Put in your amp and normalize within the analyzer
- Now open the connection and put an identical antenna on both ends
- Repeat measurement. Now you get a transmission loss that can be used to derive the signal strength in your geometry (distance, orientation)

I see the following difficulties:
- To make sense the distance between the antennas should larger than a few wavelengths (otherwise you get a lot of near-field effects)
- antenna cable losses should also be calibrated out (you could use the same cable setup with the amp).
- it is a problem if the antennas are badly 50Ohms matched, making their feedlines radiate. This must be meticulously blocked (baluns, ferrites, ...), otherwise impedance and direction info will be wrong.

I agree with all you say.  Just a couple of points to note. 
I would not normally expect to use an amplifier for antenna performance measurements.  The SA should have sufficient gain/sensitivity. 
The free space loss and the actual antenna gain would still need to be calculated external to the SA, most easily with Excel.
IF antennas are separated too far, then reflections become a more significant error.

[Wolfgang]

When your SA is a budget thing (I also worked with a RIGOL DSA815TG) sensivity/noise floor can become a problem that can be elegantly solved with a preamp, e.g.:
https://electronicprojectsforfun.wordpress.com/rf-module-gallery/the-amplifier-module-gallery/a-usb-powered-low-noise-preamplifier/

Yes, free space loss must be calculated anyway.

Antenna measurements are tricky without a chamber. A compromise has to be found in all cases. You could at least have the near field zone covered with some absorptive material.
When I was studying at Vienna University we used egg cartons covered with a layer of Shellack with iron powder in it, and then a layer of CRT repair graphite spray.

[dazz1]

Hi
While I don't expect to need an external preamp, I purchased one anyway.  I also have other uses for it.

The alternative to the egg-carton absorber is distance. If I mount the antenna on non-conductive masts (2-4 metres), I should get enough separation from the ground to reduce reflections to an acceptable level.  I can test for reflection by changing the polarization of the antenna.  If there is significant interference from reflection it will show if the polarization is changed.

[Wolfgang]

Another idea when looking at your curves: Why not use an LNA that is essential flat over the frequency range you require ? They are cheap nowadays and available to several GHz.

[dazz1]

Hi
I have a Mini-Circuits AX60-3018G-SS+ preamp.
It cost me 10x the price of the cheap Chinese preamp. 
It still doesn't have a flat gain curve and the noise figure is about twice the Chinese version.

Overall the Chinese preamp offers very good value for the small amount of money.

Dazz