Andy's website

Design Intent

I decided to spend some time to see if I could optimise the receive front-end of my 144MHz moonbounce system by reducing the system noise figure.
Up till recently I have been using a receive preamplifier based on the old 3SK121 FET - a capable device, but certainly nothing extraordinary in terms of excess noise figure or gain. Moreover, since building a PANFI, the limitations of the 3SK121 have been clearly highlighted.
The 3SK121 device achieved a gain of around 18.5dB, and noise figure of about 0.6 ~ 0.7dB. Given the class of devices available today, a re-design is clearly warranted.
I settled on the ATF-54143 PHEMT device manufactured by Avago. This device seems to be universally used by the EME community for both 144MHz and 432MHz moonbounce, and the specs look quite impressive.
So my design goals were:

• Less than 0.25dB excess noise figure.
• Less than -14dB Input Return Loss.
• Greater than 20dB Gain.
• Good IP3 performance.

Schematic Diagram

The Avago data sheet for the device (Useful Links) covers pretty much all that is necessary to get the device going.
As per the data sheet, I used an active bias scheme to keep the bias point reasonably constant over temperature (I plan to use this preamplifier at the antenna mast). I also added a simple bandpass filter to the output of my LNA design to help mitigate against large signals from FM broadcasters overloading my receiver.
I used an old version of Serenade software to simulate and optimise my design, and settled on a circuit as shown below.

144MHz Preamplifier Schematic Diagram.

Prototype Build

A PCB was made and a couple of preamplifiers built into a small die-cast box. Some measurements were taken and compared to simulation results:

Measurement Technique

I had the chance to perform some measurements of this preamplifier using a HP346B noise diode together with a HP8970A PANFI.
To make the most of the opportunity, I thought I'd try a couple of techniques to investigate noise figure measurement limitations.
Firstly, I wanted to see if a precision attenuator in front of the preamplifier would help reduce impedance variations caused by switching the 346B noise source on and off. Reducing these variations will definitely help reduce measurement uncertainties.
My results are summarised below:

Parameter	HP346B		HP346B + 10dB attenuator
	OFF	ON	OFF	ON
Zin (ohms)	51.2 -j0.4	46.99 -j0.7	49.6 -j0.8	49.34 +j0.4
VSWR	1.027	1.074	1.005	1.011
RL (dB)	37.51	28.95	52.06	45.24
Rho	0.013	0.036	0.002	0.005

The HP346B noise source shows a clear impedance shift between on and off states. The 10dB attenuator (measured as 9.95dB) has helped reduce this variation.
This impedance shift between the two noise source states adds a mismatch uncertainty, which ultimately adds to the noise figure measurement uncertainty. The uncertainty (worst case) caused by this mismatch is calculated and shown below.
The calculation is made both for the 346B alone, and together with the 10dB attenuator. Further, 2 conditions are considered: uncertainty between the HP346 noise source and the preamplifier (NS-DUT), and the uncertainty between the HP346B noise source and the HP8970A (NS-NFM).

Config	Maximum Uncertainty (dB)
	NS-DUT	NS-NFM
HP346B	0.0524	0.0814
HP346B+10dB attenuator	0.0073	0.0113

Not a huge difference, but it all adds to the uncertainties.
Please note that the attenuator (being placed before the preamplifier) is not included in the calibration loop; its effect on both indicated gain and noise figure must be calculated and compensated for manually.

Using the HP346B and attenuator combination, I performed a 2 sets of measurements for each preamplifier: the second set with an additional 0.25 wavelength (at 144MHz) length of coaxial cable between the attenuator and preamplifier. Any variation in indicated noise figure between these 2 sets of measurements will indicate a sensitivity to the impedance presented to the preamplifier.
Note that the extra losses induced by this 0.25 wavelength cable (to both the indicated gain and noise figure) must be accounted for.

LNA	Cable Length	Gmeas (dB)	Gcomp (dB)	NFmeas (dB)	Tmeas (K)	Tcomp (K)	NFcomp (dB)
Unit #1	0	28.81		0.19	13.4
	0.25	28.77	28.83	0.24	16.65	13.10	0.190
Unit #2	0	28.94		0.17	11.9
	0.25	28.86	28.92	0.23	15.94	12.40	0.180

The table above shows results for 2 preamplifiers. The noise figures in bold represent 2 measurements taken with and without the extra 0.25 wavelength cable. The numbers agree sufficiently well to indicate that the HP346B+attenuator combination provides an adequately constant impedance for the measurement.

Results

This has been another fun little project, and I have learned a bit about the trade-offs that go into designing a low noise preamplifier. But I would like to emphasise that sub-1dB noise figure measurements are notoriously difficult to make with any degree of accuracy at the best of times, let alone using the equipment available to me.
I claim nothing other than the plots and tabulated results represent what I measured with the equipment available. Applying my measurement results to Keysite's Noise Figure Uncertainty Calculator (Useful Links), a total noise figure measurement uncertainty is calculated as +-0.208dB.
The major contributors to this uncertainty are the ENR of the HP346B (+-0.2dB) and the measured attenuator accuracy (+-0.05dB). Without access to better calibration, nothing can be done to improve this uncertainty.
This preamplifier will never be "best in class", and is not really meant to be. It's a simple design using easily obtained (read 'cheap') components. With that in mind, it still makes for a very serviceable piece of gear for moonbounce.

If anyone is interested, I still have a few more of the prototype PCBs left over.

Useful Links

• Avago ATF54143
• NF measurements by SM5BSZ
• Keysite Noise Figure Uncertainty Calculator