Thursday 27 July 2017

The Low Noise Amplifier

Although I already mentioned a Low Noise Amplifier in one of the earlier blogs, it was only wideband and I haven't found much use for it. With a few long distance flights planned ahead of me, I thought I'd improve my receiving setup and learn more about LNAs again. Examining a couple of projects on the Internet and some commercial solutions, I acquired an idea of the individual sections that constituted a plug-in solution. My intention was to design a board based around Mini-Circuits' PSA4-5043+ amplifier followed by a footprint for a band-pass filter. Initially, I looked into ready-made SAW filters, but I couldn't find any for 144MHz (APRS reception) and 434MHz (most of the balloon stuff in the 70cm ISM band) with the same dimensions. That way I would have been able to use one design for both variants. Because of that I took a different approach and designed the board with a footprint for a 3rd-order Butterworth band-pass filter. That meant I could compose a filter for the desired frequency band on individual boards, or so I thought. I used this design tool to calculate the specific inductances and capacitances. However, after soldering the boards, I found the pass-bands of both filters lying significantly off the intended frequencies. I tried tuning the filters with parts close in value, but with not much success. Eventually, I scratched the plans on making the 144MHz version and decided to do the 434MHz version properly with a SAW filter. Regarding the positioning of the filter - in front or after the amplifier, I am given to understand that a band-pass filter in front of the LNA filters out strong out of band, potentially disruptive, signals ahead of the amplification stage, however, it also decreases the signal to noise ratio of the incoming signals which may be too much for some of the already weak signals one is trying to amplify. Positioning the filter after the LNA, on the other hand, should amplify the signals without the additional losses. The downside is that strong nearby signals may lead to disruptive intermodulation products in the output signal.
These are the PCBs I designed in the end. The initial attempt with the Butterworth filter is in the image on the left entitled v2.1. The final version v2.2 with the SAW filter is on the right. The Eagle files can be found here:
LNA v2.1.brd
LNA v2.1.sch
LNA v2.2.brd
LNA v2.2.sch
The dimensions of v2.2 fit inside a 40x25x25mm aluminium case findable on Ebay. As mentioned, the amplifier used in the design is PSA4-5043+. Its range is from 50MHz to 4GHz with datasheet stated gain of 21.2dB and Noise Figure of 0.66dB at 500MHz. The following footprint is for an EPCOS SAW filter in a 3x3mm package. I chose B3710 model with the center frequency at 433.92MHz and an insertion loss of 2dB in 433.00-434.71MHz range. In front of the amplifier, there is a footprint for an optional ESD protection in the form of a pair of anti-parallel diodes. The board offers three different options of powering the amplifier's operating range of 3-5V. They aren't switchable on the go and have to be decided prior to soldering.
The first option, and the one I use, is via a bias tee network. The RTL-SDR v3 can, upon being commanded to, provide a 4.5V DC bias on the coaxial cable which is being picked up by the LNA board and used to power the amplifier as illustrated in the image above. In this case R1, R3 and the voltage regulator shouldn't be soldered.
The next option is to provide externally regulated supply of 3-5V via the + terminal. Note the PSA4-5043+'s absolute maximum rating of 6V. This time the right side of the bias tee and the regulator shouldn't be soldered.
The third option relies on the voltage regulator to provide stable 3.3V. It's input range is 3.5-6V and is expected to be supplied externally via the + terminal. The R2, R3, R4 resistors and the right side of the bias tee shouldn't be soldered.


144MHz version
As mentioned earlier, this effort hadn't worked out very well. Nevertheless, I consider the attempted tuning worth a few words and images for illustratory purposes.
I soldered the board according to the original design with the bias tee power option and prepared the aluminium case. The 3rd-order Butterworth band-pass filter comprised of two series resonant arms (both 560nH and 2.2pF) and one parallel resonant pair (680pF and 1.8nH). It was quite problematic to find components matching the calculated values. This probably introduced part of the deviation in the resulting filter.
The blue line in the chart on the left represents the response of the filter build from the original parts. RTL-SDR Scanner was used to obtain the data. As can be seen it was quite low in frequency. To attempt to tune the filter, I started by replacing individual components with close in value parts and re-measuring the response. Eventually, replacing the 680pF capacitor with 470pF and the two 2.2pF caps with 1.5pF, I arrived at the response shown by the green line. Its center frequency was about 141.5MHz which was as close as it seemed I could get. The image on the right shows the final response in a little more detail.
The filter still being slightly off in frequency showed it in direct comparison of the same signal received without the LNA (left) and with it (right). Instead of amplifying it, the v2.1 LNA attenuated the signal.


434MHz version
The v2.2 of the LNA was much more successful. Once again I soldered the board and prepared the aluminium case. The two longer sides of the PCB have stripes of exposed ground which then get into contact with the case. With the two SMA connectors added to the length of the board, it fits precisely inside the box so that the connector's grounded shield touches the side panel.
The SAW filter is placed between two DC blocking capacitors. EPCOS B3710 was chosen for the 70cm-band. However, any SAW filter with the same footprint can be used.
The final shape of the LNA. Even though the SMA connector is pressed right against the side panel which is about 1.5mm thick, with the outside ring added, there may not be enough connector left for reliably attaching a cable. Hence it is better to use the longer connectors.
This is the response of the SAW filter as measured in RTL-SDR Scanner. Much nicer in comparison to what I managed with the Butterworth filter.
Once again, receiving the same signal with (right) and without (left) the LNA. This time the demonstration shows the device working and amplifying the signal nicely.


868MHz version
Since I had some parts and boards left, and I came across a SAW filter with the right dimensions, I built one more LNA, this time for the 868MHz-band. The chosen filter was EPCOS B3725 which had the center frequency at 869MHz and insertion loss of 2.5dB within 868.0-870.0MHz range.
The response of the filter via RTL-SDR Scanner.
And reception comparison with LNA (right) and without (left).


In principle, I came across SAW filters with the right dimensions for 315MHz, 916MHz, 1090MHz, 1575MHz, 2.45GHz as well, so there are more potential version for the LNA.
Placing the LNA right at the antenna output or close to it means that the amplifier won't be amplifying the signals after they travelled meters or tens of meters inside a coaxial cable which attenuates them for every additional piece of length, but will amplify the original signal which will then be at better odds of surviving the remaining distance to the receiver.


SAW filter
While I was at it, I made a small board for the SAW filter and a pair of SMA connectors so it can be added to a wideband LNA. Ordered from OSHpark for $1.55.
The footprint is for 3x3mm DCC6C package and fits a number of filters. The ones I used were EPCOS B3710 (433.92MHZ) and B3725 (869.00MHz). The Eagle files are here: SAW Filter v1.0.brd, SAW Filter v1.0.sch.


Yagi 868-band
Since I made the 868MHz version of the LNA, I needed an antenna as well. Given that the physical dimensions of antenna elements decrease with higher frequencies, I decided for an 11-element Yagi.
I got the specific element lengths and spacings from an online calculator and used 4nec2 to model the final product. The results of the modelling are in the table and image above.
In terms of materials and mounting, I followed the same approach as with my previous antennas. Light wooden boom with the brass elements pierced through the boom. The radiator was the only problematic one. I had to improvize a plastic holder to support it. All the elements were secured in place with a drop of glue.
These SWR charts were output by 4nec2 upon modelling the antenna.
These charts are the result of RTL-SDR Scanner's measurement of the antenna's reflected power while fed with a noise source. As always the quality of my measurements and conclusions suffers from lack of proper instruments.


Radio Horizon
With the receiving setup addressed, lets see what I may expect in terms of range. Operating at about the maximum allowed power (10mW in 433.05-434.79MHz band) and reasonably low bit rates (50-100 baud) the signal should be decodable all the way to horizon provided there is a line of sight between the transmitter and the receiver. A balloon at 12500m in altitude has a geometric horizon at 399km away. However, apparently due to the atmosphere bending the propagating RF signals somewhat, it could potentially cover an area with radius of 461km. That leaves the question of range in hands of the receiving setup and its specific whereabouts.
It's apparent from the images above that my horizon is fairly limited. Taken from here, the map on the left shows the maximum distance in all directions a transmitter at 12500m and 42000m in altitude can be to still have line of sight. In floaters context, it is the orange line that is of interest. The range towards north is the most reduced since I am located on a hillside. It's a shame, because Poland is the only relatively active area in terms of balloon launches in my vicinity. The other image then shows potential spots for long distance testing around my town. That is with the tracker in my garden and me at one of the spots.