My new LNA / How to evaluate performance of changes

[eepete]

I finished building the 1st cut of my LNA. It tests just fine, and I’ve cut it into my system. I want to show the LNA and it’s relative performance, and then show the graphs1090 changes when I cut over to the new LNA.

The LNA

First the LNA overview (without trying to get into an electronics design discussion):
The goals were an LNA with the inputs and outputs facing the same direction so it could be mounted with the connectors facing down.
The LNA can be powered from the feed line, with any voltage from 4 to 15 volts. There is a small switching supply on one side of the PCB that makes 5V, and then on the top of the board a low noise Low Drop Out (LDO) regulator that makes the 4 volts for the LNA chip.
SAW Bandpass filter before and after the Low Noise Amplifier (LNA). They can be unsoldered and jumped to remove them.
Attenuator after the LNA to determine if the SAW filter pass characorisis are affected by impedance miss-match

The bottom of the PC board with the switching supply looks like this:

The top of the board with the LNA looks ike this. The circuitry is about the same width as a U.S. dime (about 18mm for my global metric friends). There is the ability to put a shield over the entire RF and 4V LDO:

Here is a wide range sweep of the LNA showing its pass band and to what degree it attenuates out of band signals:

Keep in mind that the dark blue Flight Aware USB radio dongle has a SAW filter after it’s LNA, so the final system results w/r/t the passband shape is the same as with my LNA. Note the violet trace- you can see a “bump” that looks like the top of the passband. This makes me think that the LNA is able to operate a bit (maybe 6 dB?) below the noise floor of the spectrum alalyizer, which is at -120 dBm.

One option is to remove the SAW filter before the LNA and use the most excellent Flight Aware 1090 filter. Apologies for not thinking ahead and sweeping that so it was all in the same screen picture, but here is what that filter looks like over the same frequency range:

The SAW filter is narrower, but for the purpose of attenuation out of band frequencies in the 500 to 800 MHz range, both provide a bit more than 50 dB attenuation. In my environment, I will have some interference in the 915 MHz ISM band, so with the SAW filters doing better there by 16 dB or so I would go with the SAW. Since a SAW filter is more delicate than the Flight Aware filter, for most people I would go with that Flight Aware. I plan to build one LNA without the 1st saw filter in case once this LNA is outside up in the air with the antenna it fails. One last metric: the LNA chip itself was only 12°F warmer than ambient. So I’m not worried about the chip getting so hot when it is outdoors that the noise performance declines drastically.

The experiment with the attenuator shows that the SAW filter bandpass characteristics do not change with impedance mis-matches, so I will take the attenuator out of the next board spin. The first board, with that attenuator removed, showed a net gain of about 14.5 dB. This makes the board a good candidate for being placed with the antenna so that any cable losses are compensated for. I hope to get some time on a Noise Meter at a local RF group so I can get a feel for how good the LNA really is. It’s hard to justify buying a noise meter just to debug one LNA and I’m pretty good at justifying buying test equipment.

The board goes into a 3D printed chassis with spaces to support the board, This plastic chassis will go into a die cast zinc chassis with holes for the two feed lines, and screening to keep out insects.I plan to conformal coat the final PCB (and hope that doesn’t change it’s characteristics , something else to learn…). The die cast zinc chassis will be painted white. A very short (1 meter) feed line will go from the antenna to the LNA.

The Change-Over

About 11:00 AM local time, I cut in the new LNA. While I’ll check out the Flight Aware page after 24 hours to see compare results, and then in a week to check the daily changes, here are screen shots from graphs1090.
Note that at around 8:00 AM local you can see a peak of messages as all the various commercial airlines from RDU airport (Raleigh-Durham, NC) start the day. There is also a bit of a peak (I’ve seen this on many multi-day graphs) around noon with transfers, and there will be a peak in the evening the flights get in.
The two screen shots of the graphs:

Seems clear things did not get worse. Note the change on the “ADS-B Signal Level” after the change around 11:00. The noise level goes up because with this gain, there is more noise due to signals received that were still too weak to be decoded.

Next Graph set:

The first thing to see here is that prior to my installing the LNA, my gain was maxed out at typically 49. With the 14.5 dB net gain of the LNA, the automatic gain quickly adapted and went down to about 30. I absolutely love the automatic gain of the Flight Aware stick !
There appears to be an increase in the message rates. Time will tell if that is an actual improvement or if it was just a busy Friday travel day.

Note also that with LNA adding gain to the signal such that the Flight Aware stick was not at it’s maximum gain could also mean there is really no discernible improvement in the noise floor, it’s just the improvement in gain allow the stick to receive and decode more messages. The two metrics of an LNA, improving the signal to noise ratio of a signal and providing gain, are often challenging to comprehend.

And here we see the dilemma for everyone trying to optimize their system. The received signals are constantly changing due to different amounts of aircraft in the air, and due to different weather conditions. So while it’s easy to look at a specification for pieces of the systems (Filter, LNA, Coax), and easy to measure the performance of a given component on a piece of test equipment, in the real world it can be very difficult to determine if things got better or worse.
I took screens shots of graphs1090 from a week ago before cutting in the LNA, and will look again in about a week. I will also look at the coverage charts and daily report of Aircraft Reported on the Flight Aware page for my account. Hopefully that will give me some feel for what the change in performance is with this LNA. At this point, I plan on keeping it, I am working on an outside antenna site, and when that is ready I can see how the LNA holds up when it is outside. The simplicity of just two connections with the LNA, and a metal box that is very small are quite compelling here.

I hope the LNA pix are at least interesting. There have been a lot of posts on how to evaluate performance, so I hope this will help with peoples understanding of that process. I look forward to comments on how people interpret the graphs, ideas for measuring performance and other comments. I’ll add on to this post in a week with the Coverage Graph and Aircraft Reported information from my Flight Aware account so we can see if that helps when trying to determine if a system change has helped out and if so by how much.

Long post, sorry, hope it is informative, looking forward to feedback !

[keithma]

This is a really interesting read and analysis, thank you. I’m very keen to know how well this copes with strong amateur radio signals nearby because another well known LNA struggles badly with overload and just drops out completely.

Are you doing this purely for yourself or are you thinking of making these available (even if in only small numbers because of the quantity of PCBs you have)?

[eepete]

The filter before the LNA is critical IMHO. Every “real” systems I’ve done, including designing and installing 10 amateur radio repeaters from 50 to 440 MHz, had BPFs after the cavities where the combined signal went out to the feed line that goes to the antenna. I’ll trade off 3 dB of loss for peace of mind any day.
The SAW filters are sensitive because they have a 3V DC maximum input. The Flight Aware BPF is constructed with coils and (most likely) 50V capacitors, so it should be more rugged.
On the electrical front, I have two PIN diodes back to back on the input, and a 330 uH coil to ground. My hope is the will protect the SAW filter. In my use case, I have a 902 to 928 transmitter running about 1/2 watt (26 dBm) that will be at the same height and only about 1 meter away. My first go-to if there are problems is to add the Flight Aware filter before my LNA! The first responder radios in use around here are trunking radios that operation as high as 860 MHz. And I’ve seen responders key up a hand held radio while touching the “duckie” antenna to another antenna “Just to see if it was rugged enough”.
And then we have all the cell phone towers… my closest ones are about 1.2 miles away.

After the LNA has had a week to let me guess as to how much better things got (and more importantly why), I want to do a test with a broadband noise generator and also with my two signal generators to make sure that the SAW filters are really performing well. So many intermodulation problems with repeater come from a 3rd station/repeater that was 50 miles away. Sometimes it doesn’t take much to mess things up. One must always remember that all amplifiers are mixers, and all mixers are amplifiers. And crystal filters? they are modulators on steroids and as such need to be preceded by filters.

I do think that for people experimenting with ADSB systems, one can start with the LNA without a BPF at the input built in, especially due to the availability of the Flight Aware BPF that can be added if there are problems. This lets someone experiment with maximizing their results because the price of failure is low. Have you tried the Flight Aware filter out to see how it behaves with HF ? I also plan to have ferrite cores on the outside of all coax so that low frequency signals won’t mistake the coax for a antenna.

Pretty much this is for myself, this would be a costly board to contract out. In Qty 10 the LNA chip is still $5, the SAWs are $3.5, just the switching power supply boost buck chip is at a bargain “Supply Chain Special” price of about $8. This is not a trivial hand build… And pricing for building double sided PCBs is a lot too. Even for a small PCB like this, set up charges for panelization, paste masks and other items would still be about $400 or so. I really hope I can get some noise meter time, it would be very nice to know what the effective improvement in signal to noise actually is.
I do plan to build one of the PCBs with no SAW filters so I can better test the LNAs susceptibility to strong adjacent signals. It would be nice to have a broad band LNA for both field and bench use. The part number on the LNA chip is a Mini Circuits PMA2-162LN+

Thank you for your kind comment, I’ll update in about a week after there is more information available.

[keithma]

Again, really interesting to read, thanks. I completely agree about using BPFs on repeaters, our group do the same on our boxes as well. That insertion loss is worth every fraction of a dB for the difference it makes. I also like that you’re running some on 6m, we really couldn’t find a site where we’d have enough space for the cavities for 6m!

My original LNA here was overloaded with just 1W of 2m or 70cms FM into a Diamond vertical on the same property. I can’t remember exactly how close it was to the LNA but it would have been within 20’ or so. We had the same problem with the same make LNA on one of our sites where it was co-located with a 2m SSTV replay. As soon as the transmitter kicked in, the signal just dropped away to absolute zero. We replaced it with an RTL-SDR-LNA and the problem went away.

I now have an RTL-SDR-LNA, Airspy and Pi all in a box at the top of my mast with a short coax run to the aerial which is mounted on the top of my Hexbeam. The coax to the ADS-B aerial runs alongside the coax to the 40m extension and HF just doesn’t affect it at all, even at maximum legal power.

I don’t have any filter between the aerial and the RTL-SDR-LNA on either site.

I can run 100W into my VHF/UHF vertical and again, no problems whatsoever. It seems to be utterly bombproof.

Fully understood about the costings and I’ll still be following this with great interest.

[wiedehopf]

@keithma

Fully understood about the costings and I’ll still be following this with great interest.

@eepete

Even for a small PCB like this, set up charges for panelization, paste masks and other items would still be about $400 or so.

If you’re up that high you might just want to buy this instead
https://shop.kuhne-electronic.com/kuhne/en/shop/low-noise-amplifiers/KU+LNA+10902+A+TM++Selective+Low+Noise+Amplifier/?card=1813

But i suppose if you build it yourself it’s cheaper and you also want to design something.
Oh and i have no clue how good the Kuhne out of band rejection is.

[foxhunter]

@wiedehopf

If you’re up that high you might just want to buy this instead

330g weight? What the hell did they use?

Datasheet sounds interesting:
https://shop.kuhne-electronic.com/module/mysydeshop/webforms/download_file.php?file=2382

[wiedehopf]

@foxhunter

330g weight? What the hell did they use?

N connectors are hefty, also the unit is likely potted.

[foxhunter]

@wiedehopf

also the unit is likely potted.

Yes, just read about the sealed outdoor capabilities. Also the size look impressive compared to other LNA

Thanks

[eepete]

Ah the complexities of the simple “Build verses Buy” decision…

The Kune pre-amp looks very nice. I wish the data sheet had a bit more on that the internals are w/r/t/ pre or post amplifier filter. The “both coaxes come in at the bottom” form factor is also perfect. A wide band frequency response would also be nice, since interference from cellular is a common problem everyone deals with.

My decision to build was based on that “connectors at the bottom” form factor after seeing so many pictures of amplifiers stuffed into enclosures. For me very often the build decision is also about learning. I’m doing a big project (been working on it for about 3 years or so by now) at 915 MHz so this was a good opportunity to learn about parasitics in circuit design in the 1 GHz range. I also learned a lot about SAW filters that is not in the data sheets. A lot was learned about component placement, hand assembly techniques and the like. The power supply is the same one I’m using in another PCB I have sitting in front of me that is ready for population, and this was a great way to insure that I can have a switching supply that doesn’t interfere with the RF. The ground planes and their interconnections were also confirming what I have done in the 915 MHz project.
With any luck this will also give me a wide band LNA for lab and testing use.

All hobbies are economic disasters. For what spectrum analyzers, oscilloscopes, PCs, 3D printers etc. cost you can just buy products like the Kuhne all day long and come out ahead. And that works out great until there is not a product to do what you want. Like any good geek, for me it’s all about solving problems, making things that work and learning.

It was just really nice to have a way to learn that also had an application. And the responder stuff I’m doing will use the ADSB to increase safety for drone operations in deployed scenarios.

With the LNA installed, the focus now changes to how to evaluate changes in performance of equipment in the field where it is difficult to have repeatable input to a system. Measuring performance changes in the ADSB environment certainly seems like the most complex type of system for this challenge. Evaluating the 915 MHz will be easier, I can go to a particular location from the receive site and produce the same power level signal. The data telemetry will provide access to the receive signal strength, and can log location vs. receive strength to make software that provides coverage maps as I walk around an area.
It is really hard to get all the planes in the world to fly the same pattern day after day…

So my hope with the post is to show what I’m doing because it is fun (once it works…), will maybe help people learn about amplifiers, band pass filters and cable loss. And finally, we will see what can be learned about evaluating performance of an ADSB receive system as everyone works/plays on their installation.

Thanks to @wiedehopf and @foxhunter for the links on the Kuhne ! The data sheet teaches that LNAs often use a lot of power, and it shows that the warner the LNA the worse the noise that it adds to the system. This is good to know for when people want to put the LNA in the outdoor environment.
And now I wait a week or so and see if I can determine if system performance got better or not give the existing tools of graphs, tables, and images. I don’t want to go down the road of trying to learn to develop those tools, and very much appreciate the graphs1090, ADSB and Raspberry PI expertise I’ve seen on this site. Everyone makes this easier and fun. And with any luck, this small experiment combined with other posts will improve your already fantastic knowledge of this Flight Aware based system and lead to improvements in the performance of those tools.

I can’t wait to see where this goes…

[eepete]

@wiedehopf

N connectors are hefty, also the unit is likely potted.

Yep. Potting also helps spread the heat. Another thing to learn for me is the effect of conformal coating these types of circuits. Any PCB that is in an outdoor environment will degrade with time unless it is either coated, potted, or in a sealed chassis. Making a chassis that is truly sealed is not trivial. Making a chassis with a pin hole leak that slowly fills with water is easy. Note also all the discussions on sealing connectors. All good stuff to learn and know.

[eepete]

I’ve had the LNA in place for about 7 days, so I now have graphs1090 images for “before” and “after”. Once curious artifact in the plots: you can see on Jan 29 where flights were dramatically reduced. This was the effect of the winter storm that adversely affected the Eastern coast of the US. You can see a peak before that where everyone wanted to “get back home” before the weekend.

Observations:
On the “ADS-B Signal Level” you can see the shift in the Noise plot, the lowest line on the graph. Keeping in mind that “Noise” here is a composite of RF signal noise and various signals from aircraft that were too low to correctly receive, we see that when the amount of RF into the Flight Aware dark blue SDR dongle increases, everything kicks up but the number of transmissions correctly received/“Message Rates” remains about the same. Put another way, if you have an audio recording of someone taking with noise from a jet engine in the background, turning up the volume on the recording doesn’t make it any easier to hear what the person is saying.

You can also see the effect of the gain from the LNA in the “Misc” plot where the automatic gain on the dongle (what a nice feature!) went from 49 down to 30. This shows that indeed the LNA provided more signal, again with the caveat that more signal (turning up the volume) doesn’t automatically mean more received transmissions. However the extra gain means that if the LNA was placed at the antenna, and there as a lot of feed line (coax) loss, this arrangement could compensate for and remove the effect of the redline loss. Note also that the gain being “maxed out” at 49 means that with this antenna location the dongle is working with very little signal.

The antenna is currently indoors, and I have a metal roof. The antenna is in the south west corner of the house, as can be seen by the coverage graph. So clearly my current set-up is limited by the antenna, and no amount of gain will improve anything. Note that this is a common concept seen in discussions about LNAs and the antenna location. “You can’t pull something out of nothing”.

My admittedly subjective conclusion is that the Gain of the LNA did not help me at all. However, the LNA circuit board (PCB) I made had a SAW filter in front of the LNA, which adds about 3 dB of loss. The Amplifier part of the LNA is low noise, the “L” in LNA. As such, it was able to pretty much compensate for that loss. So I have the same performance that I had with no LNA but the signal being presented to the dongle has all the cell phone tower and other RF signals filters out. The feed line is only 2 meters long, so its drop is negligible. When my final install is done, there will be 30 to 40 feet (10 to 12 meters) of feed line, This LNA will make that loss irrelevant.

Finally, my next step is to remove the LNA and insert a 6 dB attenuator between the antenna and the dongle at the same place I had the LNA installed. I’ll give that a week and then change out the attenuator to a 10 dB one. My hope is this will “help to calibrate my eyes” when examining the graphs1090 images. Somewhat subjective again, but I currently have no idea at what point a given amount of gain (or loss) affects message rates to an observable degree.

Here are the before and after from the user page

Before LNA:

After LNA:

Here are the 14 day plots from graphs1090. The LNA as cut in on Friday, Jan 28. The 29th was the storm day, and then normal traffic resumes.

Graphs set 1, 14 days of coverage

Graphs set 2, 14 days of coverage

[eepete]

Attenuator Tests

Over a 2 hour period, I placed a 10 dB, then a 6, then a 3 and finally no attenuator between the antenna and the LNA input. I tried to avoid the “morning rush” of aircraft. But of course, the number of aircraft vary all the time making a lot of these test difficult to do. The graphs are below. Remember that this antenna is inside, and “starved” for signals.

Conclusions

The surprise here was how much a 3 dB change in attenuation affected the number of Aircraft Seen. In a lot of the RF work, you tend to ignore 3 dB as being not that big of a change. This provides some insight into the impact of a coax feed line with 3 dB of loss. It was also interesting that the automatic gain on the Flight Aware (dark blue) dongle stayed at 33 the whole time. When I cut in the 10 dB attenuator, the gain went from 30 to 33. So there was a lot of signal there, but, it was mostly amplified noise. My poor, starving antenna
This is a test that need to be repeated when the antenna is outside and not so much starved for signal. In my earlier graphs with no LNA the gain of the dongle was 49. So in some ways, these results show the impact of a few dB for signals that are very small/low level.
Note in the ADS-B Signal Level that the noise bumped up a bit with each reduction in attenuation. So indeed, more signals were present. In this graph, “noise” is a blend of traditional RF noise floor and all the various signals that were too weak to be able to be decoded.

From my previous post, it looked like the LNA was a wash except that the gain and noise figure for the Low Noise Amplifier chip compensated for the loss of the SAW filter at the input and amplifier induced noise of about .5 dB. Seeing how much just a 3 dB change made, having the LNA at the antenna to compensate for feed line loss looks like a win. To see more signals and also reduce out of band signal strength is nice. Note that it is most likely irrelevant if the BPF that is before the amp is the on-board SAW filter or the external Flight Aware filter.
Still to be determined is how both filters behave with large signals in the 902 to 928 MHz band, a case of interest for my particular use case, but not a concern for most people. For most people, if you have a good antenna location and the LNA at the antenna to provide more gain and compensate for feed line loss, you have a win if your feed line loss is more than 1 dB or so. It just changes the dynamics, fragility and ease of maintenance due to having active electronics up on top of a mast. It should not be surprising that when you optimize performance there is a trade-off lurking around somewhere.

[keithma]

@eepete

In a lot of the RF work, you tend to ignore 3 dB as being not that big of a change

I’m taking a very small chunk out of your last two interesting posts (I’m suffering a bit from covid brain fuzz so struggling to get my head around everything) just to comment on that specific point.

When I see 3B loss, I don’t think it’s not that big of a loss. No matter what frequency you’re talking about, 3dB loss is 50% of the signal. I’m not trying to teach my grandmother to suck eggs, I’m well aware that you know this already

When I see an extra 3dB gain on an HF aerial, it interests me. That’s a lot of extra signal to play with and gaining 3dB on top band is a massive increase. It’s far easier to achieve at VHF/UHF but it’s still a 50% increase in signal.

I don’t know where I’m going with this post. Again, blame covid brain fuzz and the fact I’ve just had my daily dose of pre-bed pain killers.

I will take the rest of this in when I’m clear of this stupid virus.

73

[eepete]

Yeah, when one makes a statement like that context is critical. Tuning cavities? 3 dB is a boatload. Looking at BPF filter response at 20% away from the center frequency, not a problem. An I willing to trade 3 dB-ish for a BPB that filters out strong adjacent signals? sure.
I was surprised at the difference 3 dB made in this case. And in a lot of receiving scenarios, loosing 3 dB is not a big deal. I see responders holding their radios at 90 degrees (i.e. at horizontal) and everything still works. Analog systems you just get “a bit more noise” but can still copy.
I was pleased that I can loose 3 dB on the front end from the input filter and LNA noise, and still see the same performance. And with these tests, I now know that 3 dB for this particular signal is a big deal. Of course, all digital systems tend to have a certain amount of “knee” effect when approaching threshold. Miss a syllable in speech, you can put the pieces together. Drop a bit, and everything changes.
Tnx for the comment, good luck with the virus. Sounds like you need about 10 dB of pain reduction

[geckoVN]

@keithma

I’m suffering a bit from covid brain fuzz so struggling to get my head around everything

What you are experiencing is normal for half the population!

[eepete]

About a year later, I re-did the LNA board. The two big changes:

1. The SAW filter in front of the LNA chip is optional. I won’t be populating with it on.
   I have a mount for the Pi Aware filter that attaches to the case for the LNA.
2. Since the power supply was so quite, I moved it to the top of the PCB making assembly easier. There is no noise at all at 1090 MHz with the input terminated in 50 ohms.

The CAD image of the LNA now. Can feed with barrel connector, screw terminal, mini-USB or on the coax. Works with 3 to 18 VDC

Photo of the PCB (no cover on the LNA for this pix). Assembly of the RF part is “fun”.

CAD image of the PCB holder and the holder for the BPF:

Picture of the LNA holder with the filter holder bolted on to it. The output coax has a means to attach to the filter holder. (the black insert). This keeps the pressure off of the SMA connections into the LNA PCB. The filter body (orange) and the holder (blue) can be bolted down into a die cast aluminum box with the cables all coming out of the bottom:

The filter holder was a “fun” 3D design. You can see the insert into the holder to position the output coax correctly (black). I also made a 3D printed support for the cables that can bolt into the box below the filter holder. This also lets me place ferrite around the coax. It takes the stress off of the SMA connectors. Made both a 5 and 6 mm clamp, one for RF58 and one for 214 cable.

Hopefully I can find someone with a network analyzer so I can verify the .5 dB noise figure. At any rate, this should let me install an antenna and not worry about feed line loss. And I can try the system 1st without a BPF before the LNA, and then easily add the Flight Aware one in front if I need to. I suspect the Flight Aware filter is more rugged that a SAW filter which will be a win.

Hopefully another year from now, I’ll have all this up on a small tower. Will advise (with a new post) when that happens.

eepete