Failed: Cheap and Nasty SMA Right-Angle Adapter (& Site Update)

If you’re a bit of a radio equipment user like myself, you’ll find that you’ve always got an amazing array of different connector needs. For example, you might have some older TV tuners with a “PAL” connector, some satellite equipment that uses an F-connector, a spectrum analyzer with an N connector input, a few old scanners that use BNC, an HF set that uses a PL-259, a GPS which uses MCX, and some new SDRs which need MMCX and SMA just to name the most common. Then maybe you’ve also got Wi-Fi equipment with RP-SMA to deal with as well.

To stock all the necessary cable variants to interconnect each and every sort would be prohibitively expensive, so in general, we get around this with adapters and “pigtail” leads. These can be rather difficult to obtain for some combinations, especially “inter-series” where impedances are not matched, and some creative “adapter stacking” might be necessary to get the desired end result.

Having realized all of this, I’ve always looked upon the catalogues in some electronics distributors and felt rather offended that some rather “simple” adapter cost almost AU$50, or sometimes, almost AU$200! This was especially true when I found the same adapter retailing on eBay or in a variety shop for under AU$5 … so I take the cheap road and that’s the end of that. Or is it?

Is there a problem?

Many people will do exactly the same thing as I did – go to a cheap variety shop, pick up some adapters, splitters, connectors, etc and plug it in. Maybe it works, maybe it doesn’t … maybe they decide to throw in an amplifier, blame “physics” for signal loss and everything seems to be swell. Ignorance is bliss.

But in reality, RF connectors are very very precision devices. For consumer applications, where sending or receiving a signal is the aim, the requirements are not so stringent. This is because the signal coding frequently is designed to withstand some losses and effects from inferior connections and cables, the bandwidth the system needs to operate in is limited, the amount of reconfiguration of connections is limited and low cost is the aim. Many users won’t notice the difference between a few dB of loss here and there. But when you absolutely need the best performance, these things can matter.

For those who want the best DX, who may be operating transmitters and repeaters where signal loss, leakage and reflection are undesired, or doing measurements using analyzers over a wide bandwidth, performance of the connector is paramount. A connector, at RF frequencies, cannot be treated as a simple joint in the wire.

RF frequencies involve quickly alternating currents, and due to the physics of the way these signals propagate (e.g. skin effect, transmission line effects), the connector itself has to perform the function of matching impedance. To quote a water-pipe analogy, the connector itself has to be “the same size as the pipe it joins”, otherwise the water pressure will be reduced, and transients in the pressure will “bounce” back and forth from the interfaces (creating standing waves, and being resistively lost).

In order for connectors to properly match impedances and minimise frequency-related variations in transmission losses, their dimensions have to be critically controlled. The size of the pins, collars, slots (if any) and dielectric (if any) size and material are controlled to produce the best match possible.

Sadly, the low-cost options don’t control for these dimensions as well as they should.

Bad Connector! *smack*!

While playing with a new SDR which uses SMA connectors, I decided to use a bunch of right-angle adapters to make wire connections more convenient. These are relatively simple adapters – what could go wrong?


Alas, I ended up scratching my head at some very strange frequency response curves at several gigahertz frequencies, and intermittency in connections when they get “bumped”. As SMA connectors are semi precision, they shouldn’t do this, so I took the connectors apart to take a look.


One of the two looked fairly okay, although this is pretty much a generic consumer grade connector. Note how the centre socket is ovular, has some slits in it, and the dielectric doesn’t fit so well around it. It also has some traces of gold dust on the dielectric, suggesting the fit isn’t so good and it’s wearing out with each insertion and removal (as expected).


But the other one was so far off, that the mating pin had crushed the socket into the side and expanded the dielectric somewhat. That’s what happens when your dimensional accuracy gets a little too slack.

I’ve also had pig-tails made of unidentified coax for RTL-SDR style dongles with PAL to MMCX adapters, but these were so bad that the MMCX end oxidised (not real gold plated) after a few months of service, and the “coax” itself lost so much in the UHF ~600Mhz band that it went from perfect reception on another tuner to no-lock whatsoever (estimated >20dB of loss).

Another MMCX to F adapter I purchased fell apart after a few months, with the MMCX collar separating from the nickel plated body of the connector under very little stress.

After you encounter these sorts of problem-after-problem scenarios, it’s easy to understand why more expensive reputable connectors (when available) might be the better choice, or in some cases, making your own adapter cables with quality branded connectors.


As I move from being a simple minded user with simple requirements, through to a power user with more difficult requirements, through to one who is even more demanding of their equipment in performing measurements, I begin to appreciate the differences between a more expensive adapter and their bargain basement counterparts. The cheap stuff being sold on eBay might work for some basic reception and bridging the “gap”, but if your goal is to do analysis or have optimal reception, they will easily work against you.

Worse than that, there is a real potential for connector damage by mating sub-standard connectors to higher quality connectors. Agilent’s connector grade FAQ details this to more extent, but the goal seems to be to use instrument grade connectors or better with instruments, and use suitable adapters sacrificially to connect devices under test to avoid damage to instrument connectors.

Even more than that, it shows how important it is to check your connectors before mating them, and clean them if necessary to maintain performance. RF will not easily tolerate bad connections in the same way an audio signal might.

Extra: Site Update

Apologies for the recent slowness and downtime. I have been tied up with too many things, and ultimately, didn’t have the time to investigate and remediate as quickly as I should have.

I am aware of a few people who have attempted to comment, only to meet into 500-type errors. The cause is mainly due to an increasing number of disrespectful robot traffic which does not obey robots.txt and will change their user agent if you try to block them via .htaccess. They have been making >33 requests per second which easily swamps the resources available from my “low end” shared hosting, which is causing occasional time-outs which are becoming more frequent. The worst part is that they claim to have slow crawl rates, but are anything but, and they come all at once, as if scripted to do so.

My robots.txt continues to grow, although I fear it is in vain. There are many legitimate robots which I cannot block, otherwise risk losing exposure for this site. There are also many illegitimate robots which cannot be blocked, because they disobey, or they don’t even recognize themselves in the robots.txt file.

Further to this, I also have contended with many bots attempting brute-force logins to the tune of over 70,000 attempts per day which is very unfriendly. While I have had some protection (mainly against brute force intrusion), it did not protect against generated server load. Reconfiguration of the server has been done to make it much more restrictive to hopefully regain some of these wasted resources for legitimate visitors.

At this stage, even though errors are occurring more frequently, mostly around 7-9am Sydney Time (UTC+10), I still cannot justify upgrading the web hosting just yet, as it’s handling with the load on a majority basis. The traffic for this site has grown to a ginormous 455Gb per month, so I suspect I’m doing quite well with that in mind.

It’s a shame that we live in a world where more resources are wasted on illegitimate visitors than actual humans, and time has to be taken away from discovery and experimentation to ensure continuity of service. But yet, seeing as I do learn in the process, it’s not all time wasted.

About lui_gough

I'm a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!
This entry was posted in Electronics, Radio and tagged , , , , , , . Bookmark the permalink.

5 Responses to Failed: Cheap and Nasty SMA Right-Angle Adapter (& Site Update)

  1. Mark says:

    For real connector fun, check out what goes into using VNA’s (vector network analyzers). They need calibration kits/fixtures/connectors (which can cost many thousands of dollars). You have to use a special torque wrench to tighten the connectors to a specified torque. And that little wrench alone can cost over $1000. Many of these instruments use what is basically a high-end super precision SMA connector, but is not an SMA connector. Much fun arises when some bozo tries to screw an SMA connector onto one… they don’t fit, but damn if some people won’t get the hint… and replacing the instrument connector requires a full factory recalibration… again several thousand dollars.

  2. Benjamin E. says:

    Thanks for the site update, I completely agree with you. It’s a shame that those sorts of things happen. In the end, everyone, even the bots lose …

  3. sparcie says:

    I understand the bots being a pain, I think they just look for anything that responds to port 80 requests and then attempt to brute force it (even if they don’t have the relevant php, wordpress or whatever they are attacking). I’ve found them connecting to even my little home-based web server, which luckily isn’t running out of resources because it gets fewer hits. I did however write a python script to perform some basic checks on the requesting clients and ban the dangerous robots by IP address for a period of time that grows exponentially with each offense.

    You could perhaps do something similar for bots that are malicious but that depends on what kind of configuration your hosting is in and how much you can configure yourself. Unfortunately the only way to reduce the server load is really to work out a way to block them (unwanted robot traffic) at the firewall level.


    • lui_gough says:

      I found that having CloudFlare reverse proxy the site for the past year or so made a big difference, but even their browser-integrity-checking wasn’t necessarily enough. Unfortunately, as part of a shared hosting platform, there really isn’t any space for fail2ban or similar solutions because configuring the firewall is out of the question.

      That being said, the user-agent based blocking seems to work better than nothing, and robots.txt is there as a “warning” to the nice bots. Administrative features are now IP-whitelisted so that should make things more difficult for the annoying guys. Serving static error pages is a lot easier and consumes much less resources. Aside from that, I’m just “sailing along” hoping that we don’t meet too many of them at a time, so the site is accessible at least 99% of the time.

      – Gough

Error: Comment is Missing!