- LocationN48 - E11
- CountdownT-00D 00:00
This is an old revision of the document!
Software Defined Radio (SDR) has finally reached a much broader mass of people, who wanted to play with RF technology, but didn't find the time or resources to learn all necessary skills, to build a hardware based radio. Thanks to the work of the GNU-Radio and OsmoCom developer crowd, this barrier is finally gone and everyone can, more or less, directly access, what the antenna receives.
The last Mission-Log about a GNU-Radio based NFM SDR receiver pulled in a lot of people, looking for examples, to better understand GRC and to improve their own SDR projects. The real beauty about it is this: Unlike hardware receivers, which can't simply be replicated and shared, we only have to come up with good software receivers/transceivers once and then may just share them amongst each other, without any limitation.
However, the antenna itself, is still hardware and will most likely never be replaceable by software. On ##rtlsdr people often ask about antennas, because they are clearly not satisfied (and who could blame them) with the performance of the original L/4 DVB-T stub. Unfortunately, there just is no can-do-it-all-perfectly antenna, even if some despicable corporations try to market their products as such.
Other people often recommend Discone-Antennas for wideband reception, but there also are other, less known alternatives, which still are a very good compromise as a general purpose wideband receiver antenna. Not everyone has the space or possibility to put up a Discone-Antenna, so why not use an antenna, that performs even better than a Discone (at least it did here in direct comparison), is a lot smaller and looks way less “conspicuous”:
One of them was the Dressler ARA-2000, covering 50-2000MHz, designed and built in the 90's. The company died the usual death by capitalism (bought by another company and then stripped down and moved production to China). Today there are only a few of these left in the wild and are traded for unrealistic prices on $bay. This particular one was used for the Argus-Prototype but sacrificed and disassembled with the hope, that replicating the antenna will be easy, so that this knowledge would get openly reseeded into the wild, instead of being lost in some archives of a dead corporation. It would be great, if the following documentation about the ARA-2000 would inspire more people, to build their own Active Wideband Receiving Antenna (AWRA) and try to improve and evolve the concept even further or come up with completely new ideas.
In order to open the ARA-2000, the black top cap has to be removed first. This can be done with a screwdriver that is pushed under the side of the cover, prying it free. After the cap is removed, the bottom plate needs to come off next. This was a tougher job and required the use of a hot-air gun, to heat up the glue and then carefully applying pressure with a wooden rod through the center of the open tube.
The following section shows the inner structure of the original ARA-2000 assembly, without the protective white PVC tube. Each image roughly represents a 120° rotation step:
The antenna element itself is a simple quadrilateral monopole, in the shape of a wedge, with a narrow start and a wider end. For lack of a common nomenclature and a relatively close optical proximity to a log-per design, this type is going to be ignorantly called log-per-spiral. The monopole is “glued” onto a self-adhesive, semirigid, matte-white material and then rolled to a cylinder with 80mm diameter, thus forming a spiral. Unfortunately, there seems virtually no accessible background data available about the RF properties of this particular antenna design. A NEC simulation would be interesting.
The small start of the original copper log-per-spiral begins at a 25mm offset from the bottom part of the white, rolled 80mm cylinder, the wider end extends 75mm over the upper edge. After 55mm from the edge of the white cylinder, the rest of the copper is bent around the outer tube and then covered by the cap. This has probably no effect on RF properties (can someone verify this?) but is probably a way to give the whole structure more mechanical support.
Copper Element Cutting helper (1:1)
Further analysis and research regarding material and availability lead to the speculative conclusion, that this foil probably is Aslan S22 PVC lamp shade film. The non-adhesive side of the material could be very much described as a satin surface and it's clearly not paper.
PVC seems like a logical choice for this support structure material. It shouldn't interfere with the RF properties of the device and can also be used in an outside environment, where it has to withstand a lifetime of exposure to drastic humidity and temperature changes and extremes, without changing its own form or function.
The transparent outer foil with the printed grid pattern (non-adhesive), which is wrapped around the log-per-spiral and PVC foil cylinder assembly, has the same dimensions as the PVC foil (405x405mm). It's obviously the by-product silicone release liner, that was originally used to protect the adhesive side of the Aslan PVC foil. That approach is actually very neat, since these foils usually end up in the trash and were put to good use here instead.
|Material||Aslan S22 PVC lamp shade film (high probability)|
Judging by the original build quality, it seems that there is some room for tolerances. It should be possible to hack the assembly ghetto-style, out of any rigid PVC foil you can find and just glue the copper log-per-spiral onto it.
The outer cover tube is made of sturdy white PVC, to protect the inner assembly from rain, hail and UV-radiation and is also used to mount the antenna. Even after several years out in the weather, the tube still looks like new. Again, other materials could also be used here, as long as they won't interfere with RF and can withstand weather and UV-radiation. However, experience has shown, that a more professional looking antenna has a higher chance, that other people like neighbors or landlords won't raise objections to the installation. Depending on your local circumstances, that is something you should keep in mind.
The low-noise amplifier PCB is mounted directly on the bottom plate and consists of 2 cascaded MMIC Amplifiers. Although the types of the MMICs are not 100% known, DD1US speculated that they most likely are Avago (Avantek) MSA-1105 cascadable Silicon Bipolar MMICs. The specification, package type and marking (Top A, bottom H) support this assumption. The typical application circuit in the datasheet also seems to match the actual circuit in a cascaded configuration with etched PCB inductors:
Due to the venerable age of the original LNA, it is very likely, that more recent semiconductors can deliver superior performance compared to the old design. The LNA is going to be replaced by a new LNA based on Infineons BFP420 which is cheap and available and should perform equally or better. The following two schematics show typical LNA configurations for the BFP420, the left one is the most simple approach (to be tested first), the right picture shows a more refined approach, with better base/collector voltage stabilization.
Both designs should also be equipped with a 50MHz high-pass filter between the antenna and the LNA input, to increase their large-signal immunity by attenuating lower frequencies, which the rtl-sdr or OsmoSDR can't handle anyways (everything below 60MHz). Additionally, it would be worth a try to compare the following cases in real-world tests:
Somebody was also thinking in these directions:
Another possible way which would keep the price low would be using inline amplifiers, such as these one: Wentronic Inline Amplifier
The internal PCB is not that bad and it can be powered via a bias circuit. After removing the back cover you can take a look on the insides:
Here you can see why these Amplifiers get all these bad reviews. I left and right mark show long wire endings, which could lead to bridges in the circuit. In the middle there is some solder on one via at the PCB. Again this dirt could cause a short circuit. All in all the inner manufacturing was poor but nothing that couldn't be solved. After some research the MMIC that is used was found as a BGA2709 manufactured by NXP. This MMIC can be powerd with 5V and therefore directly by an USB-Port. Beside the loose solder in the case there were two little pieces of copper wire.
This is a close-up of the inside PCB and the used MMIC. Confusingly the pin-numbering of the MMIC is obviously wrog and I'm not sure if that is a mistake / stupidity or a distraction to “complicate” reverse engeneering. The marking on the top of the IC says “E3”, which fits perfectly to the BGA2709. I found this list with case markings of about 22k parts.
The schematic revealed no surprise. It seems that a Zener-diode was used to stabilize the voltage of standard 12V BIAS supplys to the needed 5V. On the input-path there seems to be a high-pass filter after the decoupling C.
In order to be used with a direct power supply you have to remove two inductors on the input and output Path of the power supply. In order to be used in different positions in the signal path I solderd two SMA plugs in the case, after removing the standard F-connectors. I attached an USB connector to the power supply line and after a short test the amplifier was ready to be closed and used.
This one was equiped with cables to be installed better.
Additional pictures of the first tests and some measurements will be done as soon as there is some spare time for those things.
To put it in a nutshell: Thes amplifiers are just amazing after you clean them and I was surprised of the performance. A comparison to LNA4all and similar projects will be done as soon as the orders arrive.
Although it won't be used anymore, for sake of completeness, here are some images of the original Dressler Bias-Tee, that was used to feed power to the LNA through the coax cable. It was supplied by a 12V power supply. It seems that the voltage feeding the MMIC's was kept constant and an adjustable attenuator (the blue part) was used to prevent receiver input overloading.
Ideally, the log-per-spiral assembly should be simulated with NEC to get a better understanding of the design principle. Afterwards the antenna should be evaluated with a network analyzer, to find out if there is any room for improvement, leading to evolution instead of simple replication. But, as Lord Kelvin already said, a long time ago:
If you can not measure it, you can not improve it.William Thomson, 1st Baron Kelvin
Since the lab has no vector network analyzer yet (it's on the Wishlist), there currently is no tool available, to be realistically able to improve the design. Therefore, the antenna element should be replicated according to the original design, because it worked surprisingly well for years. When looking at the production quality of the disassembled antenna, it seems that this design type doesn't have the usual constraints regarding precision as a resonant design would.
The following assembly guide is a conclusive mini-howto, trying to best guess the original assembly instructions, based on the disassembly and reverse engineering process:
As soon as the new LNA prototype is tested and all other relevant parts are delivered, the new prototype is going to be built and a more extensive and practically proved assembly documentation will be released.
This would also be the perfect scope for some SDR-Wideband-Antenna-Building workshops, so if you're interested in having/building one of these too, please drop a note, so that it can be planned. It should be possible to build this Antenna for less than 50EUR.
NFM GNU-Radio Receiver for RTL/OsmoSDR