Node DDG, Melbourne Wireless
In the true tradition of amateur WLAN antenna design this yagi has been borrowed from designs by people much smarter than me. Unlike the so called Pringle can "yagis" that are pretty common this is a traditional yagi based on the design principles of DL6WU with folded dipole inspired by Rainer Jaeger's excellent article "DL6WU-Yagi for 2320Mhz".
Theoretically the design should be good for 15dbi but the prototype clocked only 11dbi. We'll discuss reasons for this later.
The design uses low cost materials but took six hours to build. I'd love to see some stats of gain versus total cost (materials + time) for home built WLAN antennas. I suspect this design would low on the list, but on the plus side it looks very pretty :-)
The materials required are cheap and easy to find at your hardware store and hobby shop.
The tools were the most expensive part of this project. Listed below is what was purchased but you may not need all of it
Here is what the final product should look like. The picture was taken with a cheap web cam :-(
The first thing to do is to use DL6WU's design rules encapsulated in this excellent spreadsheet (sorry it's in Excel format only). A centre frequency of 2442 MHz (channel 6), boom diameter of 1cm, through boom elements = Y and element diameter of 3/32", gives the following element positions and lengths:
| Frequency MHz. | 2442 | |||
| Wavelength cm. | 12.28 | |||
| Boom Diameter cm. | 1 | |||
| Element Diameter mm. | 2.38125 | |||
| Element Thru Boom ("Y/N") | y | |||
| Boom Length (Metres) | 0.5 | |||
| Gain (dbd) | 13.9 | |||
| Thru Boom Correction (cm). | 0.67 | |||
| Useable bandwidth MHz | 2393.16 | to | 2490.84 | |
| ELEMENT | Boom | Distance each | ||
| Length | Position | Side of boom | ||
| REFL | 6.09 | 2.00 | 2.55 | |
| DRIV | 5.93 | 4.95 | 2.47 | |
| Dir 1 | 5.70 | 5.87 | 2.35 | |
| 2 | 5.62 | 8.08 | 2.31 | |
| 3 | 5.55 | 10.72 | 2.28 | |
| 4 | 5.49 | 13.79 | 2.25 | |
| 5 | 5.43 | 17.22 | 2.22 | |
| 6 | 5.38 | 20.91 | 2.19 | |
| 7 | 5.33 | 24.77 | 2.16 | |
| 8 | 5.28 | 28.82 | 2.14 | |
| 9 | 5.24 | 33.06 | 2.12 | |
| 10 | 5.20 | 37.48 | 2.10 | |
| 11 | 5.16 | 42.08 | 2.08 | |
| 12 | 5.12 | 46.81 | 2.06 | |
| 13 | 5.09 | 51.60 | 2.05 | |
Taking the 600mm ruler mark out these positions along the centre of the channel with the centre punch as accurately as possible.
Drill 2mm holes through the channel at the positions marked. Only drill holes for the directors, holes aren't required for the driven element and the reflector. Even though the director elements have a diameter of 2.38mm (3/32") we drill 2mm holes initially because we want the elements to have a force fit.
Using a 3/32" drill bit ream the 2mm holes on each side of the channel leaving a 0.2mm flange on the inside of each hole. The depth of the flange isn't critical just redrill each hole with the larger bit until a small flange is left See the diagram below:
The directors are simply cut from the length of brass wire and filed carefully to the exact length, that's right, to an accuracy of 0.1mm. This is easy with the caliper if you first set it to the correct length and lock it. Then file the brass rod little by little until it just fits in the jaws of the caliper. Finally file a slight taper to both ends of the rods to make it easier to push through the holes. This process takes a long time!
Once the directors are completed they can tapped roughly into place using a hammer. They are then centred by measuring each side of the director and then lightly tapping with the hammer until both sides are equal. At this point you may discover that the holes have not been drilled perfectly straight, resulting in directors that are not quite perpendicular with the channel. Use a pair of pliers to bend each side until it's straight. Using a drill press to drill the holes in the first place really helps here.
The dipole on the prototype was made from the same brass wire as the directors. It would be a better idea to use thick copper wire if you have any as the conductivity of brass is about third that of copper resulting in a less efficient antenna. Copper is also easier to bend. In my case the brass was all I had at the time and as the antenna was only going to transmit 30mW, I figured the I^2R losses wouldn't be too bad. I might rebuild it with copper and see if the gain improves much.
The diagram below shows what it should look like. Use a 12mm drill bit as a former.
The dipole on it's own will have an impedance of around 300 ohms. This will be reduced when it is placed behind the directors, but will still be around 200. To reduce this to 50 we need to construct a balun from semi-rigid coax. The coax needs to be 180 degrees electrically. Channel 6 (2442Mhz) has a wavelength in space of 12.285 cm. A half wavelength is therefore 6.14 cm in space, but the semi-rigid coax used has a velocity facter or 0.70 so the physical length required is 4.30 cm.
The prototype used 0.141" semi-rigid cut to 43mm of outer shield with 10mm of inner connector left at the ends. This needs to be bent around a 12mm drill bit or similar form until it's almost a full circle avoiding any inks in the shield which will alter the impendence. This is very difficult to do with 0.141" coax, 0.086" coax would be better.
Solder both inner conductor and shield of one end of the balun to the shield of the N connector at the point where the dipole is soldered to the shield on the N connector. The inner conductor of the other end of the balun must be soldered to the solder tag/inner conductor of the N connector allow with the other end of the dipole.
The shield of the N connector is a relatively large piece of metal and will require a large soldering iron to heat it. I found that a 25W iron is useless for this job, an 80W beast handled it with ease.
The finished dipole/balun assembly can be tested on it's own and should give a few db gain over the inbuilt antenna on your wireless card.
The reflector is simply a rectangular piece of sheet aluminium 5cm x 8cm, supported from the rear with a triangular support which in turn is bolted to the boom. See the dphotos below:
Draw a 10mm square in the centre of the reflector with a pencil and ruler, then drill a hole in the centre and file edges of the hole until you have a 10mm square hole.
Construct a support from sheet alumimium as shown in the photos and bolt it to the reflector using 12mm M3 bolts and nuts. Drill a 3mm hole in the top of the support for bolting to the boom.
The diagram and photo below shows where the dipole and reflector assemblies are placed on the boom.
First measure and mark 13mm either side of the mark made for the driven element earlier. Drill a 3mm hole at each mark on the top of the channel.
Place the two 32mm M3 bolts in the holes and screw a few turns into a single 10mm tapped spacer. Slide the dipole down the boom until it's in position and then screw in the two remaining spacers and the nuts to hold it in place. The prototype used tapped spacers but any thing would do as long as the dipole is not touching the boom. If square tube was used for the boom you will only need one spacer per bolt instead of two.
The reflector assembly is then slid down the boom until it reaches the mark made earlier. A hole for the support bolt can then be drilled into the boom and the assembly bolted on. The reflector should be about 29mm from the dipole but you can vary this to adjust the impendence.
Finally, I haven't detailed how you can fix this to your antenna mast. For the prototype I drilled two 6mm holes about 7cm apart behind the reflector. Using 6m bolts the boom can then be fixed to a plate which in turn is held to mast with u-bolts.
Once again in true wireless lan style I don't have any test gear to test gain other than a Lucent RG-1000 access point and the link quality software built in to it.
In comparison to a commercial 8dbi omni I had available it shows consistently 3db more gain. Hence I put it's gain at 11dbi. This test was only run on channel 10
Theoretically a yagi of this design could manage 15dbi so what went wrong?
Modelling a yagi with nec-2 shows that the design is fairly tolerant of errors to the position and length of the directors providing that each director is larger than the one in front. A more critical component is the balun where a 1mm change in length alters it's resonant frequency by about 50Mhz. Due to my poor soldering the balun is not quite as close to the dipole as it should be which would increase it's effective length, dropping gain. The lower conductivity of the brass dipole could be problem also.
I need to do additional testing on all channels to see if the balun is indeed too long. If that's the case the gain should be better on the lower channels. Stay tuned
Ok, I've now checked the SNR across all eleven channels and the results aren't exactly what I predicted.
The chart above shows the SNR measured at the client and access point both with the yagi and with just the 802.11b card on its own. I predicted that the antenna would probably perform better on the lower channels because the balun was a little long. Another person commented that the directors looked a bit long which would also favour better performance at lower frequencies. From the chart it looks like the opposite is true, the yagi performs better on the higher channels.
As mentioned in the intro the materials bill for this design is pretty low, about $AUD30. The labour cost however is another matter. 11dbi for 6 hours work makes the Pringle cans look pretty good :-)