Circular Waveguide Antenna for 2.45 GHz / 802.11b / WiFi / WLAN

Latest version of this document | More articles

• Overview
• Construction
• Mounting
• Cost
• Warning
• List of suitable tins
• References and links
• Range tests
• Calculate dimensions using javascript
• Calculate dimensions from a table of values
• Diagram showing dimensions

Overview

We have been experimenting with waveguide antenna, made from old food cans, to massively extend the range of 802.11b wireless networks. All that was required was fitting, in the correct place, a driven element consisting of a short piece of copper wire soldered into the centre of an N-type connector.

One of the antennas made from a J&B whiskey tin. Note - This antenna is for use with 802.11b or 802.11g wireless computer networks or 2.4GHz video sending equipment. It is not for FM / AM / SW / LW radio useage.

This was evolved primarily since the Pringle's can antenna. The Pringle's can, being cardboard, does not last long in a storm, and it is very hard to affix connectors securely. The dipole-less "yagi" bit inside is fiddly to make, and initial tests show the waveguide cans to work better.

From studies of waveguide theory, which gets complicated, it seems that a waveguide antenna or "can-tenna" should have parallel sides, be a good conductor, preferably shiny, and the end needs to be be perpendicular to the sides. For 2.4 GHz the calculations indicate that the can should have a diameter between 70 mm (millimetres) and 100 mm. These are not a "brickwall" limits, but rather roll-off points. i.e. performance will diminish increasingly beyond these sizes.

From practical use we have found that strength is a good virtue, and a fitting plastic lid is almost a must for waterproofing. See appendix for a list of cans found suitable so far.

The ARRL (Amateur Radio Relay League) say that the required waveguide length is at least two guide wavelengths - The guide wavelength is the value of L_g (in the tables of values below) and is dependent upon the diameter of the can. The smaller the diameter, the longer the guide wavelength. This suggests the larger acceptable diameters should be used so the can may be shorter. Also the larger the area of the mouth of the can, the more energy can be tranferred, so the greater the received and transmitted signal.

Construction

First we selected a can with a diameter of 96 mm. We calculated from this the value for 1/4L_g (a quarter of the standing wavelength inside the can), measured this far up from the bottom of the can and drilled a small pilot hole, then drilled the hole out large enough for a chassis mount N-type connector. It is not easy to find 16 mm drill bits in the UK, so we bought a 20 mm cone cutter. To the pin of the chassis mount N-type connector we soldered about 50 mm of 1.5 mm diameter stiff copper wire. This wire was then carefully cut to the value calculated for 1/4L_o. The edges of the N-type connector and the can around the hole were then abraded heavily with glass paper. The N-type connector assembly was then soldered in place to the can, on all four sides. It is important to get a good electrical connection between the N-type connector and the can. Also we have now sourced round N-type connectors which screw into the can (from rswww.com stock no. 112-0773) , and just requires a 16 mm hole drilling.

Inside of the can, showing the driven element. Photo not taken whilst in use.

The cone cutter will cut a perfect hole if the 16 mm washer from the connector is placed over the tip before cutting. With the tools to hand, the whole process took ten minutes.

After a couple of years of experience and learning using these cantennas, it appears that it makes sense to drill a small hole in the can just behind the N-type connector. Thus any rain or condensation which finds its way into the can has an easy route out. The hole should not affect the performance of the antenna.

Detail, showing N-type connector affixed to the can.

Mounting

This antenna has a beamwidth of around 30 degrees and needs aiming. Also the polarisation is important, that is whether the driven element inside is pointing skywards (vertical) or sideways (horizontal) - this needs to match the antenna it is communicating with. We have been mounting them around a standard 25 mm television pole with a U-bolt attached to an adjustable mounting from a television antenna shop - this allows adjustment in the horizontal and vertical plane. This has then had a short piece of stainless steel tubing clamped in, attached to the can with glue and gaffer tape, or cable ties. None of these are a very good solution. This is one point needing more thought...

A can-tenna mounted on an existing television antenna pole.

It was nescessary to aim the antenna before tightening the bolts fully, and check the polarisation was correct. This involved having the N-type cable attached to a PCMCIA (Personal Computer Memory Card International Association) card in a PC (Personal Computer) or laptop with someone down below monitoring the signal, although I have taken a laptop up on the roof - it is not adviseable. An ideal solution would be a hendheld PC with 802.11b card, and pigtail....

Note - always point a can-tenna away from you, and do not look down into the can whilst running. This recommendation is based on caution rather than known danger. Human eyes have very little cooling, and hence are the parts of the body most likely to absorb, and fail to dissipate, excess microwave energy. The can is a focussed microwave device, thus better safe than sorry.

The can connected to an N-type cable.

We maximised the signal by aiming the can-tenna roughly by landmarks, and the use of a compass, and then moved it bit by bit sideways and then up and down until the maximum signal to noise, or link quality was achieved. This required the use of a wireless tool for the PC. I used Wavemon for GNU/Linux, but most wireless card drivers have some form of link quality read-out. Depending on where the monitoring PC was, we needed extra people, or walkie talkies, mobile phone etc. to relay the signal strength after each adjustment. Once the maximum signal was achieved, we clamped up the bolts tight, and celebrated a job well done.

Cost

£5.50 for the N-type connector, and the can of your choice, plus the little bit of wire, and a tiny bit of solder. When we spent £20 on our can we got a free bottle of whiskey. ;-)

Warning

Apart from the fact it works really well, no-one has yet popped on their lab-coat and done any high-brow tests on this "homebrew twig", and of course manufacturers recommend you don't do anything which they don't recommend. Or attach non-proprietary stuff to their stuff. Of course.

Tests on first construction

Our first waveguide antenna had a diameter of 96 mm, with a length greater than 3/4L_g and was made from a can which came with a bottle of gin :-)

The antenna was fitted with a standard television 75 Ohm connector (not a proper N connector). The pigtail used was made by removing the lead from a Buffalo Extended Range Antenna and attaching a standard television 75 Ohm connector. We are aware that the combination of 50 Ohm coaxial cable and 75 Ohm connectors is not a good impedance match and would result in loss, but at the time of testing, in Portugal, that is all we had, and we were keen.

Comparison was made with respect to the internal antenna on a Buffalo PCMCIA 802.11b card. Using Wavemon (a wireless measurements tool) on a GNU/Linux laptop to measure received signal strength, noise and signal to noise ratio.

Results

Our first can, including cable loss we received around +4 to +5 dB (deciBels) improvement in received signal strength, and a +10 dB improvement in signal/noise ratio.

Looking at data tables we estimate the coaxial cable used to have a 1.5dB loss.

This antenna allowed us to maintain an 11 Mbps connection at 200 metres from the Buffalo Airstation Extended Range Antenna. We could not walk far enough with clear line of sight. We were impressed that this improvement was measured even using cheap television connectors from the supermarket (at the wrong impedance).

First mountain test

Finally we managed to get hold of some 50 Ohm N-type connectors. Unfortunately no-one seemed to stock the mating pair so we had no option than to reduce to BNC and back up again at the other end. This worried us not as we were eager.

The only cable we could get hold of was RG58/U, fairly high loss

We attached a gin can waveguide antenna to the Buffalo airstation, via 10 metres of cable, and pointed the antenna out of the window towards the HILL

Iain went to the top of the HILL with his laptop, running Wavemon on GNU/Linux, and a dogfood can-tenna, via 2 metres of cable. The top of the hill has two large multi-sector mobile phone antennas, broadcasting around 800MHz, we believe. Iain was positioned roughly 50 metres below them (slowly frying his brain!).

We had a clearish (tree tops) line of sight down the valley which measured 2200 metres point to point on the military map.

Results

We achieved a 2Mbps connection, with around 7 to 8 dB to spare, although in the excitement the figures got a little lost. Well, completely lost. The important part for us at this point was that it worked.

Having looked at the specification sheet for the cable we were using, it seems not to rate as high as 2.4GHz in the losses, the highest rating is at 1000 MHz which is 0.79dB per metre. This means that by using far too much of the wrong cable we have cost ourselves 9 to 10 dB. This is good news for our next hilltop shot with short good cables, and suggests a clear 5000 metre shot.

The can shown above has given us 16 to 17dB improvement over the antenna on the Buffalo wireless card. This is successful. We are very happy with the results of this can.

We are now awaiting some spare time and enough clear space to do further range tests. We are expecting 10,000 metres on a can to can link.

Appendix

List of tins so far found to be suitable

• Slimfast Double Chocolate - England - with plastic lid
• The Simpsons Double choc cookies - England - with plastic lid
• Douwe Egberts ground coffee - England - with plastic lid
• Baby milk formula - England - with plastic lid
• Furness Ginger Biscuits - Cornwall and England
• Golden Jubilee Beer, Robert Cain Brewery - England
• Nestlé Coffee Mate 500g - England - with plastic lid (thanks to Mark Gaved)
• J&B Rare whiskey tin - Portugal
• Larios Gin - Spain
• Stainless steel toilet brush holder from B&Q - very nice (thanks to Robert Currey)
• Any large dog food tin at a push, if you can't find anything better!

Some colourant additives in the plastic lids affect signal, so test with lid off and on, and if the signal is detrimented, use without.

References and links

• http://www.saunalahti.fi/elepal/antenna2.html - waveguide antenna theory
• http://www.qsl.net/n9zia/wireless/pics/tincanant.jpg - original inspiration
• A whole hand full of library books on waveguide theory, which I didn't note down before handing back...

Circular waveguide antenna calculator - JavaScript

See Fig1.1 for how to use values - most of them are unnescessary and slightly confusing.

• D is the interior diameter of the can
• L_o is wavelength in open air = 0.122 metres
• L_c is wavelength at lower dominant mode cut off frequency
• L_u is wavelength at higher dominant mode cut off frequency
• L_g is standing wavelength inside can

L_c = 1.706D

L_u = 1.306D

L_g = 1 / (sqr_rt{(1/L_o)² - (1/L_c)²})

Ideally for the usual operating range of 802.11b:

• Lower cut-off frequency should be lower than 2400 MHz
• Upper cut-off should be higher than 2480 MHz

Fig 1.1 - Circular waveguide antenna showing design values, click to enlarge.

Table 1.1 - wavelengths and frequencies against diameter

See Fig1.1 for how to use values

Copyleft

Author: Dave Gough

Copyright (c) 2002 Psand Limited. Permission is granted to copy, distributed and/or modify this document under the terms of the GNU Free Documentation License, Version 1.1 or any later version published by the Free Software Foundation; with no Invariant Sections, with no Front-Cover Texts, and with no Back-Cover Texts. A copy of the license is included in the section entitled "GNU Free Documentation License".