There has been several inquiries at thinkpads.com about what kind of MiniPCI antenna works "best" when attempting to upgrade a system that does not have the built-in antennas already installed from the factory. I recently found several eBay sellers offering various designs of these antennas and decided to try to answer some of those questions. In order to test and compare these antennas, this is the setup I used:
 |
 |
| This is the "control" setup; the T23's built-in antenna. I covered up the area around the right side antenna with aluminum foil so I could determine which lead was for the left side since that one was closest to my wireless access point. I found that it was labelled the "main" lead. This was the only antenna lead connected to the Broadcom card's "main" input and that same input jack was used for all antenna measurements. I've rotated the NetStumbler display so that it's easier to compare the different antennas. The scale from left-to-right is the signal strength that is being measured. "Green bars" farther to the right means a stronger signal. Since antennas are "polarized" to some extent, I've measured the signal strength in both the vertical and horizontal positions to see how much it makes a difference. The ME102 access point antennas are positioned slightly off true vertical (they are around 70-75°). The NetStumbler scale that is now up-and-down is time. I took signal strength measurements for about several minutes in each antenna position, except for the T23's lid antenna which was only measured in the vertical position. All of the other antennas were positioned at roughly the same height as the T23's using a plastic straw and a weighted foam base. This was done in order to have a fair comparison between all of the antennas. Note: In almost every case, the antennas that I am measuring CANNOT fit into the T23's lid at the same location where the factory antennas are mounted. This is either due to physical size or lead length restrictions. This analysis is only being done to see which antenna design performs best when all other variables are kept as identical as possible. In some cases I had to stand the T23 on its side in order for the antenna to remain at the same height above the table as in the factory T23 because the lead length was very short. Also, if you click on the picture of the antennas, you will be able to download a copy of the eBay auction since I saved them as a .PDF file. In about six months time they will have been purged from eBay. | |
 |
 |
| The first antenna is a very simple design. It looks like the standard coaxial cable that is attached to the U.FL plug has been stripped back and a portion of the inner conductor exposed from the surronding shield. The length of the tip appears to be about 1.25" which is just about right for what is called a "quarter-wave" antenna. In fact, all of the antennas I'm documenting here seem to be based upon this design. At the frequency which is used by the 802.11b and 802.11g standards (2.4Ghz), the wavelength of the radio signal is almost 5 inches long. One-quarter of that is 1.25 inches. The first thing that is noticable on the graph is that the signal strength is 4 to 6dBm higher than the factory antenna. This can be partially explained by the very short lead, about 23cm or 9 inches long. At the microwave frequencies used by 802.11 systems, the lead length and type of coaxial wire being used can significantly degrade the signal. There is usually a known or rated signal loss measured in dB/foot. In the vertical orientation the antenna is, on average, a couple of dB better than the factory antenna. When turned so that the tip of the antenna is parallel to the ground, but pointed away from the direction of the access point, the signal fell by about 2dB. This is during the first portion of the "horizontal" measurements. I then turned the antenna (still parallel to ground) and pointed it directly toward my access point. This poistion seemed to offer the best reception. From these measurements I would have to say that the best reception for this particular antenna is to have it pointed directly at the access point. One major drawback with this antenna is that it will NOT reach the same height inside the LCD lid if you decide to try to mount it in there and it won't reach most places in a T2x base where you can get it near the outside edge of the laptop and free from being covered by any metal which would reduce the reception. | |
 |
 |
| The next antenna is a small printed circuit board (PCB) with an etched trace that acts as the active antenna. This is the usual type of arrangement for WiFi antennas as it's very cheap and easy to make the PCB and then simply solder on the lead. Total length of this antenna is 20 inches. Since the trace on the PCB zig-zags I decided to call the vertical orientation the one where the trace climbs "up" from the end of the coaxial lead. In this case, the picture above is with the antenna in the "horizontal" position. When I measured the signal strength in the first half of the "vertical" potion of the graph, the flat side of the PCB was facing my access point. After about three minutes I turned the antenna so that the edge of the PCB was facing the access point and that's when you see the signal drop by a couple of dB for the next two minutes. Turning the antenna so that it was horizontal and with the flat side of the PCB facing the access point showed the signal strength did not change. After two minutes I turned the antenna so that the edge of the PCB was pointed right at the access point and the signal strength went up to where it was before. However, this antenna performed only as good as the built in T23 factory antenna in the best case, and slightly less in other positions. One explanation could be that the trace on the PCB is not long enough (i.e., "tuned") for the radio frequency being used with 802.11 systems. My best measurement of the trace length suggests it's only 1.1" long. The PCB itself is 0.8" long and 0.3" wide. | |
 |
 |
| This antenna design is also quite common; a large, square piece of PCB material or plastic substrate with an etched antenna trace. The size of the board is 1.2" x 1.2" and has a double-sided sticky pad so that it can be stuck in place. This particular one is actually one of two with the difference being the length of the coaxial lead. The one I measured has a total length of 38 inches. The other one was 45 inches. This seems kind of excessive unless it was designed for one of those huge Dell "laptop" machines. In the vertical orientation (as you see in the picture above), the antenna's signal strength was about the same as the previous antenna; which is to say, slightly worse than the factory T23. When turned to the horizontal position, the signal strength improved and was just slightly better than the T23. I'm tempted to take one (or both) of these antennas and cut off about two feet of coax and see if the signal gets better. If I do, I'll update this page with the results. I'd say that it would probably improve by one or two dB with a shorter lead. | |
 |
 |
| Here's an interesting design: the active antenna element is actually the small piece of metal you see between the two black pads on the right side of this assembly. The rest of it acts as the ground plane and mounting tabs. In the vertical arrangement (like in the picture), the signal strength is generally about the same as the T23 control. When rotated around to the first horizontal position (flat part of the antenna facing the access point), the signal did not change. However, when the antenna was "pointed at" (the two black pads inline with) the access point, the signal dropped about 6dB. Overall size of this antenna assembly is 1.6" long by 0.2" wide; the grey "panel" you see to the left is just a piece of sticky tape. | |
 |
 |
| And lastly, since I saw it being offered up on eBay and because I have the laptop this antenna was supposedly designed for, I also bought this Gateway 450SX4 antenna. It's 2.6" long and 0.25" wide, not counting the copper foil that's attached to the side. For the signal measurements I did not move/remove the copper tape. Perhaps that's why the signal is much poorer than the others. Or it may be due to the length of the etched trace on the PCB. The length of the coaxial lead is 18"; the second antenna in this set has a 24" lead. In any event, this antenna performed the worst out of the others. Sort of makes me think twice about opening up my Gateway system and putting this in there. | |