Recasing the Linksys E3000
Posted by khlae under Projects | Permalink | | Leave A Comment
This project came to be because of the shortcomings of the Cisco-Linksys E3000 case design, with its excessive thermals. Idle temperature of the stock E3000 case is an average temperature of 100 degrees Fahrenheit, and upwards of 120 degrees in certain hot spots, never mind the actual temperature of the parts on the inside, or the thermals encountered when this router does some work. To improve cooling, the replacement case is made of aluminum. Placing the external antennas is a side effect caused by the choice of a metal case, since such cases tend to be good RF shields, so by their inherent nature, will render internal antennas useless.
Semi-final case design and layout. This image lacks wiring for the antennas, which is not very special. The antenna layout mimics the stock E3000, except that there are an additional two antennas on the back, since my external antennas would not be dual band antennas like the stock PIFA TX/RX antennas.
Most recent render of the design. On the top, there is the Wi-Fi button, just under the light on the top. In building this project, I omitted this button, since it would not find a use in my application. This render is lacking the front panel image that shapes the glowing illumination so that one may tell what each light means (see below). The rectangle seen on the front is a piece of transparent plastic, there to make the front look nicer, else it would look messy and ugly due to the way the light pipes and my lazy light masks work.
The front panel labels. The white parts go right in front of the light pipes, so that the shape itself glows.
Whew. With all the designing and measuring out of the way, time to go shopping. My shopping list looked like this:
Case:
-Adhesive-Backed Feet
-Flat Black spray paint
-Metric Hex Nut, M4 Size, 3.2mm High*
-Metric Phillips Machine Screw, M4 Size, 12mm Length*
-Hex m4 standoff, 5mm long*
-RG58 cable **
-UMCC cable assembly (3)***
-Female SMA connector bulkhead mount (6)
-90degree light pipe, for front LEDs ( ~20mm total length)
-Straight light pipe for top Led (1)
-Aluminum box, 10x6x1
-Aluminum box cover
Antennas:
-polyolefin Heat-Shrink Tubing 3/8″ ID
-male SMA connector inline****
-Bare Copper Wire, 051″ Dia.
-Acrylic Rod 3/16″ Dia.
Total cost:
~$97 USD
* Metric sizing was chosen because it somehow saved about 2 dollars.
**RG-58 was necessary for the antenna design (a whopping 6 inches of it!), and to simplify things I also used it for the 2.4 GHz chassis wiring. Bad idea. RG-174 is much more chassis-friendly, although it is more annoying to cut and strip.
***the concept of crimping on my own U.FL/UMCC/whatever equivalent connector was not appealing, so I just bought some close enough pre-made assemblies and cut them up.
**** For future reference, right angle connectors of a little higher price are worth buying.
There are some things missing from my shopping list, like the front plastic, and material to make the button on the top. I already had that stuff lying around, so I recycled that stuff. Had I bought material explicitly to make the WIFI button, I would not have gotten lazy and skipped it. The use of acrylic as the structure for the antennas is theoretically a poor choice, since the velocity factor of acrylic is relatively low compared to alternatives. The actual impact on antenna performance is questionable. If I was to redesign and build the antennas properly, the total performance increase might reach 1db or more. On to building!
Test fitting the e3000 board. Everything going to plan and fitting nicely so far.
Close-up of one of the standoffs. Metal all the way should yield vastly superior cooling, if the heat can get to the bolts/standoffs and conduct into the casing. Careful use of thermal goop could help here. It is worth noting that the E3000 circuit boards top and bottom layers are almost entirely a single copper ground sheet, which probably serves as an RF shield, and helps cooling. As a side note, the mounting pattern for the E3000 is a trapezoid, so you can only mount the board one direction. This also makes hole measuring a little more annoying.
Main case with all its holes cut out and the protective plastic removed, I suspect it’s just posing before painting, although it needs some deburring. Camera shake attack makes everything look less good. At this point, painting commences, and while the paint dries, I get to work on things like that chassis wiring and antennas!
Know how the 2.4 GHz side of the E3000 has built in antennas? Yeah that turns out to be quite annoying. To make attaching the new cables easier, I did a potentially risky maneuver here. See the large solder blob right where the outer braid of the coaxial ends? That connects straight to the board’s common ground, which coincidently is connected to the majority of the copper traces on the bottom, something that is a quite nice feature.
What I did to make coaxial attachment simpler here, was I put some holes (3 to be exact) through the middle of the coaxial ground pads. One such hole is illustrated above. I then ran a loop of wire through the hole that tied the coaxial down to the board as shown below.
This makes life much easier, although it is possible to get dangerously close to the wrong trace doing this, and some boards might be sufficiently different enough to render this hole drilling disastrous.
What you will notice here is that the bottom of the board is mostly one common plane. That is the shared ground for everything, and in part, why we can run conductive wire through a hole in the board– the two traces we might short together are already shorted elsewhere.
Paint’s dry enough to touch! This is actually about 24 hours after painting; it seems I found a super dusty spot to stick it for a few minutes before I thought to take this picture. The dust and stuff on the top isn’t actually stuck there. Again, camera shake makes everything look ugly. It seems my hands (or image stabilizer) do not like this case. You can see some of the antenna construction in the background on the drill press.
Since at this point most of the router has been done and pieced together, all that is left are the antennas. The antenna I am holding here did not make the cut, and it was shortly gone after this picture was taken. Nevertheless, the basic design is the same. A hole is drilled through the center of the acrylic rod that allows the coaxial cable to go through the acrylic and support it, and on the other end, the coaxial is soldered to the main antenna wire, which will wrap around the plastic rod. The whole bottom then will be insulated and shielded.
Some semi final antennas just before attachment and straightening to make sure they all work right. There is one small flaw with these antenna designs: 3/16″ acrylic rods circumference is too small to match the 1/4 wavelength of 2.4 GHz Wi-Fi, and too large to match the 1/4 wavelength of 5/5.8 GHz Wi-Fi. The impact should be greater for the 5 GHz antennas (left), since they have more elements (about 9db worth), versus the 6db gain for the 2.4 GHz antennas (right).
The antenna configuration for the 2.4 GHz antennas, as planned, is as follows:
1/2wavlength line, 1/4wavelength loop, 3/4wavelength radiating element, 1/4 wavelength loop, little less than 3/4 wavelength radiating element.
The 5 GHz antennas are mostly the same, except that there are an additional two 1/4 wavelength loops, and two 3/4 wavelength radiating elements.
Antenna design was based on MartyBugs’ compact 6dbi collinear antenna.
After making sure antennas are all functional and good, it is time to finish them off. Steps for this were comprised of clean up shielding on the bottom, and insulate them. For the bottom, I used a bit of heavy-duty electrical tape, since the small diameter heat shrink I used for the rest of the antenna will not go around the bend.
The finished result, all assembled, just before first full power up. The front plastic appears a bit bowed, since it had not been fully attached as of this picture.
Action shot! You can see how the front lights turned out. Truth be told, the camera makes the front look better than it does, and hides the one flaw I made here: do not cheat and use a laser printer to make the LED masks. Unless you glue the paper perfectly, which is quite hard, you will have some visible inconsistency in the finish where the stress of supporting the plastic might warp the paper, and inconsistent glue might make some strange marks. As a fix, if one soaks the paper in diluted glue before hand, and very firmly and evenly compresses the plastic-paper-case stack until it dries, the finished result might look acceptable. When I get the time, I plan to redo the front, and actually paint the led masks on the plastic.
Rear panel of the case. The labels from the original case fit nicely on the new case, and make for some nice labels that look quite professional. The USB port looks unaligned, although it is not.
In the end, these modifications were well worth the effort. The new case is vastly superior at cooling the parts, with the average case temperature a mere 15 degrees Fahrenheit rise over ambient, and the peak rise 25 degrees, approximately a 50 percent reduction in temperature rise over stock. Wireless throughput roughly doubled, but that might have been a fluke. I have not been able to test accurately the wireless range, but from rough tests, its horizontal range has been vastly improved to far beyond nominal 802.11n supposed range.
