A long time ago … maybe even eight years ago in 2009 … a computer shop I was shopping at had a network camera on sale for the low low price of AU$30 and I decided to buy it just for fun. I didn’t expect much of it, and it turned out to be an Edmiax IC-1520DP, which was the cheaper “non-wireless” version of the IC-1520DPg. It was simple – 12V DC power and LAN is all it needed.
From day one, I already realized why it was on sale for that price – it had absolutely horrible picture quality, riddled with noise, and a fixed output size of 640×480 pixels. The PTZ features were “digital”, otherwise meaning taking crops from a fisheye image, that only made the output even worse. The interface was clunky and outdated with a limited MJPEG framerate, and the dynamic range of the camera meant that it was even hard to tell whether it was blue skies or overcast outside. Its fascinatingly low sensitivity meant that any night-vision was not possible. But for the price, I could live with it even if it was more of a convenience feature.
As a result, the camera took residence on my second-floor window sill, looking outside over the backyard and adjacent areas, running 24/7 just in case I wanted to virtually stick my head out of the window. To reduce annoyance to neighbours, I even taped over the LED indicators.
After all those years of baking in the sun, its plastic had started to yellow and the matte coating degraded into something sticky. But worse still, it was starting to become unstable, dropping its stream from time to time, and hanging up for a few minutes at a time. When I retrieved it off the sill, I immediately knew what was wrong. It absolutely stank of capacitor electrolyte.
Being the crappy camera that it was, I was prepared to dispose of it. It had done its time, and as an “IoT” type device, it’s likely to be riddled with security holes of sorts. But I decided to at least take it apart first.
Removing four screws allows the case to be removed. Internally, the camera is made of two PCBs, one which houses the camera sensor which has both pin-based board interconnects as well as a four pin wire harness, and the other handling the power and data.
The camera PCB has a thermal-pad style material draped over the top of the camera module, however, it doesn’t contact anything within the case. It may have been used as a heat spreader. The camera module is marked 1246-00000051-01Z Rev.A with a date of Week 10 in 2008.
The module has 40-pins of interconnect arranged as two columns of 20 pins. This is in addition to the four wires in the harness. Two screws secure the lens frame assembly to the board.
Under the thermal pad is a MARS MR8916A-LF, which is presumably an image processor of sorts although no information seems to be available.
The front side has some EtronTech EM636165TS-7G 1Mbit x 16 bit 143Mhz SDRAM and an SST 59VF512 (?) Parallel Flash.
The lens uses a screw-thread into a plastic retainer frame to achieve manual focus. I suspect this may have some heritage from the CCTV side of things.
The sensor itself was not identified, although it seems all the years of sun have made parts of the sensor appear brighter and more matte than the darker tree area in the backyard which results in the darker portion on the right. Image sensor damage after all these years? Quite possible. Always nice to see the gold bond wires.
The top side of the baseboard has two MP1410ES 2A step-down power controller ICs, a 25Mhz clock crystal and a single PSC A2V28S40CTP 16MiB SDRAM. There are also a number of diodes, LEDs, tantalum capacitors and ceramic capacitors which look all good. The board identifies with a code of 1244-00000449-01Z.
The rear side is dated Week 52 of 2007 and has the main guts including a Realtek RTL8650B “Advanced Home Gateway Controller” which includes 6-port switch MAC, 5 Fast Ethernet transcievers, LX5280 32-bit RISC CPU up to 200Mhz, two UART ports, USB 1.1 host controller, PCM interface, 22 GPIOs and a PCI host interface. It’s clear that this camera is built around a chip which was originally designed to power routers. The PCI interface is broken into an mPCI footprint that was not populated – this socket would have been installed with an appropriate wireless card in the wireless model. The chip is accompanied by an MX 29LV160CBTC-70G 16Mbit Parallel Flash chip. There seems to be a transistor/regulator REG1 which has been getting hot leaving some discolouration on the board. Aside from that, other major components include an M-Tek isolation transformer for the Ethernet link.
The culprit? Suncap capacitors. In fact, in terms of electrolytic capacitors, the total list is:
- 1 x 470uF 25V
- 2 x 470uF 16V
- 3 x 22uF 16V
- 1 x 10uF 16V
Seeing as it’s just a total of seven capacitors, I decided to go and do a repair just for the sake of it. Because of this, I promised myself not to order any parts and use whatever I had “left over” to get it on its legs.
The first thing was that I had no 470uF at 25V, with only 470uF at 16V. As the DC input was only 12V to the camera coming from a regulated 12V switching supply, and I was using quality Panasonic capacitors, I reasoned that a 16V rating was “adequate” but not ideal.
The next issue was that I didn’t have any 22uF capacitors at all. The closest I had were 47uF capacitors which were more than twice the rating. As the circuit only seems to use them as bypass capacitors near sensitive devices, a larger value shouldn’t hurt, although it would strain the regulators a bit more on start-up. I decided this was an appropriate enough compromise.
As for the 10uF capacitor, I didn’t have one of 16V rating, but I did have one of 25V, which provides additional margin and would be just fine to install.
Fast forward about an hour, and we’re done. New capacitors installed!
Unfortunately, the look underneath is not so nice.
The board was an absolute *expletive* to desolder. It was not designed well for hand-soldering and its lead free solder had gone really crusty, taking a lot of work with braid to clear (as the holes were so narrow that the solder-sucker bulb didn’t get far at all). I didn’t clear off the flux from the desoldering braid, resulting in the brown caramel smears all over the place. Worse, as many pads were connected to large planes of copper, I needed to apply a lot of heat to get the solder to wet the pads, which made the flux even more nasty. I decided to over-apply solder to help “carry” the heat to the board, and thus the joints look “blobby” but are mechanically and electrically sound.
Replacing the capacitors made sense – from top to bottom, left to right was the 470uF 25V (close enough to the original rating), the two 470uF 16V (way under), the three 22uF 16V (below spec) and the 10uF 16V (within spec).
Of course, I forgot that the new capacitors were mechanically taller than the ones they replaced, so the case wouldn’t close. A common issue, especially where many products are designed with cheaper caps that are somehow impossibly physically small. As a result, I drilled a few holes and reamed them until the case closed. With that … it’s case closed!
It works … just as well as it did before. Which is … not very well.
The Edimax IC-1520DP has all the hallmarks of being built around a router SoC, in the early days of wireless technology where integrating wireless meant shoving a mPCI card into your design. Considering its eight year age, it had already lived a plentiful life. It wasn’t a great camera, although it was outfitted with low quality capacitors. Despite this, and the fact it was exposed to high temperatures due to being in direct view of sunshine, it did manage to exceed the expected capacitor lifetime (eight years at 24/7 is 70,128 hours). As it didn’t need much work to revive, I decided to do so with parts I had left over, resulting in an ugly hack with some potentially ill advised component substitutions, but it does work.