|
|
a few hours of the best years of my life Sunday, November 8 2015
The weather took a turn for the colder today, though it was still mild for late early November.
At various times today, I hacked at and fiddled with my caller ID circuit, hoping to make it more reliable. Despite the jumper cables carrying signals outside the rat's nest of wire under the board, the circuit would periodically fail to identify a test call, and once it failed, it often continued to fail for subsequent calls. In desperation, I finally got around to connecting up the recommended low-value capacitors and high-value resistor around the 3.58 MHz colorburst crystal that provides the timing for the EM92547B FSK chip, but these just seemed to make the circuit less reliable, so I soon removed them. Frustratingly, I kept having problem with the pin headers I'd soldered to the generic grid board. These boards are cheap, one-sided, and lack plated vias around their holes, so it didn't take much insertion force to cause the pins to break loose from the boards, delaminating their little copper pads in an instant. To correct this, I often put daubs of thick epoxy around the base and underside of problematic pin headers, but this usually prevents me from making any further wiring changes to the part of the board now buried in epoxy. I've also been known to use SuperGlue to reaffix pins and pads that have broken free, and that was a technique I tried today. Soon thereafter, though, the caller ID circuit began to act as though its EM92547B had died. I tried a bunch of things to get it working, and then I thought, wait a minute, and I tested to see if the necessary five volts was present across its VDD and GND pins. It wasn't. It soon turned out that the SuperGlue and wicked through the board from below, followed the pins into the 16-pin socket holding the EM92547B, and had solidified around three of its pins, including VDD. Interestingly, though there had been a good connection between the socket and the pins before the glue showed up, the glue had completely broken the connection. I could see this being a useful behavior in other applications, particularly one where a short has developed that is unreachable except to a flowing liquid. In this application, though, a flowing insulator was not wanted. Fortunately, I was able to clean the pins of the EM92547B and get most of the dried glue out of the socket using a dental pick, restoring the caller ID functionality. It still wasn't perfectly reliable, but it was back to where it had been this morning. I'd only lost a few hours of the best years of my life.
Later, I decided to implement the precise telephone interface that the EM92547B datasheet prescribes. Instead of isolating the phone line with a transformer and protecting it with zener diodes, the prescribed circuit hooks directly to tip and ring, using only 0.0047 microfarad capacitors and 30 kilohm resistors to isolate the wildly non-TTL-level phone voltages. I had a great deal of difficulty finding a 0.0047 microfarad capacitor from among my many dozens of scrap circuit boards, though I eventually found a pair of them inside a small switching power supply. I terminated the new phoneline attachment as just another two-pin header that would available for experimentation, in addition to the existing transformer-based phone attachment. I soon discovered that this interface was no more reliable than my transformer-based one had been; if anything, it seemed to be a bit noisier. And it's not the sort of thing I would feel comfortable leaving connected in a thunderstorm.
Complicating matters was Google Voice, which I have been using to repeatedly make phone calls to our household land line. The service is free, calls can be made quickly from a desktop computer, and, when I disconnect the household's single phone base station, the calls are unobtrusive. Everything worked great for perhaps 100 test calls placed this way, but at some point Google Voice would act like it was about to place a call but it would fail to do so, doing nothing instead and providing no explanation for why. It was as if some Google algorithm (algorithms being the stuff of which Google is made) had determined that my repeated phone calls constituted a spamming operation or some other abuse of the free service. If I waited about five minutes, I could use Google Voice to make another one or two calls, but then it would throttle me again. It's impossible to test a caller ID system when you can only call it once every five minutes, so eventually I gave up and did something else.
For linking purposes this article's URL is: http://asecular.com/blog.php?151108 feedback previous | next |