Monday, August 31, 2015

An old Audio Spectrum Analyzer project.

Several years ago I had a rack of stereo components that I had collected over the years. Tape deck, FM unit, dedicated amps, CD player, the works. I always though it would be kind of cool to have a nice analyzer unit for the audio. Instead of buying one, I decided that I was going to build one using some already designed circuits and some interesting displays.

I had originally intended to use what I've heard called "Pinball Displays" recovered from old HP equipment. Instead of (expensive!) Nixie Tubes, these were just neon bulbs inserted into an array marked 0-9, which lit up to indicate the digit. I had planned on replacing the neon bulbs with multicolor LEDs. 10 channels of bouncing lights with the option to have rectified outputs to drive analog meters.

The analyzer channels themselves were simply copied from a consumer unit that was available cheaply. The rest of the circuitry, including the precision instrument rectifier board and dot-driver was adapted from reference designs. Everything was going to be mounted modular and connected with DB connectors for easy installation and removal.

One side of the 10-channel filter board and it's accompanying precision instrument rectifier board:

However, like a lot of things, it changed. My audio rack gave way to a laptop with a good DAC, which gave way to a streaming audio device connected to a remote network share. All of that old equipment simply became dead weight, and was either sold, or donated because it no longer held value.

I was never able to get enough of the displays to complete the project, and other things like 1N34 germanium diodes became harder to find. I decided to shelve the project shortly after selling off the audio equipment, and I've been trying to figure out what to do with these things ever since - beside just using them as a talking point on job interviews...

Wednesday, August 26, 2015

Cheap crap bought online, and how to fix (some of) it.

I've always had a few alarm signs scattered around the place, but they've faded to the point of being useless. So a few summers ago, I decided to get some new ones. These signs included a nice little solar powered LED device that you could attach to the sign's pole to illuminate it at night. Cool little things, they lasted all of a season before quitting.

The units are held together with 4 poor quality screws that had almost rusted to nothing. A rubber plug covered the switch. State of the fasteners notwithstanding, everything was still doing it's job of keeping the case sealed and debris free.

It's a simple circuit, the solar panel trickle charges the battery. A transistor turns the LEDs on at night, and I assume keeps the battery from discharging through the panel. The blue-topped switch turns the entire assembly on and off.  This switch appears to be my problem.

I was able to jiggle it with the rubber seal off, and get the LEDs to light once more. Since it doesn't appear to be corroded, I'm guessing the thing is just of poor enough quality to not work. I could care less about turning it on or off, (why is this even needed? Who knows?) and the fix was to solderblob the pins together.

No more problems with bad switches, and something that should not have been there in the first place was fixed. My signs are lit once more.

Monday, August 24, 2015

An outdoor sensor, built on the cheap.

A while back, I wanted to start adding some temperature sensors to my place. The eventual intent is to tie it all in to a control system of sorts. Eventually...

I looked at a bunch of different ways of reading data. The Raspberry Pi was the logical choice, since it supports a lot of different input methods right out of the box, including 1-wire devices. I eventually settled on a commercially available product designed for data center monitoring that I was able to purchase on the cheap from an auction site. It has a whole host of inputs and outputs, and has a built in alarm and email server, all in a nice rack mount case.

Since this thing is targeting commercial customers, the sensor prices are commercial prices. Expensive! But all the components they use are just standard 1-wire devices in special packages. Their outdoor sensor is around $100, which is too much for a hobbyist.

I set out to replicate that outdoor sensor. I disassembled one of the other sensors received with the original purchase (it was dead) to see how they wired it, and got to work.

Materials needed:

  • A Maxim DS18x20. I used DS18S20 to maintain compatibility with my device.
  • Some cable and connectors as needed. I used 4-wire telephone cable.
  • A thin metal tube, cut to about 1.5" or so, cleaned and deburred. Size for your sensor.
  • A cap to fit the tube.
  • Some heat shrink tubing.
  • Two-part epoxy. You may need to experiment to find epoxy that doesn't corrode your leads.
  • Tool Dip.
  • Some sort of mount.

  • Solder the cable directly to the sensor's legs. You'll want to have heat shrink on the wires to slide up afterwards. Make sure all the joint are clean and solid, and shrink the tubing on.
  • Put your cap on the tube and slide the assembled sensor into the tube. Ideally, you'll want to have the outer jacket protruding from the bottom of the tube. Use heat shrink as needed to provide protection for the cable jacket as needed.
  • Fill the tube with epoxy and allow to cure. 
  • Place another piece of shrink at the bottom of the tube to hold the cable in place, if needed.
  • Dip the entire assembly in Tool Dip and allow to cure.
To mount the sensor, I used a simple stainless hose clamp. Nothing fancy, but it holds the sensor tightly and keeps it out away from the mounting surface.

I'm guessing there is some small delay in the sensor's reading, due to the mass of the material around it. It's not terribly important in my application, and this setup has been working flawlessly for the past few years.

Total cost was probably $10, and could be cheaper depending on how well stocked your junk box is.

Sunday, August 23, 2015

My first production board

One of the projects I've been working with is an interface board for an embedded computer system. While I've never designed anything like this before, I've worked with board for 20+ years and have a decent idea of how things should look. The board was designed in NI Multisim, and routed with NI Ultiboard. Both of those tools were unfamiliar to me, so I learned how to use those at the same time I learned how to lay out a board.

It's of moderate complexity, but is doing quite a bit:

  • Provides power for the CPU and on board storage drives
  • Brings a SATA port up for easy access and offers two choices of storage
  • Provides customer USB ports, so they don't have to plug directly into the CPU
  • Provides several troubleshooting indicators
  • Buffers and provides ESD protection for customer-level serial outputs
In all, the most complex thing about this board was drawing all of the components from often incomplete or just plain wrong datasheets. But all said and done, this board turned out pretty nice for a first-time designer. I will admit to making a wiring error on the backside, but that's why we run a prototype before actual production. It's been corrected and all is working as it should.