Hey folks, The Old Crow here. Before I continue with my retrospective on the venerable yet still useful LM3914/15 chip, I thought I’d point out this very nice analog design reference guide available for free online from TI. It covers all sorts of stuff from op-amp configurations to A/D and D/A interfacing to circuit board trace characteristics–you name it. I keep this on speed dial on my iPad along with Hans Camenzind’s Designing Analog Chips and Bob Pease’s Troubleshooting Analog Circuits on the nearby bookshelf.
Back to analog LED stuff. In 1979 Popular Electronics, one of the best hobbyist resources of the 1960 and 1970s, had an article in the September issue about how to build your own 10-band audio spectrum analzyer. (See pdf page 58). This project used the LM3915, which used on-chip resistors trimmed to have the LEDs respond in 3dB log steps. Note that this is different from the VU meter (which is also available as the LM3916) as a VU meter has a specific response curve based on the classic analog response curve of an ammeter movement fed from a full-wave bridge and was meant to enable folks to set the peak expected signal level against the headroom of whatever was transmitting or recording the audio. Thus the LM3916 has a kind of quadratic response curve. A spectrum analyzer is a device meant to measure the energy in the various bands of a given spectrum as opposed to dialing in peak levels, and the common use in audio was to use a spectrum analyzer to set the equalization of a sound system to the acoustics of a given room. The idea was to play speech or audio through the sound system with the analyzer listening, then set the band-stops of a graphic equalizer that was part of the audio chain such that the response shown on the analyzer’s LED display grid was as “flat” as possible. This cured the various problems often encountered in room acoustics such as “muddled” bass or attenuated high frequencies. A professional audio spectrum analyzer circa 1979 cost several thousand dollars, so the appeal of this project was immediately evident.
It was about this time I started doing intern-type work for a local factory instrumentation manufacturer, who at first let me tinker without really expecting me to make anything useful, but when I started making things that actually worked (usually better than their commercial products!) the boss started to give me more serious things to work on. This is how my microcontroller-based PID controller line was born, but I’ll tell that story another day. Anyway, one of my mentors was an engineer there by the name of George Haddox who was the best “seat of the pants” engineer I’ve ever known. What I mean by this is he impressed upon me that it is not knowledge or study that marks the standout engineers, but it is tenacity. One has to be prepared to blow something up 19 times in order to get it to work the 20th time, and then be prepared to do it all over again. George saw my copies of the spectrum analyzer project pages scattered about on the workbench and asked about it. I didn’t have a way to make the PC board, I told him. We had a circuit board lab which I will go take pictures of the next time I am back that way that I was free to use, but without something to print a negative from making it meant I either had to hand-tape a new positive or somehow use the one in the article without letting the image degrade, otherwise the copper etch would be fuzzy, have all kinds of little pock-marks and copper bridging, etc.
“No problem,” George said, “if you don’t mind losing the original page.” So given this was George, I decided it was a fair gamble. He took the original magazine paper page, cut out the trace pattern image, pressed a piece piece of clear transfer plastic onto it and used a roller to press out any air bubbles. Next the half-laminated page was soaked in water until the paper pulp could be wiped off. The adhesive had retained an imprint of the artwork which was then shot under the UV light box for 2 minutes onto the (sadly discontinued) Kepro orange circuit board film to produce a perfect positive. He then shot the postive onto a second piece of film to create the working negative. Once you have a working negative, the rest is procedure. You expose a fresh piece of sensitized board stock under the negative in the UV light box for about 3 minutes, peel off the plastic cover layer, then soak the boards in a tray of developer solution until the unexposed sections of the photoresist turn a cloudy color at which point a brush is used to loosen it. What you are left with is the UV-cured photoresist describing the trace pattern of the board.
When etching circuit boards in-house, we had two big 25-gallon etchant tanks made (like all things DIY PC board-related) by Kepro Chemical Co. We also used these curious rigs hand-made from PVC pipe hold the boards vertically in rows so they could be immersed in the tanks. The etchant was ferric chloride (FeCL₃) which was guaranteed to eat through clothing, jewelry, etc. so old clothes and lab smocks were a good idea. Boards were placed in the PVC holders which were then lowered into a tank for about 25 minutes. The tanks are covered and kept at a constant 120 degrees F to accelerate the etch process. At 25 minutes you took the boards out, rinsed them, inspected them for the amount of etch progress, then flipped them 180 degrees in the holders and put them back in the tank for another 20 minutes. Then they were removed and rinsed and checked once again. Sometimes they were ready, other times (it depended on how old the etchant was) they needed a few extra minutes to complete. When running an etch bath the operator gained a kind of intuition of the process over time such that one grew used to eyeballing a board and knowing how much more (or less) time the next board batch would need.
Once the board etch was complete, they were rinsed thorough, dried off and then sheared to size using the corner shear-marks (see first image). At this point they were then drilled as the hardened photoresist tended to help guide the drill bit to the etched pad hole marks. I used a Dremel press and variable-speed Dremel tool loaded with #60 and #70 bits to drill the boards. Drilling time of course depends on board size; this one took about 15 minutes. Two boards were made as it was always a good idea to make multiple boards in case the was an etch problem with one. The 2nd image shows the actual “backup” board. This one had a few minor issues such as trace bridges and “pock marks” left by the photoresist so the other board actually got assembled. This one is still good however, just a few alterations with an Xacto knife it it is good to go.
The final step in the board casting process is to use a “photoresist stripper” solution which is essentially a weak potassium chloride solution. The boards were immersed in this solution and watched carefully as KCl is fairly corrosive and once the photoresist started to “peel up” you wanted to remove the boards and rinse under hot water as quickly as possible. About two minutes for this step. An optional step is using some kind of electroless immersion plating, such as immersion tin plating solution or immersion gold plating solution. The tin plating shown here has oxidized somewhat as this board is over 30 years old. Still, with a quick cleaning using steel wool, this board could be built today.
In part two I will demonstrate the assembled analyzer. Cheers!