Several years ago, my grandparents moved into a new house. The previous owner went way overboard on his AV setup. Surround sound built into the living room ceiling, speakers in every room, ethernet in the walls, and a full AV rack in a separate room. It was really far too complex for their needs, too far away for me to support, and the equipment was too old to be useful anymore. My grandma has always hated the rack and miles of cable, so I tore it all out for them. When I was packing it all into storage, I found this piece unexpectedly interesting.

(Ignore the dust. I actually did clean it, but it is quite old and just permanently dirty now)

This is an IR repeater. The idea is that it will pick up the signals from any IR remote and forward it to another room over a simple three wire interface.

It's also unexpectedly heavy, over a pound for this little box! I figured it would be mostly empty and all the weight is in the unnecessarily thick metal shell. After all, today you'd build this with one of those self-contained three-pin IR decoder modules and a tiny microcontroller.

Oh boy, that's a lot of stuff. What the hell is going on here?! Why is that shielding can so big? What could possibly be in there?

Oh. Oh it's gonna be like that. That's not an IR module, that's just a bare photodiode and the can is full of analog circuitry.

This must have been designed before microcontrollers were cheap and plentiful. While the lone IC has a 1999 datecode, this design is definitely from at least the early 90s if not the 80s. This type of construction with axial components mounted on-end is so very typical of old-school analog design. It also appears to have been built by hand, which is wild to see these days.

With the time period in mind, let's take a roll call!

7400 series logic: yup

LM317: present

2N2222: Two of them!

That 74HC14 is a six Schmitt trigger hex inverter.

Without actually tracing out and reverse engineering the circuit, let's take a stab in the dark at its theory of operation. We take an analog-encoded digital signal, quantize it, and push it out over the wire in a digital format. Most of the circuitry here is likely just to discriminate between coded signals and random noise. The IR detection circuit is always active, so it's always outputting noise from ambient light. The 2N2222 would be shifting the output signal up to 12V, this would help with really long cable runs.

Let's look at the other half of this system, the box it plugs into looks equally exciting:

So on the front we see six inputs, four outputs, a daisy-chain output, power (on a 3.5mm TR jack, of course), and some very interesting trimmers to set output level.

Cracking it open, we see a very familiar type of design.

First impressions; this is much simpler than the receiver. Which makes sense, we don't have to do any complex peak detection to filter IR signals from noise. All this box needs to do is decode and amplify a knownassumed-good signal.

Just looking at the topside, the function seems pretty self-evident. Gate the inputs so you only decode one at a time, then undo the original encoding and pipe it to the four output amplifiers.

But wait, those aren't amplifier transistors.

Those are LM317s! What?! Now we see why those trimmers were interesting, they're tied into the LM317 bias circuits. Before trying to decipher this, let's take a look at the rest of the active components here.

74HC30 - 8 input NAND

74HC14 - 6 hex inverter

74HC132 - quad 4 input NAND (plus a delightfully chunky axial inductor and dipped resistor packs)

We can also see that all the datecodes line up with the other module: all made in 1999.

A generic power diode (just look how uneven all the resistors are)

And another LM317, presumably for Vcc. They must have got some killer bulk discount on these.

Let's take a look at the bottom of the PCB and make some more educated guesses on what it's doing.

Just looking at how the various circuit blocks are connected, we can make some good observations:

The 7430 and 7414 are doing input detection and gating. All the output amplifiers are held in shutdown unless there is exactly one input channel active.

From there, the inputs are all wire-ANDed (since only one can be active) through one of those resistor packs, level-shifted by yet another 2N2222 and fed to the 7413 to undo the encoding and get the raw IR signal back.

From there, the signal is fed into the bias circuit of the LM317s. Effectively turning them into amplifiers. Can't say I've seen 317s used quite like this before. This drives arrays of IR LEDs to blast out the relayed signal to remote equipment. These are properly called 'IR Blasters', by the way.

Overall quite an interesting piece of electronic archaeology. I love seeing how we accomplished tasks like this in eras before cheap and ubiquitous computing. Today, this would be nothing more than a microcontroller and a few lines of code and it would likely cost an order of magnitude less than this beast.