Text messages over walkie-talkies

2010-07-05T11:55:00.000-07:00

For my 12th birthday, my dad got me a pair of Radio Shack FRS walkie-talkies. They never did get much use after our initial range test (2km, LOS) and have been sitting around in various drawers ever since. Here's what they look like:

So, what to do with a pair of unused walkie-talkies? Send text messages with them, of course!

Step 1: take 'em apart.

Notice all those nice pads to solder onto? I ended up using power, speaker, and the power switch pads. Transmission is hooked up with a microphone jack. Channel select, etc are just left static.

The cases got a bit trashed in the process of taking the circuit boards out. It would have been nice to keep them around, however. It could have been possible to just plug into the radios' microphone and headphone jacks, they could have been driven exclusively via these. Then, the radios could be disconnected from the text transceiver circuit and used "normally" for other activities.

A bit of damage to the case.

Step 2: Build a circuit for interfacing.

Here's a bit closer up:

Top left is a 5v linear power supply. Top center is an Atmel AVR ATMEGA88PA MCU. Top right is a MAX232 for converting TTY voltage levels to RS232 levels so this can be plugged into a computer via a DB9 serial port. The IC in the middle breadboard is a common single supply op-amp. To the left of it is a passive bandpass filter.
Not shown are a MOSFET and a JFET. These were used for powering on the radio and transmission enable, respectively.

Step 3: MCU programming
So, how do you convert text like this (01110100011010000110100101110011) to go over the air? Turn the bits that comprise letters into different frequencies, of course. Alternatively, you could do OOKing with Manchester encoding, or really whatever else you want.

One of the MCU's timers is clocked via rising edge waves from the op-amp. By sampling the number of edges received (the timer's value) at regular known intervals, the frequency of the received signal can be computed. Sampling is done three times in the duration that it takes to send one bit. Thus, at least one of three samples will always be part of a complete bit and not a transition to or from another.

The algorithm for receiving is as such:
Sample the frequency.
If the start bit is received (only sent at the beginning of a letter), wait three individual sample periods and then sample again. (Because the start bit came through cleanly, if we wait for three sample periods, we should have another non-adultered part of a bit.)
Bit bang that first bit into a 8 bit datatype.
Repeat until we've banged in 8 bits.
Send that data out to the serial port as a character.

Transmitting is a fairly simple operation of bit banging a character received on the serial port and toggling an I/O pin at a rate for a period of time depending on the bit.

These radios really were a bit of a black box, and strange things happened during assembly. Everything was hunky-dory on the bread board. The radios would turn on via MOSFETs nicely, they'd stop transmitting when I set my P-channel JFETS high, etc. It was only once everything was soldered down onto stripboard that problems cropped up. I had to tie the mic port high to keep the radios from sticking in transmit (which could sometimes be alleviated by touching the bare antenna with my finger,) etc. The passive bandpass filter which was necessary on the breadboard caused the radios not to produce a sine wave once on the stripboard (except for one time that it did for a few seconds...) all sorts of odd things. I'd like to put most of the blame on the old electrolytic capacitors these radios have.

In a nutshell, this project took way more time than I anticipated due to erratic behaviour of the radios. (And also because I did not have an oscilloscope when I started.) I had to do a lot of guess-and-check experimentation to find things that worked. I'm loath to open up the project boxes to do any tweaking. They work, I'm happy, and I don't ever want to mess with them again.

Now I just need to find a problem I can solve with these....

One of two identical boxes.

Another shot.

Super exciting operation, I know. Text goes in one box, text comes out another. Wohoo.

Some thoughts on efficiency:
I ended up with a baud rate of 10. Not super speedy, but fine for some applications. I never did squeeze the radios to see how many separate frequencies I could shove into smaller and smaller timeslots. I would not be surprised by a 4x speedup or more by this alone.
ASCII is not the best choice for a space-efficient character encoding. I transmitted the check digit, and never computed it. Plus I did not use any compression schemes (e.g., one frequency and timeslot for '11' instead of taking up two timeslots by transmitting '1' and then '1'.) There are plenty of things that could be done to speed this up and cram quite a bit of text in one second.

Why buy it when you can build it?

Text messages over walkie-talkies