Speech
Einstein did "thought experiments" => engineers do "design thought experiments" :) Before simulation or breadboarding the engineer thinks about how to solve the problem, maybe with something new factored in. Idea (which many engineers don't like to confront) is to make things more cheaply.
This was one of the first ways for an experimenter or a ham to add speech to an electronic project:
https://en.wikipedia.org/wiki/General_Instrument_SP0256
I have one of those chips (SP0256A-AL2 from Radio Shack):
 |
| I did take it out of the package and misplaced the doc |
Here is the Radio Shack datasheet for the 276-1784 SP0256A-AL2 chip:
https://archive.org/details/bitsavers_gispeechGe56AAL2datasheetRadioShack2761784Apr1984_15859898
I can't recall ever using the chip. There are still a few of them on Ebay but the prices ! Also, the phoneme based voice synthesizers didn't sound that great. So the SP0256 isn't really an option for the 21st century project designer other than for nostalgia.
I did a search on talking code practice oscillators on the ARRL periodicals archive and search and found this:
It is a relatively small microcontroller based unit but uses 5 DIP ICs. I could see modernizing the design to 3 DIP ICs - that doesn't sound like much but it's a big savings in pins, from 116 pins to 30 pins:
 |
| Sorta complicated |
The way I would do it in a design thought experiment is with a 14 pin PIC 16F1764 which has an on-board 10 bit D/A, an 8 pin EEPROM and an 8 pin opamp. The serial EEPROM can be huge:
 |
| relatively cheap, just a good socket for a 28 pin DIP would be almost as expensive ! |
I haven't examined the code for the QST speech CPO but I suspect the same hardware used for the speech is used to create a sine wave sidetone with shaped keying for the actual Code Practice Oscillator. Here are the zipped files for the QST article:
https://www.arrl.org/files/file/QST%20Binaries/staton.zip
Of course the software would have to be completely re-written for a PIC which would be a big undertaking. To simplify things I'd probably use a brute force approach taking advantage of the big serial EEPROM to hold uncompressed voice or sound samples captured from PC WAV files and spit them out via the 10 bit D/A. Of course this would have to be tried first with a shaped sine wave as a proof of concept. A quick calc: an 8 megabit serial EEPROM could hold about 62 seconds of speech with an 8 kHz sample rate and 16 bit samples (throw away the 6 lowest bits).
8000000 bits / 8000 samples per second / 16 bits = 62.5 seconds
62 seconds should hold 10 digits, 26 letters and special characters and a tone. But the neat thing to me if it works is that the 3 chip speech board could be a springboard to enable other voice projects.
Anyway, just an idle thought for now.
Best Regards,
Chuck, WB9KZY
http://wb9kzy.com/ham.htm
