1978 was very early in the 'personal computer' era. I'd guess that the use of a VCO was modeled after early synthesizers like the Moog. (It used a 1-volt per octave scheme ... changing the external voltage input from 2 to 3 volts moved pitch up an octave.)

The 76477 could take an external voltage (up to 5v) (Pin 16 via J2) for pitch - maybe the design's intent was to be able, when needed, to inexpensively generate external voltages, e.g. through a pot. The much-more powerful General Instrument AY-3-8910 arriving in 1978 sank any hopes the 76477 had.

https://en.wikipedia.org/wiki/General_Instrument_AY-3-8910

Interesting how much analog architecture was used. Today everything is processed digitally with ADC/DAC at the edge. Those old 8-bit MHz-class processors and memory chips lacked the horsepower to synthesize sounds without disrupting gameplay.

The Apple II has a single register that toggles the voltage to a speaker, allowing 100% CPU-driven construction of any square wave you want. I think other systems did something similar on cassette ports. That's an example of the power available to an 8-bit class CPU without any extra hardware - you can do a lot with PWM and high-frequency attenuation and such but definitely limited.

1-bit sound construction was discussed earlier this week on HN:

https://news.ycombinator.com/item?id=23144469

A lot of game sounds were found by placing writes to the speaker in various game event loops.

The result was lean sounds, but they also made a lot of sense.

All sound on the raspberry pi is 1 bit PWM.

(Unless you're using the HDMI sound channel.)

Even for a sound chip, analog was unusual. a lot of the arcade games from that era and for another 10-15 years would often dedicate whole CPUs, such as Z80s and 6809s to sound. e.g. run some classics in MAME and it displays the sound chip(s) used in each game. Some of them have as many or more cycles used than the game CPU :)

the tradeoff this chip made was fewer transistors and cost in exchange for not being very music, or able to be tuned well. it really was more a sound effect chip then a music chip.

This has me thinking of the era when the LaserWriter printer had a significantly faster CPU and more RAM than the Macs it was paired with. The complexities of interfacing with the analog world...

The LaserWriter had to run the Postscript interpreter, which required a lot more CPU power than just generating Postscript files like the Mac did.

I love this kind of stuff. I’m always amazed about how small we can operate.

It's also amazing how much smaller transistors have become since 1978. The 76477 sound chip had about 200 transistors, and now chips have billions of transistors.

Obviously transistors have gotten way smaller, but it's not entirely fair to compare transistors in a sound chip to something in a microprocessor.

Anything that's driving a speaker is going to have pretty large transistors. Additionally the power stage of chips generally takes up some room. It'd be a lot more comparable to compare similar basic sound output chips... I bet they don't have more than a few thousand transistors even today.

And yet this chip wasn't designed to drive a speaker. The reference designs I've seen show using external transistors to actually drive the speaker.

There was the SN76488, if you wanted to drive a (small) speaker directly.

Author here: that's the mobile link, so the images are small. Click on an image for a larger one.

http://www.righto.com/2017/04/reverse-engineering-76477-spac... is the non mobile link

Changed from http://www.righto.com/2017/04/reverse-engineering-76477-spac.... Thanks!

Thank you so much for your blog, Ken. I get tremendous joy from reading it. I look forward to each post. I wish we could archive it for future generations to read; I guess our best hope for that is the way back machine / archive.org.