Theremin

Now!

Well now that I am grown up, it is finally a chance for me to find out how actually the game loading of ZX Spectrum works! If I can replicate that at least in a primitive way, then I could do some cool stuff with it, like make sounds with encoded text or images and later load them using my ipod or better a cassette player.

It took me quite some googling to figure out the principle. I registered on the Sinclair nerds forum and they gave me some good tips an pointers to sources.

So here every new bit (random in this case) changes (or not) the frequency of a square wave every 10ms. I couldn't really figure out at what rate ZX Spectrum exactly encoded binary data, but in the paper I read it says that typical values for one element in FSK modulation are between 5 and 22 milliseconds. So in an element of 10ms there are 10 cycles of a 1000hz wave (square wave in my case, but it could as well be a sine wave or a saw i guess) and 20 cycles of 2000hz.

Here is the recording of that patch.

And a youtube video of some dude who recorded the actual game being loaded

If you got bored reading this, then try playing some ZX Spectrum games here:

http://www.zxspectrum.net/

Lets create my own ZX Spectrum cassette tape!

Encoding  binary data in sound is a bit more complicated than I showed above. There are various ways and some of them are better at certain things than the others. But the main goal in this business is that actually not a single bit should get lost, otherwise an error will occur. And the real challenge actually seems to be not encoding itself, but interpreting encoded information. But it's a kind of a closed loop, in order to properly interpret the sound, you need to encode it properly :)

So what I am trying to do is to encode and decode a piece of text as well as possible!

I took the following quote from Aldous Huxley book "Brave new world":

Mother, monogamy, romance. High spurts the fountain; fierce and foamy the wild jet. The urge has but a single outlet. My love, my baby. No wonder those poor pre-moderns were mad and wicked and miserable. Their world didn’t allow them to take things easily, didn’t allow them to be sane, virtuous, happy. What with mothers and lovers, what with the prohibitions they were not conditioned to obey, what with the temptations and the lonely remorses, what with all the diseases and the endless isolating pain, what with the uncertainties and the poverty—they were forced to feel strongly. And feeling strongly (and strongly, what was more, in solitude, in hopelessly individual isolation), how could they be stable?

And converted to binary numbers here:

http://www.roubaixinteractive.com/PlayGround/Binary_Conversion/Binary_To_Text.asp

Then I got a lot of 0s and 1s. Once i had my binary representation, I converted it to sound in the patch I made.

And there you go! I have my recording. It's 1 min. 56 sec. long.

So now that I have the recording, I can play it back analyze it and load the text from it. For that I wired another max patch which seems to be doing the job:

Not so bad for the first try. But obviously an error occurred and screwed up the rest of the text. After an eternity of debugging I am convinced that the problem lies in the interpreter side. And that problem is:

1. analysis is too slow

2. it's hard to keep the input syncronized

No wonder there are various techniques which i didn't mention before of how to make sure that the bits don't get lost. One of them is produce kind of opening and closing sounds for every 5 bits.

It basically just means that the computer is sort of reminded (very often) that it's gonna receive 5 bits. In this way it's easier to keep things in sync.

But for now the solution is to encode and decode the information at a much slower rate. But then it means that the above piece of text would take more than 10 min of sound. Not very efficient, is it? :)