A CD player from scratch

[bsquared]

Originally Posted by Orgasmic Troll

I think the hardest part of this is reading the data off the CD. I think you're right that it will have to be optical, and precisely reading the CD will also be difficult.

Yes, this is what I've been trying to say. As retina says, there is a smooth polycarbonate layer which prevents mechanically reading the pits, and in any case the pits are too small to reliably sense mechanically. The only way is optically. This is done with a semiconductor laser and a photodiode, both of which require fabricating diodes. Maybe you can do it with a gas laser of some sort, but you have to pay attention to the wavelength in order to be able to read the pits and also detect the reflections. Diodes first require extremely pure silicon and then precise doping, too much and the material degenerates into a conductor and not enough and it's still an insulator. To do this you need a huge infrastructure for the industrial processing.

Once you have the laser/detector, you have to precisely control their location with respect to the disk - location needs to be held to tolerance of about a micron or you'll be bouncing around between tracks. And you have to spin the disk with a precisely varying speed or else your data will be read incorrectly... if the rotation slows somewhat then streches of constant intensity can be mistaken for more 0's or 1's then they really are. I'm not in the CD fabrication business, but to me this level of precision implies all sorts of feedback sensing and real-time processing in order to implement the control loop... even if we are reading data at 100x less than real-time. Some of this technology you may have once you've built your laser and photodetector though.

Originally Posted by Orgasmic Troll

I think the data on the CD is such that the error correcting code is just a few added bits (Someone correct me if I'm wrong) so they can just be ignored. I believe that aside from the check bits, it's just the amplitude of that sample in binary. It seems like that would be "easy" enough to implement mechanically to record to wax. As I said before, recording to wax removes the problem of needing a loudspeaker, or for the device that reads the data to have to read and convert in real time.

I think that the data is encoded fairly heavily. This implies you'd need to reverse the 8-to-14 modulation as well as the reed-solomon encoding in order to get at the raw data... there are only something like 192 bits of music in every 588 bits of data.

I don't understand the bit about the wax... how can you use a wax recording to hear something without a speaker?

[10metreh]

A slight variation:

We've all heard of the Boltzmann Brain. So, assuming the proton does not decay, how long will it take for a CD player to spontaneously appear in the universe?

[retina]

Originally Posted by 10metreh

A slight variation:

We've all heard of the Boltzmann Brain. So, assuming the proton does not decay, how long will it take for a CD player to spontaneously appear in the universe?

Perhaps they are spontaneously appearing everyday. But the problem is for you to be at the exact spot where it appears. Perhaps better to ask "how long will it take for a CD player to spontaneously appear near where I am located?".

[xilman]

Originally Posted by bsquared

This is done with a semiconductor laser and a photodiode, both of which require fabricating diodes. Maybe you can do it with a gas laser of some sort, but you have to pay attention to the wavelength in order to be able to read the pits and also detect the reflections.

It is much easier to build a gas laser than a diode laser. A "simple" HeNe laser has about the right wavelength. At least, to my eyes it looks to be very similar in colour.

I once saw a SciAm article on how to build a gas laser at home pretty much from first principles. Ok, that one worked far out in the IR and used something common like CO_2 as the lasing medium, and assumed the availability of electronics and glass-blowing equipment, but the technological requirements were much, much lower than that needed to grow and cut your own LED crystal.

I wish I could find that article again. It was quite a few years ago now and almost certainly preceded the appearance of the CD.


Paul

[Visu]

Originally Posted by xilman

It is much easier to build a gas laser than a diode laser. A "simple" HeNe laser has about the right wavelength. At least, to my eyes it looks to be very similar in colour.

I once saw a SciAm article on how to build a gas laser at home pretty much from first principles. Ok, that one worked far out in the IR and used something common like CO_2 as the lasing medium, and assumed the availability of electronics and glass-blowing equipment, but the technological requirements were much, much lower than that needed to grow and cut your own LED crystal.

I wish I could find that article again. It was quite a few years ago now and almost certainly preceded the appearance of the CD.


Paul

Might be the one in the Sept 1971 issue...

[nibble4bits]

Well are you aware that the CD's format is based on.... video tapes? The 44.1 KHz sample rate is actually related to the timing of some VCRs.

Let's say that you just want to make a CD player out of stuff you have access to but without relying on taking apart existing players or laptops or... In other words, something you could make in your garage using the same kinds of tools available before the invention of FPGAs and microcontrollers.

Problem 1) Laser and photodiode to read the data... Now nothing says it has to be a laser diode. The reason they use laser diodes is the expense, power consumption, reliability, and size. The wavelength isn't as critical as you might think simply because of tolerances.
2) Even with an optical head, you still need a mechanical tracking mechanism to keep on the right path. Luckily, there's a nice little sine wave hidden in the format that in theory can be mechanically reproduced/regulated by an adjustable cam. When you fall out of alignment, the transistors would force the timing back because the sine waves don't match. At a slow enough rotation speed, in theory this would be practical even without access to custom chips or FPGAs, or indeed even a CPU.
3) The timing of the reads only matters if you're needing real time and error free decoding of the bits. With batch post-processing, the raw analog data could be converted into the correct bits to hand off to the decoding stages. The distance between each bit can be determined after reading the raw optical (analog) data.

Problem 4) Removing the NRZI layer can be done with the kinds of chips you can get from the 74xxx or 40xx series, but is easier done using a microcontroller which I pretty much ruled out in the first paragraph. Doing this using some electromechanical device is a pain but relays would work as well as transistors for making gates. Heck, you could even use water flow control valves. It depends on how far you want to take this "primitive" thing aka masochism. :P My idea is to make it do automated batches of reading/conversion since a human isn't immortal or infinitely patient.
5) You have timing/sync data, audio data and so on that has to be decoded from the 588-bit frames. Removing ECM would be ideally done with some kind of table stored on a ROM of some sort (this avoids the full CPU by making a simple special-purpose processor/state machine). A slow mechanical device would roughly resemble a music box's wheel, to give you a hint as to how silly we can get.
6) Error correction and detection. This you would love to avoid but may need if you're using a homemade mechanism, even if the CD is clean.

Problem 7) You have the 588-bit frames decoded to 24 samples each, now how to make analog output for later listening? The normal CD player builds up 2352 bytes of audio data from 98 of those 24-sample frames, but we don't have to do that! Since the audio sectors are played in order, we actually have the same information in the same order, just in smaller chunks. We can handle 24 samples at a time and have our DAC output to a (slow) write head. Once it records the data, the drive has to reseek the correct frame. Slow but it should work in theory.
8) The audio playback from the analog medium should hopefully be a lot easier than the previous long process! Everything from film to magnetic tape to vinyl/metal disks has been used for analog media.


Now I admit that making this stuff with the right tools and parts and knowledge is hard enough and without it would be pretty much impossible. I just thought about just how long it would take to extract just 1 second of audio using such a haphazard setup and it's not pretty even if it would work reliably and automatically.


::EDIT:: Haha look at that picture at the end of the page. http://www.tentlabs.com/Products/diycd/index.html

[Orgasmic Troll]

great reply! I didn't know about the VHS connection, that's amusing

The only thing I'm seeing here that I wish to further elaborate on is the transformation from a digital signal (regardless of the encoding, which is another bridge to cross) to an analog one. Since my most recent class in electronics that covered digital to analog conversion was back in the 1980's, this is a grossly simplified and woefully out of date explanation, but should at least help folks to understand the process; but first, some simple assumptions:

1.) The analog signal has an amplitude (the measurement from the most positive signal to the most negative) range of +7 to -7 (volts, milivolts, whatever). This gives us a good working range.

2.) The method of converting from digital to analog is exactly opposite in application from converting analog to digital (it's really not, as far as the electronics go, but it's easier to explain analog to digital)

Ok, We have a simple analog signal, which is a standard sine wave at 100Hz, or 100 cycles per second. A fairly low frequency sound. The graph below shows a single cycle of that sound. Ignore all the letters and numbers for now. They come into play in a bit.

http://www.geekcavecreations.com/images/sineWave.png
{It seems there's no provision for posting images, so here's a link, instead}

Now as our signal is fed into the converter, it measures the voltages at regularly spaced intervals of, let's say, 1/1,000th of a second each. That's 10% of the cycle per sample.

(I know, a LOT of numbers here. Bear with me, because I want to be accurate here)

36Now, we know that the range of the signal is from +7 to -7, and anyone who's had some electronics, or even basic physics can tell you that the value of a sine wave is a function of the sine of it's current "angle" (hence, called a sine wave) multiplied by it's max value, you can easily arrive at it's value at any given point in time. However, the circuit doesn't know all of this, so it directly measures the voltage. Cheater!

Anyway, the first measurement is zero, at the very beginning of the signal. This measurement is converted into a 4-bit number 0000, then the next measurement is made (letter "a" on the graph), which is 4.11 volts. It's rounded to the nearest whole number, since this is a primitive device, and stored as binary 0100.
The next measurement is made ("b"), resulting in 6.66, again rounded to 7, giving binary 0111, and so on, and so on, back to zero, halfway through the cycle. Negative numbers are handled simply by firing off the first bit of the binary number, so -1 is 1001, -4 is 1100, etc. The binary sequence for the entire cycle would look like:

0000 0100 0111 0111 0100 0000 1100 1111 1111 1100 0000

Now, to decode from digital to analog, the reverse happens. The first block (0000) is fed in, and the converter outputs zero volts, then the next block is accessed, and a voltage of 4 is output, and so on, etc.

By understanding this aspect of the problem, you can see that simply transferring the data from the digital format to a sound generating device (e.g. amplifier & speaker) would result in gibberish.

Ok, that being said, I'll also try to answer the overall question, as well.

Given an average productive lifespan of 80 years (I'm being generous here, I know), and either a pair of astronauts (for future generations of work, and completely ignoring the whole genetic diversity issue), or a female one (parthenogenesis is highly unlikely for a female, but impossible for a male), and given a doubling of population every generation (ambitious, ain't I?), I'm guessing at no less than three generations, and no more than 6, to create a functioning device to obtain and understand the instructions placed on the CD.

[xilman]

Originally Posted by DaveMorton

... or a female one (parthenogenesis is highly unlikely for a female, but impossible for a male) ...

This looks like an example of Clarke's First Law.

Paul

[kuba]

I think that there's no need for semiconductors of any sort to read a CD. All of it could be vacuum tubes. No need for lasers either. All you need is a strong source of light, say a sodium lamp, and a monochromator to select a good line from said sodium's emission spectrum. Heck, with sodium's two major "lines" you could probably work around it somehow without monochromation. Perhaps there are sources better than sodium.

An analog controller with vacuum photodiodes could be used to do servo tracking and pickup, no need for anything digital there. The processing of the bitstream requires a frequency reference (not a very good one, RC oscillator would do), and a way of servoing the spindle to maintain the transition frequency at the pickup's output. You'd need a bunch of logic to do the EFM decoding, frame detection, and pulling of the audio channel(s) out of each frame.

If, due to the size of the circuit, you'd need to keep things slower, another approach may be to use an imaging pickup where there's an array of photodiodes on a variable-radius curve (a mechanical setup), and you illuminate the CD and image the pits on the photodiodes. Probably it'd read one frame at a time. The tracking could be done purely mechanically if you're lucky, you'd probably index the disc using some mechanism. A frame at a time could be "rendered" a sample at a time by using a counter and a ramping current to a write coil on iron wire. The iron wire would then be played back at a high speed.

Generally speaking, you'd need all of the following technologies mastered to some extent: woodworking, brass and copper, blown glass, cast iron for machinery, steel for cutting tools (perhaps sintered ceramics would work OK), ceramics for isolators, abrasives, crucibles and capacitors, machine tools for brass working (steel and iron you'd grind and cast only), grinding optics, some chemistry process capacity for metallurgy, oil extraction, development of phosphor since you'd need a CRT for a rudimentary oscilloscope, some plastic polymers for wire insulation, something for varnish for windings in transformers and inductors, some analog test equipment and standards (a VOM), writing implements and some kind of paper, some math references (log and sin/cos tables) and tools (slide rule), vacuum pumps for pulling tube vacuum and vacuum deposition vacuum (for multilayer caps, I'd think). I don't know yet if it'd be possible to avoid dealing with aluminum and electrolytic capacitors. I'd bet on ceramics being sufficient, as long as high-K ceramics could be worked out quickly.

All that stuff could be *probably* pulled off by a couple of very dedicated individuals who had someone else as a host to do all the house chores, and who had knowledge of exact step-by-step instructions for putting it all together, including locations of natural resources, etc.

Pretty much it could be doable in practice only if the team was dropped onto that planet who had an exact, day-by-day plan of action, and didn't have to worry about food, board, health etc. Myself, I'd probably go halfway and say "screw it, I've got plenty of toys to play with already, who cares about reading CDs".