I’ve heard it said that a CD player read a little ahead (and behind) to make adjustments in the output, but Googling does not produce a link that says anything about this “reading ahead” business. (Maybe “oversampling” is the wrong term?) Or, if it does, it is lost in the technojargon. Can anyone help me link to more info on this, or at least give me the proper term for this? Thanks!
I don’t know all that much about CD players but the terms you are looking for are most certainly “read ahead”, “caching” and “buffering” and not oversampling.
Oversampling is related to the fact that digital data isn’t continuous. The output of your CD player might be something like 250, 200, 150, 100 (it usually won’t be a straight line like this but I’m trying to make a point). You can see by looking at it that the original signal was dropping by a constant rate at that point in the music. Your CD player won’t put out anything close to a straight line, though. It will stay at 250 for its entire sample period, then it will make this harsh step down to 200, where it will stay for one sample period, then step again to 150, etc.
Your CD will have a filter on it which helps to reduce that harsh stepping, but it is very difficult to make a filter that can smooth out the stepping without cutting out part of the original signal.
If the CD player uses oversampling, it looks at the data and tries to fill in the gaps to make a smoother signal. Instead of jumping from 250 to 200, it will jump in smaller steps. If you are doing 10x oversampling, it will make 10 intermediate steps in between. Instead of a harsh 250 to 200 jump, it would go 250, 245, 240, 235, 230, 225, 220, 215, 210, 205, 200. This also runs at 10 times the speed, so that your 250 and 200 occur at the same time as they would have in the original version without oversampling. Now the jumps aren’t quite so harsh, and it is much easier on the filtering. You end up with a smoother signal that is closer to the original at the output.
Oversampling does look ahead in the data so that it can figure out the intermediate steps in the signal. It actually looks a bit ahead and behind to try to make the intermediate steps smoother, and isn’t quite as simple as my straight line example above. It’s basically trying to make some intelligent guesses about what the original signal looked like, since it only has samples and is missing the in-between sample parts of the original signal.
As groman said, CD players also use caching (aka buffering) so that if they get an error, they can read the disk again and hopefully will get a good read before they need the data.
ETA: ECG’s excellent post wasnt’ there when I started. Then I got interrupted for an hour. …
There is something in CD player tech called oversampling, but it is not at all the same thing as “read a little ahead (and behind) to make adjustments in the output”.
Unlike an analog tape or LP, the data which makes up the audio stream is not laid down on the CD in strict chronological order. It also contains a lot of extra error correction (“ECC”) data.
If you want to think about the read-&-play process at this level of detail, you cannot think of the CD player as just reading numbers off the disk and converting each to an analog voltage which is amplified & sent out the output wires.
What really happens (IIRC, it’s been a few years since I read the design specs) is a block of data is read which has a bunch of address and ECC information in it. Then the next bunch of data+ECC blocks are read, followed by an end block. Then that whole blob of data is confirmed to be consistent, and if necessary some of the audio data is changed based on what the ECC says it should have been.
Then that cleaned data is sent into a Digital to Analog Converter (“DAC”) to generate the analog voltage which is the audio signal. That voltage is then ampllfied a bit & sent out the line level outputs. Or in teh case of an all-in-one boombox or car audio unit, is sent on to the power amplifier to be amplifed up to a level to drive the speakers.
The metering of data *into *the DAC stage has to be done with a precise clock so the music comes out at the correct tempo/pitch.
The read-from-disk-and-apply-ECC processes upstream do not happen against a precise clock like that. There is a pretty big buffer between the two stages and the goal of the upstream process is to keep the buffer full enough that the DAC stage never runs out of data. So the upstream process starts & stops as necessary to keep the buffer mostly full. It can also re-read a given data block the next time the disk comes around if necessary.
In cheap players the buffer might be 1/2 second. In better ones, or those for portable use, the buffer might be 5 seconds. So what you are hearing *now *may have come off the platter 5 seconds ago.
And this is where interpolation comes in. Particularly for portable players, it’s quite possible that entire block sets as described above come off the disk too badly mangled to make sense out of. So when all else fails, the player assumes any given sample is simply the average of those around it. This is pure guesswork, but for most music most times this guess is good enough for human ears. For maybe 1/10th of a second’s worth. if the gap gets bigger than about that long, the player will fail back to momentary silence until a fresh block-set can be successfully read & decoded.
Oversampling has almost nothing to do with any of the above. The original data is laid down at 44.1 KHz. In other words, there are 44,100 samples (per channel) in one second of music. That’s all the data there is, and there ain’t no more.
The idea behind oversampling is that by doing the DAC process at a higher rate, say 88.2, you can interpolate a fake value between each actual value. And given that most music is fairly continuous at this time scale, the fake value will be very close to the actual value that would have been there had the data been recorded at a higher sampling rate.
And so by sending an 88.2 Khz data stream into the DAC stage you get a supposedly smoother & more real-sounding analog result.
When CDs first came out, the 44.1 KHz rate was a compromise to get enough minutes of audio on the physical disk given the bits / square inch limits of the then-new laser reader technology. Many audiophiles (read anal retentive reactionaries) thought the music sounded jaggy, like the way a too-much zoomed digital picture looks.
Oversampling is/was a way to hide that. And to sell a more feature rich (ie greater markup) product. Which made both audiophiles aand their dealers happy.
The state of the art in digital music recording has advanced a boatload since the CD format was fixed. As a reult, clever though it was, today it’s pretty darm primitive. If there was ever an effort to create a new CD audio standard we’d do things a bunch differently. As a practical matter that won’t happen. What will & is happening is treating the CD like a dumb file storage media and putting all the decoding in filetype-specific and installable software onthe player