ijor wrote:It is not a nice though, it is required to get the desired result using this technique. If you just make the transitions fixed on the middle range, then the result would depend on the drive.DrCoolZic wrote:But it add a nice touch: a slowly sliding bits that will for sure result in fuzzy bits ...
In fact it is not necessary. As you are aware (because I think you reproduce that it in Pasti) drive rotation speed is far from beeing constant and vary randomly and therefore if you have a transition on cell border this will result in random value returned.
You might argue that it is impossible to make sure that these transitions will be positioned correctly at the extreme border of a cell due to the fact that beyond the small random variations of rotation speed of the drive the "center" value is quite different on different drives. But this is where the DPLL of the WD1772 comes into play. I do not know if you have looked at the us patent that describes it, but it explain quite well how the DPLL does phase corrections (on top of frequency corrections) that will center the inspection window on the transition stream perfectly (please refer to my doc or patent for more info). As I have explained I have implemented the DPLL, as described in the patent in my prog, and indeed it does that very efficiently. This also explain why very often when you have this kind of "fuzzy bits" technique used you will find what I call "fuzzy bits pairing". That is two very close transitions: one at the extreme begining of the inspection window, and the pairing transition at the extreme end of the window. By using this pairing neither the frequency nor the phase correction are affected.
Actually if you look at the actual values used in DM (that I provided in my doc) they are mostly 0 or 1 (extreme left) and 9 or : (extreme right) but no sliding transitions. Actually i believe that this is probably more efficient than the "sliding transitions" that would affect the DPLL phase/freq correction.
Ps the other method for getting fuzzy bits are much more difficult to realize and I do not think were used for Atari diskettes?
ijor wrote:I didn't read it detaily, but if you are talking about using track misregistration, or physical alteration for producing weak bits, then AFAIK not.
The two other methods described are: slightly moving the head while writing (misalignement) resulting in weak signal during read, and writing 2 tracks between tracks (tracks overlaps). These two methods are resulting on "analog" variation of the signal and are probably more difficult to control/predict and this why I do not think they were widely used.
What is interesting is that the patent describes the concept of the fuzzy/weak bits (acyually it even provides a flowchart to verify ramdomness of the read) and incidentally indicates some techniques to acheive this.