There is no need for surface scanning. The drives do this when reading as part of the reading process, but only for the sectors requested by the read. It is not true that they willy-nilly access passing sectors if they have no business there. A drive access only sectors requested by the host. If it encounters a read error or seek error it will retry a number of times and then mark the sector as bad. A bad sector will not prevent subsequent read attempts. It will prevent write attempts.
RAID controllers often include a scrubbing option which periodically reads an entire drive allowing the drive controller or the RAID controller to reallocate marginal or bad sectors. Sectors may be reallocated on read even when the data is good or at least recovered.
There is also a big difference between logical and physical mapping.
There is a logical to physical mapping table and at least on older drives, spare sectors are distributed over the entire surface. High end drives that used a dedicated servo track and supported low level formatting could optionally include a spare sector on every track lowing the number of reported bad sectors but also sacrificing capacity.
Raid comtrollers can scrub : but the drive firmware needs to suppport it. You will need SAS drives. Your runof the mill ide or sata can't do that.
Dediated servo tracks no longer exist. That has gone away with the diskpacks. Damage to the servotrack was a catastrophy as the entire disk became unusable , true the head for servo was read only but a mechanical damage on that track was a disaster. In diskdrives as we know emfrom the PC world , servo data is interspersed with user data. That is the base operating principle of the drive.
There is a servotrack created by the formatter to lay down the servo wedges. Once the wedges have been written that servotrack is no longer accessible. It has never been accessible by the drive.
Here is how a drive is 'formatted' in the factory. You have a surface with no information on it. So how do you place markers ? The drive mechanics is not precise enough and only works if servo mechanisms can stabilise the head.
Well. Enter the formatter. An additional head is loaded on the outer rim of the drive. Speed control of the motor doesnt even work as we need to measure interval of servo wedges... Which have not been written yet.
So the formatter writes a single 'ping'. The motor controllr is told :disengage your servo , go in open loop and simply drive based on your crystal frequency.
The formatter now reads the ping and measures the time between two pings. This gives it accurate rotational speed (crystals drift enough so motor speed is off enough to have an impact on data retrieval). The formatter now tunes speed up or down by writing control register in the motor controller. Once the ping is seen at the required rate a square wave is now laid down. Every 'click' (o e to zero or zero to one) marks a potential sector change( not all tracks have the same amount of sectors. Track on the outside are longer than tracks on the inside )
We now have a precise physical marker coming from the disk that can tell the electronics : write now.
The formatter now grabs the headstack mechanically and moves it using its serve mechanism with anoptical encoder. So this is like a stepper motor with very fine precision. We walk the tracks and everytime there is a tick detected we write a servoburst and the track/sector datablock.
Ance this is done we can retract the head that was reading the ticks, close that opening, decouple the headstack and close that opening as well. The drive is now sealed and ready for surface test.
The servo mechanisms now all work : the servoburst controls both speed of the motor and timing of write and read operatins and the datablocks tells us where we are. Amplitude detectors keep us centered on the block. Now we write the remaining empty data with a pattern (not just all zeroes... We need changes there so we can do amplitude detection to find the center of the track)
Anyway. Like i said, i can go on and on about this. There are now event techniques where a formatter is no longer needed. The electronics that recovers data is so smart and adaptive that speed fluctuations are not a problem. It can adapt at will. Think of it as a serial port where the baidrate should be 9600 but can drift from 9000 to 10000 at random without data corruption. That is what the signal coming from the head looks like. The data samplers can adapt on the fly. You only need a track position but that can also be done because now we have velocity control on the voicecoil.
Home
Help
Search
Login
Register
