English subtitles for clip: File:Harddrive-engineerguy.ogv

1
00:00:04,000 --> 00:00:07,000
Hard drive teardown

2
00:00:07,000 --> 00:00:10,000
flying heads, voice coil motors, amazingly smooth surfaces & signal processing

3
00:00:10,000 --> 00:00:17,000
series 3 engineerguy videos

4
00:00:17,000 --> 00:00:23,000
A home computer is a powerful tool, but it must store data reliably to work well, otherwise it's kind of pointless, isn't it.

5
00:00:23,000 --> 00:00:25,000
Let's look inside and see how it stores data.

6
00:00:30,000 --> 00:00:32,000
Look at that: It's marvelous.

7
00:00:32,000 --> 00:00:35,000
It's an ordinary hard drive, but its details, of course, are extraordinary.

8
00:00:35,000 --> 00:00:38,000
Now, I'm sure you know the essence of a hard drive:

9
00:00:38,000 --> 00:00:41,000
We store data on it in binary form - ones and zeros.

10
00:00:41,000 --> 00:00:43,000
Now, this arm supports a "head"

11
00:00:43,000 --> 00:00:45,000
which is an electro-magnet that scans over the disk

12
00:00:45,000 --> 00:00:48,000
and either writes data by changing the magnetization of specific sections

13
00:00:48,000 --> 00:00:50,000
on the platter or it just reads the data

14
00:00:50,000 --> 00:00:53,000
by measuring the magnetic polarization.

15
00:00:53,000 --> 00:00:54,000
Now, in principle, pretty simple,

16
00:00:54,000 --> 00:00:58,000
but in practice a lot of hard core engineering.

17
00:00:58,000 --> 00:01:02,000
The key focus lies in being sure that the head can precisely

18
00:01:02,000 --> 00:01:03,000
error free

19
00:01:03,000 --> 00:01:05,000
read and write to the disk.

20
00:01:05,000 --> 00:01:08,000
The first order of business is to move it with great control.

21
00:01:08,000 --> 00:01:11,000
To position the arm engineers use a "voice coil actuator".

22
00:01:11,000 --> 00:01:14,000
The base of the arm sits between two powerful magnets.

23
00:01:14,000 --> 00:01:17,000
They're so strong they're actually kind of hard to pull apart.

24
00:01:17,000 --> 00:01:18,000
There.

25
00:01:18,000 --> 00:01:20,000
The arm moves because of a Lorentz force.

26
00:01:20,000 --> 00:01:23,000
Pass a current through a wire that's in a magnetic field

27
00:01:23,000 --> 00:01:25,000
and the wire experiences a force;

28
00:01:25,000 --> 00:01:28,000
reverse the current and the force also reverses.

29
00:01:28,000 --> 00:01:30,000
As current flows in one direction in the coil the

30
00:01:30,000 --> 00:01:34,000
force created by the permanent magnet makes the arm move this way,

31
00:01:34,000 --> 00:01:36,000
reverse the current and it moves back.

32
00:01:36,000 --> 00:01:39,000
The force on the arm is directly proportional to the current

33
00:01:39,000 --> 00:01:40,000
through the coil which allows the

34
00:01:40,000 --> 00:01:43,000
arm's position to be finely tuned.

35
00:01:43,000 --> 00:01:45,000
Unlike a mechanical system of linkages there

36
00:01:45,000 --> 00:01:49,000
is minimal wear and it isn't sensitive to temperature.

37
00:01:49,000 --> 00:01:53,000
At the end of the arm lies the most critical component: The head.

38
00:01:53,000 --> 00:01:57,000
At its simplest it's a piece of ferromagnetic material wrapped with wire.

39
00:01:57,000 --> 00:01:59,000
As it passes over the magnetized sections of the platter

40
00:01:59,000 --> 00:02:02,000
it measures changes in the direction of the magnetic poles.

41
00:02:02,000 --> 00:02:06,000
Recall Faraday's Law: A change in magnetization

42
00:02:06,000 --> 00:02:08,000
produces a voltage in a nearby coil.

43
00:02:08,000 --> 00:02:10,000
So, as the head passes a section where the polarity

44
00:02:10,000 --> 00:02:14,000
has changed it records a voltage spike.

45
00:02:14,000 --> 00:02:16,000
The spikes - both negative and positive - represent a "one"

46
00:02:16,000 --> 00:02:19,000
and where there is no voltage spike corresponds to a "zero.

47
00:02:19,000 --> 00:02:22,000
The head gets astonshingly close to the disk surface

48
00:02:22,000 --> 00:02:25,000
100 nanometers in older drives, but today under

49
00:02:25,000 --> 00:02:27,000
ten nanometers in the newest ones.

50
00:02:27,000 --> 00:02:30,000
As the head gets closer to the disk its magnetic field

51
00:02:30,000 --> 00:02:32,000
covers less area allowing for more sectors

52
00:02:32,000 --> 00:02:35,000
of information to be packed onto the disk's surface.

53
00:02:35,000 --> 00:02:38,000
To keep that critical height engineers use an ingenious method:

54
00:02:38,000 --> 00:02:41,000
They "float" the head over the disk.

55
00:02:41,000 --> 00:02:44,000
You see, as the disk spins it forms a boundary layer of air that

56
00:02:44,000 --> 00:02:48,000
gets dragged past the stationary head at 80 miles per hour at the outer edge.

57
00:02:48,000 --> 00:02:52,000
The head rides on a "slider" aerodynamically designed to float above the platter.

58
00:02:52,000 --> 00:02:56,000
The genius of this air-bearing technology is its self-induced adjustment:

59
00:02:56,000 --> 00:03:01,000
If any disturbance causes the slider to rise too high it "floats" back to the where it should be.

60
00:03:01,000 --> 00:03:04,000
Now, because the head is so close to the disk surface

61
00:03:04,000 --> 00:03:07,000
any stray particles could damage the disk resulting in data loss.

62
00:03:07,000 --> 00:03:11,000
So, engineers place this recirculating filter in the air flow;

63
00:03:11,000 --> 00:03:14,000
it removes small particles scraped off the platter.

64
00:03:14,000 --> 00:03:18,000
To keep the head flying at the right height the platter is made incredibly smooth:

65
00:03:18,000 --> 00:03:23,000
Typically this platter is so smooth that it has a surface roughness of about one nanometer.

66
00:03:23,000 --> 00:03:26,000
To give you an idea of how smooth that is: let's imagine that this section is enlarged

67
00:03:26,000 --> 00:03:31,000
until it's as long as a football field - American or International -

68
00:03:31,000 --> 00:03:35,000
the average "bump" on the surface would be about three hundredths of an inch.

69
00:03:35,000 --> 00:03:38,000
The key element of the platter is the magnetic layer,

70
00:03:38,000 --> 00:03:41,000
which is cobalt - with perhaps platinum and nickel mixed in.

71
00:03:41,000 --> 00:03:43,000
Now this mixture of metals has high coercivity,

72
00:03:43,000 --> 00:03:50,000
which means that it will maintain that magnetization - and thus data - until it is exposed to another powerful magnetic field.

73
00:03:50,000 --> 00:03:52,000
One last thing that I find enormously clever:

74
00:03:52,000 --> 00:03:57,000
Using a bit of math to squeeze up to forty percent more information on the disk.

75
00:03:57,000 --> 00:04:04,000
Consider this sequence of magnetic poles on the disk's surface - 0-1-0-1-1-1.

76
00:04:04,000 --> 00:04:06,000
A scan by the head would reveal these distinct voltage spikes -

77
00:04:06,000 --> 00:04:09,000
both positive or negative for the "ones".

78
00:04:09,000 --> 00:04:13,000
We would be easily able to distinguish it from, say, this similar sequence.

79
00:04:13,000 --> 00:04:16,000
If we compare them they clearly differ.

80
00:04:16,000 --> 00:04:20,000
Engineers, though, always work to get more and more data onto a hard drive.

81
00:04:20,000 --> 00:04:22,000
One way to do this is to shrink the magnetic domains,

82
00:04:22,000 --> 00:04:25,000
but look what happens to the voltage spikes when we do this.

83
00:04:25,000 --> 00:04:28,000
For each sequence the spikes of the ones now overlap and

84
00:04:28,000 --> 00:04:30,000
superimpose giving "fuzzy" signals.

85
00:04:30,000 --> 00:04:33,000
In fact, the two sequences now look very similar.

86
00:04:33,000 --> 00:04:37,000
Using a technique called Partial Response Maximum Likelihood engineers have developed

87
00:04:37,000 --> 00:04:40,000
sophisticated codes that can take a murky signal like this,

88
00:04:40,000 --> 00:04:45,000
generate the possible sequences that could make it up and then choose the most probable.

89
00:04:45,000 --> 00:04:49,000
As with any successful technology, these hard drives remain unnoticed in our daily lives,

90
00:04:49,000 --> 00:04:51,000
unless something goes wrong.

91
00:04:51,000 --> 00:04:53,000
I'm Bill Hammack, the engineer guy.