This is the third installment of a series on magnetic devices and media used in computer systems. In this article we will look at the storage media, primarily tape and disk, and how signals are represented on that media. To better understand this I will compare the differences in audio tape, as most people are most familiar with that, and digital tape.
In the first part I mentioned the bar magnet - a piece of iron with one end having a magnetic north pole and the other end being the south pole.
The media for recording is made of very small pieces of magnetizable particles, and for the most part these pieces are just like small versions of the bar magnet. Somewhat rounded - sort of cigar shaped. The term for describing these is 'acicular' - which means 'needle shaped.
We need this form because we have to have something which is can be magnetized and have a size and shape which is readily magnetizable, and one whose state can readily be determined. We use these particles to induce a magnetic field into a transducer - which in this case is a magnetic head.
Magnetic tape first made it appearance in Germany in the mid-1930s as developed by BASF. In the US much work was done at Armour Research - later IIT - Illinois Institute of Technology - and one of the most significant people there was Marvin Camras - who helped develop the particles and held over 500 patents in the field.
[For anyone who wants to learn almost anything you could ever imagine about magnetic recording, in all formats, audio, video, and data, I can unhesitating recommend the "Magnetic Recording Handbook" by Marvin Camras. Published in 1988 by Van Norstrand Reinhold - out of print - technically oriented - but should be available through some larger libraries.]
To describe it crudely, magnetic tape is nothing more than rust on tape. Magnetic tape consists of a base - the carrier for the 'rust'. Today we see different forms of plastic material for the base. However the base could be any flexible material capable of being magnetized. Some of the early experiments were used long metal bands.
In the US the first tape seen by the general public was a paper base with a magnetite coating. Then came plastic and the ferric oxides [rust] became the standard. Over the years tape technology has evolved to a very high level - but at the heart it is really just a long band of flexible magnetizable material.
The characteristics we rely on in the magnetic media are coercivity and retentivity. The first - coercivity - is how much energy we have to use to force a change in the magnetic state. [You might remember that from the word 'coerce' - meaning 'to force. It would be the magnetic equivalent of The Godfather making it an offer it could not refuse"]
A low coercivity material is known as being soft magnetically, while a high coercivity is considered to be hard. A low coercivity material can have it state changed by quite low magnetic fields. Early floppy disks could even be erased if you set a telephone on top of the disk and the phone rang. The magnetic field generated by the telephone bell caused this. [For the younger readers old phones had real bells driven by electromagnets to cause the clapper to vibrate. A disk with low coercivity could be erased by the field generated by the bell].
The other property is retentivity - how much magnetism remains after the magnetizing force is removed.
Most people are familiar with the tapes in their audio cassettes. While data tape and audio tapes are both forms of magnetic material their design is quite different. I'm going to try to explain the difference because that should make the entire process easier to understand.
Remember that the shape of the particle is needle-like or bar shaped. Also recall that to generate a current in a wire we can move a magnet past a wire [or in tape playback past a head] to generate a current. It doesn't matter whether which pole goes first [north or south] but we do have to have it change in order to generate a current. [Remember that a constant north or south pole would generate a short burst when it first started and then would fall to nothing].
In our magnetic 'paint' these particle are aligned in random directions. That means that some of the particle are going to pass under the head parallel to the head gap and since both the north and south poles pass under the head as the same time no current will be generated by the parallel particles.
The random order doesn't matter that much in data as we are recording the the domains to their maximum [saturation], and by magnetizing enough particles we will be able to generate a current. However orientation matters greatly in audio tape. the major reason is the signal we are recording and reproducing.
Analog recording places a magnetic representation of the audio wave form on the tape by varying the current and direction of the current going to the recording head. The highest output will come from particles that pass longitudinally under the head. We also will be magnetizing these at various levels so that when we pass the tape past the head an image of the original signal will be generated. [This is where the name analog recording comes from.]
Particles oriented at any other direction than perpendicular to the tape head will not contribute as much energy. What this means is that the output will be lower than it would if all the particles were oriented along the path of the tape.
The effect of this is a poorer signal to noise ratio with resulting hiss which you are no doubt familiar with in cassette recordings. In audio we also have to apply the voltage between certain levels - a minimum and a maximum - in order for an increase in the recording level will be at the same relative level on output.
In analog tape, while the coating [slurry] is still wet, the tape is passed through a magnetic field so that most of the particles will be oriented in a given direction. This is why just because a tape is the same size/width you can't always use it in any devices. Old video tape was 2" wide and used rotating heads that moved across the tape - so the tape was 'aligned' vertically. Using this in an audio 2" machine would give poor performance because most particles were oriented almost 90 degrees from optimum.
In digital we don't worry about this. If you think about this in relation ship to floppy disks - since these are 'pancakes' cut from large sheets of tape - random orientation is the only method which would work.
If you've ever had to push a car to move it, you know it takes more force to get it moving, but once moving you can get going faster and faster up to a point. It takes more force to get it moving, and once you get to a certain speed no matter how much harder you try you aren't going to get it going any faster.
Magnetic media works in much the same way - a little current and nothing happens, and as the current increases the amount of magnetization increases until it reaches a point where no matter how much you force it, no further changes will occur.
There are points just after the lowest point and before the maximum where the changes occur in a straight line. We call this linear response. If we try to record at too high a level the signal becomes compressed as we have reached the maximum output. We have most often distorted the signal by overdriving the amplifiers also. To low a recording level and we just get the noise from randomly oriented particles.
What we are trying to do in audio is to make the tape itself as transparent as possible so we hear only the recording and not the tape. Recording at too high a level increases the distortion and good audio should always be under 1% and maximum should be 3%. Signal to noise ratio should be at least 60db.
However in data we don't really care about distortion, and we want to record as 'loud' as possible, but within limits. We call this saturation recording.
In digital we can accurately reproduce the original signal with distortion well past 30% and signal to noise ratio in the order of 20db. You would NOT want to listen to an audio reproduction at that rate. All we have to do is determine if there is a pulse there or not. We may send the pulse to the head in a nice square fashion but we don't care what it looks like when we observe it on a display tube because all we want to know is if there is a pulse there or not. It it looks like there is a pulse we can rebuild one that is perfectly square if we wish to.
The ruggedness of data signals - whether there is a pulse or not - is why morse code was used in early message transmission. Even if the noise and static on a radio receiver was so bad you couldn't understand the voice, you could hear the tones that made up the dots and dashes. Other methods turned the transmitter on and off which was also readable where other means were not.
Now that you understand how the signals are stored lets look a bit more at the coercivity/force in regards to this.
If the coercivity is lower then it is easy to change the state. It's easy to magnetize, easy to erase. But there are other problems that low coercivity brings about.
If you put too much signal into the recording head you can erase part of the signal you just recorded so you have to space the pulses far enough apart so you don't effect 'self-erasure'. The lower the coercivity the easier it is to self erase. This limits the amount of signal we can store.
If you try to put the data pulses too close together the second pulse while being recorded can help erase the first pulse because while the magnetic field is concentrated at the gap it exists outside this are.
Another property of magnets - that I neglected to mention in the first part - is their interaction. A north pole of a magnet will attract a south pole of another. Opposites attract, and likes repel. One north pole will try to repel another north pole.
This gives rise to a phenomenon known as 'bit-shifting'. If two like fields are placed close together they will tend to move away from each other. In a medium where the bit-density is constant - as in tape this defines how much data can be placed in a given area. In low-coercivity medium the bits much be further apart and less data can be saved.
Therefore the amount of data we can put on tape or disk is a function of the coercivity and the size of the particle. Think of the difference between a wall built with 8" cement blocks and one built with bricks - with the blocks representing a single magnetic particle.
If the particles [bricks] are made smaller they can be closer together. We have to use a higher coercivity to as not to erase the adjacent particle and we also have to have a higher retentivity - more remaining magnetism - because the particles are now smaller and have to do more work.
We can help this along by maximizing the amount of magnetic material by a process called calendaring. This is accomplished by forcing the tape through very high pressure rollers so that if we compare it to paint we have more pigment and less binder. This also give the tape a highly polished surface.
This is needed to ensure high data density also. If the tape were not smooth the peaks in the oxide could be high enough so that the valley in the oxide would be further away from the head-gap than the width of the headcap - and the field changes would be effectively invisible to the reproducing head.
We continue this process of making particles smaller and smaller until we reach a limits. One limit is because the particles are ferric oxide. That is they are part metal and part oxygen. If we could get rid of the oxygen and have only the metal remaining we would be better off.
This was the step to metal particle tape. This is tape where the domains are made of particles of pure-metal instead of an oxide. Now we can make data pulses smaller as we can get more output from pure-metal particles than we could with particles that are not completely metal.
As we try to get more and more data onto the medium we see that now that we have metal particles the only way we can get more material to magnetize is to put the particles as close together as possible, and a way to do this is to get rid of the binder - the part of the tape paint that holds things together just as the carrier in paint holds the pigments.
That brings us to the media which we use today - that is called metal tape. It really is just metal particles clinging to the tape in the same way metal is plated on plastic toys to make them look like metal. Still small domains but with no binder gluing them together so that essentially everything can be magnetized
Now the only limits are how small we can make the magnetic domains. These particles become rather small. So small in fact that even though they are metal they can break. Different manufacturers approached solving this problem in different ways. Sony coated these small particles in ceramic which gave them strength and also protected against oxidation/corrosion. Oxidation [rust] was an early problem with metal tapes - and if it rusted we'd be back to the original tapes.
That wraps up this overview of the magnetic material, the next articles will deal with how we put all this together to store and retrieve data.
An editorial observation at this point:
During the process of developing metal and metal particle tape - manufacturers such as Sony and Fuji were having problems making this technology work as it should. But the did not abandon their work. While this was going on many American based tape manufacturers were pushing the current tape technology to the limits but changing their formulations of the oxide - a method often called doping - by adding various other elements to the oxides.
As with any technology there are usually finite limits and the manufacturers were more comfortable pushing the current technology than embracing a new technology - eg exchanging oxide technology for metal technology.
Once the metal medium was perfected the traditional manufacturers were left far behind. Manufacturers such as 3M - who revolutionized the world with the standard Scotch 111 audio tape in the early 1950 from the orignal Armour research have now all exited that field. Ampex also exited the field as did many others who stayed with 'traditional' methods.
The old saw 'innovate or die' seems to apply here.
Got something to add? Send me email.
More Articles by Bill Vermillion © 2011-04-29 Bill Vermillion
The camel has evolved to be relatively self-sufficient. (On the other hand, the camel has not evolved to smell good. Neither has Perl.) (Larry Wall)