What do you call the area on a hard disk smaller than a sector?

Solution 1:

I believe the term you're looking for is "magnetic domain", "a region within a magnetic material which has uniform magnetization" (wp). Hard drive designers are always trying to reduce the size of the magnetic domains.

But.

First, "channel codes" are used: The 0s and 1s recorded on the drive are not the same as the 0s and 1s you write and will eventually read. Sawdust is correct about how 1s and 0s are recorded, but there's more: The drive recovers clock pulses (so it can know where to expect a flux reversal, if there is one) from the flux polarity reversals, but cannot do so from stretches where there are no reversals.

This can be a problem. It's entirely plausible that someone might write an entire sector - 4096 bits with 512-byte sectors - of all 0s! Which would (if recorded simply) have no flux reversals. Due to irregularities in rotation speed, among other things, the drive would likely "lose its place" long before the end of that sector.

So the data to be written is actually expanded into somewhat more bits, using a channel code that ensures there will never be more than some number of non-flux-reversals written in a row.

I don't have a reference for the channel codes used in modern hard drives, but you can get a sense for how it works by looking up the "eight to fourteen modulation" ("EFM") that's used on CDs. Under EFM, each group of eight bits (which have 256 possible combinations of 0s and 1s) gets converted to a sequence of 14 bits (16384 combinations, but only 256 of them are valid codes). The sequences within each 14-bit code are chosen so that there are never more than a few - I think it's three - non-flux-reversals (0s) in a row. They're also chosen so as to reduce the bandwidth of the signal. Sounds bizarre, but it's true: By recording more bits you can get away with fewer flux transitions. For example, eight bits of all 1's would require eight flux reversals without a channel code, but instead can be recorded as 14 bits with far fewer than eight flux reversals.

Now, think about the very first bit written to a sector. Let's assume it's a 0. Where is it? Thanks to the channel code, the first bit actually written to the sector might well be a 1!

Incidentally, talking about CDs is not as off-point as it might seem. CDs use a similar scheme to that described by sawdust: The beginning or end of a "pit" marks a 1, a place where a pit could begin or end, but does not, is a 0. Just like flux reversals.

Then there's error correction. Error correction involves additional data stored with each sector. In the past the drive would read the primary data field + the ECC data of the sector, and if any errors were detected (for example, by reading one of the many "should not exist" channel codes), it would use the ECC data to correct the errors.

No more. Modern data densities are such that errors are more or less expected. So the ECC mechanisms were strengthened so that far more of the errors are correctable.

Yes, this does mean you have to record more bits, but it's a net win in terms of capacity.

The result, though, is that we can't really say that an individual bit, even a bit of a channel code, is recorded in a specific location, because the ECC data is as vital to recovering the bit as the channel code. And the way ECC works, the "influence" of each bit on the ECC data is spread over many, many bits of the ECC data. (This principle is called "diffusion.")

So, where's the bit? Well, it's sort of spread around. Change one bit in the input and there will be changes in the flux reversals in many places in the sector.

If that seems strange, wait until you learn about PRML, which stands for "probable response maximum likelihood": even the waveform recovered from the head, in which the drive looks for flux reversals, is interpreted statistically. But that doesn't have much to do with "where the bits are".

Solution 2:

I'm talking about the spot that stores 1 bit of data, the tiny little area that stores either a 1 or a 0 magnetically

Technically the magnetic particles do not store "either a 1 or a 0". That is simply nontechnical folklore to dumb-down the concept of magnetic storage. It's flux reversal that determines bit value, with the requirement that reading begins in the gap consisting of zeros. See this answer for more on digital magnetic recording techniques.

platters, which are doughnut shaped, highly polished disks.

"Doughnut" is not the correct adjective to use. "Doughnut" is synonymous with torus, and neither have any flat surfaces.

Each platter has a series of tracks going around it,

Tracks are concentric circles on the surface(s) of the platters.
The concept of cylinders needs to be mentioned.

This means that each physical sector on the hard drive has 4096 transistor like “areas" that can be either magnetised or not magnetised.

This is an inaccurate description. The magnetic recording is not like a "transistor" (e.g. a switch). The magnetic coating of the platter surfaces cannot be "not magnetised".

Any area that is magnetised represents a binary 1 and any area that is not magnetised represents a binary 0

This is inaccurate. The magnetized particles are polarized in either of two directions to create flux reversals to determine bit states. No flux change indicates the same bit state as the prior bit. A flux change indicates that the bit is the inverse of the prior bit.

what do you call the "area" on a hard disk that makes up the sector?

The "sector* is actually composed of an ID record and a data record.
The data record typically consists of a leading sync byte, the payload data bytes, and the ECC bytes.

On some types of HDDs, such as the old, floor-standing Storage Module Drive (SMD), the removable disk pack used a prerecorded servo surface to provide the bit timing and cylinder/track positioning. This prerecorded timing signal was derived by reading the dibits on this surface.

From a SMD reference manual (for the CDC BJ4A1 and BJ4A2):

Dibit is a shortened term for dipole bit. Dibits are prerecorded on the servo surface during manufacture of the disk pack. Do not confuse the the servo surface with the pack recording surfaces.

Dibits are the result of the manner in which flux reversals are recorded on the servo tracks. One type of track, known as the Even track, contains negative dibits. The other type of track, the Odd track, contains positive dibits.

But dibits are not the name you are looking for.
The most appropriate term that I could find is cell, as in:

The length of time required to define one bit of information is the cell.

Note that this definition refers to time rather than magnetic particles.

Solution 3:

I have worked for disk manufacturers and dealt with the hardware plus firmware that reads, writes, and formats the data. There is no name for anything smaller than a sector. However, a sector does not need to be 512 bytes. I have worked on systems that had sectors that ranged from 64 to 8192 bytes.

As others have mentioned, it would really help to know the audience. The OP's proposed explanation is wrong in many ways. I'd want to know the audience before proposing an explanation. For what it's worth, the Wikipedia article for disk sector, https://en.wikipedia.org/wiki/Disk_sector, has a reasonable lay-person's explanation.

Something that's missing from the Wikipedia article on disk sectors is coverage of the parts of a sector. Most disks are what we call soft sectored disks. Unfortunately, "Soft sector" redirects to the floppy disk article. They have an article on hard sectoring (https://en.wikipedia.org/wiki/Hard_sectoring) though it's incomplete as older hard disk drives also were hard sectored. Rather than holes in the media they used either small magnets mounted on the spindle, or a part of the spindle that stuck out a fraction of an inch and had holes much like the holes on a hard sector floppy diskette, or dedicated one platter surface that was pre-recorded in the factory with sector and clock marks. Hard sectoring simplified the logic needed to figure out when you could start reading or writing data.

Hard disks manufactured since the early 1980s are soft sectored. Soft sectors have the following components:

• Preamble - this is a special sequence of bits that whose pattern never appears in data.
• Header - this contains the sector and track numbers. In some of the disks I worked on we also recorded the head number here too.
• Sync - this is a special pattern much like the preamble. It exists because
  • It takes a finite amount of time to inspect the header data to see if this is the sector we want to read or write.
  • It takes a finite amount of time to switch the head from read mode (to read the header) to write mode (to write disk data).
  • The speed of rotation is not constant ask the disk gets older, it gets hotter or colder, or the power supply voltages change.
• Data - The data starts immediately after the sync pattern. When writing a sector we read the header and then write the sync and data. When reading we read the sync and using that can detect the start of the data. There are many ways to record the data. Non-return-to-zero (see Wikipedia) is a common method. Early disks used longitudinal magnetic recording (LMR) (see Wikipedia), while modern disks use Perpendicular Magnetic Recording (PMR) (see Wikipedia)
• Following the data are Cyclic redundancy check (CRC) (older disks) or Error Checking and Correcting (ECC) (newer disks) code bits.
• Following the CRC/ECC is the leadout pattern. This is much like the sync pattern and is so that the disk controller knows it's hit the end of the data. If it reads the leadout sooner or later than expected then the controller knows there was a glitch in the process.
• There is a bit of padding after the leadout. Nothing is written here. It exists in case the disk was spinning slightly faster than normal at the time a sector was written. We don't want to overwrite the following sector's Preamble, much less its header, sync, or data.

So, getting back to the OP's question, while there is no name for things smaller than a sector there's still quite a bit there.

Some of the disks I have worked on do sector blocking and deblocking. For example, we could be using 1024 byte sectors on a particular zone of the media (see Zone bit recording (ZBR) on Wikipedia) but the outside world only sees 512 byte sectors. Essentially, for each zone we use the most efficient on-disk sector size. I use the terms "sector size" and "internal sector size" meaning that while at times we dealt with things smaller than a sector they were still called sectors.

Solution 4:

ollimpia, I would replace the latter part of your explanation with:

"can store 512 bytes that each have eight bits. This means that each physical sector on the hard drive holds 4096 bits of data. The platters are coated with a special material that can both reliably retain a magnetic polarity and allows the polarity to easily be changed. The data is stored using combinations of north-to-south and south-to-north magnetic polarity."

I deliberately did not provide a name such as "spot" or "area" for the bits on the media. Neither word is wrong but they are not a perfect fit either. I also deliberately did not spell out the translation of 4096 data bits into the polarized "spots" on the media.

The reason I'm avoiding words such as "spot" or "area" is that when reading the data we don't read the magnetic polarity but rather sense the shift from one polarity to another. Thus we are looking for either a "shift" or "no-shift" to know if we are dealing with a 0 or 1 bit.

The reason I avoided saying there is a one-to-one translation between data bits and what is written on the disk media is that we can't go for too long with "no-shift" as we might loose track of where we are. We use the shifts to stay in sync. A disk drive translates sequences of data bits into slightly longer sequences of bits that are used on the physical media. The sequences used on the media are designed so that we never go too long with "no-shift" regardless of what the user data contains.

Group Code Recording (GCR) is a common method for encoding the data can be explained as using five bits on the media to record each four bits of data. That's not a perfect explanation as the disk is looking at the shifts in polarity and not the bits. If you look at the tables on https://en.wikipedia.org/wiki/Group_code_recording you will see sequences of zeroes and ones. A zero "no-shift" and a one is "shift." The four data bits "0111" could be encoded as "10111". We read "10111" from left to right and when writing this to the media we will polarize the media as: 1) north-to-south (shift or no-shift depends on the last bit of the previous nybble) 2) north-to-south (no-shift compared to the previous bit) 3) south-to-north (shift compared to the previous bit) 4) north-to-south (shift compared to the previous bit) 5) south-to-north (shift compared to the previous bit)

Earlier I explained the parts of the sector with the preamble, sync, etc. The preamble, sync, etc. are recorded using shift patterns that do not exist in the GCR translation tables. Usually they are longs strings of shifts or no-shifts. For example, 6250 GCR RLL will never have more than seven shifts in a row meaning our special patterns can be eight or more shifts in a row. 6250 GCR RLL will also never have more than two no-shifts in a row meaning we can use three or more no-shifts as special pattern that will never exist in recorded user data.

As technology improves we are able to have longer runs of "no-shift." This has lead to encoding systems that are more efficient than four data bits encoded as five on-disk bits. The extra efficiency has been used to both increase the available storage but also to add error checking and correcting (ECC).

Other technology improvements are to take advantage of that can tell a south-to-north shift apart from a north-to-south shift and "analog recording" in that they are able to vary the intensity of the polarity as a way to squeeze extra information onto the media.

Thus, while the disk in a Macbook Pro appears to be a digital storage device engineers that design read/write heads and the coating applied to disk platters are working with analog signals.

If you are interested in math then look up "finite field arithmetic" and "abstract algebra" both of which are used in designing what are known as channel coding systems.

Rather than "doughnut shaped" I would say that the disk platters look like CD or DVD discs made from metal or other hard materials. Blank platters ready to be installed in a disk have a hole in the middle just like what you see on a CD or DVD.