Magnetic media notes
Technical theory
Linear tape heads
Note: a good deal of this section applies to reel-to-reel tape, compact cassette, and many other tape formats.
Heads and tracks
A lot of tape uses multiple tracks side by side, which are format-specific conventions of splitting the width of the magnetic material.
For example, cassette tape typically has four separate channels: stereo, and two sides. Stereo cassette players heads typically have two coils, for both tracks of just one side. These two tracks are to one side of the tape, which is also why you physically flip the cassette in the other way: it puts the other two tracks on the active side of the head, also going the other way.
Eight-track has, well, eight tracks, and all go the same way.
These tracks are typically used as eight mono tracks, though there are variations.
Reel-to-reel tape had more variations, and is a separate discussion.
There are many variants, though many of them specific to a line of products, or sometimes even a single one. For one example of a variant on cassette tape, the Tascam Portastudio series uses cassette tape as a four-track(verify), one way, which given that it was a consumer-affordable(-ish) 4-track in the eighties was pretty amazing.
Gaps and fancy versions
Heads are conceptually a ring of conductive material used as an electromagnet - with a (very small) gap in that ring.
Driving current into this creates a magnetic field that fringes out at the gap, and will change the magnetization of magnetic materials.
Similarly, playback is induction into that gap.
The gap is small in part because it functionally only really does anything at its edge,
and a smaller gap lets you control the magnetic field more precisely: the smaller the gap and and the closer the medium, the better the frequency response can be (for a combination of reasons, but one roughly analogous to sample rate).
While the workings of a head means you could DIY one in theory, good audio response means you want a gap on the order of micrometers, which is impractical to make yourself.
The record head and playback heads are the same principle.
Yet their workings involve different currents, so different ideal impedances, and involve somewhat different circuits. The details of the gap size also differ.
- for playback a smaller gap gives you a better defined magnetic readoff, so a somewhat stronger signal.
- For recording you can deliver a little more flux for stronger magnetization (the wider gap doesn't lower fidelity, because the imprint is left by one edge of the gap).
So while you could use one head/gap for both reading and writing purposes, and it works decently enough that simpler, cheaper machines do this, for practical reasons it works better to have separate and record heads.
There is a discussion whether you care to have a separate record and playback head.
The central point is that if you use the same gap for both record and play, that's necessarily a compromise in gap size, and separating them avoids that compromise.
With one head being erase (see below), this is often referred to as 3-head (erase, write, read) versus 2-head (erase, write+read) setups. But that's confusing naming, because of the two-gap read+write heads, which are basically the same quality audio, and less aligning bother.
Note that one side effect of actual 3-head is that it may have a switch to let you listen to to what you are recording (useful as a quality check during copies, also lets you make a tape delay machine of it), while this doesn't seem to be an option on two-gap heads (presumably it'd magnetically couple so not be useful this way)(verify).
https://www.mwit.ac.th/~physicslab/hbase/audio/tape.html
Erasing
If you wrote audio to a tape that already had audio, it would blend the signals, sort of like double-exposing a photograph (would probably vary a little with the kind of bias used(verify)).
When you overdub that's the point, but in general you want it to either always wipe as it's writing, or you want the choice.
In any case, units that can record also have an erase head which, roughly speaking,
randomizes the magnetization of the tape, by using an above-audio rate signal, fairly strongly.
Erase heads are separate in part because they act more widely than the write, they work better with larger gaps and at higher current (and frequently have two gaps, basically for thoroughness).
Simpler erase heads, that are essentially just a permanent magnet, are seen in cheap units (likely to also be DC bias, not AC bias), but this saturates the tape one way and is less ideal for the same reasons DC bias is.
Bias
The earliest attempt at recording applied the baseband input signal to the recording head, as it came in.
Baseband here meaning 'put the actual audio on the head's transformer, as-is'.
Very roughly speaking, if you think of a waveform going around its average, the result would be a magnetic field going one way and the other, so one half of that wave would lead the tape to be magnetized in one direction, the other part of the wave magnetization the other way.
This certainly works,
but it turns out the magnetic material isn't quite linear used every which way.
In particular, things don't quite do what you tell it around that middle of the wave, where you would ask for little to no magnetization on the tape
and the polarity flips. (...for multiple reasons. In part because the magnetization you apply is weaker because when intended signal is weaker. This seems to be one of the reasons for a moderate hysteresis effect around there)
If you e.g. tried to put a perfect sine wave on there, it would be close, but a little wibbly around its middle.
This amounts to an uneven frequency response, and introduces some soft distortion, in particular for lower frequencies.(verify)
So it's certainly sound, but it's not quite right.
"If that zero crossing is part of the issue, why not shift it away from that zero point?"
This is a simple fix, yes. It is the basis for DC bias. Add a constant voltage to the signal you put on the head. As long as we can keep that waveform above zero, we never run into that flippy hysteresis, and as long as we don't saturate the tape (=hit the point at which it will not retain more magnetization), we don't run into that issue either.
This does give a much more linear response, and lessens the distortion from the around-zero non-linearity.
It also means you can use less of the range of the magnetic medium, meaning less dynamic range and more noise than you probably could get from that magnetiztion.
Also, it necessarily means net magnetization, which turns out to give some other issues (noisier? why?(verify)).
"Okay, what else?"
We ended up primarily with AC bias.
It turns out that adding a higher-than-audio-frequency signal while recording has an effect is not only a decent way of wiping the tape, it also leaves net-zero magnetization. So what if we add such an erase-like signal to the signal we record?
The placement of the waveform is still around zero, with polarization both ways as it was without any bias.
But it turns out there are physicsy reasons that means that mean less hysteresis than just baseband, and with little net magnetization. It ends up being a great tradeoff that avoids the biggest downsides of DC bias and of no bias.
Notes:
- (something between 25kHz..150kHz(verify), apparently 40kHz..80kHz is ideal?, and at higher current than the recording current(verify))
- (note that often, the same same oscillator is used for both erase head and bias signal, but with different currents).
- The ideal current to apply varies per #Tape medium type, which is part of why tape medium is either a setting or detected (as in cassette tapes)
Most systems use AC bias because it works better,
though some recent cheap models re-adopted DC bias to save a few bucks.
See also:
https://www.eevblog.com/forum/projects/cassette-tape-record-circuit-bias-osc/
Tape medium type
Type I (ferric / normal)
Type II (chrome, but often pseudochrome)
Type IV (metal)
The breakable tabs are for write protection - prevents recording once removed.
Has effect on
- bias (e.g. metal types need more energy)
- EQ
For the compact cassette format, the the notches on the cassette can make players/recorders detect this.
Other players didn't have that detection and made you toggle switches.
Some fancy recorders also let you fiddle with the amount of applied AC bias,
because too little makes for more distortion, and too much means less high-frequency response(verify)
On azimuth
The positioning of the head relative to the tape matters, in all directions.
When mounted in a machine, most directions (distance to tape, roll along tape) of a head are fixed and/or well controlled.
Azimuth ends up being the main one you may need to fiddle with.
Basically there is one screw that pushes one side of the head across the tape.
- the larger effect is that the head can be to the side of the tracks, sideways across the tape
- when it's mostly right it also matters slightly in that if it's particularly rotated, you get a very small delay between tracks, which if mixed works out as comb filtering in the high frequencies
For combined-R-and-W heads, misalignment means general tapes will sound worse here, and tapes recorded on it will sound worse elsewhere, but tapes recorded and played back within this one misalisystems will seem pretty good.
For separated R and W heads you have different combinations of reads fine, records poorly, "other person's recordings work differently".
You can always adjust your head for a specific tape, and if you have a factory-recorded tape, for tapes in general.
You can e.g. count the screw-turns it takes to go from equally bad to either side, and then aim for the middle.
This works decently, though isn't overly precise.
There is a more precise method if you can forcibly mix the channels to mono, as the phase effect difference creates the comb-filtering-like in the higher frequencies, which gives a narrower optimal position where it audibly sounds better. So basically find any factory-recorded tape (see e.g. second hand stores). (Metal tapes are slightly better for this purpose due to better high-frequency response(verify)).
The tiniest drop of locktite may be a good idea to fix it in place for longer.
http://www.endino.com/archive/cassettes.html
For the electronic fiddlers
Playback is a magnetic field carrying baseband audio, so you can play with this pretty easily, and for low-fi it doesn't have to be precision engineered either, see e.g. https://www.youtube.com/watch?v=MMD-d-1erk4
But you may care to do it a little better than that.
The minimal circuit for playback is mostly an op amp, both for the voltage gain to line level, and as a buffer.
Playback amplification may well also equalize, in which case you'ld like to add a filter via a few more resistors and capacitors. Look for tape head pre amplifier circuits', and know there are ICs to help. https://www.youtube.com/watch?v=wn772K9cm9E
Recording is more complex.
The recording head's inductance matters (affects frequency response, you basically have to adjust for that) and recording current matter, as does the bias current, as does keeping the bias away from the source. (you can, alternately, mix the bias in with the signal before amplification, though this needs better amplifier specs)
See also:
Strategies of using tape
Linear tape
In linear tape, information on the tape is in one long line along all of the tape (well, up to), dragged past the head or heads.
In analog audio tape, the strength of the stored magnetic field is the signal you get out, without any added translation.
Do an image search for magnetic view audio tape if you like more visuals.
The most basic tape might use much of the width of the tape for a single signal - for comparison, not entirely unlike a magnetic wire recorder.
Later we figured that we don't need much width, and could design to store multiple tracks. Multiple tracks takes a little extra care, such as wider tracks so that the tape doesn't have to be positioned very precisely, but adding enough wide separation between tracks so that it wouldn't crosstalk between them.
(Yes, you could put things closer together - linear serpentine does exactly that, but it has to spend hardware and engineering on keeping exactly on the right tracks. On simpler linear tape like cassette and reel to reel, the only guide you need is the bits that physically guides the tape through.)
Limits
The last means we're not using all that much of the magnetically available area. But that's not even the largest issue.
It turns out there is an upper limit to how well tape magnetizes for a fast-changing incoming signal.
That rate in change (which is effectively the bandwidth, whether spent on data or audio sampling frequency) is limited, first by the head design not being too simplistic, but soon after by the tape's mechanical speed.
For audio, our demands (~20kHz bandwidth) on typical tape
work out at a mechanically reasonably low tape speed when you use it for linear recording.
Consider that cassette or reel tape can easily store 30 minutes to a side.
A handful to a few dozen centimeters per second (using the tape faster is costlier just because you use more, but worth it for professional use).
Video on the other hand has much higher bandwidth demands. VHS uses ~3MHz (and isn't all that good).
Sure, the tape is specifically better at this, but only so much. Sure, we could pull the tape past that much faster, but that becomes mechanically impractical for a few different reasons, like that the speed means easier damage (or waiting longer for smooth braking), running out of tape very quickly (or having to make the tapes huge). Also, just how much faster you need for, say, video is impractical.
So people looked at other methods.
Data lives more in a "it gives us as much as it gives" are but we have expected more from tape data backup for a long time as well, and arguably the only reason tape backup is still relevant at all is that it does still live up to that.
Multitracking
One way to add more content is to make the to add more channels alongside each other.
For the same amount of data, making the tape wider is also easier than pulling it through faster.
In audio you could call them completely separate.
- Say, stereo cassette tapes were essentially four tracks: left and right one way, left and right the other way.
- 8-track has eight, one waym abd
- Reel to reel tape exists in multitrack variants that give you amounts in the 4..24 range (the last only on chunky 2" tape), but the upper range is already pricy (and starts to approach the difficulties linear serpentine was aimed at).
Some methods might use a few.
- An audio multitracker might play back some or all, and record one (or more).
- And yes, this also blurs the line somewhat between linear tape and linear serpentine.
- You could say that linear serpentine is this pushed to the limits that you need costlier hardware for.
- Also that it is pushed so so many tracks that you can't even read off all content at the same time
Other methods might let you split a single signal into all these tracks, and then merge them together somehow in playback.
Either way, this works fine up to a point, though it puts higher requirements on aligning the tape correctly.
If you don't, then in audio you lose quality, in data it just stops working before that.
Linear serpentine is usually considered its own thing for that reason - it needs servos and feedback to work at all.
Helical scan
Another way is helical scan, as used e.g. in most video tape, and in DAT/DDS. These systems place tracks on diagonal strips. This doesn't increase the density much over many-track, and is more complex head construction - the head itself is angled, it requires tracking and reconstruction, the head spins faster than the tape passes.
The last means that you get more effective speed across the tape, without having to pull the tape faster.
What comes out is a single high-rate thing (a factor ~150 more than most basic audio tape, and still a few factors over most many-channel linear tape(verify)).
It also does help that reassembling your signal is usually a single standard and the player's/head's job, not historical creativity as with a bunch of many-tracked linear tape.
https://en.wikipedia.org/wiki/Helical_scan
For tapes for data backup, practical things like physical size of the archive matters.
Both helical and linear serpentine are now roughly at their engineering limits, and unsurprisingly their density is similar to each other. The reasons to choose one over the other for e.g. backup are largely pragmatic ones - while for other uses (VCR, DAT, analog audio) they are mostly historical.
Linear serpentine
With that wowed introduction of helical scan, you might assume linear is dumb, helical is best, and e.g. the LTO backup tape that is still in use must be using helical.
Actually, LTO uses linear serpentine, which is a fancy way of saying:
- it uses so many tracks that it has more tracks than it has heads
- it has a way to move the heads to specific tracks (with precision - the width of tracks is on the order of dozens of micrometers)
The engineering is a lot more precise than in classical linear tape, and you need to feed the tape past the head multiple times to even read, or write, all of its data, but if you care about the most density per tape, this works out better than helical scan (and much better than linear).
For some reference,
- LTO-1 was 8 heads and 384 tracks
- LTO-8 is at 32 heads and 6656 tracks
Yes, that means writing a tape to capacity actually requires, depending on the LTO version, something between fourty and a hundred-fifty passes of the tape's length.
So this is not at all a random-access system, and has rather high read and write latency. Yet for most backup and archival uses, that's rarely an issue, the price is hard to argue with, and it's also engineered to last longer than many other common storage media.
Audio and video formats
Reel to reel tape decks
Some practicalities
Tape width
Probably most common on home machines is 1/4" (6.35mm, usually called 6.3mm).
1/2", 1", and 2" inch also exists (12.7mm, 25.4mm, or 50.8mm) wide but is more specific.
Reel sizes
Leader tape
Leader tape (Vorlaufband, Aanloopband) is plastic tape of the same width as the magnetic tape that you would splice-and-tape at the start and end of a tape.
This avoids you having to crumple a bit of useful tape.
Also, machines could stop playing when they sense a lack of magnetic tape, instead of winding out. You could also use that mechanism to separate individual tracks, though presumably mainly radio stations did that(verify).
Rotation speed, reel size switch
Compact Cassette
Being a common format, this was often called 'cassette tape', or 'audio cassette'.
But when you want to distinguish it from one of the many similar formats that contained tape, you can use its more official name, Compact Cassette https://en.wikipedia.org/wiki/Cassette_tape
DCC
Digital Compact Casette (DCC) seems conceived as
- a successor to analog cassette tape
- designed for somewhat easier adoption: the size and head assembly are similar enough that DCC devices could be designed to play (but not record) the older analog cassette tapes.
- something of alternative to DAT,(some even calling it S-DAT (Stationary-Head Digital Audio Tape), to distinguish it from the rotary head nature of DAT tape),
- and a competitor to MiniDisc
The tapes are two-sided (with all players having auto reverse), each side having 9 tracks of information that will be combined into one stream of information (8 for the audio, plus one for extra information).
Even with those 8 tracks, there is only limited bitrate (also consider subcode information for emphasis, SCMS), error correction, and 8b/10b encoding, so (acoustically-aware) data compression was used (called PASC, which would later be called MPEG-1 Audio Layer I). So yes, unlike DAT, DCC was a lossy format.
The ninth track could, on prerecorded media, repeat an index of tracks, with names, positions, and side. Seeking around still wasn't very fast, though.
Despite being more targeted at the consumer market than DAT, it never took off,
though like DAT, it was moderately popular in some professional setups.
Released in 1992, it was discontinued in 1996 as it seemed to displace none of the mentioned formats. The recording industry / RIAA legal dealings didn't help either.
DCC recorders could input from S/PDIF, which was often output from CD or DAT or such,
so it counted on sample rates like 32 kHz, 44.1 kHz, 48 kHz.
https://en.wikipedia.org/wiki/Digital_Compact_Cassette
DAT
DAT (Digital Audio Tape) was one of the first digital audio recording products, introduced in 1987.
(sometimes R-DAT, Rotating-head Digital Audio Tape, apparently to distinguish it from dcc?)
DAT recorded audio onto magnetic tape with a helical head, for more data density than linear tape at reasonable tape speeds (like video recorders).
Digital audio recording wasn't unique at the time (there were some niche systems at the time that recorded digital audio e.g. onto video tapes), but DAT was originally intended for more general use.
...which it never did, probably for a good part due to legal action from the RIAA, but also in part because it was expensive.
For some historic context, audio CD was released a year or two before DAT was;
Where CD seemed poised to replace vinyl,
DAT (and/or minidisc) was perhaps intended to do the same for analog compact cassette tapes.
DAT never took off at home or to sell albums, though.
Lobbying and legal threats from the RIAA probably had a lot to do with that - RIAA was afraid that an easy recordable format would kill their sales.
That, and the DAT setup was fairly pricy.
So DAT ended up mainly used as a tool within radio and the recording industry, and that use went away when recording studios moved to more convenient formats - like solid state recorders, and computers.
It was an seemingly still is used in more professional film/video recording, presumably in part because you can write out the same timecode you used on video, and avoid a lot of searching later.
More technically
- helical scan head
- also implying the tape is always on the head
- and you can be reading while seeking (though that sounds like a garbled version, not a high-pitched version like in analog tape).
- and players could assist finding pieces of silence/audio
DAT data and metadata
See also:
SCMS
DDS
Digital Data Storage (DDS) is DAT tape extended to be used for computer backups. It's digital, why not use it for data, y'know?
The first generation of DDS tapes were basically just regular 60-meter DAT tapes,
had the same roughly-1.3GB worth of storage on that tape,
and DDS hardware was mechanically quite close to DAT.
Later DDS variants increased length and density.
There are 7 generations of DDS (the largest option 160GB), but note that this incremental change means there are backwards compatibility details to consider.
(This (and many other backup tape formats) have mostly been replaced by LTO, which has been around for a while but is still very much relevant, because the price-for-capacity is still better than platter or flash.)
Is DAT and DDS the same tape?
The 60m DAT tape versus first generation of DDS, 60m DDS tape? Basically yes.
It's the same mechanism, and it's just data, the use of the tape is basically the same.
Though the interpretation of that data is different, you could use either tape for either.
Some DDS drives were capable of reading audio off DAT, though it seems few people ever did that.
Note that the tapes are used a little differently.
DAT is often used accessed in a mostly-linear way (long recording, long playback)
DDS seek around more frequently, using it more intensely. Which wears out tape a little faster, which seems to be why people say DDS must surely be built to a higher standard. I'm not how true that is, and how much of it is still confirmation bias and such. But it is a reason to consider avoiding used DDS tapes for DAT.
So can I use DDS tapes in my DAT machine?
tl;dr: If you stick to ≤ 60m (1.3GB) DDS-1, yes.
Longer tapes may well work, but aren't guaranteed - 90m DDS-1 and 120m DDS-2 probably still work (DAT audio tapes was sold in these lengths too), longer may not.
You can get into backwards compatibility and other specs, but probably me main reason is mechanical: DAT players's motors may only have the torque to deal with DAT's 60m of length (DAT never changed much).
Professional decks may have more leeway here when they're a little overspecced, and because of DAT's history, many decks are relatively professional, so YMMV and you may be fine with 90m DDS-1, DDS-2 and maybe DDS-3, but you should really expect DDS-3 and later will not be working.
Like in reel-to-reels, a good test is to rewind and forward it completely - because fast forward (unlike playback) is constant rotational speed so this makes the highest demand on the motor torque, higher than regular playback.
If feeding it through to both ends doesn't seem to struggle, then it's probably fine.
(and note that it will stop fast-feeding before it stops being comfortable playing(verify))
Video8
https://en.wikipedia.org/wiki/8_mm_video_format
DV
https://en.wikipedia.org/wiki/DV_(video_format)
Data formats
Floppy heads
Floppy heads are conceptually quite similar to tape heads.
Roughly compatible, even. this neat audio-floppy hack wires the floppy head to a cassette tape device.
Note the read/write slit leads the two erase slots to the side
There are differences, primarily in the way it deals with tracks.
It seems typical floppy heads:
- on the track: read-write gap
- to the sides: tunnel-erase gaps (well, typically tunnel style. There also seems to be a straddle type, which is alongside rather than behind, and requires slightly different timing)
- these tunnel heads are not strictly required, but were practically quite useful:
- wipes the spillover that may appear between tracks. Makes for more stability. Also helped interoperability a bit - without this, reuse of disks in different standards would be likely to react to off-track magnetization that was old data but would work as noise (and could not be altered in that other-type drive because it's off-track)
Note that where tape erase and write at the same time, floppies do this in two passes:
- erase the sector (stronger amplification on the R/W head)
- write the sector (regular R/W amplification)
Perpendicular recording existed, but was never very common (used e.g. in 2.8MB (4MB raw) floppies).
This required the magnetic medium to have higher coercivity,
and apparently implied a pre-erase gap leading the read/write gap (like how erasing works in tape. Not sure whether this was necessary or just a design decision).
It seems the R/W heads are typically a center-tapped coil (i.e. coil pair on one core).
I'm guessing this is because bit values 0 and 1 are magnetized in one direction or the other, and it's slightly easier/cheaper to drive this way. (verify)
A read is probably across the pair, because stronger signal. (TODO: read up, this must be moderately common knowledge)
Opening a head and seeing two coils is usually the RW part, and the erase part(verify) (the center tapping of the former is hard to see)
As such, you typically have 5 wires and they'll be:
- shield - grounded, and not connected to other parts at the head size (other than large bits of metal around it - also making it easier to find this one with a multimeter)
- common between all coils
- R/W 1
- R/W 2
- Erase
Making a table of all possible combinations's resistance (just because they're different lengths of wire) typically helps find which is which, though you'll be thrown by there possibly being a diode and resistor in there.
The best explanation on how writing bits works that I've found sits in a Shugart service manual (SA400).
See also:
- http://ohlandl.ipv7.net/floppy/floppy.html
- http://ohlandl.ipv7.net/floppy/DS011332.pdf
- http://www.osiweb.org/manuals/MPI_B51-B52_Product_Manual.pdf
- https://nfggames.com/X68000/Documentation/Floppy%20Drives/Shugart/SA400%20Minifloppy%20service%20manual.pdf
- http://www.retrotechnology.com/herbs_stuff/drive.html
- http://jlgconsult.pagesperso-orange.fr/Atari/diskette/diskette_en.htm
- https://cdn.hackaday.io/files/20185863595040/NEC%20FD1036%20Floppy%201985.unlocked.pdf
- restoration/DIY (has some practical notes)
- https://www.retrohax.net/commodore-1541-floppy-drive-fixing-chaos/
- http://deramp.com/downloads/altair/hardware/8_inch_floppy/Drive%20Restoration%20Drive1.pdf
Floppy formats and interfaces
Loading data from tape
Unsorted
SPARS code
A three-letter code to signify that
- recording: Analog or Digital
- mixing: Analog or Digital
- mastering: Analog or Digital
This mattered before studios were doing digital recording and mixing -- using computers as is ubiquitous today --
in part because reel to reel cannot be pushed to 16 bits of pushing-the-noise-down quantization even with fancy noise reduction.
So 'ADD' and 'DDD' meant 'this CD is probably a little lower noise than it would be if it said AAD'
It was meant for varied formats, but ended up mostly used on CD (for which the last is always D).
SPARS actually withdrew this, because a lot of recording and mixing is mixed digital and audio, so this isn't really representative.
Record labels continued to use it, probably figuring that this was a nice marketing label.