Magnetic head notes

From Helpful
Jump to: navigation, search
This article/section is a stub — probably a pile of half-sorted notes, is not well-checked so may have incorrect bits. (Feel free to ignore, fix, or tell me)

Linear tape heads

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 one side. This is also why they're to one side rather than in the middle of the head, and why you physically flip the casette in the other way: it puts the other two tracks on the active side of the head.

Eight-track has, well, eight tracks. Typically used as eight mono tracks (though there are variations), and only goes in one direction.

Reel-to-reel tape had more variations, and is a separate discussion.

Gaps and fancy versions

This article/section is a stub — probably a pile of half-sorted notes, is not well-checked so may have incorrect bits. (Feel free to ignore, fix, or tell me)

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.

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).


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.


This article/section is a stub — probably a pile of half-sorted notes, is not well-checked so may have incorrect bits. (Feel free to ignore, fix, or tell me)

The earliest attempt at recording applied the baseband input signal to the recording head.

Intuitively (and roughly): if you think of a waveform going around its average, this would correspond directly to the tape being magnetized one way or the other.

This works, but it turns out the magnetic material has a moderate hysteresis effect, which means the magnetic response is not very linear particularly around zero (in part because the magnetization you apply is weaker because when intended signal is weaker), which messes frequency response and distortion, particularly the low end.

A simple improvement is DC bias, which means adding a constant voltage to the signal you put on the head, to use the region where the tape gives much more linear response.

This lessens distortion from the non-linearity, but leaves net magnetization, which is noisier than necessary (why?(verify)). It also means you can use less of the range of the magnetic medium, meaning less dynamic range than possible.

We ended up primarily with AC bias, which refers to adding a higher-than-audio frequency AC signal (something between 25-150kHz(verify), apparently 40-80kHz is ideal?, and at higher current than the recording current(verify)) while recording.

This doesn't alter what audio is recorded, but leaves very little magnetization, making for pretty good linear response, so avoids most downsides of DC bias.

The ideal current to apply varies per tape type (e.g. metal types need more energy), which is one reason tape recorders detect it via cassette notches - or have it as a setting.

Some fancy recorders let you fiddle with the amount of applied AC bias, because too little makes for more distortion, and too much means less high-end response(verify)

Most systems use AC bias because it works better, though some very cheap models re-adopted DC bias to save a few bucks.

See also:

On azimuth

The positioning of the head relative to the tape matters, in all directions.

When mounted in a machine, most directions (distance, roll) of a head are fixed and well controlled, so azimuth ends up being the only one you need to ever fiddle with.

Basically there is one screw that pushes one side of the head across the tape.

when it's mostly right it matters slightly in that if it's particularly rotated, you get a very small delay between tracks that works out as comb filtering in the high frequencies
it usually matters much more that you are moving away, sideways, from the position of the recorded track on the tape

For combined-R-and-W heads, misalignment means general tapes will sound worse here, and tapes recorded on it will not worse elsewhere, but tapes used within this one systems will seem to mostly perfectly.

For separated R and W heads you have different combinations of reads fine, records poorly, and cases of interchange.

You can always adjust your head for a specific tape, basically just fiddle until it stops sounding muddled or weird. But this is pretty inconvenient. (we used to do this with our Commodore 64 tape deck for tapes that came from a friend, tweaking and repeating loads until it worked)

Ideally all heads are aligned the same way, and in particular you want to record in a standard position.

Fairly-decent aligning the R head is easy - if you have a factory-recorded tape. You can e.g. count the twists it takes to go from bad on one side to bad on the other side, and aim for the middle.

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 by it sounding 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.

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.

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.

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:

Helical scan

Most audio tape is used linearly. Since there is an upper limit to how well tape magnetizes a faster signal, data rate is eventually limited primarily by the tape's mechanical speed.

This is fine for audio, our ear's frequency response asks for what works out as a reasonably low tape speed - consider that cassette or reel tape can easily store 60 minutes.

Getting higher frequency response was interesting e.g. for video.

Spinning a lot faster faster brings in some mechanical trouble, so people looked at other methods on still-slow tapes.

One is making it wider and adding more channels. This is e.g. what eight-tracks (and some reel-to-reel) has over cassette tape, tends to use more of the tape surface, is still easy, and is great if you like storing separate things.

Another way is helical scan, as used e.g. in VCRs and DAT. These systems place tracks on diagonal strips. This doesn't increase the density much over many-track, and is more complex head construction (angled head, requires tracking, spins faster to get more effective tape speed than the tape's mechanical speed) but what comes out is a single high-rate thing (a factor ~150 over audio tape, and still a few factors over most many-channel(verify)) and you wouldn't have to separate and reassemble your signal because that's the head's job.

Tapes for data backup, practical things like physical size of the archive matters. Both helical and linear are now roughly at their engineering limits and unsurprisingly their density is similar, and 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.

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.

This whole thing is ~4mm wide, the black stripe ~0.4mm, the three slits ~0.1mm each
Note the read/write slit (leads), and 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)
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 RW amplification)

Perpendicular recording existed, but was never very common (used e.g. in 2.8MM (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 two separate things than it is to drive one coil both ways. (verify)

A read is probably across the pair, because stronger signal. (I can probably verify this looking at some)

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.

See also:

  • mostly head-related:
  • wider tech notes
  • restoration/DIY (has some practical notes)

Loading data from tape