How The Mellotron Works

Very simplistically, the sound generation technology of the Mellotron is very similar to what you’ll find in your hi-fi’s cassette deck (assuming you still have a tape deck in these days of CD!). Or, as somebody once put it, "rust being moved past a magnet". So to understand the Mellotron a little better, let’s consider a cassette deck analogy for a minute. For all you technology freaks out there, to whom this is old hat, please bear with me as I wish to start at first principles to ensure that anybody reading this can understand the Mellotron.

The basic principle of cassette replay is to encode a magnetic representation of sound onto magnetic tape and then, to recreate the original sound, pass the same tape over a magnetically sensitive replay head to induce a current into some electronic circuitry. The current induced into the head by the moving tape will be analogous to the original sound recorded.

Within your cassette deck, the basic transport mechanism and sound path is:

  1. A tape feed spool;
  2. A tape head, and an associated sprung felt pad (within the cassette itself - go on, have a look) to press the tape against the head;
  3. A rotating metal spindle, called a capstan, and a rubber wheel called a pinch roller;
  4. A tape take up spool.

Playback of the recorded sound is brought into effect when you press the play button on the cassette deck by the mechanism of the deck:

  1. Moving the playback head up from its rest position to press it against the tape;
  2. Pressing the pinch roller (and thus the tape) against the rotating capstan;
  3. Revolving the take up tape spool (and on some decks the feed spool is also motorised).

A common misconception with tape decks is that the take-up spool drags the tape past the head. Actually, it’s the capstan and pinch roller that perform this, which is why they’re close to the actual heads. The take up spool is turning to take up the tape that has passed the head, as without this you’ll end up with a bit of a tangle. Finally, a cassette tape has four tracks upon it, with two tracks for stereo playback in the forward direction and the remaining two for the other side. Thus a replay head actually needs to sense the recording on two tracks for stereo reproduction.

So, how does this cassette talk map over to the Mellotron?
I’m going to talk about a particular Mellotron model for this description: The Model 400 or the most successful version. There are some differences between this and earlier models, but more on this later.
If you imagine a Mellotron keyboard, we actually have a very similar transport in some respects to the cassette analogy, but it’s also quite different.


First, under each key you’ll find a piece of tape with three tracks, each with a sound recording on it. There are 35 tapes; one for each key. Usually the pitch of the sounds on each tape are associated with the key that it is under, but this dose not have to be the case.
Under each key is a tape head. Each key also has a felt pad attached to it, and also has its own pinch roller.

All the Mellotron tape heads are only capable of mono playback, so why the three tracks on the tape? The tape heads are mounted on a sliding assembly, which allows the them to be laterally moved across their respective tapes to align with one of the tracks. This is controlled by a selector switch on the control panel which allows the user to select one of the three sounds recorded on the tape. Think of it as a patch change facility. You could also set the tape heads so they were straddling two tracks, and this would give you a blend of sound

The first real difference we find between the Mellotron and our cassette deck example is a continuously rotating capstan, which runs the whole length of the keyboard and which is driven by a high torque electric motor. Yes, we’re talking about a two foot feat of precision engineering. Think about how easy it is for a "tiny" dirty capstan to chew your cassette tapes, and then think about how "true" the Mellotron capstan has to be to ensure that it dens’t do the same on 36 tapes, especially for the tapes furthest away from the capstan motor.


Click Here for a diagram
Click Here for a full size view
Click here for the TIFF image

On a cassette deck we have the feed and take up spools, whereas on the Mellotron all the tape ends are fixed within a metal frame. The frame itself is removable, so you can change frames and thus change to another set of sounds. If you’re really good you can do this within a few minutes, and some bands used to perform such changes in the middle of a live set. One common misconception on the Mellotron is that it has tape loops. It dens’t; the tapes are actually single sections, each six foot long. More on this later.

When a tape is at rest, it is threaded through a vertical pulley system, which "folds" the tape about four times to reduce its depth, before being threaded through the tape head/pressure pad and the capstan and pinch roller. The pulley mechanism is sprung.


Click Here for a diagram
Click Here for a full size view
Click here for the TIFF image

When a key on the keyboard is pressed, the key brings its felt pad and pinch roller, and thus its tape, into contact with its tape head and pinch roller, and the tape is drawn past the head to recreate the sound on the tape.

The tape itself once past the head is collected into a collection bin.
With the length of tape and the speed of the mechanism, you have about eight seconds worth of sound before the tape is stalled and the sound dies. So you then have to release the key and allow the time for the sprung pulley system to pull the tape back.


Click Here for a diagram
Click Here for a full size view
Click here for the TIFF image

As you can imagine this means that the Mellotron imposes a certain playing technique: basically avoid those long drone notes, and your chord progressions need to be planned to have totally different notes in successive chords (forget all of your chord inversion tricks for smooth progressions). Tonal clusters are also best avoided (even two handed chords of more than eight notes), because the torque of the capstan motor on some Mellotrons couldn’t handle that many pinch rollers being simultaneously pressed against the capstan, which resulted in a lovely pitch drop as the capstan slowed down!

So there you have the basics of how it works, now lets think about what made the Mellotron so attractive in its day, and even now. The immediate attraction must surely have been that it was a polyphonic keyboard capable of playing back "natural" sounds. Remember, the Mellotron was introduced in the early sixties and not even mono synths were around then let alone poly synths, which didn’t see the light of day until the late seventies (remember the Polymoog?).

Until the Mellotron, keyboard players were stuck with organs and pianos. So imagine the excitement when they found they could have an instrument that could play back "real" sounds, without the hassle of hiring in a real string section. Of course, playing back string sounds on a Mellotron that were recorded by "real" players incurred the wrath of the Musician’s Union (same old story then – it was nothing new in the seventies with the "synthesiser threat" or the eighties with the "sampler threat")!

As you can probably guess, you could have any sound you wanted on those tapes. As well as an extensive library of stock Mellotron tapes sold by the manufacturer, some Mellotrons were used for sound effect "spotting" (think early Doctor Who and the BBC Radiophonic workshop); some artists even had custom tapes made to their own requirements. Paul McCartney had a Mellotron loaded with bagpipe sounds so that his band could play "Mull of Kintyre" live, and Steve Hackett had a tape frame with his own voice recorded on it.

One of the big things about the Mellotron was the fact that it used tape strips and not loops. But why? Surely, if the system had used tape loops you wouldn’t have "run out of sound" whilst playing a key? True, but as we now know, the human ear/brain combination is a strange device: give it an audio bandwidth full of information and it ignores most of it! MP3 compression uses this fact to good effect by stripping out the parts of the spectrum we ignore. Of importance to us (and our tape loop question), is the fact that our brains take the start of a sound and then sort of stop listening to it, as it’s already decided what we’re listening to. So as long as you accurately reproduce the startup transient of a sound, you’re 90% there on the authenticity trail.

The Mellotron gave the recorded sound in its entirety, including the all important start up transient and then eight seconds of the sound’s natural response. For example, a guitar sound would startup and decay as expected. You couldn’t do that with a tape loop. But we still haven’t considered the character of a Mellotron. For all that technology based upon original recordings of real instruments, it had an ethereal quality and rich sound character all of its own, and never really gave a totally faithful reproduction of the source material. This was partly due to the fact that each tape was a separate recording in its own right. Think of a string player in the recording sessions for the tapes: he played a note for eight seconds; stopped; played the next note for the same length of time; and so on. All done with probably no reference to the previous notes. Thus, each recorded note had its own unique undulations in pitch, timbre and volume. Put those different notes together in a chord and you’ll end up with a very rich and naturally varying phasing of pitch, timbre and volume between the notes akin to a natural string section.

Then, the bandwidth of the Mellotron tape system was only around 8KHz, so that made for a more mellow sound than typically produced by the source instruments. Finally, the playback mechanism itself imparted something on the sound. Think of the changes in torque on the capstan as notes are pressed and released affecting the pitch of the notes being played back, along with the general "wow and flutter" characteristics of the entire mechanism, of which the capstan is the only common component. Just to add to the fun, each key had individually adjustments for the pressure (felt) pad and the pinch roller, which probably had at least a small effect on the playback characteristics for each note.

As I mentioned earlier, I deliberately picked the Model 400 for the description. Out of interest some earlier Mellotrons had a slightly different tape mechanism. Instead of the ends of the tape being fixed in a frame, they were wound onto feed and take up spools. The tapes were actually made up of six splices, with each splice having its own set of three sounds, giving eighteen sounds per tape. A "bank" was selected by the mechanism spooling the tape to the start of a particular splice. Apparently this system wasn’t reliable enough and was renowned for mangling tapes during change over, and was dropped on later systems in the quest for more reliability and to save weight and cost.

A Comparison between the Mellotron and Digital Sampling.
Hopefully, by the end of this "How it works" section, you’ll now be amazed at what the Mellotron was capable of in its day, and is still capable of now. To emphasise this in a totally different manner, let’s compare the storage capacity available in a Mellotron to that of both general computers and digital samplers, as this provides a very interesting comparison of the growth of digital storage capacity over the years compared to the Mellotron’s tape based storage capacity.

So to start off, consider the Mellotron and the amount of sound it stored. For this comparison, let’s consider the Model 400 Mellotron as this was the most popular model.

The basic tape system of the Mellotron allowed for eight seconds of sound, over 35 keys that’s 280 seconds worth of sound per track. If you consider the three sounds stored on each tape, that’s actually 840 seconds worth of sound stored within a Mellotron.

For a MK2 or FX series Mellotron, with six sound banks and a dual manual keyboard, the total storage capacity was over 10,000 seconds of sound. Now to consider the digital equivalent of a Mellotron’s storage capacity, take the Mellotron’s 8KHz bandwidth and double it prior to sampling (remember, you need to sample at double your signal’s bandwidth for accurate reproduction). Also, remember the Mellotron was effectively a multi-sample playback device, with a sample for each key, so we won’t try and save space by playing the same sample over several octaves. Now let’s assume we are using eight bit sampling. Double the capacity figures given below for the sixteen bit sampling of Mellotron sounds. The amount of memory needed to store a single sound of eight seconds on a single key at 16KHz is 128,000 bytes of memory.

For the three sounds, on each tape strip, that’s equivalent to 384,000 bytes of memory.

For 35 tape strips, that’s equivalent to 13.4M bytes of memory needed to storage the digital equivalent of all the sound stored within a Mellotron. For a MKII or FX series Mellotron, the equivalent is 160M bytes of memory!

Now let’s contrast that to the storage capacity found within various computers through the ages. As there were no home computers in the sixties, I’ve picked a high end mainframe that was about when the Mellotron was first released.

Time Line System Storage Capacity Bytes
1962 Mellotron 13,440,000
1962 Ferranti Atlas Computer 49,192
1980 Sinclair ZX80 1,024
1984 Sinclair Spectrum 16K 16,384
1987 Commodore Amiga 512,288
1990 386 Based PC 4,194,304
1995 First Pentium Based PCs 16,777,216
2000 Modern PC 67,108,664

Thus, you can see, the Mellotron had a thirty five year lead over the typical storage capacity that we have in a standard desktop PC.
Compare that now to commercially available digital samplers. For the purpose of this exercise, let’s consider the stock memory capacity of the sampler, with no additional RAM installed.

Time Line System Storage Capacity Bytes
1962 Mellotron 13,440,000
1976 Prototype Fairlight 32,768
1981 Emulator I 131,072
1984 Emulator II 524,288
1988 Emulator III 4,194,304
1994 Emulator IV 8,388,608
1998 Emulator ESI 4000 4,194,304

As you can see, even now the base memory capacity of samplers, has yet to exceed the "base" storage capacity of the Mellotron.

If we now move back to the EX series synthesisers, the stock memory is 1,048,576 bytes of memory (or 524,288 words as its sampling is 16 bits). If you take a fully loaded EX with 1MB base memory, 64MB DRAM and 16MB FLASH memory, that’s 97MB of memory or 48.5MB worth of 16 bit sound samples. But remember if we are sampling at the higher rate on the EX of 44KHz, which gives us around 1,000 seconds worth of sound. Thus, for a fully loaded EX, sampling at 16 bit resolution and at 44KHz, we can get the equivalent of one Mellotron into its sample memory (remember, the total sound storage of a Mellotron is 840 seconds). When thinking about that, just remember that the Mellotron predates the EX by about 40 years (or 55 years if you also count its predecessor, the Chamberlin).

So to round this technical tour off...
some people see the Mellotron as a mechanical behemoth, that long outlived its day. I hope I’ve persuaded you to see things differently, and that you actually now see the Mellotron from a different perspective. Namely that it was ground breaking long before synthesisers were even commercially successful, and that to have a sample playback device at least thirty years before digital sampling could affordably match the Mellotron’s sound storage capacity was simply awesome!