Spin-Echo Storage. Spin-echo storage is a form of storage based on 

nuclear spin phenomena. Although there are some serious disadvantages 

and problems in adapting spin-echo storage to a digital computer in a 

practical way, it has been possible to come surprisingly close to a 

useful system through clever design. Spin-echo storage is particularly 

interest in that, so far as is known, it is the only computer component 

which has been seriously considered that utilizes nuclear phenomena of 

any kind. 



An appreciable knowledge of theoretical nuclear physics is 

required for a thorough understanding of the spin-echo mechanism. 

However, the basic points can be comprehended fairly well from the laws 

or ordinary classical physics if a few "facts" can be accepted without 

question. One of these facts is that all fundamental particles such as 

protons seem to be spinning at a constant velocity. The spinning 

positive charge of the proton acts like a small electric current passing 

around a small loop of wire and produces a magnetic field that is called 

the magnetic moment of the proton. If the proton is placed in an 

external magnetic field, the axis of spin or the direction of the 

proton's magnetic moment will precess around the direction of the 

external field. The situation is vaguely analogous to that of a spinning 

mechanical top. When the top is not in the vertical position its axis 

precesses about the vertical. In the case of nuclear particles the 

frequency of precession is called the "Larmor precession frequency" and 

is dependent upon the gyromagnetic ratio of the particle and the 

strength of the applied field, but is independent of the angle between 

the spin moment and the direction of the external field. For protons in 

a magnetic field of 7000 gauss, for example, the Larmor precession 

frequency is about 30 megacycles per second. 



There is a force which causes the direction of spin to tend to 

line up with the direction of the externally applied magnetic field. 

However, a proton will remain in its initial orientation for a period of 

time before it suddenly falls into alignment with the field. The time 

required for most of a large number of protons to align with the field 

is called the "longitudinal relaxation time" and is in the order of 20 

milliseconds for the protons in the hydrogen atoms in the molecules of 

liquid water. The storage action takes place in a time which is short 

compared with this relaxation time. It may be assumed for purposes of 

explanation that the proton spins are initially in alignment with the 

field, but when they are reoriented they will remain in their new state 

for a substantial period of time.



The actual storage medium is the orientation of the spin magnetic 

moments of a large number of protons, and ordinary water is a 

satisfactory substance for supplying and holding the protons. When the 

storage unit is in operation, the spin orientation of any single proton 

is never known, because the strength of the magnetic field is not 

measured with an accuracy sufficient to determine the spin orientation 

after a few cycles of precession. In fact, inhomogeneities in the field 

cause the precession frequency to be slightly different for protons at 

different positions. Because of these differences, the actual 

orientations assume the appearance of a random distribution very quickly 

after a pulse is stored. It is not only possible to achieve binary digit 

storage with this kind of a distribution, but the apparently random 

distribution is necessary for the storage mechanism to work. The steps 

involved in the process of storing a series of information pulses and 

subsequently recovering them are outlined in the following paragraphs. 

The apparatus needed for spin-echo storage is indicated in Fig. 6-16.



A container for the liquid is mounted between the poles of a large 

magnet, and a coil surrounds the liquid with the axis of the coil at 

right angles to the direction of the field produced by the magnet. The z 

direction is indicated as parallel to the magnetic field and the x 

direction is parallel to the coil axis. The y direction is perpendicular 

to the plane of the paper. The magnet is so designed that the strength 

of the magnetic field is nearly uniform through the sample, but the 

storage mechanism depends on its not being exactly uniform. It may be 

assumed that all of the proton spins are initially oriented parallel to 

the z axis. If a current is passed through the coil, the direction of 

the resultant magnetic flux will be shifted away from the z axis by a 

small amount, although it will remain in the xz plane. The magnetic 

moment of the protons considered collectively will then precess about 

this resultant direction. If the current is made to alternate at the 

Larmor precession frequency, it can be shown that the direction of this 

magnetic moment (spin vector) will execute a spiral, the angle between 

it and the z axis increasing linearly with the product of the current 

strength and the time.



After the alternating current is removed, the spin vector will 

precess about the z axis with a constant angle between the spin vector 

and the z axis. Since the magnitude of the external field is not 

exactly the same at all points in the liquid, the precession frequency 

will vary somewhat from one proton to the next. When the spin vector of 

each proton is considered individually, it will be observed that each 

vector has a component which is constant in the z direction and another 

component which is rotating in the xy plane. The individual components 

rotating in the xy plane will soon get out of phase with one another 

because of the differing frequencies; in fact, they will become randomly 

distributed in the xy plane. A burst of several cycles of high-

frequency current can be used in this manner to store one bit of digital 

information. The problem is now to recover the bit of information.



Recovery is achieved by applying a "reading pulse" to the coil. 

The reading pulse is a burst of high-frequency current the same as 

before except that the amplitude and duration are chosen such that each 

spin vector is "spiraled around" to a position which is rotated about 

the axis through an angle of 180° from the position the vector would 

other wise have. The effect of the motion can be visualized more easily 

by assuming that the pulse in the coil is only a small fraction of a 

cycle in duration but of such extremely large magnitude that the field 

from the magnet is relatively negligible and that the resultant field 

direction is in line with the x axis. The spin vectors will then 

precess around the x axis, and the product of current and time is 

chosen to provide for 180° of precession. This rotation about the x 

axis inverts the phase relationships among the xy-plane components of 

the spin vectors. At the termination of the reading pulse the spin 

vectors again precess about the z axis in the same direction as before.



If the reading pulse is applied at time t after the information 

pulse, then at time 2t all of the spin vectors will again be in phase. 

This condition can be sensed by the voltage induced in the coil. During 

the time that the vectors are out of phase no voltage is induced, but 

when they become in phase an alternating voltage with a frequency equal 

to the average Larmor precession frequency is generated in the coil by a 

magnetic coupling action. This induced signal is called the "echo." The 

process has been likened to a group of runners racing on a circular 

track. It is assumed that the runners are started at the same point and 

that they all run with different but constant speeds. After several laps 

they will be randomly distributed around the track. If they all suddenly 

turn around and race in the other direction, each with the same speed as 

before, they will all return to the starting point at the same time. The 

time required for them to return to this in-phase condition is equal to 

the time which elapsed between the start of the "race" and the turning 

around. The analogy would be slightly more accurate (but less realistic) 

if each runner, instead of turning around, were suddenly shifted from 

his current position to the corresponding position on the opposite half 

of the track, where the halves are as determined by the diameter which 

passes through the starting point. 



The storage of the second and all successive information pulses is 

accomplished by the same process used for the first one except that the 

motion of the spin vectors becomes complex and is not easily described. 

Although it is somewhat of an oversimplification, the second information 

pulse may be visualized as operating on those components of the spin 

vectors which remain parallel to the z axis after the application of 

the first information pulse. Note that the information pulses (or 

absence of pulses for stored 0's) will emerge as a mirror image of the 

input signals, that is, in the opposite sequence from which they were 

entered.



An alternative storage procedure provides for the return of the 

information pulses in the same sequence in which they were entered. A 

"start pulse" is applied to the coil to rotate the spin vectors 90° away 

from their initial position, which is parallel to the z axis. All of 

the vectors are then parallel to the xy plane and are precessing about 

the z axis. Before applying the information pulses to be stored, enough 

time is allowed to elapse for the vectors to become randomly distributed 

in the 2y plane as a result of the slight inhomogeneity of the external 

field produced by the magnet. When the information pulses are applied 

the effect is to tilt the array of spin vectors so that some of them 

have components in the positive z direction and others in the negative 

z direction.



To recover the information a "recollection pulse" is applied which 

causes these z components to be rotated into the 2y plane. These 

components again precess about the z axis with various Larmor 

frequencies. Because the tilting effect gave z components to individual 

spin vectors that had certain frequency relationships to each other, it 

turns out that their different precession frequencies will cause them to 

become in phase at time t = ti + T, where ti is the time after the start 

pulse that the given information pulse is applied and T is the time 

between the start pulse and the recollection pulse. In a sense, an 

information pulse selects spin vectors that have a certain amount of 

phase separation from each other and remembers them by giving them 

components in the z axis. The amount of phase separation which occurred 

in time ti (and which may be many revolutions of separation) can then be 

returned to zero in an equal length of time when these z components are 

returned to the xy plane. The recovery of the information in the 

original sequence as provided by this scheme is generally more adaptable 

to computer use than the mirror-image sequence. The storage of 1000 

pulses in normal sequence has been reported when using 2 cubic 

centimeters of a glycerin-water mixture for the storage medium. The 

delay time was about 1 millisecond. A serious and probably crucial 

disadvantage of spin-echo storage is in the fact that information cannot 

be re-entered while it is being recovered, as is possible with ordinary 

delay line storage. In fact, after the information has been stored and 

recovered it is necessary to wait for the spins to again become aligned 

with the external magnetic field before new information can be stored. 

This amount of time is roughly the longitudinal relaxation time which 

was mentioned near the beginning of this section. The amount of 

information which can be stored and also the time duration for which it 

may restored is limited by the "transverse relaxation time," that is, 

the time during which the spin vectors will stay in the proper phase 

relationships with respect to each other as they precess. These 

relaxation times are functions of the physical properties of the medium; 

solids, incidentally, are not suitable for storage because their 

transverse relations times are extremely short. Many other disadvantages 

and problems related to spin-echo storage could be mentioned, and it is 

therefore not a lucrative approach to a digital computer storage system 

in spite of its attractiveness from an academic viewpoint.