 In the first film of this series, we examined the basic properties of magnetic cores. In this film, we'll be dealing with three types of core transfer loops. The single diode transfer loop, a split winding transfer loop, and the inhibit transfer loop. The simplest of the three transfer loops has already been described. The single diode transfer loop. In the single diode transfer loop, we assumed that both cores were initially in the binary zero state. A non-dot input pulse switched core A to the binary one state without affecting core B. Later, a dot pulse applied to the shift winding cleared core A back to zero. The flux change in core A produced an output there. The pulse traveled along to the input winding of core B, entering it on the non-dot side and switching core B to binary one. Thus, at time t zero, we stored a binary one in core A. And at time t one, we transferred that binary one to core B. The selective isolation of core A from core B was achieved by placing a unidirectional device, the diode that gives this loop its name, in the circuit. The diode allowed current to flow in the circuit in one direction only, into the non-dot side of core B's input. But we can work many variations. For example, we can turn the diode around. Now the situation is reversed. Now the diode will allow current to flow around the circuit in the opposite direction only, into the dot side of the input winding at core B. To demonstrate the loop as we have it here, we'll assume that both cores are initially at binary one. A pulse is applied to the dot side of the input winding of core A, switching it to the binary zero state. The change in flux in core A induces an output on the dot side of the output winding. But the diode stops the flow of current in this direction. Core A now holds a zero. Core B remains as it was, at one. However, if a non-dot pulse is applied to the shift winding of core A, it switches the binary state there back to one. The change in flux induces an output in the direction the diode allows. This pulse enters the input winding of core B on the dot side, and core B is switched to binary zero. In effect, we have transferred a zero physically from one core to the next. We have used the single diode transfer loop two ways. First, we were storing a binary one in core A and transferring it to core B. Second, by reversing the diode and changing the input to and the transfer from core A prime, we used the loop for storing and transferring a binary zero. And in both cases, the transfer was physical. That is to say, it was done by producing an output pulse that switched the next core in line. And in both cases, we were able to control the time of the transfer. The input pulse at time t zero placed the bit in storage, and it was held there until the shift pulse at time t one transferred it. The single diode loops have many applications in automatic data processing systems, but there is a limit to the amount of control they provide. Suppose, for example, there were several inputs coming into the first core of a loop. In this way, there are three inputs here. The one at the left will set the core to binary one. The top and bottom inputs will clear it back to zero. And as the diagram indicates, whenever the core is cleared to zero, an output occurs. Switching core B to one, but this means that either one of the dot inputs can bring about a transfer of the bit from A to B. The bottom one or the top, tighter control is frequently necessary. We may want the shift to occur only when the bottom pulse is applied. That is, we want the transfer of information to be conditional. As the diagram shows now, the other two inputs can be used to switch core A back and forth between the binary one and the binary zero states. But an output will occur only under the condition that a binary one is stored in core A and a shift input is applied at the bottom. Let's put the time symbols in the sequence we want. We should have the dot pulse at the top come in at time t one. If this pulse is applied, it kills the binary one that had previously been placed in the core. That is to say, it resets the core to zero without providing an output pulse. The shift pulse comes in at time t two until the shift pulse occurs without the occurrence of a kill. There will be no transfer of information to core B. The circuit we have here is the second of the three we are considering, the split winding transfer loop. This logical diagram explains what it does. But we'll need a schematic diagram of a split winding transfer loop in order to see how its work is done. Start with core A. We'll add two input windings first. The input on the left is connected so that the current flows into it on the non-dot side only. This is the one that our logical diagram showed as coming in at time t zero and setting core A to binary one. At the top, current comes in on the dot side, setting core A to binary zero and the time is t one. There is also an output winding at core A. And like those in the single diode loops, this output has more turns in it than the input windings have. Two diodes are used in the split winding loop. As you see, they're placed so they oppose the flow of current coming from either side of the output winding. Finally, we'll put an outlet into the circuit between core A and the lower diode. We'll call it point Y, and it will serve to conduct current out of the circuit. The reasons for all this will be clear after we set up core B. First, we'll wrap two small windings around core B, an upper winding and a lower winding. The polarity is the same in both cases as the dots indicate. The two small windings are joined together at point X, and this gives us the split winding from which this loop gets its name. All that remains is to connect the split winding into the circuit and provide it with an outside source of current. Before we go into the complete operation of the loop, let's see how the split winding itself works when a pulse is applied. The current travels up to point X, where it divides into two branch currents. The upper branch comes into the dot side of the upper half of the split winding, producing a negative flux in the core. At the same time, the lower branch of the current comes into the non-dot side of the lower half of the split winding, and the result is a positive flux. If the two branch currents are equal, then the two magnetic fluxes will be equal and cancel each other out, leaving the core in whichever state it was initially. But if the branch currents are not equal, then one or the other of the magnetic fluxes will prevail. One more point. The pulse we've been applying is the shift pulse for the loop, and its time designation is T2. To demonstrate the operation of the split winding transfer loop, we'll clear both cores to binary zero. Now we'll apply the three input pulses in sequence. At time T0, a pulse comes in the non-dot side of the left winding, and core A is switched to binary one. Ordinarily, this switch would produce a pulse coming out of the non-dot side of the output winding. But the lower diode prevents a flow of current in that direction. Core A now stores a binary one. Core B has not been affected. At time T1, a pulse is applied to the dot side of the other input winding of core A, switching the core back to zero. Again, the result ordinarily would be an output pulse on the dot side this time. But the upper diode stops it, and nothing happens at core B. With both cores at zero, we go on to the next step. The shift pulse at time T2. As we saw before, this current goes to point X, where it divides. And the branch currents enter the two halves of the split winding. The upper branch current comes out on the non-dot side of its half of the split winding. The lower branch comes out on the dot side of its half of the winding. And both of them move along to the diodes. The diodes conduct in this direction, and the two branch currents flow toward core A. But the lower branch does not get there. Instead, it flows out of the circuit at point Y. The upper branch, however, does reach core A, entering its output winding as a dot pulse. Since core A is in the binary zero state, that is, its magnetic flux is already negative, an additional pulse of dot current will have no practical effect on the core. There is a counter EMF induced in the winding when the pulse is applied. But because the change in flux is so slight, the counter EMF is negligible. As a matter of practice, however, a small coil is inserted into the lower connection to provide a counter EMF there that will balance the counter EMF in the output winding. Therefore, the upper branch pulse meets no more impedance than the lower branch does. The impedances are equal, and so the branches are equal. At point Y, this upper branch current joins the lower branch and flows out of the loop. The result is no change in either core. Core A remains at zero because the upper branch current is applied to its winding as a dot pulse. The upper and lower branch currents are equal, and therefore core B also remains at zero. Because equal currents there produce opposing but equal magnetic fluxes, which cancel out. So far, we have not transferred information from A to B. We need to make a change in the sequence of inputs. Again, at core A, we'll apply a non-dot pulse, T zero, switching the core to binary one. But this time around, we'll skip the dot pulse at T one and leave core A at one. Then the shift pulse is applied. To understand what follows, we must look at the output winding of core A. With the core holding a binary one, the residual magnetism is in the positive direction. The upper branch current comes in as a dot impulse, a negative magnetizing force, and switches the core. But as one result of the complete change of flux, a large counter EMF is induced in the winding. The impedance, now presented by the output winding, is high. In fact, since this winding has more turns in it than any of the others, the impedance here is the highest in the loop. Its effect on the upper branch current is to make it smaller throughout its course. Let's go back to the moment when the shift pulse was applied. The current divides at point X, but in this situation, it divides unequally. The upper branch current is relatively small. The lower branch is large. The upper branch current working alone is able to switch core A from binary one to binary zero. At core B, however, the upper branch is working against the larger lower branch. And magnetic fluxes that result are also unequal. The weaker negative flux is overcome by the greater positive flux, which switches core B to binary one. To summarize the operation of a split winding loop, we can switch core A back and forth without affecting core B. But when core A stores a binary one and the shift pulse is applied, the one is transferred to core B. And all of this is shown in simple form in the logical diagram. Incidentally, to illustrate the split winding, the shift arrow is placed here at core A in the logical diagram and not at core B. Because this shows its function to clear core A back to zero and at the same time put a one in B. The third transfer loop, the inhibit transfer loop, is really a variation of the split winding transfer loop. It offers the same kind of isolation between the sending and receiving cores plus an additional means of control over the transfer. We'll need only two inputs to core A, the input pulse that switches core A to binary one and the shift pulse. But we'll be using three cores. The third core, C, has its own non-dot input and it shares the shift pulse with core A. If core C is at one, the shift pulse will clear it back to zero and the change in flux will of course produce an output. But this output from C serves a special purpose. Whenever it occurs, it will prevent or inhibit the flow of current from core A to core B. And this action gives the loop its name, the inhibit transfer loop. In order to transfer a binary one from core A to core B, three conditions must be met. Core A must be at one, core C must be at zero, and the shift pulse must be applied. In other words, assuming as usual that the three cores are initially at zero, if there is a non-dot input at A and no such input at C, then the shift pulse will transfer a one from A to B. But if the input does come in at C, there will be no transfer. Schematically, this loop resembles the one we've just been studying. There's the same split winding arrangement to which the shift pulse is applied, the same diodes, placed so they oppose the flow of current out of either side of the output winding, and the same outlet connecting into the loop at the non-dot side of the output winding. The differences are all on the left. We won't be using the same kind of output winding. Core A has to be moved up to make room for the third core. These two, A and C, have identical input windings and identical output windings. Now the output windings are unique. Their dot sides are on the right, facing core B in both cases. And their non-dot sides are joined together. We extend the Y outlet to meet the common non-dot side, connect the dot sides into the circuit, and our schematic diagram of an inhibit transfer loop is complete. Notice that each core is isolated. Core B is not directly affected by what happens at A or C because the diodes oppose the flow of current from A and C. The diodes also isolate A and C from each other. A change in the flux of A will not produce an output to C because the lower diode will not conduct in that direction, nor can there be an output from core C up to core A because the upper diode will not conduct. Only one kind of transfer can be made in this loop. It will occur when a shift pulse is applied to the split winding and divides unequally into a small upper branch and a large lower branch. This is very much like the split winding loop itself, but the inhibit loop has a different way of making the branches unequal, as we shall now demonstrate. All three cores are cleared to zero. A non-dot input is applied to core A, switching it to binary 1. And for our first operation of the loop, core C is also switched to 1. Now the shift pulse is applied. And the question is whether the branch currents into which it divides at point X will be equal or unequal. The answer depends on the impedances they encounter. Look at cores A and C. Both are at 1. In both, the magnetic flux is positive. Both output windings receive branch current on the same side, the dot side. And the windings themselves are identical. Consequently, the impedances are equal. If the impedances at core A and C are equal, the shift pulse will divide into equal branch currents. At core B, the equal branch currents set up equal but opposing fluxes, which cancel each other out, and leave the core at binary zero. At core A and at core C, they come in as dot pulses and switch the cores to zero. The two branches join at point Y and flow out of the circuit. Thus, the shift pulse cleared cores A and C, but it was unable to transfer the one from A to B. To achieve a transfer, we'll apply the input to core A, but not the input to core C. This switches A to 1 with its flux in the positive direction, leaving C at zero and its flux negative. A dot pulse entering the output winding of A will meet a high impedance, but the impedance at C will be low. Under the conditions we have here, the binary one in core A can be transferred to core B. When the shift pulse is applied, it divides unequally. The upper branch, which comes up against the high impedance at core A, is small, the lower branch, which has only the low impedance at core C to overcome, is large. The result at core B is the same as it was with the split winding loop, a weak negative flux and a strong positive flux, and therefore B is switched to the binary one state. The upper branch current clears core A back to binary zero. The lower branch has no practical effect on core C because it is already at zero. And again, the branches rejoin at Y and flow out of the circuit. In this operation, by keeping C at zero, we were able to transfer A1 from A to B. Would the reverse be true? Suppose we kept core A at zero and put a one in core C. Would we then be able to transfer the one from C to B? No, because the impedances would be reversed. The high impedance at C would result in a small lower branch current, and the upper branch would be large. The upper branch enters the split winding at B on the dot side. Therefore, B would stay at zero. The function of C in this loop is to inhibit the transfer of information from A to B when the inhibiting action is desired. The three transfer loops we've been examining, the single diode, the split winding, and the inhibit, have an amazing number of applications in automatic data processing equipment. One device that is widely used is just a simple extension of the single diode loop. Four or more cores are strung together in a series of single diode loops. The shift pulses are alternated. That is, one pulse serves cores A and C, and the second pulse serves B and D. This is a magnetic shift register, which provides a temporary storage for binary information. Here's how it works. At time t zero, an input pulse applied to core A places a binary one there. The first shift pulse at time t one moves the bit to core B, and the second shift pulse at time t two advances it to core C. We now have one bit stored in the register, and we can store another one by applying an input pulse to core A again. The capacity of this particular four core register is two bits, but there are registers that can hold as many as 50 bits or more at a time. Essentially, a register accepts information when it is available and delivers it when it is needed. Two rounds of shift pulses will send the information into the output from core D. There goes one bit and there goes the other. A magnetic shift register can act as a speed buffer. For example, it can receive information from a relatively slow source, like a teletype writer operating at standard transmission speed, and then deliver it to a device that operates at high speed, like a magnetic drum. We've been looking at a serial register. It handles information serially, one bit at a time. But registers that work in parallel are put together by combining split winding loops with single diode loops. In fact, almost all the functions of control, logic, and arithmetic are achieved by using these three loops, singly or in combination. They have many advantages. They are small and lightweight, so that a lot of them can be put onto a circuit panel. And they are rugged. The single diode, split winding, and inhibit transfer loops are, in a very real sense, the building blocks of an automatic data processing system.