I found a very good detailed description of high speed core memory in the CDC 6600 training manual ? copy available on bitsavers at http://bitsavers.trailing-edge.com/pdf/cdc/cyber/cyber_70/fieldEngr/6014740… . Chapter 4 describes the memory. It includes schematics, and a pretty detailed description of how the circuits work. Those memories are a bit more complex than usual; there are two inhibit wires rather than one, and the inhibit does not cover the whole core plane but only a part of it (i.e., there are several independent inhibit current paths). It doesn?t say why. My best guess is that this is to keep the inductance of the inhibit wire similar to that of the X and Y wires, so the same driver design works with roughly the same timing. This memory is much faster than most other memories of the same era: 1 microsecond full cycle, in 1964. paul

On 2015-Mar-11, at 10:45 AM, Rick Bensene wrote: Speaking very generally, pulse transformers were often used in the discrete days as a cheaper means of providing some circuit functions compared to doing so with multiple transistors. -- Regarding the AL-1000 specifically: schematic ref: http://www.cs.ubc.ca/~hilpert/eec/calctd/CasioAL1000.pdf, page 21-22. The transformers in the sense amps are formed as a single primary & center-tapped secondary. The floating primary is essentially a cheap way of providing a differential input. (Be it transformers or ICs, the differential input is used with the sense loop - in contrast to grounding one side of the sense loop - to avoid detecting (to reject) common-mode noise and induced signals produced by all the other electrical activity on the address lines.) The center-tapped secondary is then used in essence for the "absolute-value" function: one or the other side of the secondary will produce an equivalent pulse for a pulse of either + or ? polarity from the sense loop / primary. Each side of the secondary has a one-transistor amplifier, then combined in the AND gate - actually functioning as a negative-logic OR - to produce the output indication of a sense-pulse having been seen. The "absolute-value" function is necessary because the typical threading of core memories has the sense line going through half the cores in one direction and half the cores in the other direction relative to the address wire currents. This results in a read-pulse on the address wires producing a relative-positive pulse for half the cores and a relative-negative pulse for the other half. That is, if A & B are the two ends of the sense loop, then the sense amplifier must produce an output indication if either: V(A-B) > Vthreshold or V(A-B) < ?Vthreshold == V(B-A) > Vthreshold Similarly, the address-line drivers use a center-tapped primary configuration as a cheap way of providing the bi-directional drive current for the address wires (like push-pull audio). Note in contrast the inhibit drivers - which only require uni-directional current - do not use transformers. -- In other core (and other digital) designs, pulse transformers were used to simplify the binary decoding functions in address decoding, by putting n*n transformers into a matrix, driving them by 2n drivers along the matrix axes, and taking pulse outputs off each secondary. This is much better described graphically than in prose, but for example, in a core memory system I have, 8 of the 15 address lines need to be decoded to 1-of-256. It does so with two sets of 16 pulse transformers formed into two 4*4 matrices. Each of the 4 axis of 4 lines is driven by simple 2-bit to 1-of-4 decoders. The 16 secondaries from each little matrix then drive each end of one core plane axis of address wires, thus forming them into a 16*16 matrix. Each address wire is thus selected as 1-of-256. That accounts for just one axis of the core plane, a similar scheme is used for the other 7 address bits for the other axis of 128 wires, to achieve the net 1-of-32,768 memory word decoding. Pulse transformers could also be used to isolate and float a signal, so it's voltage level relative to ground could be shifted to something more amenable to another circuit element. For example in nixie anode drivers, the digit pulse of say 0-to-5V relative to ground could be shifted up to 180-to-185V to trigger the anode driver. Keep in mind that in audio and telephone, transformers were used for some similar objectives: producing the differential drive and input for balanced lines for common-mode rejection. Pulse transformers are still used in twisted-pair ethernet line xcvrs, for differential input with high common-mode voltage rejection.