Answer 1—Delay-line memories in general are a way of storing information by converting an electrical signal into some other physical phenomenon, allowing it to propagate to another location where it’s converted back to an electrical signal. If the output signal is then fed back into the input, the signal can be preserved indefinitely. This works best if the signal is digital, in which the feedback mechanism can regenerate a clean signal on each pass, eliminating the effects of noise and timing jitter.

Many early computers used acoustic delay lines of various types, where the electrical signal was converted into mechanical vibrations in, say, a column of liquid mercury or a steel wire. Delays of a few milliseconds would allow a few thousand bits of digital information to be stored.

There are several important consequences of storing information this way, including the fact that the information is only available in a sequential fashion, and accessing a particular item incurs a latency on the average of half the total delay. This makes delay-line memories particularly suitable for certain applications such as video display refresh.

Another example of delay-line memory is the dynamic MOS shift register, an early form of IC memory. This type of memory is slightly different in that the data gets regenerated at every step of the shift register, but clocking must be maintained continuously because the data is stored as charges on tiny capacitors. Charge-coupled devices (CCDs) are similar, but they store continuously variable amounts of charge instead of binary values, and are subject to the accumulation of errors.

In a sense, magnetic media such as drums and disks are a kind of delay-line memory. Although the signal, once recorded, doesn’t need to be actively refreshed, physical motion is required to do the recording and to read the data back again, incurring the same issues with sequential access and latency. Magnetic bubble memory falls into this category as well.

Contributor: David Tweed