The two parts of a compound printing plate. The projections on the right-hand plate correspond to the holes in that on the left. Each plate made a different coloured impression. © Science Museum/Science & Society Picture Library
The process of making the compound plates so that they were a
perfect fit was central to the success of Congreve’s process. His
1820 pamphlet gave little away: ‘A second plate is then combined
with the first by a process which it is not necessary here to
detail…’ . His
subsequent patent also mainly concerns the printing process itself,
although brief mention is now made of the plates’
…either by casting one metal plate into another, the metal of the lower plate being in that case softer and easier fused than that of the upper plate, or they may be fitted of two metal plates of equal hardness, either by hand or by casting each part in separate moulds duly adjusted.
Recent analysis by Dr Maureen Greenland suggests that pairs of plates were cast, poured from the same melted metal, implying that the second process was used.  Not describing the process prior to having patent protection is understandable. That it subsequently appeared to be relatively straightforward seems curious at first sight, given that the exact registration of the combined plates was critical, and that Congreve had already emphasized that preventing forgery relied on banknote design, ‘…which can only be imitated by the same process as that by which the original note is created’ .
If the original note relied on a special-purpose machine, it could be considered more secure than if it used a hand process that was more widely available. That said, finishing the plates by hand so that they were an extremely close fit still required a very high degree of skill, so Congreve was evidently happy to rely on this skill, held by a relatively small number of people.
Making the pairs of plates fit together ultimately became the job of a large pantograph engraving machine, recently acquired by the Science Museum. The machine was originally used by Donkin for planing the strainer plates used in paper-making machinery . Once converted to compound-plate manufacturing work, it cut the apertures in the upper plate, and projections in the lower plate, as well as the lettering and design on the combined plate, from a large master template.
The pantograph machine is unusual in two respects. First, rather than moving a feeler across a static template, it is arranged so that the feeler is stationary and the template itself moves, attached to the pantograph, and that is attached to the reproduction being made, which is again moving around beneath a fixed-position milling tool mounted on the bridge above.
A side view of the geometric lathe built by Bryan Donkin. The tool holder is on the right, the headstock and mandrel that held the plate being engraved is in the centre, and the mechanism to produce the geometric patterns is on the left. © Science Museum/Science & Society Picture Library
This is exactly the opposite practice to that used in most other copying machines.
Secondly, the machine appears to date from some time after 1820, and may have come into use making compound printing plates as late as 1868 . The pantograph was a well known means of making exact copies of two- or three-dimensional shapes. One source states that Donkin visited James Watt in January 1814  – Watt’s garret workshop at his Birmingham home contained two sculpture-copying machines, embodying the same principle. The most likely explanation for the later date of this machine is that it was only brought into use as demand for compound printing plates rose beyond that which could be satisfied by earlier methods.