Percy Ludgate

Percy Ludgate

Accountant and designer of a proposed analytical machine · 1883–1922 · Ireland

Ludgate's 1909 paper offers a valuable exercise in reading a design on its own terms. Its arithmetic representation, storage mechanism, and control arrangements differ from Babbage's. The machine was proposed rather than built, so its specified capacity and speed estimates must not be presented as measurements of working hardware.

What the historical record establishes

Percy Ludgate was born in Skibbereen on 2 August 1883, not 28 August, and died in Dublin on 16 October 1922. Recent archival research identifies schooling, civil-service employment, and later accountancy training. His father taught shorthand. The older picture of an entirely uneducated inventor without access to colleagues is misleading. Royal Irish Academy biography

Ludgate described his initial design as independent of Babbage, while acknowledging that subsequent study of Babbage helped him. He also thanked C. V. Boys. Independence at the beginning does not mean absence of later intellectual assistance. His paper was published in 1909 in the Scientific Proceedings of the Royal Dublin Society. Primary paper

The record is not a story of complete disappearance until 1971. Boys reviewed the work in 1909; the paper was used in a patent dispute in 1960; Brian Randell's 1971 study brought renewed historical attention. The name “Irish logarithms” also has a contemporary origin in Boys's response, rather than being a recent nickname. Updated historical account, 1909 review record

Read the mechanism before applying modern labels

Ludgate's store used shuttles carrying sliding metal rods whose positions represented digits. The shuttles were arranged in cylindrical boxes; each digit was not a concentric cylinder. He proposed 192 variables of 20 decimal digits, with signs. Formula paper combined operation control and operand selection. His index scheme converted single-digit multiplication into addition of specially chosen indices and a product lookup. These indices were explicitly not ordinary logarithms. The proposed dimensions were 26 by 24 by 20 inches. Design description and tables

Terms such as storage, arithmetic unit, and instruction selection are useful functional descriptions. They do not prove that this machine had a modern stored-program architecture, binary arithmetic, or transistor-like mechanisms. A control medium can direct a sequence without being writable numerical memory.

Worked mechanism: multiplication by an unusual encoding

The following calculation uses Ludgate's digit indices but is an original teaching example. It models the arithmetic mapping, not every physical movement of a proposed mechanism.

Decimal digit d	0	1	2	3	4	5	6	7	8	9
Index I(d)	50	0	1	7	2	23	8	33	3	14

To multiply digits a and b, add I(a)+I(b) and use a table mapping the sum to a×b. For example, I(6)+I(7)=8+33=41, and table entry 41 gives 42. For 4 and 9, the sum is 2+14=16, which maps to 36. Zero has special index values; I(0)+I(7)=83 must produce zero. This mathematical lookup includes such cases without claiming how every exceptional index was handled physically.

Consider 37×24. Expand by decimal place:

(30+7)(20+4)
= (3×2)×100 + (3×4)×10 + (7×2)×10 + 7×4
= 6×100      + 12×10     + 14×10     + 28
= 600 + 120 + 140 + 28
= 888

The four index sums are 8, 9, 34, and 35. Their corresponding products are 6, 12, 14, and 28. The encoding handles digit multiplication; positional weighting, accumulation, and carrying are additional responsibilities. Treating a lookup as a complete multiplier hides those responsibilities.

The encoding must satisfy a consistency condition: if two digit pairs have the same index sum, they must have the same product. Distinct pairs may legitimately collide. For example, (1,4) gives index sum 2 and product 4, while (2,2) also gives sum 2 and product 4. A collision between products 4 and 6 would make the lookup ambiguous and invalidate the scheme.

The full 100 ordered digit pairs can be checked against that condition. It is a finite verification of a table property, not proof that a mechanical device would meet its timing, friction, tolerance, or reliability requirements.

What a specification does and does not establish

The proposed rectangular envelope has volume 26×24×20=12,480 cubic inches, or 12,480/1,728≈7.22 cubic feet. Calling it two cubic feet is an arithmetic error. Even a correct envelope calculation is not the total operating footprint: access, support, and maintenance may require further space.

Similarly, 192 twenty-digit variables provide 3,840 decimal digit positions before counting signs or control hardware. That count describes representational capacity. It does not determine information throughput, energy use, or working reliability. These are separate questions for an implemented system.

A design paper can establish an ingenious organization of computation while leaving manufacturability unresolved. Preserving that distinction gives Ludgate credit for the work documented by the evidence without inventing a completed machine.

Exercises

1. Use the table to compute 58×6. Show the two digit-index sums, decoded products, place weighting, and final carry. Explain why you must not add the indices for 58 as though 58 were a single decimal digit.
2. Show that (1,8) and (2,4) create a permissible collision. Then explain why assigning index 7 to digit 4 instead of 2 would create a problem using the pair (1,4) and another pair.
3. Evaluate this museum label: “Ludgate built a two-cubic-foot computer holding 192 twenty-bit words and remained entirely unknown until 1971.” Identify four independent errors and rewrite the label using only defensible claims.

Compare your reasoning with the worked solutions

1. For 5×6, indices 23+8=31 decode to 30. For 8×6, indices 3+8=11 decode to 48. Thus 58×6=30×10+48=348. The 48 contributes 8 units and 4 tens; adding those 4 tens to the first partial product gives 34 tens. The encoding domain is the ten digits, so it contains no entry for a two-digit operand such as 58. Multi-digit multiplication requires decomposition and place handling.
2. Both (1,8) and (2,4) have sum 0+3=1+2=3; both products equal 8. This collision is harmless. If digit 4 had index 7, (1,4) would have sum 7 and product 4. But (1,3) already has sum 0+7=7 and product 3. One lookup entry could not correctly return both. The scheme's correctness depends on the entire mapping, not merely distinct labels for the digits.
3. No completed machine is established; the paper describes a proposal. The stated envelope is about 7.22 cubic feet. The capacity uses decimal digits, not bits. Contemporary review and the 1960 patent episode contradict total obscurity. A defensible label is: “In 1909 Percy Ludgate published a design for a programmable mechanical calculating machine with a proposed store of 192 twenty-digit decimal variables. It was not built; later historical research clarified its design and reception.”

Source limits

The linked primary paper is a modern OCR transcription. Its tables support the arithmetic analysis; the Royal Irish Academy biography incorporates more recent archival findings than older summaries. Historical uncertainty about practical construction remains uncertainty, not evidence that the design would have succeeded or failed.