Dreyer Fire Control Table

Mark III Dreyer Fire Control Table, as it appeared in 1918, front view

Dreyer Fire Control Tables were early mechanical computers meant to process data to permit a ship to engage a distant target with heavy artillery.  They were fairly involved pieces of equipment, and grew more intricate between the invention of the first prototype table in 1911, their first deployment on dreadnoughts in 1912 (?) and the creation of the Mark V table in 1918.   The dreadnoughts of the Grand Fleet relied on Dreyer equipment at Jutland to convert sporadic and imprecise estimates of range and bearing into workable firing parameters.  Sadly, the level of success was spotty, as visibility conditions made such systematic data collection too occasional for performing the sort of graphical analysis on which the Dreyer depended.

The Dreyer FCTs were literally sturdy iron tables fitted with a number of fire control devices tied together by rotating shafts, bicycle chains and other linkages, worked by 7 or more men simultaneously as a corporate endeavor.  They were housed deep within the ship in the Transmitting Station (TS) located beneath the armored deck.  By the time the Mark V table was fitted in HMS Hood, as many as 30 people might occupy the TS, working the table and its many ancillary devices or serving as liaisons to the fighting positions of the ship.  It is fair to suggest that the Dreyer and its environment and attendants resembled a premonition of the mission control centers of the Apollo program 50 years later: a human/machine system on a broad scale to factor down a torrent of real time data from sensors to create a manageable environment for exerting command at a rate a human could tolerate.

Evolution and History

I'm going to gloss over this a bit, as the information is hard to find in a single place, and I'm not with my materials.  In a manner similar to many Royal Navy weapons and systems, the various generations of Dreyer table were denoted by Mark numbers in roman numerals.  Oddly enough, the series not only did not start with Mark I, but it did not always seem to increase. 

The first table deployed on dreadnoughts were Mark III tables.  A Mark III* variant was soon created while the Mark IV and IV* were being developed.  As these evolutions were underway, a variation called the Mark II was developed to incorporate elements of the rival Argo Aim Corrector system, and a downsized Mark I table was produced to fit the smaller TSes of the earliest dreadnoughts.  The last Mark was the Mark V table, intended for several of the ships then under construction, but in the end shipped only in HMS Hood (and lost with her in the Second World War).

Separate from the advances in the various Marks of FCT, each type enjoyed an incremental maturation process during the war as deficiencies and new possibilities revealed themselves through use.  Typically these changes increased the automation level of the tables and reduced opportunities for operator error.  Most importantly, such refinements offered a chance to solidify thinking on the systems that would supplant the Dreyer tables in the post-war years.

The Battleship as a Computer System

It is not possible to study the Dreyer tables without developing a familiarity with the ship-wide art of fire control, the process of calculation and articulation by which the shells can be made to rapidly and continually fall in a pattern around a distant maneuvering enemy.  The Dreyer's role was akin to that of a CPU within a modern computer system, and its "socket" was the Transmitting Station (TS).

Examining a battleship as a computer system is the best way to develop this understanding.  Just as a CPU processes input received via keyboard, mouse, and network adapter, the Dreyer table was given data on what the enemy appeared to be doing.  And just as a CPU might generate output on a screen or printer, the Dreyer's output peripherals were high power naval guns.  Without paying too much attention to the input and output devices, let's examine the TS's role in the system by drawing a circle around it and observing which inputs and outputs cross this membrane.

Mark III Dreyer Fire Control Table as it appeared in 1918, rear view

Principle of Operation

Almost all the Dreyer tables had the same basic composition:  a dumaresq with a range clocknestled within, and two paper plots where observational data of the target's range and compass bearing were visually plotted in real time.  The various models added refinements which produced a more intimate level of integrity to the operation and flow of data, but even the earliest Dreyer tables could do fairly well if the operators were attentive and well trained. As in any computational endeavor, much relied upon the quantity and quality of data provided to those crunching the numbers. It was to prove that this factor was where the failures which seemed to accompany the Dreyers' most visible showing lurked.

The Range Plot

Primary amongst these were coincidence-type rangefinders (RF hereafter).  The important aspect of their data was that it was subject to errors which increased with the square of the actual range being estimated and upon a variety of other factors. 

The Dreyer table allowed estimates of a number (up to 9, in some cases) RFs to be plotted over time on a large semi-automated "Range Plot" (seen on the right in the images above).  Operators would position a special typewriter laterally along a gauge marked off by ranges in 25 yard increments.  When the index pointed at the range being reported, the key would be struck and a symbol imprinted on a wide scroll of paper unfurling under a motor's action across a large plotting table.  In this way, as time passed the range plot would contain a time-based sample set of ranges. 

The important fact is that in truth, the enemy is only at a given range at any point in time, and so one could imagine this being plotted on the paper as it unrolled.  Though this would be useful indeed, this is not what God gave the boys in the TS -- they only had a point cloud that their faith in Saint Barbara told them must surround the unseen curve.  A statistician would say at this juncture, "Give me the points, and I'll run a linear regression on them to come up with a linear approximation of the range as it changes over time."  The Dreyer solution was the seat-of-the-pants equivalent to this approach:  someone would lean over the table and look at the points.  Looking through a special grid of parallel wires which he could cause to slant at any slope across the paper, he could align the grid so it appeared, to his eye, to best approximate the trend of data over time, and read off the corresponding "rate of change of range".  This measure, which is usually referred to as "range rate", or "rate", is the time-based derivative of range.  At this point, let's leave off further description of the Range Plot and leave it where we have it:  a place where range estimates can be visually recorded over time and a determination made to the slope and intercept of the range versus time curve.

The Bearing Plot

In a similar fashion on the left side of the table, bearing data was plotted on the Bearing Plot over time.  Like range cuts, bearing cuts were triggered down from instruments aloft (for practical discussion, by reading a step-by-step transmitter atatched to the base of an open rifle sight through which someone was peering at the target).  This data arrived in a relatively coarse 1/4 degree granularity, but it did not hit the bearing plot in this form.  For subtle reasons, relative bearing is not helpful in this aspect of computation -- compass bearings (or any absolute direction) to target are desired. This was obtained by adding the output of a gyrocompass repeater to the relative bearings, and this had the effect of rectifying the relative bearings to compass bearings (by adding them to the course of own ship). 

But even this measure did not impact the bearing plot paper.  Unlike ranges, where knowing the value is of utmosst importance, in this part of the Dreyer operation, the only essential demand was to grasp the "rate of change of bearing" (or "bearing rate"), and not bother with knowing the compass bearing itself.  Given this, plotting 0-360 degrees within reasonable physical bounds was difficult, and so Dreyer wisely chose to plot just 10 degrees across the face of the paper (a clever spiral cylinder underneath the page provided the desired "wrap around" plotting so a full compass face would wrap 36 times across the sheet.  Just as on the range plot, a grid of wires was aligned with the apparent trend of dots triggered onto the bearing plot (completely automatically from the triggerpress aloft), and an indicator on this grid told the operator what bearing rate this corresponded to.

At this stage, your head is probably spinning.  Mine was.  The sad part?  We're hardly started.  To review, we've covered the range plot and bearing plot:  where they obtain their data from, and what they produce.  The range plot is informed by 1-9 RFs and delivers an estimate of range and range rate, and the bearing plot is informed by a single bearing indicator with gyrocompass help and produces an estimate of bearing rate.

The Dumaresq

As I cover in a video, the Dumaresq was helpful in relating a range rate and "speed across" (sometimes referred to as "dumaresq deflection", that component of relative target motion orthogonal to the range rate) to the heading/course of own ship and heading/course of the enemy.  Since we can trust that by reading a gyrocompass and Forbes Log (speedometer) we can ascertain the former, the Dumaresq becomes a device to relate range rate and speed across to enemy speed and heading.

If one knows the range to the enemy (or has an estimate they'd put some faith in), bearing rate is equal to tan(speed across / range).  At this point, you're probably saying, "oh gee... TRIG?!?".  The thinkers of the time, Frederic Dreyer amongst them, understand this reaction, and had an accomodating answer.  To help the team track this and facilitate tying the dumaresq into the bearing rate estimated on the bearing plot, a graph of speed-across curves was printed onto a metal plate (X axis was bearing rate and Y was range) and wrapped around a cylinder to save space.  When the grid measuring bearing rates on the bearing plot was adjusted, an index swept across the X axis of this cylinder.  When the gun range (to be spoken of later) was adjusted, the cylinder spun, setting the correct Y axis position.  The speed-across being suggested by the bearing rate and gun range was then visible under the cursor.  I think it is important to pause here and reflect upon the fact that a complex 3 dimensional relationship has been printed into metal and reduced to an almost one-dimensional display area.  That's not shabby for the prehistory of computing.   This practice would reach a new high in the creation of the Wind Dumaresq, which had a large etched cylinder under its dial plate.  By referencing these drums (one in early Dreyer tables, later two, to provide both dumaresq and sight deflection -- about which more later), the Cartesian and angular relationship of relative enemy motion could be compared. 

In addition to the basic function of a dumaresq, the Mark III Dreyer FCT's Mark VI dumaresq had a number of other features.  One was that a gyrocompass repeater on the back headboard of the table could drive the orientation of the own ship heading armature and thus keep this vital and oft-changing parameter continually updated on the Dreyer table.  The benefits of this feature will be discussed in depth later.  The dumaresq also boasted a pair of variable speed drives:  a bearing clock and a range clock.  The bearing clock was primarily used within the dumaresq, and it could allow a constant bearing rate to be dialed in (this was a separate adjustment from the bearing rate handle near the Bearing Plot), and this would cause the enemy bearing on the dumaresq to slowly alter at a constant rate (within a range of +/- 15 degrees per minute, though I should check this as I see Rob Brassington has depicted a +/- 60 dpm envelope in his model). 

The other clock was the range clock.  Its speed was set  by a turning a handle on the right side of the dumaresq, and its current setting was indicated by a small index that moved within a slot along the line of bearing in the fixed dial plate of the dumaresq.  This arrangement was a clever visualization based upon the fact that the projection of relative enemy motion onto the line of bearing is the range rate, and by scaling and orienting the clockspeed indicator in this fashion, a helpful cue was realized:  when the dumaresq was properly set to enemy speed and heading, setting the clock to the proper speed was achieved by simply matching this indicator to the dumaresq's pipper.  But, unlike the bearing clock, the range clock's influence was not manifested within the dumaresq itself, but at the range plot and beyond.

"The range plot?!?", you might say, "but we've just covered that bit!"  And you'd be right.  But what I find makes the Dreyer table so interesting, like many such devices, is that it encompasses several feedback loops that defy a structured explanation.

Output of the Range Clock

The range clock's constant speed output went into a differential device called the Spotting Corrector, whose gearing multiplexed it out to three further destinations.  One was a worm screw arranged laterally across the range plot to which a rider was fastened bearing a pencil which touched the paper at approximately the same vertical position where the range cut typewriter's marks were made.  Imagine the range clock being set to zero, and the worm screw therefore being motionless.  In such a case, the pencil in the rider would trace a vertical line lazily up the plot, as the paper scrolled by at 2 inches per minute.  Now, imagine the range clock being set to a speed of positive 500 yards per minute (positive rates were dubbed "opening", and negative ones "closing").  In such a case, by careful agreement with the scale of the range plot, this would cause the pencil to be dragged laterally across the paper (to the right) so it's position on the range scale would alter 500 yards each minute.  The combined influence of this motion and the paper's own vertical traction at 2 inches per minute would cause a diagonal line to be traced, depicting a hypothesis of the range to the enemy increasing at 500 yards per minute.  In this way, for each setting of the range clock, the range to the enemy could be automatically modeled in a linear manner.  Alterations of the range clock's speed would cause a new slope to be realized, and what is a curve but a series of infinitely short straight lines altering continuously in their derivatives?  Well, this is a helpful realization that a sporadically-corrected series of linear range rates are a fair foundation for modeling any evolving range relationship.

I mentioned that the Spotting Corrector was a "differential".  I'm not very automotive, so this bears some explanation.   A differential is actually an adding machine of rotating shafts.  One shaft in this device is the output shaft of the range clock.  The other is the spotting correction handle on the corrector.  The operation was subtle.  When the spotting correction handle was rotated, a spring-loaded dial atop the corrector indicated the amount of the range correction being applied.  The output shaft to the plotting pencil was not affected by this motion, but the other output shaft was. 

The other output shaft of the Spotting Corrector did three things.  First, it drove a cyclometric digital display ("cyclometric" is an old term... imagine an old car's odometer, in which numbers on a series of cylindrical wheels indicate a count-up total in digital form) to indicate Gun Range (plotted range plus the total spotting correction).  Second, the Gun Range was also conveyed to the Bearing Plot side of the table where it was used to rotate the deflection drums to the current gun range.  You can see now, perhaps, how self-involved is the data flow within a Dreyer Table!  Third, it leapt up into the overhead via a flexible shaft to drive an indicator hand on a clock-like master range transmitter.  From this, it suffices to say that the Gun Range was able to propagate to the director and to the gun mounts, as I have to draw the line somewhere as to what a single essay covers!

The last primary detail about the range clock and the range plot was the "Tuning Handle" and "Pedalling Clutch".  At the outset of the engagement, the position of the range pencil and of the zeroed-out Gun Range Counter would be at an arbitrary place (perhaps in the center of the paper and the gun range envelope).  When the first range cut was reported, as it was the most concrete data on hand, one would want to drive the pencil and the gun range counter to this range immediately.  The Tuning Handle accomplished this.  But later, once the battle was underway and the guns were firing, one might notice that the range pencil had drifted out of the trend of range cuts being reported.  It would be comforting to see the pencil in the midst of the range cuts, and so the temptation would be to just twist the tuning handle until it was right there in the cloud.  However, if one did this, one would be altering the gun range, and this would be a mystifying and perplexing jolt of reality merely to suit a visualization desire in the TS.  The pedalling clutch (so named because in early designs it was a foot  pedal, but in deployed tables it was actually more like depressible handle lever. When the pedalling clutch was pressed, the tuning handle's motion would alter the position of the range pencil but leave the gun range counter and range master transmitter undisturbed. 

To put a bow on the range side of the table, the Range Plot allows semi-automated display of discrete range cut data (typewritten onto the plot) to be examined alongside a pencil-drawn linear hypothesis of range versus time, and a handle by which spotting corrections can be added to the plotted hypothesis to produce a "Gun Range" for use elsewhere on the table and by the gunners so intently waiting for the best guess of the target's range.

TO BE CONTINUED

Nitty Gritty

Rangefinder data was provided across electro-mechanical transmitter/receiver pairs.  The Royal Navy was deeply invest in the use of "step by step", or asynchronous communication.  This was essentially a digital data network of a dedicated type, wherein the transmitter and receiver, which resembled an old car's odometer, would be "synchronised" or "lined up" so they agreed with each other.  From there, changes in the range being signaled would produce pulses of electricity (1 every 25 yards change) which would cause a motor in the receiver to rotate its digit wheels by the same increment as the transmitter had been altered in registering the change.  In this manner, provided no misstep occurred which might cause a discrepancy, continual working of the transmitter would remotely be signaled in a fairly clear manner in the TS. The design of the transmitters was elegant:  in general, the RF team only had to work their rangefinder to have its range transmitted below, and when they were satisfied that they had merged their prisms on the target ("made a range cut"), they'd press a trigger which would cause a light to appear at the receiver or a shutter to flip open to expose