The Micrometer

A two-monthly review of the engineering of time

Issue 9   January - February 2018

Yet another (impending) mainspring failure

I seem to spend a lot of my time considering and repairing or replacing mainsprings, which I suppose it is not that surprising because it is about the only component in a clock that is highly stressed.  It is also something which while it appears simple, is in fact a highly complex piece of engineering.  We all use them, yet I wonder just how many of us could size a mainspring from a (lack of) knowledge of the demands made on it to drive the train, the escapement and any ancillaries such as advancing cams or lifting strike levers?  I couldn't for one.

     The photograph below shows a 38 mm wide x 0.40 mm thick mainspring from an English fusee bracket clock.  It is a 20th Century replacement that had been cropped (not very prettily) and re-drilled (again not very prettily).  It is now showing signs of impending failure as can be seen from the two cracks enlarged in the inset, and at least one other tiny crack in which a piece of fluff is trapped.  The person responsible for this did not de-burr the cropped end, but critically did not smooth off the inside of the newly-formed hole, so leaving stress-raisers that could (and did) initiate cracks.  Sure, the mainspring might have cracked anyway, but unusually there is no sign of grossly distorted metal at the hook position.  To me, this points to the impending failure being simple crack initiation from an edge discontinuity.

     To minimise the probability of crack initiation in a newly-formed hole, I always draw-file the inside of the hole along the line of the spring's thickness, even going to the trouble of rubbing the draw-filed edge with an oval burnishing broach not only to minimise further the size of crack initiation points but also to introduce a slight compressive stress in the hole edge*.  Whether it does any good, I don't know, but I do not believe it can do any harm.


*  Introducing a compressive stress into the skin of a stressed steel can delay the onset of fatigue crack initiation.  In structural engineering applications, this is most commonly done either by shot peening or by a needle gun (high-frequency impact treatment).

The 'G&B' stamped on the spring is interesting; was it an original mark by the 20th Century manufacturer which, before cropping, would have been above the original hole, or was the last repair shop proudly advertising its (not very pretty) work?  I guess we will never know.



Clock case repairs

One common form of damage to English bracket clock cases is to the side frets, which tend to be pushed in when the clock is lifted.  This is not helped by the fret retaining beads being in poor condition, and an example is shown in the photograph below.   In this clock the bezel escutcheon had also been damaged and lost, perhaps by the key being left in the lock and also caught while lifting the clock?

     To repair this damage invariably requires what remains of the glazing bars to be removed and new ones made and fitted.  I find pinning them in position with 1.2mm diameter x 10mm long headed pins satisfactory (not least because they are removable), both the bar and case being pre-drilled to 1.1mm.  Access being severely restricted, pre-drilling the case generally requires twiddling a 1.1 mm drill held in a pin vice to about 6mm depth after marking-through from the new retaining bar.  By pre-drilling, the pins can finally be pushed in position with a pad of hardwood, so avoiding the need to use a hammer or punch.

     In this clock the side frets were of very soft zinc (old, if not original), the bent zinc responding well to rolling flat with a wooden rolling pin.  But to reduce the possibility of future finger-damage caused by lifting the clock (seen in the above photograph), a two-piece glazing bar was screwed with No.3 (2.5 mm) brass wood screws inside the case so that a shaped pad of timber supported the top back of the fret where the fingers are most likely to cause damage.  The rather exaggerated sketch of a view from inside the case (above) hopefully clarifies the arrangements I used.

     The replacement key escutcheon (the style selected to match the rear door escutcheon) was fixed in place with Araldite epoxy adhesive before making good the surrounding mahogany veneer.


Whether the bars are for retaining frets or glass, the fret- or glass-retaining bars can be much the same.  Another technique that I have used for making an arched-top glass retaining bar is to laminate it from four strips cut from 0.7 mm thick birch ply sold for model aircraft construction.  In the case illustrated below, the door frame itself was used as a mould, the glass recess being lined with cling-film before clipping the strips into position with multiple toolmakers clamps until the glue had set, making sure any joints in the strips were well staggered.  Once set, the U-shaped ends of the glazing bar were trimmed to clear the lower glazing strip, and the visible edge given a domed outer edge before drilling to take small brass wood screws (patinated), staining and waxing.  Strips of veneer could be used instead of plywood to make the construction less visible as a modern intervention, but cutting and bending long lengths of veneer without splitting is difficult.

A third option (at least for glass) is to use putty as was sometimes done.  A fourth which is sadly too often seen is to use wire pins or even carpet tacks directly against the glass, fret or silk covering.  It is a technique of which I do not approve as it invariably ends badly.



An interesting 'lubricant'...

Invariably I use Swiss synthetic oils (and very occasionally grease) as lubricants for horological work, even though if one filled up one's motor car engine with it, it would cost more than the car itself.  Cleanliness is, of course, important in order to minimise the entrainment of solids that may affect the oil's performance.

     But in a recent count wheel striking longcase from the early 18th Century serviced a mere four years ago according to the owner, the movement appeared very dry with little oil on any of the pivots, and just a mucky smear of a grainy grease-like substance on the pallets.  On setting the movement on my bench for a closer look I discovered the problem (or at least one of them); the movement had been liberally sprayed with a 'lubricant' I have not seen advertised in the horology catalogues (photograph below).

     Yes, you've guessed it: the movement was covered with a heavy dusting of sawdust, which was almost certainly from the recently-sawn ends of a replacement seatboard.  Coupled with the use of cross-headed wood screws not at all sympathetic to an 18th Century clock, the seatboard ends had been cut off probably to get the seatboard to fit inside the hood.  With the movement still secured to the seatboard, the ends had probably been cut off with a coping saw by simply laying the movement face-down on a blanket or sheet of bubble-wrap.  To readers of The Micrometer, further comment is unnecessary.

     There is one other thing of interest in the main photograph and that is the position of the count wheel secured to the strike train great wheel.  I think this is a less usual position than being outside the backplate, but I may be wrong.



... and repairing a fault to reveal a second fault...

This same clock had a motion work post not only slightly bent but with the end missing (snapped off at the taper pin cross hole).  This meant the motion work could slide forward slightly until it was arrested in its slide by the lifting piece, the significance of which I will explain later.  So I set about making a new post from a length of 5 mm square key steel, which is greatly facilitated if one has a self-centring 4-jaw chuck and, for drilling the taper pin cross hole, has made the lathe cross-slide drilling spindle I described back in the November 2016 issue of The Micrometer.

     With the new post fitted, the clock stopped on test, the stoppage appearing to be fairly random but fairly 'crunchy' when one went to reset the minute hand to time.  And then I noticed the hour hand was out of synchronisation, the fault being that the motion work pinion was not only badly worn at the circumference at which it was now forced to mesh with the hour wheel but that it was heavily tapered along its length.  In short, its engagement with the hour wheel teeth was unreliable.

     Measuring the outside diameter showed it to be 7.0 mm at the rearward (unused) end and 6.3 mm at the outer (used) end.  After fitting the now straight replacement post with a taper pin, the pinion had not only a shallower depth of tooth engagement but also was prevented from sliding forward; consequently it could now only mesh at its small, heavily worn and bruised 6.3 mm diameter.  At this diameter the depth of tooth engagement was so shallow that the two most damaged pinion teeth butted the tip of the wheel teeth rather than meshing with them, thus stopping the clock.  Something had to be done if the clock was to return to service.

     I could have moved the post, but it is never easy to plug and re-drill a hole in a plate.  Alternatively I could have bent the post to improve the depth of meshing, but I do not like fixing faults by creating a second fault, though arguably this would be the conservationist's approach.  It was also the solution chosen by the previous repairer, which is perhaps how the end of the post came to be missing.  Another option would be to reverse the pinion so that it meshed with its mating wheel at its larger diameter by silver soldering on a spigot at the small end; after truing and bringing to size in the lathe this spigot could then be pressed into the motion work wheel.

     I decided to do something else, and cut a new parallel pinion in CZ131 yellow brass to an outside diameter of 7.0 mm*, taking the opportunity to bore the hole in the wheel central to its outside diameter by lightly gripping it in the soft jaw lathe chuck (a 3-jaw chuck fitted with soft jaws specially machined to the outside diameter of the wheel).  Looking at the photograph, the new assembly is now retained by the taper pin in the new post.  Towards the top can be seen the lifting piece that previously arrested its forward slide, while the replaced tapered and rather worn and bruised pinion is shown in the inset.

     If you look very closely at my new pinion (arrowed in the inset, upper left), you might also see a slightly raised boss (5 thou or 0.13 mm) to the outer end, which I machine to give just a tiny bit of clearance to miss any burring or slight bending to the taper pin ends occasioned when the pin is pressed home.     So, returning to the tapered pinion, I wonder why it was made like this?  Was it poor workmanship and poor quality control?  Was it cast, the taper representing the moulding 'draw'?  Perhaps the taper allowed the best diameter to be selected from a longer length of tapered pinion stock (bar) to suit the actual centre spacing (the centres being hand-marked, as can be seen in the photograph from the scribed diameters on the plates)?  I don't know the answer, but had the clockmaker fitted the pinion to the wheel the other way round, all would have been well.


* The more curious student of horology might wish to calculate the module of the pinion (hint: the data apparently no longer available from their web-site, the addendum of a 1.0 mod 6T pinion cut with a  PP Thornton cutter is 0.855 mm). Answer at the bottom of the page.



... but there still was a problem...

This time with the date ring rollers, which I have attempted to illustrate in the photograph below.  The edge of the movement frontplate is to the left, the back of the dial plate to the right, the date wheel carrying Pin 'A' in the centre left, and (one of the two) mushroom-shaped stepped rollers supporting the date ring just to its right.

     As is not uncommon with a hand-beaten sheet of brass, the dial plate was not particularly flat.  Moreover, ranging between 0.9 mm and 1.6 mm, the date ring thickness was uneven.  These two variabilities conspired to ensure that the length of the date ring actuating pin (Pin 'A') needed to be clear of the back of the dial plate by at least 1 mm if it was not to press against it and stop the clock.  With the date ring roller arrangements fitted, this meant that the date failed to change when the pin was presented to the thinner parts of the date ring, the pin either sliding behind the date ring teeth or (again) stopping the clock by jamming against the back face of the date ring teeth.

     Without considerable work (and not a little damage) I could see no realistic way of improving the respective flatness and thickness of the dial plate and date ring.  The only solution I could come up with was to machine and fit new grooved (rather than shouldered) rollers to ensure the date ring was held clear of the back of the dial plate by 1.5 mm.  One of the replacement rollers is shown in the lower left inset to the photograph, the running face running against the back of the dial reduced to 6 mm diameter for a length of 0.15 mm to lower the frictional (rotational) torque.

     As an aside, I am not sure the mushroom-shaped rollers were original as the pale-ish brass seemed to be of very high quality (homogeneity, a view perhaps reinforced by the shouldered steel screws also being of excellent form and in excellent condition.  The owner perhaps considering a functioning date display being somewhat optional, maybe a previous repairer found it easier to make simple stepped rollers than more awkward-to-machine grooved rollers?

... and a final reflection

Finally, the more conscientious horology student might find it worth reflecting on the above as a 'case study'.  What primary defect(s) caused the clock to stop?  Did I undertake unnecessary work?  Is there anything that I should not have done (or done differently)?

     For what it is worth, here is my hindsight assessment on the above three points:

  • Sawdust: Probably not the primary cause of the clock stopping, but its presence might tend to draw oil away from the pivots and/or get into the pivot holes.  And most mechanisms respond favourably to a good clean, so I think removal of the sawdust was necessary for the long term future of the movement.
  • New post and pinion: The worn and tapered pinion was clearly a significant contributor to the clock's failure to perform reliably. To axially locate the forward position of the motion work using the lifting piece is not good practise, not least because the wheel teeth rubbing on the lifting piece would represent at best increased friction.  That it could also slide rearwards so that the pinion no longer reliably meshed with the hour wheel would guarantee unreliability in service.  This is perhaps more important for count wheel striking than rack striking, the former requiring intervention by the owner should the clock miss a strike.  I am comfortable that both a new post and a new pinion were the only options for reliable service.
  • New date rollers:  Given that reliable date operation was important to the owners, I could see no other realistic solution to the problem.


     Given the owners' requirement for a reliable, fully-functioning clock balanced with limiting my intervention to a minimum, I think I am happy with what I did, the clock performing reliably while on test.  The one criticism I have of myself is that I didn't fully identify all these defects before I started the repair work.  Apart from hurt pride, this could, if working to a quotation rather than an estimate, result in an unexpected cost increase that, without their formal agreement, cannot legally be passed on to the owners*.


* To protect myself from this, I invariably work to an estimate to which I add an allowance for unforeseen defects, charges against the allowance only being used as incurred.  It also protects the owner from exorbitant quotations as he or she only pays for what I do.

Connection to UK BT phone lines and Broadband


I have often been mildly irritated by having to hang an ugly ADSL* filter 'dongle' (photograph below) from each BT telephone socket wall plate, which seems to me to be very untidy.  With the fast-becoming essential requirement for homes to be connected to the Internet, I do not really understand why unfiltered telephone sockets are still supplied or fitted, but perhaps I am out of step with other UK consumers who increasingly have abandoned land-line telephones in favour of their cell-phones.  Or perhaps I need to upgrade to BT Infinity vDSL which, I think, does not require filters on every telephone socket.  The problem with BT Infinity is, as I understand it, that one has to have a mains power socket near to the master socket, and I will not countenance the standard installation of cabling from a distant power socket nailed to the skirting board.  In my view, all cabling should be run in conduits, and in a domestic situation, this conduit should be buried in the plaster.

     But to return to my more basic Broadband set-up, I recently investigated the situation and, with a bit of trawling, found that filtered ADSL wall plates are available, though filtered secondary sockets are not available from BT.  Some are filtered wall plates with just a telephone socket, while others have both a telephone and Internet socket on the face.   I have used both successfully, though with one warning: buy those of anonymous manufacture on-line without a backing box and they will be damaged in the post if only packed in a padded bag.  This is because the bond between the copper track and the substrate on the printed circuit board is poor, and any squashing by the Post Office in transit results in the components being pushed down and breaking this bond, which in turn breaks the copper track.


*Asymmetric Digital Subscriber Line.  Apparently the asymmetry of this type of Digital Subscriber Line (DSL) favours downloads from the Internet at the expense of uploads to the Internet which, unless one is hosting a web-site(s) (i.e. information is primarily being accessed from your computer), is what the bulk of consumers require.


     This explanation of the cause of the damage  is beyond the comprehension of the suppliers of such products, who just send a replacement packed in the same way so that it also gets squashed and broken.  The second time this happened to me, I got to work with a soldering iron and a few strands of copper wire to repair the printed copper tracks - life's just too short to keep on explaining (i.e. just write-off the cost to experience and never use that supplier again).  They don't believe you might know what you are talking about, anyway.

     The photograph above shows one of our telephones connected to the repaired filtered wall plate.  This telephone is a vintage Bakelite GPO 232 telephone introduced in 1935 that was still supplied into the 1950s.  This telephone has been modified to work with modern telephone exchange lines, and has a modern bellset incorporated within (in the original, the bellset was a separate unit generally fixed under the base).  It works very well, though one does not have the advantage of a hash (#) key.  We call it the 'Grace Kelly' phone as it is similar to that used in Alfred Hitchcock's 1954 film 'Dial M for Murder', though perhaps fortunately, ours does not have the Maida Vale telephone number.  To keep the ageing Bakelite looking good, I give it an occasional light rub with Liberon black patinating wax, which I do not think does it any harm.

     A more reliable wall plate is perhaps that made by Pressac Communications (photograph above), though finding a supplier of small quantities is not easy, and even when a supplier is found they seem to be frequently out of stock.  They are boxed and so do not suffer from postal damage.  But for those annoyed by the 'dongle' the quest is probably worth while.  The only down side that I can see is that if the faceplate of the single telephone-only socket does not make it clear that an ADSL filter is incorporated, so a future home owner may fit an ADSL dongle, though having two in series is unlikely to affect the working of either telephone or Broadband, and may even reduce the possibility of cross-talk.

     Finally, if you intend to fit and remove telephone cables to Insulation Displacement Connector (IDC) sockets* reasonably regularly, it is worth investing in a metal IDC fitting tool, as the plastic ones supplied really do only last for a few uses.  The two types of tool are shown in the photograph alongside two Pressac ADSL wall plates.


* Connections that do not require the removal of the insulation on the wire before insertion into the connector.  Pressing the insulated wire into the terminal block automatically cuts through the insulation to provide the necessary metal-to-metal contact.

Web statistics

Readers may be interested to see the 2017 statistics The Micrometer's hosting service collects.  These are definitely not collected by any underhand methodology or cookies placed by me; they are merely an anonymous indication of the IP addresses that have viewed The Micrometer.  As the IP addresses are assigned at the time your computer is connected to the Internet, the collected data cannot identify individual persons or home addresses; at best all it can do is identify the general area of the UK (or World) in which the IP address is probably located.  It's a bit like the first part of a post code (zip code); neither it nor I can identify a specific home let alone persons.

     The chart above shows the results after exclusion of the 'bots' which crawl the Internet (such as Googlebots (benign) and spambots (malicious)).  The February statistics (which are low) were only collected for 12 days (out of 28) because of a hosting service glitch.  At around 160 unique Readers per month, the overall numbers are low, but it is heartening that many Readers visit The Micrometer more than once.  Advertising could perhaps increase these numbers, but as I receive no income from the site (which costs me in excess of £600 per year in software subscriptions alone), increasing the number of Readers is not a primary aim.  The primary aim is the pleasure I get from producing a Journal.

     The least useful statistic is 'hits', most of which are internally (and unknowingly) created every time a link to an image or piece of HTML, CSS or JavaScript code is accessed by a Reader.  In other words a 'hit counter' as sometimes seen on web-sites appears to be nothing more than a way of puffing up the web-site to suggest it gets far more Readers or Visitors than it actually does.

     Finally, it is worth reiterating that The Micrometer is a periodical, as I find there is nothing more depressing than web-sites which just silt up with information year on year or (even worse) just sit there years after the writer's first flush of enthusiasm has passed.  The down side of my approach is "miss it and you have missed it", so if you think you may want to reference an article in the future, the best thing to do is download the download and/or the print-friendly version and save it on your computer.  But a request: if you do download anything, the downloads are for your personal use only and not for distribution, publication or commercial exploitation in any form whatsoever.



And how is The Micrometer written?

The Micrometer is written using Adobe Muse, which effectively is a powerful front-end application to simplify code writing in HTML (HyperText Mark-up Language) supplemented by the use of CSS (Cascading Style Sheets) and JavaScript.  Taken together HTML, CSS and JavaScript (not to be confused with Java) are 'the triad of cornerstone technologies' used by the World Wide Web.  Adobe Muse was chosen over 'free' web developer software partly to give greater freedom of layout (freedom in design) and partly so as to avoid any possibility of pop-up advertisements or other interference by the supplier (free suppliers have to fund their software development and maintenance from somewhere).

     Each page of The Micrometer is made available in three formats: desk-top, tablet and smart phone.  Optimisation is for desk-top viewing, and some of the images may be a little small when viewed on other platforms.  The logic of this is to allow an initial browse on whatever platform is convenient, while allowing reversion to the desk-top for greater study and printing on paper if needed (e.g. engineering drawings for workshop use).

A new project

Four years ago I undertook some tests which allowed me to watch a going barrel mainspring unwind, photographing it at each turn and making a few suggestions as to what I saw.  The results were written up and published in Horological Journal, January to April 2015 but disappointingly for me the articles received no comment whatsoever.  The principal findings can perhaps be summarised by the over-arching statement that, compared with watch makers, clock makers (with remarkably few exceptions) appear to be well behind in high-performance mainspring design.

     But recently my enthusiasm was rekindled by some correspondence with watch expert David Boettcher, and I resolved to pick up my testing tool again.  I have just completed the first two sets of tests, one set on a barrel-wound mainspring and one set on an open mainspring.  Both springs (from the same manufacturing batch) were wound round an identical arbor and were of identical width and thickness, though each spring has been tested at a wide range of lengths.  By way of illustration, I include two sample plots of torque ( v. number of winding turns from unwound to fully wound at spring lengths of 1000 mm which, for this thickness spring, is very close to the optimum fill length for the barrel.  Equations (curves) have been fitted as shown by the grey lines, the first having a 3rd order polynomial fit and the second a linear curve fit.

     Not yet being confident I really know what the results show, for the time being I shall refrain from commenting in any detail on the shape of the curves or comparison between them.  But as a taster readers may wish to ponder on why the total available energy from the barrel-wound spring from full winding to full unwinding appears to be about 24% lower than that from the open-wound spring (the energy being calculated from the area under the torque v turns curves).

     Another thing I am currently pondering is whether the conventional presentation of the torque-turns curves is confusing.  What one is interested in is the delivered torque, viz. the reduction in torque from fully wound to the clock stopping, and therefore does this suggest that the fully-wound torque/turns point should be the 'zero' point.  Noting the maximum torque of both springs is identical (and they appear to have an identical pull angle and hence couple (photograph below)), the graphs might read better if the 'zero' point was defined as the maximum torque fully wound.  In the barrel wound spring, this might lead more naturally to the statement that the torque does not show a marked increase in torque over the last turn of winding but shows a marked decrease in torque as it starts to unwind due, perhaps, to its pre-load from installation in the barrel?

     Perhaps one could also suggest that the linearity of an unconstrained* open mainspring is simply a function of the spring radius at its state of winding, which is not the case for a mainspring pre-loaded into a barrel**.


*   Unconstrained either by a barrel or by any movement pillars that might interfere with its unwound shape as would be the case in a real clock.


** As every structural engineer will know, linear beam bending theory says σ/y = M/I = E/r  (viz. assuming no further post-yield permanent set after first winding, couple M is proportional to reciprocal radius r).   And as experienced structural engineers will also know: if stuck for ideas, it can help if one starts thinking strain rather than stress ( ε = σ/E ).

     For those wishing to exercise their maths, the expression fitted to the data for the 3rd order (barrel-wound spring) curve is  y = 946*x³ - 9574*x² + 41348*x + 3626  and for the linear (open spring) curve  y = 10650*x + 4651

     If any readers of The Micrometer interested in the science and engineering of mainsprings would like to offer any greater illumination or suggestions for further tests, I would be happy to engage with them and keep them up to date with progress.


Following a suggestion by correspondent Derek Hall, this time I include an embryonic Index to past articles published in The Micrometer which can be accessed via the Archive page.  I hope it is of use, but please let me know if and how you feel it could be expanded.

Next time

A high-speed tailstock drilling attachment to suit the Myford lathe.

     There will also be an article on the design and construction of safety pallets for John Wilding's English regulator.  Designed to limit the force that can be applied to an escape wheel tip in the event of the movement running down or the maintaining gear malfunctioning, I think deadbeat enthusiasts will find my approach innovative.  It even works while setting up the escapement, thus protecting the escape wheel teeth at this potentially vulnerable time.

     I had hoped to give readers a 'sneak preview' but sadly I spilt ink over the photograph of the safety pallets...

Guy Gibbons




Module calculation


PCD of 6 tooth 1.0 mod pinion = 6.00 mm

Outside diameter = PCD + (2 x addendum) = 6 + (2 x 0.855) = 7.71 mm

Therefore: module of a 6 tooth pinion with an outside diameter of 7.0 mm = 7.0/7.71 = 0.91


A 0.9 mod cutter would be an accurate enough choice for this application.


(Note: showing something approximating a full ogive tooth tip shape, my diagram is just that.  PP Thornton cutters for 6 tooth pinions give a more rounded 1/3 ogive tooth tip shape as shown in my earlier photograph of the newly cut pinion.)

Copyright (c) 2018  G E Gibbons

The Micrometer

A two-monthly review of the engineering of time

Issue 9   January - February 2018

The Micrometer

A two-monthly review of the engineering of time

Issue 9   January - February 2018

Copyright (c) 2018  G E Gibbons