Back Up Next

Continue final finishing of escapement, begin balance springs testing - June / July 2016


The pendulum cross arbors and their spacers are now completed.

Notice in this photo that Buchanan uses a numbering system to match up parts wherever there are similar pieces that may be confused upon reassembly.


Another example of the numbering system. Even the individual numerals are beautiful. Next is an example of a finished screw. The photos tend to make them darker than they really are.


The first photo shows the diagonal bracing for the pendulums. Next the spacers are assembled.

The finished front pendulum and escapement supports are shown here.

The barrel shaped worm gears are beat adjustments for the grasshopper escapements.



Buchanan appears to be trying a different approach here for the bluing process. Remember how a modern torch was used in the prior process. Here he has gone to an open flame burner. So the process is now completely in keeping with the old methods.


The front end pendulum assemblies are now complete and ready to accept the grasshopper escapements.


Buchanan now proceeds onto the grasshopper escapements. While he previously indicated that these were largely in their final finish form, they were not completely so.


We now turn to the steel components. First, all of the jewel escapement stones must be removed; these are the bird’s ‘beaks’. The jewel pivots are protected within the body of the part and do not need to be removed.


The remaining escapement components which comprise the allegorical bird’s bodies, the feathers are finished. Notice how slender the curvaceous escapement pair appears in the first photo. Next the parts are drying after lacquering. Buchanan writes: We have 6 small brass cocks to go and the component parts of the escapement will be complete. Hope to have that done today. I am blueing screws and finishing the incomplete pillars that hold the escapements to the beat adjusters.

The pair of allegorical bird escapements complete with tail feathers and comb on the heads are now finished.


The second photo shows the complexity of the grasshopper joint. A dozen components comprise this one small system.


Buchanan now goes through the steps to make the escapement pillars. These are the last components to complete the escapements. They have a fairly complex shape requiring not only conventional lathe turning, but manual filing using both round and square filing buttons. Filing buttons are used to guide the file into a particular contour, usually a round or flat profile. The buttons are custom made just for this step. The first two photos show the initial turning.


Next the flat filing buttons are used to mark off the tabs where the mounting holes are located behind the flat buttons on each side of the pillar.


Much of the original round flange is rough cut away as seen in the first photo. Next the flat buttons are used to fine-shape the mounting tabs.


In the first photo the round buttons are used for the fine-shaping of the round portion of the pillar. The second photo shows a before and after shot in relation to the end flanges and the use of the filing buttons to achieve this. The turning of the pillar body, though, is already done on the piece to the left.

The next two photos show this piece reinserted into the completed escapements and pendulum support rods. The combination of color and finish is visually compelling.


Now Buchanan tackles the large balance frames. Each one is nearly 24 inches tall with a lot of internal complex patterns. Buchanan writes: Both frames are filed up to sanding point. I did an opening count and there are 85 opening in each balance frame or 170 openings. If I can polish one opening in 20 minutes (some will be quicker and some longer.) I have 3400 minutes or 56 hours to go. Then there is the outside and the back and front face. I was asking myself why the filing was taking so long. I have the first opening polished! 169 to go. The final finishing steps take a substantial fraction of the time it took to make these pieces in the first place. While these parts are especially large and complex in their layout, the machine is filled with such parts.


I was curious how Buchanan was finishing the frames so I asked him what the procedure was. Was it with a stick charged with grit? Or paper on a shaped surface? There are so many concave surfaces of differing configurations I’m curious how this is tackled.


These photos explain how it’s done. There are five grades of sanding sticks, red-400 grit, purple-600 grit, blue-1000 grit, green-2500 grit and yellow-5000 grit. By the time one get’s to that grade it is a pretty fine polish. The other thing that is being accomplished here is that the geometry of the curves is being subtly altered to a better refinement.


I have the first of the small panels polished on the first frame. The blue marker is for a reminder as to which part has had a certain grit.  First as I use 250 grit I mark each facet as I finish then I remove it with 400 grit then as I finish 100 grit I remark each section then remove it with 2500 grit and a final polish with 5000 grit. That is 32 facets times 5 grades of paper.  I have 3 more small panels to go and then the inside of the first frame is complete.

Buchanan writes: I am looking for the name of the Greek, Hindu, Eskimo, or what ever god of frustration. This will be the official name for this clock. One can completely empathize with his predicament.


Now the last step is performed. The two large flat sides that compose the front and rear of the balance frames are now tackled. The first photo shows the filings left behind on the right leaving a shadow of the frame itself on the left. Next is a close up that shadow on the wood board.


Buchanan glues the sandpaper to a thick sheet of glass to ensure that the paper remains perfectly flat and the final finish is as close to an ideal plane as possible.


Notice the pile of brass dust on the table top. This area is just below the table top of the custom polishing machine he made last month. Next a shot of Buchanan’s hands after a hard day’s work.

This photo shows the end of an eight hour session. There is a prodigious amount of used and broken sanding sticks on the floor. Notice the pile of brass dust under the filing machine’s table in the upper right corner of the photo. Hope that balance frame doesn’t drop out of the vise!


The balance frames are now brought to their final finish. These were later lacquered.


The frames and escapements are now ready for assembly. In the next photo the assembly is complete. Unlike in a conventional clock, the pendulum is not readily separable from the escapement. They are an integrated unit, excepting the escapement wheels themselves.

Rear three-quarter elevation. Both sets of triple antifriction wheel support systems are seen in their full complexity.

Front three-quarter elevation. Again, complexity and the visual impact of the various colors are brought to the fore. This photo better represents the color of the blued screws. It is the lighter, 'electric blue' seen in watch work that I was after. While not to everyone's taste in the horological world, I think it works well in this application; particularly because these screws are very large.

A view from the inside, rear. Notice how the everything appears to radiate out from the center axis, especially the six straight rays that follow the longitudinal orientation. At the 45 degree angle from these it appears that the metal has peeled away like a delicate sliver of wood under the application of a wood planer tool. All edges are sharp as they should be under strict standards. yet the entire polish gives a glowing, 'liquid' look. Next month the second balance will be finished.


Buchanan now turns to the springs for the balances. With the generous collaboration of Douglas Drumheller we begin the process of bringing the machine from a visual attraction to an accurate timekeeper. Or at least as accurate as can be accomplished in such a complex mechanism.

After Buchanan read Mr. Drumheller’s article there were several responses back and forth. Buchanan and Mr. Drumheller chose to respond with additional dialog within the original email and with responses in different colored font. The original letter from Buchanan to Mr. Drumheller's  is in black. Mr. Drumheller's  response is in red. Buchanan's next reply is in blue and Mr. Drumheller's rejoinder in Green. I've included this dialog to give the reader a taste of how ideas, questions and solutions come about through collaboration. There will be several such collaborations on the rating of the clock.

I have read the article and I think I can follow Mr. Drumheller in his reasoning.  I need adjustment on the following points.

I have known of your Harrison for some time and must compliment you on it. It  is beautiful.  It would surely be unique to actually navigate by means of a Harrison Replica. It’s my dream, but of course it’s hard to get buy in for the cruise and documentary. At this point I don’t think it will happen. Anyway I’m off to my boat in the Great Lakes in a few weeks to dream about such things.

I found your paper fascinating and enlightening.  I have always understood how the temperature compensation worked on H.1 but had not thought about the effect of moving the point of attachment on it until I had started to make my adjusters. Your paper answers a number of queries I had. Thank you.

I have a query about Page 3 middle paragraph. ”An alteration of as little as 1/2mm will MAKE A DIFFERENCE OF SEVERAL MINUTES A DAY.  This I would understand to mean an adjustment on our outer adjusters on the balance or on the verier adjusters as they are now, but with no temperature compensation fitted.  And not a change in semiarch but a change in spring rate.  I see also the need for an accurate way to measure the arc of our balance, as we will need to monitor every factor that can change.

Almost all of this observed change in rate is due to the presence of the temperature compensator which alters the springs to a highly nonlinear behavior. If the springs themselves are perfectly linear and the cheeks are perfect circular arc this adjustment will have little effect on the rate of your system, which does not have a compensator. Yes you are correct when you take the temperature compensators into account.  I agree with you. I missed this until I had read through the complete article.

Page 4 last paragraph, first line. I must misunderstand something here but I never thought of a helical spring as linear. The result of 4 helical springs in a Harrison system may be linear. I do not have any real life experience with this yet to be able to comment. Engineers think of them as linear. Of course a clockmaker looks at very tiny deviations from linearity as very important. I’ve not measured the force deflection of one of these springs, but my guess is that it will be hard to measure the deviation from linear. Now if the spring is grossly extended it will exhibit a strong deviation from linearity. Measuring the rate as you change your adjusters will give a very good measure of the deviation provided the amplitude is held constant. I think I could plot some sort of deflection Force graph for my springs.  I will see what I can manage.  I have a mill with a digital readout on the z axis  if I suspend an overweight weight on a spring  but support the weight on a sensitive balance , as I drop the knee of the mill I will get a decrease in the reading on the balance. This will be very useful data. This system should have no friction involved anywhere. Only the weight of the spring coils to take into account. I am wondering how we could measure the force of deflection for the complete balance assembly.  I suggest a series of coast-down tests, which will yield the rate vs. the amplitude. You’ll also get Q.  The version of MicroSet that I hope you purchased has the capability to measure two things with the photogate laser. First it measures the period of the beat, and simultaneously it measures the time that the flag, which you mount on the balance, blocks the laser beam. These measurements yield values for the rate and amplitude during operation of the balances. You can repeat the test with differing amounts of initial stretch of the springs. I suggest you use the procedure outlined in my HSN article which is attached. This procedure requires using the MicroSet to capture every beat so the amplitude can be extracted from these readings. If you send me the Microset files that you generate, I’d be willing to help you reduce the data. (I suggest starting with a flag of measured width that is about 1/16” to 1/8” wide.) Often a one point calibration is needed to determine the extent to which the laser beam diffuses around the edges of the balance flag. This is accomplished by comparing a Microset measurement of amplitude to a direct measurement of the swing amplitude.

Page 6 second paragraph. Last line.  I need convincing that a change in outer band length produces a change in the  0max  Note 3  I have experimented with this on My No.1 clock and found that shaping the cheeks can have a major effect on isochronism. But my cheeks were on a balance with an arc of over 180 degrees. So we are comparing apples with pears. I do not intend to imply that changing the band length changes Omax. Instead I was pointing out that with the temperature compensator installed the average stiffness of the springs can be changed either by changing the position of the compensator pads or by changing the amplitude of the swing. Yes I agree. That is why the compensator is such a bad device. It destroys the isochronism of the balances. It was one of Harrison’s worst ideas. Yes I agree. I also do not mean to imply that you can’t tune the clock by altering the shapes of the cheeks. Rather I suggest that such an alteration of shape, while important and useful in many applications, is too weak to overcome the huge anisochronism produced by the temperature compensator. Yes I agree.

 Page 14, last paragraph.  I agree with this paragraph totally .I think I recall reading that on the return trip  Harrison applied a calculated correction which gave an extremely accurate rate.This is very important to me. Is there any chance you might remember where you saw that. I thought you would ask that. I will see if I can find it. It is not in Gould or Sobels book. Perhaps in Such Mechanism. I found a short reference to a correction after a trip with H.4 but that is not what we want

These comments are my uneducated observations. Any enlightenment will be gladly accepted . I am not wanting to criticize but want to understand better what we both are working towards.  I found the description of the arc measurement very interesting.

I really appreciate your interest, and I think that what you are doing on this project is marvellous. Thank you. If this discussion produces a better clock or a better understanding of what we are doing it is wonderful.

I have also a few more ideas on temperature compensation.

I’d be interested in hearing them when you are ready.

I better get on with the polishing. The 6 small frames are finished and lacquered, polishing screws now. My dad was a tool & die maker---steel dies for pouring molten zinc and aluminium. The die faces had to be polished to mirror finish so that they would release the castings properly. On occasion he would unintentionally polish his fingernails to paper thin. Mom would help cure that with her fingernail polish.

I spent most of my early working life making plastic injection molds. Polishing tool steel is no fun, I sandpapered my fingertips through on my first clock. Mr Frank has kindly financed the full suite of sensors for my Microset timer. I have tested it briefly but do not know if I can use if for accurate amplitude measurement.

You need the version of Brian Mumford’s MicroSet hardware and software that allows you to turn on the amplitude display in the computer graphics. The resulting plot with the amplitude shown as a blue line will not be scaled properly, but we can solve that problem.

 I think it has some feature like that.  If not I would most certainly need to ask you for more detail about your method. I use the MicroSet with the photogate laser. The sensor needs to be shielded from ambient light. During my initial tests I didn’t do this and I first noticed problems after I realized that the readings shifted as I walked past my clock wearing a bright shirt. The best calibration of the amplitude scale is obtained by comparing a MicroSet reading to a direct measurement of the balance amplitude. It’s a one-point calibration.

I am working on finishing my balances to a point where I can start an accurate analysis of their behaviour. I will, no doubt, call on your expertise at this point.  My only previous experience relates to my first serious clock which has a horizontal bar balance, two Harrison type springs and a constant force escapement. I will attach a photo or two of it.

Thanks for the great photos. They make me jealous.

You may be interested in our website: www, Mr Frank’s chronology of his clock on his website is superb.

Mr. Drumheller makes the final line with the following humorous comment:  

We should call this the Rainbow Discussion. At some point we’ll run out of colors. Doug


Buchanan now begins to address the fitting of the new balance springs. He writes: I have attached a few photos of my spring force / stretch apparatus and the data gathered and also a graph. I was rather pleased with the results as well as with the degree of ‘linearity’ of the first test spring. I have only the main balance frames to polish before I can fit my springs and start a little serious testing. See graph below.    


Buchanan writes to both Mr. Drumheller and me:  I have now taken the force readings and inverted them and then added the two values together to see what a two spring system would look like. The yellow line is the result. This gives us a weaker restoring force towards the centre of the balance oscillation. This should counteract somewhat the escapement impulse. I am going to have to find some way to measure my impulse force!  

Mr. Drumheller wrote back:

Very clean data! I’ve reworked the spreadsheet and attached it. As you will see I was able to play some with your data. I would also suggest you reread my HJ article with this data in mind.

 If you look at the spreadsheet you’ll see that I’ve added a few columns and a plot. I’ve changed the units of your measurements to lbs and inches. The key information that we need is a plot of the force (lbs) vs. the deflection (in), which is included as the gray line in the plot. The slope of that plot is called the spring stiffness or often the spring rate. This curve is very close to a straight line. When absolutely straight the spring is linear. My yellow curve is such a linear fit. While the gray line has a slope that varies very slightly from one end to the other, the yellow line has a unique single-valued slope. Its value is listed in Cell G2. For reference the value of the stiffness of a balance spring in my replica is given in H2.

Column F contains the deviation of the gray line from the yellow as a percentage of the maximum measured force. The deviation is about 1% maximum. This is a real deviation only if you feel your force measurement errors are less than a percent. If the measurements were made while the deflection was increased we might try taking measurements as the deflection was decreased to check if the deviation reverses sign.

This is the effect of one spring acted on the balance. The effect of two linear springs is obtained by simply doubling the values of the forces in Column D. (The effect of two nonlinear springs is approximately the same with a small complication.) This in effect doubles the stiffness listed in Cell G2, and indeed the two springs do work together to create a restored torque that acts to return the balance to the neutral position. The yellow line that you created in your plot by added a reversed data column does not apply to this situation.

 The nonlinearity of your spring is very slight being a small fraction of that of a pendulum acting with the same amplitude. Now you should notice that for each balance one of the springs will extend while the other contracts. The nonlinearity causes the stiffness of the stretched spring to increase very slightly while the stiffness of the contracting spring decreases slightly. I think these two effects will nearly cancel to make the clock isochronal except for the action of the recoil escapement which will cause gain as the swing amplitude increases.

 I’ll be interested in your comments and will try to respond in a timely manner over the next few weeks.

Good Morning Mr. Drumheller.  

I realise that I made a mistake by just inverting and adding the data. I have no problem working in pounds and inches but my instruments are mostly metric.  I checked the measurements when relaxing the spring and it was very consistent but I will conduct a proper check in both directions. I was also pressed for time when I took the readings so I think it would be good to repeat the measurements again. Thank you for your comments. As you can see I have not yet given the spring system much thought. I would also like to perform the same type of force measurement with the springs fitted to the clock with and without the escapement operating. I can also easily vary the impulse force on the clock although the supplementary arc of the balance is rather limited as there is  not much space before the two balance balls touch each other.

Mr. Drumheller writes back:

I think the biggest accuracy issue you will have is temperature compensation. I’m playing a game with my replica where I am trying to redesign the Harrison temperature compensator using the technology available to him. It’s very difficult to figure out something that will retrofit onto H1. I’d think you might employ invar springs, which should do the trick.  Without it the error is about 7 s/day /deg F, which is all caused by softening of the steel springs as I now realize that thermal expansions of the balances are already self-compensated.  All this other stuff we are talking about will fall into the < 5 sec/day category. I also think you will have some problems holding the arc steady because the balance arbors roll on anti-friction wheels. Dust becomes an issue. In fact I now believe the anomalous excursions in arc that I describe on Page 499 and Fig. 7 of my HJ article are due to dust. It doesn’t happen often as the clock is in a glass case that is hardly ever opened.

This last observation about the springs by Mr. Drumheller will turn out to be exactly correct. We will see how this is resolved in the next few months.

Back Up Next