What caused this amplitude shift? There is a small drop in temperature and if you look at amplitude v. temperature over the last month (since the weather station was moved into the pendulum chamber) [link] you see multiple episodes of correlation between amplitude and temperature, but these are probably due to changes in drag (temperature rise causes density drop). My theory though is that a temperature rise also indicates a change in air flow so that warm air is now rising causing amplitude increase. Maybe a good idea to seal up the pendulum chamber more properly.
10 Jan 2011 [06:35]
ADJUST: +300 ms/day to 29800 - curious that ths is 3000 heavier than it was before the barocompensator was removed. Mysterious. Surely the same device returned to the same location should need the same compensation mass.
08 Jan 2011 [11:40]
ADJUST: -3400 ms/day to 29500 to compensate for the mass of the barocompensator
08 Jan 2011 [11:02]
Barometric compensator reinstalled, in the same place it was (up near the pivot) because I want to give it another chance. Thanks to Hilary for clambering around behind the clock, all witnessed by The Strassers (Mike, Lou, Peter, Andrew) and Jonnie Hunt.
06 Jan 2011 [14:00]
Mark has measured the expansion coefficients for both the aneroids at 0.03 mm/mbar - so they are working fine. In fact this is 40% more sensitive than I had assumed, so it is a mystery why the barocomp is not working. I have a theory that there is something funny about the way the flexure of the pendulum suspension works, so I am going to have a go putting the compensator further down the pendulum, away from the flexure. This is what Philip Woodward did ... so maybe there is a good reason. It will mean having to screw up the regulation nut under the bob again - which is what stuffed up the temperature compensator last time. Ah well, it all makes work for the working man to do.
06 Jan 2011 [10:00]
... the observant amongst you will have noticed that the clock sped up by 3 seconds retrospectively. This is not relativity in action - it is simply that the clock is actually sounding half a second fast at the moment, so I have brought the logger in line with the real world. Yes, there is a real world!
05 Jan 2011 [21:30]
So the barocompensator is removed. This means that I now have to do what I used to do and check the pressure forecast at BBC weather [link] because I know that the clock loses -8ms/day per mbar of pressure rise. The barometric presssure is due to rise by 20mbar over the next 6 days so lets see if the clock slows down by 120ms/day.
05 Jan 2011 [19:00]
ADJUST: +5500 ms/day to 32900 now that the barocompensator has been removed
05 Jan 2011 [18:55]
The old computer as been reinstated until we can figure out compatibility of NiDAQ drivers with OPENSUSE 11.3 Many thanks to Tim Gale for his help! Useful to record that NiDAQ drivers are updated using sudo updateNIDRIVERS `uname -r` and also to update to the server use lftp -f uploadData.lftp .
04 Jan 2011 [14:35]
Removed the barometric compensator for bench testing - this required stopping the pendulum. Thanks to Hilary Costello for climbing around behind the clock.
04 Jan 2011 [12:00]
Datalogging suspended due to replacement of computer - the old one did not have enough memory to run the webcam reliably. Thanks to Tim Gale and Anson Cheung for all their efforts
02 Jan 2011 [13:20]
ADJUST: +400 ms/day to 27400 thanks to Mark Rainer. The barocompensator is not doing its job. I will remove it on Tuesday and give it a good look over.
31 Dec 2010 [11:30]
It is hard to tell how well the barometric and temperature compensators are working because temperature and pressure change at the same time. This is where the monitor page come in [link] because an equation for working out drift from temperature and pressure data can be tested. The equation I am testing is: Going = -120*(Temp-6)-8*(Pressure-1013)-40*dTemp/dt -60) and it fits pretty well. Here is a direct link to the drift estimate and the actual drift - a good fit [link] and the XY plot [link] shows a straight line. If this equation is right then it means that the variation of going with temperature is -120 ms/day per deg C [-520] , with pressure is -8 ms/day per mbar [-8] and with rate-or-change of temperature is 40 ms/day
31 Dec 2010 [11:29]
per degC/day . The values in [ ] are for an uncompensated pendulum.
23 Dec 2010 [13:55]
ADJUST: +100 ms/day to 27000 thanks to Hilary Costello and Alice Cicirello
21 Dec 2010 [12:30]
ADJUST: -500 ms/day to 26900
19 Dec 2010 [09:38]
Bedmaker Effect, interim conclusion: 1: opening the door in the room below on a cold day causes the air flow in the pendulum chute to change (a cold downwards air flow is replaced by a warm upwards air flow) because of the chimney effect forced by the tall heated staircase in the building. 2: as the air warms up, the density and viscosity of the air both fall allowing the swing amplitude to increase. 3: The change in direction of air flow (downward to upward) causes the Going to drop (the clock slows down) because upward drag on the pendulum causes an apparent reduction in g.
19 Dec 2010 [09:37]
Bedmaker Effect (continued) 2: the Going changes more dramatically [link] . What is interesting is that there is a step change in going of about -900ms/day and this occurs straight away, not at all related to the gradual temperature rise. This is clear in the scatter plot [link] (ok, there is a gradient at the expected slope due to thermal expansion, but this is after te initial step occurs. It was Dwight Elvey who made me look again at this, and I think he is right - the observed change in going could well be due to the change in air flow up the chute. If descending cold air is replaced by rising warmer air then the sudden change in direction of air flow will cause a change in Going. This has been looked at in the Theory section of the website, and in Section 4.6.4 and
19 Dec 2010 [09:36]
estimate was calculated as -300ms/day per 1m/s of upward air flow. This suggests that the change in upward air flow this morning should have been around 3m/s - which is pretty fast. Certainly measurable. I will have to invest in a digital anemometer now!
19 Dec 2010 [09:36]
Bedmaker Effect (continued) So, here are some observations: 1: amplitude changes linearly with temperature, as can be seen in a scatter plot [link] and the fact that the rise curve and the fall curve follow each other (ie there is no hysteresis) really suggests that the temperature sensor responds quickly to temperature change and that the amplitude measurement is due to air temperature change. The gradient is about 0.8mrad/C which agrees really well with theory for how drag varies with air density. Nothing definite yet, but the evidence for this theory is mounting.
19 Dec 2010 [09:35]
Bedmaker Effect, just opening the door (no bedmaker!) on a cold morning, -4C outside. Heating on inside, really toasty. So, opening the door causes a chimney effect, forcing warm air into the pendulum chute. This temperature rise can now be seen [link] because the weather station is located in the pendulum chute itself. I have made some observations and these are in separate comments above.
18 Dec 2010 [12:00]
It is good to see that the clock has settled down to where it was. [link]
18 Dec 2010 [12:01]
Conclusion on temperature sensitivity of sensor: It seems that the sensor setup is not temperature sensitive over the range of temperature changes observed in the Bedmaker Effect (below 1C change). The key indicator is seen here [link] where the amplitude change when a hairdrier is blown onto the sensor is in the wrong direction.
18 Dec 2010 [09:39]
Expt 3: [link] Now a hair drier is used to warm up the sensor, and at the same time part of the pendulum shaft is warmed up (it is not possible with a hair drier to target the sensor being so close to the pednulum shaft). We see that there is a small change in indicated amplitude, and this is downwards consistent with the negative temperature sensitivity of the sensor as found in expt 1. But most significant is a very large change in going (slope of the drift) to -1700ms/day. This is consistent with lengthening of the pendulum due to thermal expansion, and it is a large effect because the hair drier is quite hot.
18 Dec 2010 [09:33]
Here [link] I bumped the pendulum while moving the temperature sensor into the pendulum chamber. The idea now is that the temperature indicated on the website is now the temperature inside the pendulum chamber rather than in the clock case.
18 Dec 2010 [09:21]
Expt 2: [link] The same again, but the soldering iron is in place for about 5 minutes. This time there is a bigger effect on amplitude (about -0.2mrad) but again no significant effect on going. The thing to note is that the indicated amplitude change is negative, so that the effect of rising temperature is to reduce the indicated amplitude, but temperature change does not appreciably change the indicated rate of Going. The air temperature that the soldering iron produces around the sensor is HOT - very much higher than anything that the central heating will produce. It seems then that while the sensor is indeed temperature sensitive it cannot explain the Bedmaker Effect.
18 Dec 2010 [09:17]
Experiments, in the light of helpful comments by Dwight Elvey who rightly points out that the optic sensor is probably temperature senitive. See [link] . Expt 1: placed a hot soldering iron under the sensor for about 30 seconds (wanting to warm up the sensor without warming up anything else). Note that the amplitude drops by about 0.05mrad while the going (slope of the drift) is unchanged.
17 Dec 2010 [09:05]
It is good to see a prediction come true! Interesting to note, with a longer period of the door being open, that the amplitude looks to be increasing exponentially by about 0.4mrad at a time constant of about 20 minutes. The going does not show any decay - just a step of about -450ms/day. This reinforces the theory that the change in going is due to a change in length of the outer tube due to thermal expansion - corresponds to a temperature change of +1C. The puzzle to figure out now is a mechanism for the observed change in amplitude that corresponds to a change of +1C.
17 Dec 2010 [07:40]
I asked the bedmaker to leave the door open for 20 minutes today [link] . Another concern: is the barometric compensator working [link] ? The blue data points should form a horizontal line - but they are following the -8ms/day per mbar lines exactly as we had before the compensator was installed.
15 Dec 2010 [14:50]
ADJUST: +200 ms/day to 27400 - not getting this nailed as quickly as I would like. Hopefully then clock will drift toward zero over the next few days and then I will make a final adjustment in the new year.
11 Dec 2010 [07:40]
The Bedmaker Effect is absent - as expected, it is a Saturday. But I have a theory for what has been going on, and the sums sort-of stack up [link] . Essentially, warm air is forced into the pendulum chamber causing amplitude to increase due to reduced drag and going to decrease because the outer tube of the pendulum warms up - it takes a while for the temperature compensation to reach equilibrium. A test to do on a cold morning next week will be to leave the door open for long enough for an equilibrium to be reached.
10 Dec 2010 [09:30]
ADJUST: -900 ms/day to 27200 should stabilize now that the draught-up-the-pendulum problem is fixed
10 Dec 2010 [07:45]
cleaning lady leaving the door open. Hopefully over the weekend there will be no such glitches at 07h45, nor next week during which I have asked her to leave the door closed. Perhaps, as a treat, I will let her open the door next Friday!