rom: Bob Holmstrom 

To: psnlist 
Sent: Sat, 29 Sep 2012 23:56
Subject: re Something old with something new




Randall,

Thank you for raising the issue of mechanical properties again.  I would li=
ke to add to the argument that better mechanical design is important also. =
 The last years of my career were spent designing laser pattern generators =
used to make "masks" for the semiconductor industry and now I edit the Horo=
logical Science newsletter for readers interested in the science of timekee=
ping.  Both endeavours involve electronic and mechanical design.  It is int=
eresting to me to see how many newcomers to both endeavours want to "solve =
the problem" with sensor / electromechanical solutions.  Education on good =
mechanical design for precision mechanisms seems to be limited to a very fe=
w schools and to specialist symposiums.  My "clock nut" friends are more li=
kely to look for historical solutions than newly minted engineers even in t=
he day and age of Google which has made the information easier to find.

Hi Bob,=20

    Thanks for your Email. I always find the differing points of view of ot=
her people building similar types of equipment quite interesting !=20
    Sorry, but I doubt if either silicon or diamond could be of use to us. =
Seismometers need to be very stable, robust and shock resistant, so we tend=
 to use metal construction. They also have to be very sensitive, typically =
measuring movements down to 10, or even 1 nano metre.=20

Assembling a "stable structure" is an impossible task, but it is possible t=
o do better with good design.  A good overview of the required design and a=
ssembly techniques is covered in R. V. Jones "Instruments and Experiences" =
- fasteners though holes in one object attaching it to another object is no=
t stable.  Tutorial sessions at precision mechanism symposiums provide addi=
tional material. A Google search for "design guidelines for thermal stabili=
ty" yields a bit of useful information. Most if not all materials undergoin=
g a temperature cycle do not return to their initial dimensions!  Some mate=
rials do better than others.  Bob Matthys, in his book "Accurate Clock Pend=
ulums" (Oxford University Press) evaluated many materials using the period =
of a pendulum as the measurement means (useful to better than .01ppm) - The=
 best material he found was "aluminium silicon bronze" - e.g. it came back =
to its initial dimensions better than other materials.  Clearly a limited e=
xperiment, but a place where "amateurs" can contribute.
   =20
    Do you have the reference for R V Jones' paper, please ?=20
    Aluminium silicon Bronze has too high an expansion coefficient at 18 pp=
m / C Deg, for it to be easily thermally compensated, but it could be used =
as a component in a 'grid iron' type pendulum. I don't, at the moment, have=
 access to Bob Matthys's book to be able to comment further. $140 for 276 p=
ages is very expensive.
    Some suitable materials for very high precision clock pendulums are Inv=
ar, Super Invar (when stabilised), Zerodur glass, fused silica and carbon f=
ilament rod / tube. Two part Acrylic glue can be used for sticking glass, q=
uartz and carbon eg Holdtite ST3295. We don't need such accurate periods in=
 seismometers, but we do need very low loss / low error precision suspensio=
ns
    Check out the Littlemore Clock made by E T Hall http://www.hsn161.com/H=
SN/hsn_article.php It used a 2 second Invar pendulum and mass, a triangular=
 Agate bearing / Agate Flat suspension and it was enclosed in a high vacuum=
 chamber which was mounted on a 12 ton concrete base ! The mean total varia=
tion over 50 days was +/-10 milli seconds =3D 2.3 10^-9 error. He did have =
some problems in strong winds with a large nearby Walnut tree. Someone in t=
he USA bought it when Teddy Hall died. You might try to track it down ?

Much of the concern in seismometers seems to be the about the non-perfect n=
ature of springs (or other mass suspension pivots, etc.).  I have attempted=
 to suspend a pendulum so that it has the period vs amplitude predicted by =
theory.  It will not surprise you that springs, knife edges, rollers, etc g=
ive less than perfect results.  The best I have done is to use porous graph=
ite air bearings.  http://www.newwayairbearings.com/  By best, I mean that =
the period vs amplitude error ("circular error" in clock vernacular) was th=
e best.  An image of the pendulum and a bit of the results are at http://ww=
w.hsn161.com/HSN/airclock2.html  It is not clear that this form of suspensi=
on is of any use for a seismometer, but perhaps one of the long period pend=
ulum configurations could be implemented with air bearings.

    The suspension springs on vertical seismometers are very highly stresse=
d and the lever system makes the position of the mass extremely sensitive t=
o any small changes in the tension. The springs are composed of small hard =
metal crystals, in a semi random orientation and the displacement signal is=
 greatly amplified, so you are likely to observe signals if any slips occur=
 in individual crystals.    =20
    I can't think why you might have had any problems with crossed roller b=
earings. These and spherical bearings rolling on flats, crossed foils and m=
ore especially figure of 8 rolling foils, can ALL give excellent results. I=
 suspect that your suspension spring may have been too thick.
    The highest precision pendulum clocks  (Riefler, Shortt, Fedchenko, Den=
t) all use small, precise, stable angles of swing, to reduce any variations=
 in the 'circular error'.=20
    The Russian Fedchenko clocks use a three foil suspension layout to comp=
ensate for the circular error over small angles. Another method is to mount=
 two pairs of small magnets on the mass and the frame, so that increasing s=
wing amplitudes give increasing repulsion. The bob then needs to be magneti=
cally screened to eliminate interactions with the Earth's and other magneti=
c fields.=20

    The NBS found that when two precision pendulums of the same period were=
 mounted on the same wall (80 tons?), they tended to beat together and to i=
nterfere with each others' motion. To get them to beat totally independentl=
y, it was necessary to mount the swing axes at right angles. Is the high se=
nsitivity of pendulum clocks to the stability of their wall / frame mountin=
gs appreciated by your enthusiasts ?

Bob Holmstr=C3=B6m
Editor Horological Science Newsletter
www.hsn161.com =20

    Regards,=20

    Chris Chapman

=20
=20
=20
=20
=20
=20
=20






































































































rom: Bob Holmstrom <holm=
stro@..........>



























To: psnlist <psnlist@..............>














Sent: Sat, 29 Sep 2012 23:56














Subject: re Something old with something new
































































































Randall,





























Thank you for raising the issue of mechanical properties again.  I wou=
ld like to add to the argument that better mechanical design is important a=
lso.  The last years of my career were spent designing laser pattern g=
enerators used to make "masks" for the semiconductor industry and now I edi=
t the Horological Science newsletter for readers interested in the science =
of timekeeping.  Both endeavours involve electronic and mechanical des=
ign.  It is interesting to me to see how many newcomers to both endeav=
ours want to "solve the problem" with sensor / electromechanical solutions.=
  Education on good mechanical design for precision mechanisms seems t=
o be limited to a very few schools and to specialist symposiums.  My "=
clock nut" friends are more likely to look for historical solutions than ne=
wly minted engineers even in the day and age of Google which has made the i=
nformation easier to find.





























Hi Bob, 



























    Thanks for your Email. I always find the differing point=
s of view of other people building similar types of equipment quite interes=
ting ! 





    Sorry, but I doubt if either silicon or diamond could be=
 of use to us. Seismometers need to be very stable, robust and shock resist=
ant, so we tend to use metal construction. They also have to be very sensit=
ive, typically measuring movements down to 10, or even 1 nano metre. 





























Assembling a "stable structure" is an impossible task, but it is possible t=
o do better with good design.  A good overview of the required design =
and assembly techniques is covered in R. V. Jones "Instruments and Experien=
ces" - fasteners though holes in one object attaching it to another object =
is not stable.  Tutorial sessions at precision mechanism symposiums pr=
ovide additional material. A Google search for "design guidelines for therm=
al stability" yields a bit of useful information. Most if not all materials=
 undergoing a temperature cycle do not return to their initial dimensions! =
 Some materials do better than others.  Bob Matthys, in his book =
"Accurate Clock Pendulums" (Oxford University Press) evaluated many materia=
ls using the period of a pendulum as the measurement means (useful to bette=
r than .01ppm) - The best material he found was "aluminium silicon bronze" =
- e.g. it came back to its initial dimensions better than other materials. =
 Clearly a limited experiment, but a place where "amateurs" can contri=
bute.














    













    Do you have the reference for R V Jones=
' paper, please ? 









    Aluminium silicon Bronze has too high an expansion coeff=
icient at 18 ppm / C Deg, for it to be easily thermally compensated, but it=
 could be used as a component in a 'grid iron' type pendulum. I don't, at t=
he moment, have access to Bob Matthys's book to be able to comment further.=
 $140 for 276 pages is very expensive.













    Some suitable materials for very high precision clock pe=
ndulums are Invar, Super Invar (when stabilised), Zerodur glass, fused sili=
ca and carbon filament rod / tube. Two part Acrylic glue can be used for st=
icking glass, quartz and carbon eg Holdtite ST3295. We don't need such accu=
rate periods in seismometers, but we do need very low loss / low error prec=
ision suspensions













    Check out the Littlemore Clock made by E T Hall http://w=
ww.hsn161.com/HSN/hsn_article.php It used a 2 second Invar pendulum and mas=
s, a triangular Agate bearing / Agate Flat suspension and it was enclosed i=
n a high vacuum chamber which was mounted on a 12 ton concrete base ! The m=
ean total variation over 50 days was +/-10 milli seconds =3D 2.3 10^-9 erro=
r. He did have some problems in strong winds with a large nearby Walnut tre=
e. Someone in the USA bought it when Teddy Hall died. You might try to trac=
k it down ?




























Much of the concern in seismometers seems to be the about the non-perfect n=
ature of springs (or other mass suspension pivots, etc.).  I have atte=
mpted to suspend a pendulum so that it has the period vs amplitude predicte=
d by theory.  It will not surprise you that springs, knife edges, roll=
ers, etc give less than perfect results.  The best I have done is to u=
se porous graphite air bearings.  http://www.newwayairbearings.com/ &nbs=
p;By best, I mean that the period vs amplitude error ("circular error" in c=
lock vernacular) was the best.  An image of the pendulum and a bit of =
the results are at http://www.hsn161.com/HSN/airclock2.html  It i=
s not clear that this form of suspension is of any use for a seismometer, b=
ut perhaps one of the long period pendulum configurations could be implemen=
ted with air bearings.



























    The suspension springs on vertical seis=
mometers are very highly stressed and the lever system makes the position o=
f the mass extremely sensitive to any small changes in the tension. The spr=
ings are composed of small hard metal crystals, in a semi random orientatio=
n and the displacement signal is greatly amplified, so you are likely to ob=
serve signals if any slips occur in individual crys=
tals.     













    I can't think why you might have had any problems with c=
rossed roller bearings. These and spherical bearings rolling on flats, cros=
sed foils and more especially figure of 8 rolling foils, can ALL give excel=
lent results. I suspect that your suspension spring may have been too thick=
..













    The highest precision pendulum clocks  (Riefler, Sh=
ortt, Fedchenko, Dent) all use small, precise, stable angles of swing, to r=
educe any variations in the 'circular error'. 













    The Russian Fedchenko clocks use a three foil suspension layo=
ut to compensate for the circular error over small angles. Another method i=
s to mount two pairs of small magnets on the mass and the frame, so that in=
creasing swing amplitudes give increasing repulsion. The bob then needs to =
be magnetically screened to eliminate interactions with the Earth's and oth=
er magnetic fields. 



















    The NBS found that when two precision pendulums of the s=
ame period were mounted on the same wall (80 tons?), they tended to beat to=
gether and to interfere with each others' motion. To get them to beat total=
ly independently, it was necessary to mount the swing axes at right angles.=
 Is the high sensitivity of pendulum clocks to the stability of their wall =
/ frame mountings appreciated by your enthusiasts ?





























Bob Holmstr=C3=B6m














Editor Horological Science Newsletter














www.hsn161.com  <=
/a>



    Regards, 



    Chris Chapman