]\gt ; : [ 1055 ] XXV .
On the nination othe of the rnometer by Experiments with \mdash ; Part .
T. R. ROBINSON , D.D. , F.R.S. , Received June June 17 , 1880 .
preceding Part Philosophical I gave the results obtained by anemometers attached to a whirling machine , which substitutes motion through the air for real wind .
If the air were quiescent this method would be nearly unexceptionable ; but the whirling gives the air a vorticose motion for which it is impossible to make an exact allowance , and therefore some uncertainty these results .
In the conclusion of that paper I expressed an opinion that greater certainty might be obtained b.y comparing two anemometers , similar and equal every respect except friction ; and stated that I would endeavour to carry this into effect .
I propose now to give an account of my attempt to do so .
( 56 .
) The instruments used , and their alTangement , are described in paragTaph ( 51 ) .
The situation in which they are placed would be a good one but for the dome of the west equatorial , which in some points of the wind may interfere with its full action on one or the other of the instruments .
The diameter is ; the height of its summit above the platform is ; that of the arms of the instruments being 16 ' .
The horizontal distance of its centre from the Kew instrument , bearing from it S.S.E. , S. The distance from the experimental one ( E ) is 23 ' , and its bearing S.W. .
S. The distance between and .
Of course when the wind is S.S.E. , will be less acted on than , and vice versed , but probably the difference will be much less than that caused by fluctuations of the wind itself .
When the wind is E. W. the eddies caused by the windward anemometers may perhaps reach the leeward , but ) I think , to any great extent .
( 57 .
) The chronograph record of each experiment at first entered in groups during which , the velocity of , was nearly uniform ; and , the number of turns made by each ument , was an integer .
The length of the chronograph helix gives the time ; it is measured in eighths of an inch ( as the Obsel.vatory possesses scale of eightieths ) and when divided by the length of a second on the same scale , we have the number of seconds .
As the chronograph in its present situation is exposed to considerable variations of temperature , its rate is not as regular as it at Rathmines , but the second-space was determined each day of observations .
The average in * For facility of the numeration of the graphs and tables is in scqucncc to of Part 1 .
DR. T. R. ROBINSO ON THE DETERMINATiON winter is ; and ' times so deduced are certain to less than .
Latterly the time was noted by a watch .
( 58 .
) It was soon found that the method proposed in paragraph ( 52 ) is not available , for the wind is never uniform long enough to make two successive experiments fairly comparable .
It was therefore necessary to use that of paragraph ( 53 ) .
Assuming such values of , and , the constants of equation III .
as will give V very nearly equal to V ' ( the accented letters belong to E ) , we may correct them so that the mean V'\mdash ; V may vanish .
This assumes first , that however the wind may vary in the course of an ation f one instrument to another , yet if the time be sufficient it omes to each of t.hem with an equal amount ; its deficiency at one part of the time being made up by its excess at another ; and secondly , that the V computed for a mean value of will be its own mean value .
( 59 .
) As to first of these assumptions , I have come to the conclusion that if an observation lasts for nine or ten minutes , the average action of the wind on the two instruments will be nearly equal , though during portions of the time it may vary very much .
This may be illustrated by the following table , which contains a set of and taken with the normal frictions at and , which are and ; these were taken September 17 , under unfavourable circumstances , for the wind was S.W. The and ought to be nearly equal , for the difference of the friction will only diminish by TABLE Total time ; mean ; mean ; mean These show plainly both the variation of wind at one anemometer and the difference at the two .
In No. 14 , ; in No. 10 , it is .
These represent and .
If we look to the column , at No. 35 we find , at No , 62 THE rANTS OF CUP ; fourteen , and nine are tive .
But if we divide theIn into four consecutive and nearly isochronal groups the discordance is less .
8 .
Time 5 .
4 .
1.133 The extreme range here is instead of , grouping them in pairs There can be little doubt that the total means nearly correct , these values of differ from the mean one by and .
In general , be less than this ; and if it be observe ] by inspecting the chro1lograph while cm vation is proceeding that the ratio of A to varies notably : a longer time should be taken .
( 60 .
) As to the second point it is easily shown that no great errors arise assuming that V is truly given by the mean .
The mean V of a sel.ies is , taking the time into account , SVT Now the first of these .
In ments like where is small , if we develop the radical in powers of , the second telm becomes and as the term may be neglected the mean of radical velocity .
When is large the simplest course is ( calling the radical R ) to compute , or what comes to the same thing being the time-space , and this with the computed with the mean .
Taking at ndom No. X. of } .
whose , we have for the separate )'oups whose A and are nearl .
uniforlt ] Here the do not range widely , and I take September 16 , 1878 , .
T. R. ROBINSO IIIIII .
IVXIVIII.ction Even here the is not of a nature to interfere with the determination of the constants , though in such work terms like V. and XI .
had better be omitted .
If it were thought necessary the exact computation is not difficult .
( 61 .
) At first the additional friction was applied by the brake-levers , and was measured by the process described in the note to paragraph ( 19 ) ; but it was soon found to be irregular on account of the rusting of the cast-iron disc on which the rubbers pressed .
This could not be prevented in the present location of the instru- meant .
The rust wore off in the course of an experiment and filled the pores of the cloth on the rubbers .
Yet more , it became evident that the constants which in the whirling xperiments had V\mdash ; W pretty fairly , fail totally here : for instance , with the set last given they give , the difference being far too great to be caused by any error of the friction .
( 62 .
) I intended to remove the uncertainty caused by the rust by substituting for the iron disc one of bell metal of the same diameter ; it is , however , some 20 oz. heavier , and the normal friction of is now grains , and its moment But while it was being prepared it occurred to me that instead of measuring the brake friction first and assuming its permanence during a series of observations it would be better to record and measure it the entire time of each observation .
PRONY'S brake afforded a ready means of effecting this , and was thus applied : a ring of iron an inch deep and inch thick is divided into two semicircles held together by screws tapped in the lugs ; when these are removed it can be got on the axle , owered to the disc , and is made to clamp it with any required pressure by relylacing and tightening the screws .
The ring has an which carries an arc of the same depth concentric with the disc and of radius .
It is obvious that when the anemometer is turning , a cord attached to this arc will be pulled by a force moment of the ring friction at 8 This pull is measured by a spring balance which I made with one of the olock rings described in paragraph ( 26 ) .
THE CONSTANTS OF THE CUP ANEMOMETER .
63 .
) It consists of an iron spindle long , turning on a small toe below , and above in a brass collar carried by a transverse piece of wood supported on two uprights .
It has a projecting piece to which the inner end of the spring is attached by a screw secured by a check nut ; the outer end is fixed to one of the uprights .
On the top of the spindle is screwed a disc of mahogany diameter and thick , on the edge of which is turned a groove in which the thread that connects it with the brake arc is wound .
On this disc is fitted one of the papers used with my original anemometer , which has its circumference graduated to half degrees , and is covered with circles apart , every tenth one stronger than the rest .
By pulling the cord the spindle is turned and the spring tended , the number of its revolutions is shown by a tell-tale fixed to one of the uprights , and the degrees by a pointer .
( 64 .
) It is thus used : tighten the clamp screws so that when the arm is held fast the anemometer shall turn without coming to a stop ; pass the cord of the balance through a hole in the remote end of the arc , and tighten it till the increased tension keeps the arm nearly in the same position , then secure it to a pin provided for the purpose .
In state of things it is evident that the tension corresponding to the angle through which the balance has been turned , is the moment of the friction at 8 from which the moment at the cups is known , to which must be added the friction .
The brake-ring weighs 14 oz. , which would increase the friction a few grains , but this was obviated by hanging an equivalent weight on the relieving apparatus mentioned in paragraph ( 51 ) .
The ring was at first lined with cloth , but as it slightly abraded the bell-metal , I removed it and used the iron surface icated with lard .
( 65 .
) The relation between the tension of the spring and was thus obtained : the balance being clamped to a table its cord was passed over a pulley ; nineteen weights in regular succession from 2 oz. to 36 oz. were hung to it ; and to eliminate the effect of friction the disc was turned a few degrees in advance and in rear of the positions of rest , when they were attained , the mean of the 's was taken as that due to the tension .
From ten to thirteen sets were taken for weight .
I had expected that the.tension would be very nearly , , but this spring such is not the case ; at the beginning grains , at 4 rev .
it , and the change is not uniform ; so I formed from the series an interpolation table with argument , from which is easily computed by a formula analogous to that given by STIRLING for equal intervals .
In carrying out the work I was met by an ullexpected difficulty : friction applied in this way is not constant ; and I found that in strong breezes ( when the wind is always fitful ) the arm oscillated more than , the utmost range which the opening of the iron box ( paragraph 2 ) permitted unless the friction was reduced .
Tbese oscillations made it necessary to have a record of the tension , which was provided by clockwork moving a pencil from the centre to the circumference of the paper at the rate of per minute .
This traces an irregular sector from which the mean is easily obtained .
But , besides this , it is necessary to reduce the oscillations below , so that they all may be recorded .
This was effected by connecting with the an auxiliar balance , so that its action would begin only at the minor limit of the arm 's motion , and increase , so as to prevent it from reaching the major limit .
It consists of an iron tube diameter , containing deep of mercury .
In this is immersed a rod of iron diameter , reaching to the bottom , and with a cross piece at top resting on the tube ; from this cross piece descend two ires carrying weights just sufficient to balance the flotation of the mercury .
To the top of the rod is attached cord , passing over a pulley to the arc .
It is easily shown that if the rod be raised an inch the cord will be pnlled with a force-weight of a cylinder of mercury diameter and high .
( For this also I formed an interpolation table , but in it is nearly insensible .
) To use it the spring balance is tended till it keeps the arm near the middle of the opening of the box ; the arm is then pressed back to touch the box , the cord is looped on the pin aiready mentioned and shortened till the cross just rests on the , and the read which gives the zero of auxiliar .
Deducting this from the mean , we have , the measure of the auxiliar tension .
* The largest oscillation which I have observed under this arrangement was ; the wind was moderate , V being only .
This is equivalent to a change of tension grains at cups , nearly of the entire tension there .
I cannot account for the great ularities of this friction , but I believe similar facts have been observed on a large scale in applying PRONY'S brake to machinery .
The extent and frequency of the oscillations do not seem to follow any regular law of V or , though they evidently are related to them .
( 67 .
) The process described in paragraph ( 58 ) gives for each observation an equation containing three unknown quantities , , and two unknown variables , V and V ' , or being the difference of wind at the two instruments .
It is shown by Table .
that may be considerable for a few seconds , but when the time is a few minutes it is probably confined within the limits .
Even when ( as in the whirling experiments ) we know V approximately , and have not V ' to consider , the coefficients are so related that it is impossible to get accurate values of the constants by the usual methods of elimination , and here the difficulty is still gleater .
I have therefore thought it best to assume probable values for and , and .ermine so that the mean may vanish .
In the first approximation to this , supposing the true value of the wind at , we have being Hence .
* A far better mode of retarding the motion of would be to have on its shaft a sheave connected with an apparatus like that described in raph ( , so that the instrument in revolving might draw up the driving weight .
The moment of this at the cups wollld be constant and accurately known , and the observer would only have bo continual ] unwind the cord .
Unluckily , this did not occur to till the series of Table XXI .
was nearly completed ; and I was unwilling to repeat the measures , for , owing to deficiency of wind , that series had occuI ) several months .
CONSTANTS OF CUP If the number of observations be sufficient , , and we have .
This will give an nearer the truth ( not exact unless be insensible ) , and a second computation will in general be sufficient .
When the constants give the mean , the V so obtained must be very nearly , as shall be shown ently .
( 68 .
) First as to : in the case of , we have the measure of it given in paragraph ( 27 ) .
These must be reduced by some hypothesis as to the action of frictio the first part of my paper I assumed that the momentum of the cups carried them past the point of equilibrium , induced to this by the small value of given by nlin .
squares ( paragraph 39 ) , .
This gives .
My preliminary work with the two instruments showed that this was too small , I recurred to the more natural supposition that the cups stop when the wind 's force This gives at Bar .
and Therm. For cups it is , very nearly in the proportion of he areas .
I know no means of determining whether this constant varies with ; the individual measures seem to show that it does not change with .
The lateral pressure on the upper bearing of the shaft causes a resistance as , and will diminish ; but the probable value of its coefficient is , which may safely be neglected .
The change of , if it exist , cannot have much influence on V ; for .
Taking I. and X. of Table XXI .
, whore is a mini m and maximum , we have and ( 69 .
) As to , if there be no resistance as , except what appears in the the equation of motion is , from which we see that , the coefficient of .
This is its major limit ; if we by , the equality of and V ' may still be preserved by diminishing .
But the value of is a little decreased : so the , or , in the case of Such diminution can only be affected by an expenditure of power in driving .
before the cups , or throwing it outwards ; and I tried to find a limit by making them revolve in quiescent air .
For this purpose I mounted four forms of on a vertical spindle driven by HuYGHEN'S maintaining apparatus , and noted the time moment at cnps .
The resistance was always more than twice the action of direct wind on the convex sides , and I think its excess may be taken as the extrerne possible value of in the * I tried to measure it by the spring balance , but the oscillations were too extensive to permit any continuous observations .
By noting the time , and counting the turns of while the ions of were clear of the sides of the box , I got values of and ; but the possible ; of wind must be remembered , negative direction .
It would give for ; but I think this action must be small in a current of wind moving with nearly three times the velocity of the cups .
It is found to increase as the diameter of the cups and the length of the arms diminish ; for it gives , but for ( to be soon described ) .
This supposition would give smaller values of ; for No. II .
of Table XXI .
, where V with is , the difference is , and the true value is certainly between these .
I will use as certainly known .
( 70 .
) Fox , as and are similar and equal , it must be the same in both , and the means of obtaining it are explained in paragraph ( 67 ) .
Here I need only show how its first approximation is got : Supposing , we have ; but as in is small , and may here be neglected , we have , and the sum of the equation becomes ( VII .
) from which , is easily found .
When is not similar to the process is simpler : the reading of gives V , and we have , whence ( VIII .
) Both these fornuulae are defective from omitting , but are near enough to begin with .
( 71 .
) The following table gives the results of the comparison of and , which is equal similar to K. The second column gives the wind 's direction ; he third air 's density ; the fourth the time in seconds ; the two next A and , the number of turns made by and ; the seventh of or ; the two next the velocities of the centres of and ; the two next the computed velocities of the wind ; and the twelfth V'\mdash ; V. V and V ' were computed by the formulae ; V ' when CONSTANTS CUP These were computed with and ; , which being divided by , we have ; ; limit of For It is evident from the values of V'\mdash ; V that the constants do not ) ' with or ; but that their differences are casual owing to difference of wind at the two instruments .
They differ when the 's are nearly equad : For instance , I. VIII .
differ by .
and XIV .
by 1441 , and IX .
and X. by ; and that such differences of wind may exist for some time is shown by .
where during the first 321 seconds , and during the it is The minus values predominate during S.W. as might be expected from paragraph ( 56 ) .
This and are larger than those given in paragraph ( 40 ) , mely , and , whioh give for the limit This is due partly to my having then used an only two-thirds of I believe to be its real value , partly to the uncel.tainty of the fi.ictions employed of , and partly to the defect of the method of minimum squares in such a case .
* As these ihstruments are generally constructed to register , their roadings should cted by subtracting of the recorded Reducing the first 21 of Table X. by formula XIII .
, and with my present values of and I get , and the limit The 's used in computing these constants were certainly inaccurate .
I measured them in the plane of the centre of the anemometer , but as the disturbance of the air will be as must be less in the upper semicircle than in the lower , while it acts with less mechanical advantage in lessening .
It must also be kept in mind that any measure of is an average one , and that it may have very different values in parts of the air vortex .
( 72 .
) Ill , the cross remaining the same the cups were set at from the axis ; it is my No. III .
In it the constant for is half that for , and the normal friction is double .
With the approximate and , the results are given in the following Table , in which the densities are omitted as involved in TABLE XXII .
S(V\mdash ; V ' ) which divided by Se ' , we have , and the limit .
These are larger than those of The results obtained in paragraphs ( 38 ) and ( 39 ) would make it less , but in the present work it is the rule that diminishing the length of the arms increases ( 78 .
) In the inch cups were fixed distance from the axis : too near for good work , but I wished to see the effect on .
With its approximate values and , I computed Table XXIII .
THE CONSTANTS OF THE CUP ANEMOMETER .
The sum of .
Hence and the lirpit .
The residue is a little too ; but I did not think it necessary to pursue the approximation farther .
The great increase of is remarkable , and I think shows that when the cups are so near the axis of rotation they disturb the regular action of the wind .
Even with the arms this effect is sensible .
( 74 .
) .
I now fixed the cups on the cross at from the axis .
angement is similar to the cups at 24 and I thought that the same might serve for both , but it was far otherwise .
The measures in paragraph ( 27 ) giye for cups at the normal pressure and temperature .
For the first ten observations ; but as these are oz. lighter than the ones , I thought there was too little pressure on the toe , and changed the relieving weight from the .
to .
This made the friction .
I computed with and Table XXIV .
NDCCCLXXX , 6 X TABLE XXIV .
The sum of V\mdash ; V ' and the limit .
This and shows not only that anem same constants , but also that length of the arms ; for here it ( 75 .
) .
The cups were 4 from the axis being 26 75 ; 2 be less .
At first it was mount that I thought its friction mounted it on the spindle ( VI .
, XXIII .
) computed Table Sc ' , so excess of above that of is very remarkable , neters must be equal as well as similar to have the depends on the diameter of the cups as well as the , greater than in , though the arms are longer .
as far out on the cross as possible , the distance greater than that of , and I expected its would on the axle of , but it moved so much slower than was too much for the small cups .
I therefore mentioned with friction , and with its results XV .
TABLE XXV .
* Time short because it was al the end of the chronogra : ) sheet , CONSTANTS OF THE CUP ANEMOMETER .
omitting the first five , Se ' .
Hence and and and the limit .
This result surprised me , for the friction was so small that no irregularity of it could have any sensible influence , nor does it seem probable that the pressures on the surfaces of the two sets of cups are in any other ratio than that of the surfaces .
The is actually larger than that of The five first were computed with the final .
They give V\fnof ; m rather too snuall , but in three of them the wind was S.W. ( 76 .
) .
I now placed my old anemometer , cups 12 arms , on the axle of E. The of these cups ( if as their area ) and their .
With the second approximation , , I recomputed the and of Table XXVI .
TABLE XXVJ .
It will be observed here , as in Table XX .
, that is sometimes greatel , sometimes less than ; the near equality of the constants of the cups to those of makes the irregularities of the wind manifest .
The giving .
and double weight ) , therefore and the limit .
This is a little less than that of ; which shows that fluence of the diameter of the cups is felt even here , oyerpowering the effect of the ( 77 .
) I shall conclude with a few remarks on the preceding results .
The process by which the of is determined seems liable to but two ctions : *In my paper ' On the Cup Anemometer ' ( Trans. R.I.A. , vol. xxii .
, p. 170 ) I have mentioned some trials to measure .
As the 's given there were doubtful , I have recompnted them with constants and the friction of that instrument .
The six give for ( at llormal pressure and temperature ) , agreeing fairly with that given in the text .
I may mention here , as further proof of the unsteadiness of the wind , that on one two cups of this anemometer so that all the convexes were opposed to the wind ; I remain at rest , but they were in continual oscillation through many degrees , so that in limitod there must have been differences of able to overcome a friction of 53 grains .
the assumption that , and the possibility that the wind errors are not eliminated .
As to the first , I have shown in paragraph ( 69 ) that it cannot be far astray ; even were the extreme diminution of it which is mentioned there to occur , the error for would only be ; but such is not likely to be the case , and may be accepted as a major limit and one not far from the truth .
As to the second it is certain that in a sufficient number of observations the \mdash ; errors must balance each other ; but it may be a question whether the XXI .
of Table XXI .
were enough .
Still , it is evident , from inspection of the column that there cannot be any large outstanding residue .
I have pointed out the defects of the situation .
Could I have had my wish I would have established the instruments on spars 20 ' high , erected on a level ground away from any influence of houses or trees , and used a better mode of applying friction to .
I also regret that no strong gale occurred during these experiments to verify the formula for a very large .
Under favourable circumstances , I think this mode of determining the best that is known .
I have stated reasons for distrusting the results obtained by the whirling machine , and as yet no unexceptional mode of carrying an anemometer through the air in a right line has been devised .
Even could we get a locomotive which could travel without disturbing the air through which it passes ( and perhaps the new electric motors might effect this ) , and a line of rails certainly screened from wind , there would still remain the doubt whether the pressure is the same when a body is moved through a quiescent fluid or a current impinges on the body .
( 78 .
) The results obtained with other anemometers show that is a function both of , the length of the arm , and , the radius of the cups .
I subjoin its values .
15@80 2 2 If we take Nos. 2 , 3 , and 4 in which the cups are equal , the dependence of on is manifest .
In No. 3 it larger than No. 2 , and in No. 4 .
This is partly due to the air 's escape before the convexes being less easy as the circle described by them is less .
Such a fact is strikingly shown by the whirling experiments ( paragraph 69 ) which I made in search of a minor limit for , in which I found the resistance and 79 for the three respectively .
This was in quiescent air ; but a similar though much smaller effect must occur in the actual working of an anemometer .
Its influence can be obtained only by experiments , such as the present .
thought to test this by removing two opposite cups in .
As in this case there is only one cup oaoh semicircle at a time , the probability of their mutual disturbance was small .
A set of ten gave the HE CONSTANTS OF THE CUP IETER .
( 79 .
) It is more difficult to account for the similar dependence of on the size of the cups ; priori , there seems no reason why small cups should be more resisted than large ones , but such is evidently the case .
Unluckily I did not place the and cups at the same distances as the 9 that the effect of the cups on is mixed with that of R. I tried to eliminate the latter by inberpolating for the values that the would have at the 's of Nos. 1 , 5 , and 6 , but this co not be done very exactly from the three values .
However , this gives for the less than for the ; for the 4 in No. 5 , greater , and in No. 6 , .
The only way in which I can conceive the possibility of such an occurrence is the existence of powers of and in the factors , which express the mean effects of wind on the concave and convex surfaces of the cups .
In equation IIT .
I suppose the mean to be that of the centre of the cups , and that the mean impulse and resistance act at these points .
But this is not necessarily the case .
The effect of the resultant on an element of the cup is ( 1 ) the square of that resultant ; ( 2 ) as the perpendicular pressure on the element ; ( 3 ) as the resultant of that pressure perpendicular to the plane of the cup 's mouth ; ( 4 ) as the distance from the axis at which the projection of that resultant meets the arm ; and ( 5 ) as the magnitude of the element .
Of these five factors the first contains and Now as the element which contains and also enters the fifth , so that the difFerential may contain and .
As to the second we ignorant of its formula , and it is pretty certain that it will depend on powers of sine and cosine of incidence and ( at least the concave ) on the curvature .
If we knew its exact form we could integrate the differential which they form get the impulse resistance for a given , and multiplying this by , and again integrating from to we should find their mean values .
Of the terms in this last integration those which have as a factor disappear ; ( the surface ) will be a factor of the others , among which be the three first powers of ; and these may produce the change of ( 80 .
) In paragraph ( 41 ) I inferred from the work with the whirling machine that with cups the is the same for and arms ; but what precedes shows that this is not the fact , and that each type of anemometer has a special .
1 would therefore suggest to meteorologists and opticians the propriety of confining themselves to two types : one for fixed instruments , the other for portable ones .
the first the Kew type should , I think , be adopted ; if the determination of its constants , given in paragraph ( 70 ) , be not thought sufficiently exact , there would be little difficulty in making more obseryations like those described there , and under less than in No. 4 , but so large as to make it evident that there must be some ausc of the in xeased value of .
T. R. THE favourable circumstances ; but I think they would make very little change in my number .
For the portable instrument , the only one of which I have expenence is CASELLA'S cups and arms , and I found it very convenient : its should be determined as above .
Some such arrangement seems necessary to ensure a uniformity of velocity measures .

