One of the first things in science, or any other matters disseminating information, is the open assumption that all measures and statements have been produced in such a manner to avoid any bias or dishonest practices. Here we live in the true hope that observers, both present and past, are not giving any deliberate errors. We do normally assume, and have faith that these kind of mistakes are just unfortunate transcription errors or widespread slip-ups. If such dishonest methodologies were indeed found to be true, this makes the task of sorting out real changes in true double star positions or other related information an absolute nightmare! (Let us here assume that it is not!)

The wholesale problem of finding and understanding various discrepancies within any double star data is certainly a difficult and challenging one. As a rule-of-thumb, observers should always check and question their data using AT LEAST three different reliable sources, and if anything else, as many sources as practicably possible. I.e. Use original sources, say from the; IDS 1961, WDS 1984, WDS 2001, WDS 2003 WDS 2006 and the latest, WDS 2011.

Alternatively use the original published measures, or sensible or logical options, taken the expertise in from the many catalogues, journals or even other common independent references.

There is much work that can also be done by detailed investigations through all the available catalogues. Fortunately for the observer, I so think the most serious problems existing say twenty or thirty years ago have already started to fade. Yet in regards many of the lesser issues there is much an visual observer can do on cloudy nights, which can usefully contribute to the science of double, multiple and binary stars. Most do not require vigour mathematical or astrophysical expertise and know-how; but often only requires clearing the head, applying some common sense and the acquired ability to see the unusual or incorrect information. Once found, these can be notified to the appropriate astronomical authority so the same pitfalls can be avoided into the future or assessed in some distant future time or date.

To explain this concept better, perhaps it is best to speak from some useful experiences — using my own perspective as a southern observer over many decades. In my humble view, experience has regularly revealed many shortcomings in common data or of the chronicled available measures. I also think this same judgement also probably applies equally to the northern pairs as well, even though the historical basis and longer history might not pose exactly the same troubles, problems or issues. So it is likely best here to talk from my own experiences and give some examples, labels, and discuss my own standpoint.

Some Southern Experiences

We find much of the earlier visual pair data was then based much on the data within the Index of Double Stars (IDS) (1963) and all the data quoted, in say the three volume Burnhams Celestial Handbook or E.J. Hartungs Astronomical Objects for Southern Telescopes (AOST1) (1968), etc. For example, the majority of Hartungs AOST1 pairs are measures made in 1960, though some of these quote the earlier measures. Today (2012), this is now fifty years or more years ago!

Even today, many common texts on visual double stars still gives this now almost antique data. In the southern pairs, much of the information of these works usually come from one single source — the R.T.A. Innes great southern work; Reference Catalogue of Southern Double Stars, published in 1899 in the Annals of the Royal Observatory at the Cape of Good Hope. Here Innes sets out the general framework of the behaviours and available observations made prior to 1899. This 316-page document has most of the common southern pairs with useful and critical evaluations of positional or scalar errors, gravitational association, colours, magnitudes, with many general relevant historical statements of interest. It contains 2,140 pairs listed in right ascension order by hours, all roughly south of −25° declination. Several important updates appear in the first two decades of the 20th Century. These texts then set the scene for the next sixty or seventy years. When you read either Burnhams or Hartungs, you become quite acutely aware of the origin of data and comments they are stating> Hence, the distinct limitations when observing the pairs for yourself. As time progresses, these older discrepancies become even more obvious.

Although both these amateur references became very popular, especially due to the glorious writing style and overall authoritativeness, and moreover, the sheer number and variety of objects, they also do have their drawbacks. As you may be already aware, this was at the time about these were the only useful common references available. For double stars, the main detraction was often just using two isolated data points to access variations or changes.* This often meant that when, say, trying to predicting any future (or past) separations and/or position angles, become or was nearly always fraught with some kind of intrinsic or intractable problem. In some cases they are just plainly wrong.

* Historically, for the southern observers, much of the modern double star data was produced during frantic three periods. These were between the 1890s, around about 1928 and then again in 1955. In roughly the late-1970s these problems were decidedly quite notable when looking through the telescope. This was because many of the discussed pairs were significantly out in their position, brightness or against the current published data. Furthermore few data were available to find those pairs who needed closer inspection or had significantly moved.

In the example of the mainly wide Dunlop pairs, which Ive been slowly studying and reassessing, we find more often than not, that the data available on these were and remain very poor. It was not usual, for arguments sake, that the last observation being made on some Dunlop pair was something like 1910 or 1928! This was during a period for observers that was more than fifty years later! Needless to say the information was an observational minefield!

Southern Double Star Examples

Perhaps the classic example of all were the standard pairs presented in the annual BAA Handbook and many standards of systems with presumably known widths — specifically advising for calibrating simple types of double star micrometers. This was with the notable southern double placed at the apex of the Southern Cross, γ Cru / Gamma Crucis / DUN 124 AB (1.83V & 6.45V magnitude) with the quoted width of 111 arcsec. PA 31° as measured in 1919. After the WDS 2001 the separation was listed as 125.4 arcsec PA 26°, using the Hipparcos data of 1991. Because of Dunlops reputation for having poor separation and PA measures the 1826 observation of (93.4″, 46°) was declared as plainly wrong. To show this dramatic effect of this conclusion, some useful references like Megastar 4.0 (for example, published in 2000) still used the 111 arcsec value of 1919! (Fixed in Megastar 5.0) A micrometer calibration would have had its own R-value consistently about 12% too low! Hence, every future measure using this as a calibration star would be out by this constant amount or, if the observer ended-up using several different pairs, would be left with just confusing estimations of this R-value.

Since about 1984, as stated in the introduction of the Washington Double Star Catalogue (WDS), the general detail available about many individual pairs —both north and south — has increased enormously. In  Gamma Crucis case, perhaps the most significant changes of all were the 1991 positions from Hipparcos astrometric satellite — giving the additional useful stellar parameters (other than just Separation and P.A.) including accurate positions of the individual stars, the sometime individual parallax(es) and the proper motion(s).

I.e. Had the proper motions been known for DUN 124 AB of A=+028 -264 and B=+005 -014 — it became clearly obvious that this optical pair was far from fixed. Even worst, in the main 2000 epoch FK5 frame reference, the PPM (South) (Positions and Proper Motions) star catalogue, the immediate successor to the SAO (Smithsonian Astrophysical Observatory) star catalogue, had assumed the measured proper motions for A and B were exactly the same!

Comments on the Changes in Current Versions of the WDS

One of the biggest difference is between the different versions of the WDS. This catalogue is now being updated continuously — all without formal notification — and it is important to state the Version being used. (I have used things like WDS03 or WDS03 Sept, but the USNO never gives such changes in these terms. Perhaps we should request or just use a version number base on the download date? I.e. WDS031510 (year, day, month).

Users of the WDS will find significant differences with the data in say the still standard WDS84 — and still information from here is often quoted in many software and written texts. I have found that I have had to use two of the latest WDS (WDS12 June) AND the WDS03. The reason is that the USNO has changed the FIRST and LAST DATE of OBSERVATION.

I.e. HJ 4573 (See Example 1 Below.) The WDS 01 (and WDS 84) gives the positions for 1913 and 1924, while the WDS 03 gives the positions of 1834 and 1924.

This can be even more confusing when the FIRST DATE is earlier and the LAST DATE is later than the previous version of the WDS !


Below are several different examples I have been working on. They are just mainstream investigations from the commonly available information or data. (Each have been added to an observing list for further study.)

Example 1

HJ 4573 13137-5616 (Cen)
10.0 /11.0 mag.   
HJ 4573 (7.6″, 57°) (1924) WDS03
HJ 4573 (7.7″, 57°) (1913) WDS84
HJ 4573 (4.0″, 57°) (1834) WDS03

This pair looks like it should by in the order of 10 arcsec along PA 57°. HJ 4573 is unresolved in GSC and single in both the Tycho and Hipparcos catalogue.

Conclusion: This pair requires new observations and
should be added to a neglected list.

Example 2

HJ 4590 13332-7734
Cha (6.5 and 9.3 mag) F6V 
1837  137°  25.0″  WDS 03  
1880  134°  22.4″  WDS 84**
1991  133°  22.4″  WDS 03  
2000  133°  22.1″  WDS 12  
NOTES: HJ 4590 A is variable S Cha, B CPD-76°769.
** Russell 1880.413 134.9° 22.41 HJ 4590 A: pmRA = -367 pmDec=-153 mas per year HJ 4590 B: pmRA = -351 pmDec=-118 mas per year HJ 4590 A: Parallax = 28.45± 1.13 mas HJ 4590 B: Parallax = 52.09±39.91 mas * mas (milli-arcseconds)

Here the WDS84 only gives the 1880 position and nothing else but n=3 (Number of observations), while the WDS03 has fixed up the problem by quoting the first and last observations 1837 (J.Herschel) and 1991 (Hipparcos) n=11. In WDS12, n=13.

This is likely an optical pair based on the differing pmDec. The problems with the error in B parallax makes this conclusion uncertain. Using the primarys parallax of 28.45±1.13 mas. gives the distance as 34.14pc. or 114.6±4.56 ly. At this distance, the stars true separation of 25″ corresponds to 878 AU (1.3×1011 km) and an estimate maximum period is about 1.8×104 years. Using the 6.5v and 9.3v component magnitudes, the absolute magnitudes are about 3.8 and 5.6. Using the Mass-Luminosity Relationship (MLR) the solar masses are about 1.2M⊙ and 0.7⊙, respectively. (Agreeing well within the given Spectral Type)

Theoretical Catalogue Limit (TCL) Heinze (1979) for attachment is 173″ at 34.14pc. Where;

TCL= 10^(2.8 − (0.2 × Δm)) or

TCL= 10(2.8 − (0.2 × Δm))

Assuming linear motion, the prediction for this pair in the near future is;

YEAR  Sep ″   PA (°)
2000  21.93   133.83
2005  21.84   133.73
2010  21.74   133.63
2015  21.65   133.53
2020  21.55   133.44
2050  20.99   132.85
2100  20.05   131.88

Visual observation on 23 June 2003 using 10.5cm. f/11 (190×) estimate 21 arcsec and 134°.

Conclusion : Confirms the pair is acting as expected.

Update Comment: October 2008

Since writing this text in 2006 the U.S.N.O. has published on its site the preliminary Catalogue of Rectilinear Elements of many such optical pairs or systems. This includes notes and a useful ephemeris prediction for the next twenty-five years. Using the resources as the repository of double star measures, these linear elements are determined by weighting the various available measures based on the accuracy of each known measure.

This is far more detailed than the example here, but it does emulates the way for assessing which pairs to measure or study.

Example 3

If I find significant change that cannot be explained, I get all the positional information from the USNO and plot all these values against the one in question. For a broad example, at the moment I have found some problems with the H.C. Russell pair R71 AB;

R71 AB / SLR22 BC (07144-4440) is a white triple star some 8.6′E (PA 99°) of L2 Puppis. R 71 AB is the brighter of the two that contains the 9.6v and 10.2v (9.61V & 9.77V) magnitude stars separated by 15.9 arcsec along PA 256°. R71 suggests it has shown some change in the positions, reducing in PA from 288° to 256° and 18.4″ to 15.9″ in the last 125 years. (2004) Russell, incidentally, gave the magnitudes as 10,10.

R 71 was discovered at Sydney Observatory by H.C. Russell just before midnight on the 1st April 1879 (1879.248), measuring 15.39″ along PA 77.465°. It seems that he probably just evaluated the brighter component incorrectly so the position angle had 180° added to make the new PA as 257.465 or 257.5°. However, the WDS03 gives this 1879 separation here as 288° and 18.4 arcsec which does not match at all the Russells Sydney Observatory List of New Double Stars that appears written in the 1881 publication. At the time of writing the original measures are being obtained from by the USNO, but it does seem that something is quite wrong with the WDS03 figure or that another observation was made in 1879 by another observer that preceded Russells own discovery. In the earlier versions of the WDS, the 1879 position is given as 15.2 arcsec along PA 256° — matching Russells observations. His position angle has likely been precessed from the time of the original measure. Another possibility is that this is the 1935 measure and not the 1879 one, as the older versions of the WDS state the same scalar measures.

SLR 22 finds that the R71 B is again double. C is 11.2v (10.90V) magnitude being separated by 1.7 arcsec along PA 257°. Since discovery by R.P. Sellors at Sydney Observatory in 1896, the position angle has reduced from 266° to 257° while the separation remains fixed.

Common proper motions here suggests that SLR 22 is likely a true binary star but the brighter A component is almost certainly an optical companion. All three stars form a straight line that is proportioned by the ratio of about 9:1. Given spectral classes are A5, A5 and A7III, respectively. The whole system contrasts nicely with the orange-red colour of L2 Puppis in the field, being easily visible in 7.5cm.

Example 4

As for the example of Mu Cygni, the differences between the two WDSs are obvious;
21441+2845 Mu Cyg
WDS01 Desig Cmpnts PMag SMag Sep1 Sep2 PA1 PA2 Yr1 Yr2 Sp STF2822 BD 6.1 6.9 206.1 204.8 54 53 1902 1909 A5 STF2822 AB 4.5 5.9 5.6 2.0 109 307 1823 1997 F6 STF2822 AC 4.7 11.53 5.3 68.2 263 289 1878 1987 STF2822 AD 4.5 6.9 217.4 198.3 61 46 1823 1991 F7
WDS03 Oct
Desig  Cmpnts PMag SMag   Sep1  Sep2  PA1  PA2 Yr1  Yr2  Sp      pmRA/Dec pmRA/Dec
STF2822 BD    6.1  6.9   206.1 204.8   54   53 1902 1909 A5      +230-218 -001-055
STF2822 AB    4.75 6.18    5.6   2.0  109  312 1823 2002 F6V G2V +277-251 +277-251
STF2822 AC    4.7  11.5    263 289 3  5.3 68.2 1878 1987         +277-251         
STF2822 AD    4.75 6.94  217.4 197.5   61   45 1823 2001 F7V F2V +277-251 -001-055
Cmpnts =Components;
PMag= Primary Magnitude
SMag = Secondary magnitude;
Sep1 & Sep2 = First & Last Separation;
PA1 & PA2 = First & Last Position Angle; 
Yr1 & Yr2 = Year First and Last;
Sp = Spectral Class 

Note : Here the D star is optical while the others are almost certainly associated. Also the differences are in the magnitudes. WDS01 is visual v magnitude while some of the WDS03 Oct is now visual photometric (V) magnitude. Some of these brightnesses can differ by 0.4 or 0.5 magnitudes depending on the particular colour of the star in question.

STF 2822AB (21441+2845)

YEAR    PA     Sep.    
2002   309.9   1.811   
2003   310.7   1.793 **
2004   311.4   1.774   
2005   312.2   1.754   
2006   313.0   1.735   

** NOTE : This is a Grade 4 Orbit (Hei1995)
(Measure in 2002 : 312° ; +2°)

NOTE: Statistically it has been known for sometime that for moderate to wide pairs the repeatability of observations need only be once every twenty to thirty years. As most of these will have periods between hundreds if not several thousands of years. It is only the close ones, or those possible few with highly eccentric orbits, that are needed to watch out for. Us amateurs should mainly concentrate on pairs between 2 arcsec and 4 arcsec for any changes. This is because the Hipparcos satellite had restrictions within this range and could not measure these ones accurately if at all. More importantly these are the ones that can show significant change in motion depending on the orientation and eccentricity of the orbit to our line of sight.

Some Comments on Binary Star Orbits

As for binary star orbits, it is the grading of precision that is useful here. In the general old grading system (its simpler to describe) Grade 1 (like Alpha Centauri) can be calculated to high precision. Where as a Grade 3 is intermediate, and a Grade 5 is considered poor.

Differences between the good (seeing a number of orbits) and the bad — and usually only a part of the orbit in partial arcs. Yet it does depending on when the orbit was written, where it is not unusual to find expressed some range of dates in which the data is useful. I.e. 1975 to 2040 etc. This is used in multiple systems where orbits are constantly changing I.e. By apsidal motion, etc., causing small variations changing elements slightly from one orbit to the next.

Of the 1789 binary systems with orbital parameters appear in the 6th Orbit Catalogue. The grading is as follows;

  Grade    No. 
Grade 1     52
Grade 2    199
Grade 3    307
Grade 4    483
Grade 5    453
Grade 8      8
Grade 9    254
Grade 8 : Interferometric Orbits (high quality)
Grade 9 : Astrometric Orbits (low quality)

The average period of these stars is P=370.84 years and the mean size of the orbit is a 1.217 arcsec. A mean eccentricity for the orbit e is 0.523.

Of these;
Binaries  Period
 (No.)    (yr.)    
   77     >1000    
  105   500 to 1000
  233   250 to  500
  436   100 to  250
  245    50 to  100
  684  0.01 to   50

Whilst I do agree that many orbits are incomplete and may not give an exact position in Sep. and P.A., those for Grade 1 to 3 accurate enough to quote masses to reasonable accuracy. The problem is with refining the orbits for these differences are essentially minor.

To elaborate more on Toms statement;

Few if any orbits are known with complete certainty (even those that one sometimes sees quoted as being reliable). The great majority are pretty conjectural, and are constantly being revised by the few brilliant people who really understand the extremely difficult art (for that is what it is) of orbit computation.

As W. Heintz says in Double Stars (1970) p.53;

Presently orbits of nearly 700 objects are available (Note : This is now double this), covering a considerable range of periods and other parameters, and naturally differing also with regard to reliability. Not too many cases, perhaps 25 (Now 52), meet first class standards, so that elements can be definitive.

These have; Good quality in observations, One-and-a-half revolutions covered, Semi-major axis large enough to make error small, and all are mainly nearby systems.

Two or three hundred (now 558) are reliable; the elements are substantially correct, and only minor corrections are expected. Many pairs with largely though not completely observed ellipse, and close pair with residual scatter qualify for this category.

[Total 610 or 39.7% (Not counting astrometric binaries) accurate enough.]

About 300 (Now 925 or 60.3%) may receive the label preliminary.
— Substantial changes may be forthcoming, particularly for orbits over 300 years. (515)
— In short period orbits only with limited or inconsistent observations
— Even the basis of mass can be made by a3/P2 even though a and P may not be well known (dynamical parallaxes)
— Orbits are of statistical use.

About three-quarter (75%) of the orbital periods are in the range of 20 to 25 years (Now 874 or 48.9% — Things are improving!) The very close pairs are too infrequently observable, while most of the long period ones are still too indeterminate… Finally there are a small group of indeterminate orbits… voided as misinterpretations by subsequent measurements, and some spurious or faulty computed.

Several points of interest to the discussion regarding radical changes in orbits include;

— One third of binaries have eccentricities exceeding 0.6
— Most orbits are obtained from stars within 100 parsecs of the Sun (Limit now about 200pc.)
— An improvement to orbit accuracies and a narrowing, of the parameter ranges has now occurred from the more accurate Hipparcos parallaxes. I.e. a semi-major axis in AU is calculated by a = a in arcsec / parallax in arcsec.

More Comments on Discrepancies

As for Toms statement;

Detecting these discrepancies, which doesn t always require sophisticated equipment, is part of the fun of double-star observing! If published, it can also be scientifically useful. I think you will find that there are several other members of this group who have had similar experiences to my own in this area.

I agree 110%. You can have more fun researching the discrepancies and errors you find from your own observations. Although the mathematics in working out orbits is likely beyond yours (and my) capabilities, it is the discovery of those system that can break the normal view of things.

Bills point is very valid here in the sense that a sudden change is PA might be real or not. If anyone does find a true error (such as unexplainable changes in either separation or position angle) that they cannot explain — then dont be frightened to talk about it. There are enough observers here to point you in the right direction and if you do find something, they have both the experience and likely the necessary reference material. Even if you dont discover something of significance — at least you can correct the error at its source!

Regards and Clear Skies!


(That inspired the text above from the S33 Group.

Message: 15
Date: Mon, 20 Oct 2003 21:27:38 -0000
From: Bill Green
Subject: Data Discrepancies

I have noticed variations in data for many doubles. I can surely understand minor differences but some are gross discrepancies. For example; the chart in PJs recent article on Autumn Double lists the PA of Alpha Piscium as 277°. The observing list of doubles at the Astronomical Society has it at 50°! There are many others with equal differences.
Am I reading something wrong or just plain missing something?
Bill Green

Message: 18
Date: Mon, 20 Oct 2003 23:56:27 +0100
From: Thomas Teague
Subject: Re: Data Discrepancies

I seem to recall reading recently that there are some deliberate errors to be found here and there in some amateur observing lists, designed to catch out those who report observations they havent actually made! I cant vouch for this, though.

However, it is very common to find discordant data about double stars. With experience one gets to regard all such data with considerable reserve. As a rule, popular amateur texts dont bother to cite their sources. One therefore doesnt know whether they come from predictions based on computed orbits that may have been completely discredited, or from observational data that may well be seriously out-of-date. With some pairs, there can be considerable relative motion in just a few years. Indeed, near periastron, even pairs with quite long orbits may show remarkably rapid relative motion — a good example right now is Gamma Virginis, which will show extraordinary changes in PA over the next few years. With stars of more or less equal brightness, moreover, one often gets disagreement as to which component should be regarded as the primary, leading to 180° discrepancies. Sometimes, people carelessly give data for AB which actually turns out to be for AC or some other combination of components.

In some handbooks, such as the annual BAA Observing Handbook, you will find lists of binaries printed with orbital predictions in which angular separations are given to 0.01 arc seconds. It is tempting to assume that these are accurately known data, which can be taken on trust. Nothing could be further from the truth. Few if any orbits are known with complete certainty (even those that one sometimes sees quoted as being reliable). The great majority are pretty conjectural, and are constantly being revised by the few brilliant people who really understand the extremely difficult art (for that is what it is) of orbit computation. In practice, one quite often finds that a casual glance in a small telescope is enough to reveal that a seemingly devastatingly accurate orbital prediction is in fact grossly out. I found this myself with Mu Cygni a few years ago using a little 2.5-inch refractor (just eyeballing — no micrometer), and later found that my conclusion had been independently confirmed by Bob Argyle using one of the Cambridge refractors in conjunction with a filar micrometer. Detecting these discrepancies, which doesnt always require sophisticated equipment, is part of the fun of double-star observing! If published, it can also be scientifically useful. I think you will find that there are several other members of this group who have had similar experiences to my own in this area.


Last Update : 4th October 2012

Southern Astronomical Delights © (2011)

For any problems with this Website or Document please e-mail me.