This journal article documents letters to the Editor exchanged between the GIA cut study team and the cut group who wrote an article in The Australian Gemmologist
The original article from Volume 23, Number 1, January – March 2007 is not reprinted here, but a similar article was posted here on this websites on-line Journal.
Here is the GIA’s letter to the Editor:
The recent work by Sivovolenko et al. (“Diamond cut grading: Unintended consequences and solutions,” First Quarter 2006, pp. 447-454) mentions our work on Cut Quality grading in round-brilliant-cut diamonds (RBCs). We found several inaccuracies in this article related to our research and cut grading system, and therefore felt we should take this opportunity to clarify some points for the record.
We agree with Sivovolenko et al. that the current GIA and AGSRBC grading criteria give different relative results for some sets of proportions, but we disagree with the reasons they give for these differences. What is not adequately mentioned by the authors is that there is much overlap in the cut grades from the two systems, and that many of the remaining differences in grades may result merely from the fact that our system uses five grades, while the AGS system uses eleven grades. We note that the diamond appearances in their figures 3(3) and 3(6) are brighter, not darker, than the same diamond resting on a gray background in figure 3(2); this would indicate to us a result counter to the one provided by the authors. We disagree with the authors’ contention that the environment used in our research led to the favoring of “steep/deep” diamonds over those with “shallower” proportions. When diamonds stray too far in proportion space to either of these areas they are deducted for in our cut grading system. We also believe, but cannot confirm without further information from the authors, that the areas being contested are areas at the borders of our cut grade categories, and would contend that areas at grade borders are often contested in any evaluation system.
Three points in the section titled “Some Perceived Problems” warrant comment. First, as stated in reference iv (again, the Fall 2004 issue of Gems & Gemology), we knew the system needed more than just brightness and fire metrics because those two factors alone gave an agreement rate of only 58% ; in contrast, the finished GIA cut grading system for RBCs showed 92% agreement between human observations of overall appearance and system assignment of grades – the same agreement rate the observers showed with each other. We would also like to emphasize that much of our observation testing was devoted to “overall appearance” and not just to brightness and fire as stated.
Second, the Cut Group may feel that 38 million proportion combinations is “excessive and un-workable for rough planning,” but both Sarin and OGI have incorporated the GIA cut grading system with this number of data points, and diamond manufacturers have been using these tools successfully for eight months.
Third, the authors’ point about the difficulty of using the same approach for fancy shapes as we used for rounds is a valid one. We will need to examine the feasibility of using our cut system for RBCs as we enter our next stage of research on fancy shapes. However, we also believe that an accurate and comprehensive system for grading the cut quality of round brilliants should not be jettisoned merely because its wholesale application to fancy shapes is uncertain. Future research (including many more observation tests) will show whether our current system is applicable to fancy shapes, or whether particular shapes might need their own adapted cut grading system.
In contrast to the statement that GIA or AGS “have yet to find an adequate metric for scintillation,” footnote 3 on page 454 of the article shows that the lack of a specific scintillation (sparkle) metric was based on the comments we gathered during human observations when we asked why one diamond had a superior overall appearance to another. People spoke a great deal about “dark-zone distribution” but very little about relative sparkliness. Also note in this same footnote we state that during our research we found that a diamond’s relative sparkle was accounted for in our brightness and fire metrics.
The authors state that “green spectral flashes are rarely seen in a diamond,” and assert that a very small aperture and a small light source at a great distance are required to photograph such a flash. As described in reference xi (Fall 2001 issue of Gems & Gemology), although our model conditions use a point light source and a hemispheric detector, both infinitely far from the virtual diamond, our photographs of fire used real diamonds, an actual hemisphere 40.6 cm (16 inches) in diameter, a hole for the light 0.95 mm in diameter with a fiber-optic light 20 cm (17.5 inches) behind the hole, a 50 mm lens and f/# of 16, yielding a camera aperture of 2.8mm, which is not “very small”. This reference also showed the similarity between our modeled results and photographs despite the differences in conditions. Both photographs and human observations of hundreds of round brilliant diamonds revealed a variety of expressions of fire. Most samples displayed all spectral colors, including green. We agree completely with the authors’ recommendation that readers try to reproduce this for themselves: take a round brilliant, rock it back and forth under spot-like illumination and look at the fire. See whether any of the flashes are green – our experience indicates that some will be.
There are also incorrect statements and implied assertions that do not address our work directly. For example, the earliest diamond reports containing proportions were Diamond Certificates issued by the AGS starting in 1936, and the earliest effort to grade cut from proportions was made by GIA in 1938, not in the mid 1950s as stated by the authors. Also, to the best of our knowledge, ASET (both program and viewing device) was developed by AGS and Dr. Jose Sasian; the article implies that GH was involved in its development.
In addition, we noticed technical mistakes in this article, such as almost identical captions for figures 5 and 6, and a nonsensical illustration for figure 8 with a caption that appears to belong to figure 6.
We wish Sivovolenko et al. all the best in their research intentions, but hope that any resulting articles are vetted conscientiously.
The GIA Diamond Cut Group*
* The GIA Diamond Cut Group consists of:
Ronald H. Geurts
Al M. Gilbertson
T. Scott Hemphill
John M. King
Mary L. Johnson
Thomas M. Moses
Ilene M. Reinitz
James E. Shigley
Thank you for giving us the opportunity of responding to the letter of critique sent to the Editor from the GIA Diamond Cut Team concerning our paper in the April-June 2006 issue of The Australian Gemmologist.. This response from our Cut Group* represents a rare opportunity to engage in open dialogue with the GIA Diamond Cut Grading research team, and we also whould like thank them for this opportunity.
With regard to differences between the GIA’s and American Gem Society’s (AGS) grading systems; we have choosen only to highlight examples of extreme grade differences between the two systems to avoid the very argument outlined in the GIA Cut Grading research team’s letter “that there is much overlap in the cut grades from the two systems”. We have taken the GIA team up on this challenge by comparing the overlap between the AGS’s 11 grades and GIA’s 5 grades for the common ‘Excellent’ table size of 57%. Comparing that overlap is difficult because GIA arranged its pavilion axis opposite to the generally accepted chart format that AGS (and other researchers) have traditionally used. However, Peter Yantzer, the Director of the AGS Lab, kindly provided us with the AGS ‘candidate’ charts in the GIA’s format (as displayed at www.diamondcut.gia.edu/charts/ ). A comparison of the 57% table size charts showed that GIA ‘Excellent’ proportions (outlined in black on figure 2) had only 26 good AGS0 and AGS1 matches, and 31 bad matches. Included in these bad matches were several very bad matches—specifically 7 AGS 5s and 4 AGS 6s that would be graded Excellent by GIA (see table 1 and figure 1 for details). We consider this to be a poor match and, in addition, confusing for both consumers and the industry. The situation is worse for table sizes below 52%, where GIA has no ‘Excellent’ proportion sets. For example, at the 51% table size AGS has 17 AGS0 and 22 AGS1 ‘candidate’ proportion sets that GIA would grade as Very Good, Good or Fair.
Table 1 The good and bad matches between different grades of GIA and AGS ‘candidate’ grades.
(The authors deliberately excluded AGS 2 because in a 5 v 11 grade comparison no significant conclusions can be drawn.)
|Good matches||Bad matches|
|GIA Ex & AGS0||13||GIA VG & AGS1||12|
|GIA Ex & AGS 3||3|
|GIA EX & AGS1||13||GIA Ex & AGS 4||5|
|GIA Ex & AGS 5||7|
|GIA Ex & AGS 6||4|
|26 good matches||31 bad & very bad matches|
1. The AGS ‘candidate’ proportions by pavilion and crown angles for table size 57%. The black overlain shape contains the GIA ‘Excellent’ proportions used to produce the data in table 1.
We would like to clarify the comments made concerning the diamond in figure 3 of our paper, with a 57% table size and comparatively steep crown (35.3° ) and deep pavilion (41.4° ). This diamond was given a grade of ‘Excellent’ cut by GIA (you can check its position on the chart in figure 2), but it did display very unattractive leakage just inside the table when viewed over skin or on a dark background. This leakage was not apparent when the stone was resting on the light coloured (18% gray) tray used in GIA’s observation-based survey.
The GIA team had noted that diamonds appear different when observed (and also photographed) on backgrounds of differing shade. But, after personal communication with a key member of the GIA team, at the GIA Symposium August 2006, we do not believe they understand fully why.
We will show that the way our eyes adapt to a brightly lit and light colored background reduces our ability to make fine distinctions between diamonds. GIA’s Common Viewing Environment (CVE) that was used in their survey environment had a background and stone trays of a shade of 18% gray, with twin diffused fluorescent tubes mounted less than 33 cm (13 inches) above the diamonds. This causes an observers eyes to ‘overexpose’, creating a bad environment for discerning differences between diamonds of differing brightness. The authors of this response spent a little time with the commercially available Diamond Dock® version of the CVE at the GIA Symposium in August 2006, This confirmed that the viewing environment is quite similar to a diamond dealer’s twin fluorescent desk lamp and viewing position. Yet, when asked to evaluate the brightness of a diamond, buyers and dealers usually do not use the desk lamp for this purpose. Instead, they typically examine diamonds under the edge of the desk, with their body turned away, using ceiling lighting near a window or even in a hallway under spot lighting.
At the Symposium, the Cut Group authors conducted a brief experiment with GIA’s Diamond Dock® CVE and took the two photographs shown in figure 2. We apologise for the quality of these images; but we were stopped by a GIA Director who protested that we were conducting experiments “to use against GIA”. Note that our eyes, and a camera with auto exposure, adapt to the background brightness in a similar manner. What we ‘saw’ was indeed similar to the photographs shown in figure 2.
With the twin fluorescent lamps turned on (Fig. 2, LHS, when looking at these sets of coloured cubic zirconia , which were used for the poster presentation of Sergey Sivovolenko and Yuri Shelementiev, the princess and oval cut stones appeared brighter than the ideal cut rounds. With the lights turned off and only with ambient room illumination (Fig. 2, RHS), the relative brightness of these various cuts gave the opposite results, the results one would usually expect to see, for the round stones appeared brighter. It is indeed unfortunate that as the stones were adhered to black cards, as this somewhat lessened the overall effect of the bright background illumination that we were seeking to demonstrate.
Fig. 2. These coloured cubic zirconia of different cuts were photographed inside GIA’s common viewing environment Diamond Dock®. The photograph on the left was taken with the fluorescent grading lights turned on. For the photograph on the right the lights were turned off and only the ambient room light was illuminating the stones. Note that with the lights on the oval and princess cuts appear brighter than the well cut rounds. The results are the complete opposite when the brightness grading lamps are turned off, with the rounds appearing brighter. The reason for these differences is that the camera’s auto exposure adapts to the background in a similar way as our eyes.
The GIA authors’ contention that they have struck a balance between shallow and deep diamonds does not concur with the widely accepted Tolkowsky standard. Holding the pavilion angle at 40.8°, the GIA ‘Excellent’ grade’s average crown angles for table sizes between 52% to 62% are around 1.5° steeper than Tolkowsky’s 34.5° crown angle. Alternatively, holding the crown angle constant, GIA’s Excellent proportions have an average pavilion angle that is 0.3° deeper than that recommended by Tolkowsky. Clearly the GIA cut grading system favours deeper diamonds than the generally accepted optimum proportions.
The GIA team wrote in their letter: “we state that during our research we found that a diamond’s relative sparkle was accounted for in our brightness and fire metrics”. Unfortunately, to date, GIA have never reported any metric that measures a “diamond’s relative sparkle” or scintillation. As their brightness and fire metrics had only a 58% match to the observation results, and since the GIA have never developed a sparkle or scintillation metric, how can they claim a 92% match between the metrics and observation results? This seems to be a circular argument? It would appear that GIA’s first 10 years of virtual diamond modeling study have been of little or no value. This is unfortunate. For GIA does not, as they agree in their letter, have the tools to develop cut grading systems for fancy shaped diamonds.
Our reference to 38 million proportion combinations as being excessive was related to fancy shaped diamonds. Such a data base would be thousands of times larger and more complex than that for the round stone. Also it probably would be unworkable for rough diamond planning.
The round data base is relatively simple, and the vast majority of the 38 million data points are of no use to manufacturers who are only interested in the boundaries of Excellent and Very Good cuts. After GIA released its cut grading charts in February 2006, our information is that sales of the expensive Facetware® in Sarin and Ogi rough diamond planners have all but ceased.
We did not claim that green is never seen in polished diamonds. What we said was “rarely”. We too hope that all readers will do some simple observations in normal lighting too to discover for themselves how rare green flashes are. It would appear that the GIA team has confused what the eye can see when a sparkle of light from a diamond shines onto a surface, such as a dome, with what the eye would see if the dome was removed and an observer looked from this position with their naked eye. Most flashes we see with our eyes appear white because the spectral colours would have merged together. To help explain why green is rare we have paraphrased a good explanation of ‘fire’ by Anton Vasiliev that was published 12 years ago in a work still available in full at www.gemology.ru/cut/english/rainbow/5.htm )
Consider the observer, which GIA did not in their published studies. For this observer with a pupil diameter of 3 mm, a viewing distance of 25 cm creates an angular size of 0.7° to allow for the possibility of light entering the eye. For an observer to see a ‘pure’ spectral colour, a gem must have a high ‘fan’ of angular dispersion. But because green is in the central region of the spectrum, there is a high chance that green will be mixed with other colours, and appear whiter (see figure 3, left side for an example). For this reason, the greater the distance between the gem and the observer, the more pure the ‘fire’ one can see.
Now consider the type of illumination , which GIA also did not do in their published ‘fire’ studies. To create a full spectrum, a prism should be illuminated with a light that has a very small angular size—such as that from the sun, a fiber optic source shining through a pin hole, or a small lamp at a very great distance. In contrast, when a light is larger and closer, as in most indoor situations, the light beams fan out and cause the spectral colors to mix, degrading the purity of the observed color. Hence, good green fire requires illumination of the gem from a light source of small angular size. The source’s angular size affects the color purity in a similar way as that of the the size of the pupil in the observer’s eye..
Fig. 3 These images show OctoNus Effective Total Angular Size (ETAS) DiamCalc models for various viewing conditions. (Click on the pictures to open larger images)
The image on the left is based on the information provided in the GIA letter to the Editor and shows the full spectrum of flashes that could be seen on the surface of a dome. On the lower right of the left side image a small circle shows that part of the flash that could enter a 4 mm pupil if the dome was removed and an observer looked from this position with their naked eye. Under these circumstances the flash would appear white because the spectral colours would merge together. If the observer moved a little up and to the left they would see a reddish flash, and if they moved down slightly to the right, the flash would appear blue as in the example shown here www.gemology.ru/cut/english/grading1/5.htm. The image on the right shows the probability of the type of flash an observer might see from 25 cm if the source of light was 10 mm in diameter and was positioned 2 metres away from the diamond. In reality (with the dome removed) the probability of seeing green is very small.
As we have shown in figure 3, the illumination and observer are just as important as the diamond itself. We believe the GIA used inadequate illumination models for their virtual studies, and to the best of our knowledge, they have not studied or published any work concerning observer’s eyes or their vision systems.
The diamond on the left of figure 4 has been photographed in office lighting and shows predominantly blue coloured flashes, with some red and yellow. As we have explained green flashes are harder to see if the light source is large in size, and relatively close to the diamond. The diamond on the right shows some green; for it was photographed in partly shaded direct sunlight, effectively parallel rays of light. In addition, both pupil and camera apertures’ are small because we are outdoors in very bright daylight. Our narrow pupil aperture allows us to see some green flashes without the merging of yellow and green or blue and green. An additional explanation can be found here at www.gemology.ru/cut/english/grading1/4.htm
Fig. 4. The 15 ct diamond on the left has been photographed in office lighting where green will rarely be seen because the lights in the office have a large angular size. The 9 ct diamond on the right shows some green; for it was photographed under a broad leafed tree in partly shaded direct sunlight, effectively parallel rays of light. In very bright daylight both pupil and camera apertures’ are small which allows us to see more green flashes without the merging of yellow and green or blue and green. Both photographs were taken with the same Canon Ixus camera, set to auto exposure mode, with the diamonds positioned on author GH’s fingers.
After the close of the GIA Symposium in San Diego all four authors were able to experiment with (but not photograph in) the GIA Diamond Dock® as we all attended a GIA Cut Grading Course.
We noted different instructors assessed ‘brightness’ using differing lighting/observer angles with the light box (and other light sources, such as the small fluorescent lights on microscopes. When asked about the correct lighting/observer angle, the instructors could offer no consistent or quantifyable recommendation.
In the original Diamond Dock® Users Manual, a recommendation is made that the observer views diamonds at an angle of 45° to the vertical plane, which we believe was the same angle observers used for the survey testing of 70,000+ observations that formed the basis of GIA’s new Cut Grading System. Anecdotal information from volunteers who participated in GIA’s observation testing seems to confirm that observational tests were performed in a seated position that would have subtended an angle of 40° to 45° between the stones and the centre of the lights (depending on observer height).
When the incident light from the fluorescent tubes comes from this angle, slightly shallower diamonds have less brightness in Diamond Dock®, whereas Tolkowsky proportioned and deeper diamonds appear brighter. We were able to demonstrate how important this viewing position is to two GIA instructors, who by their own choice, viewed diamonds from a much higher view point than 45°—one where the incident light would have been subtending an angle of approximately 20° to 30° relative to the vertical and the light source. We had with us two diamonds of similar colour, clarity and diameter. One diamond of 1.19ct, with proportions very near that of Tolkowsky’s, had been graded Excellent by GIA Gem Trade Lab. The other shallow 1.16ct stone graded Very Good (a border line Good when using Facetware™). From the higher viewing angle, both instructors preferred the shallow stone for brightness (and were split one each for ‘fire’ using the instrument’s LED lighting only.
Figure 5 shows ETAS data based on images produced from three dimensional models of these two stones made with DiamCalc® along with the actual steep deep diamond used in our original article. This demonstration can be used to explain why the shallower stone appears brighter when the incident light comes from a higher angle; for the shallower diamond draws and returns more light from directly above, while as round diamonds get deeper they draw more light from lower angles closer to the horizon. Readers are invited to use similar modeling software at this website: www.cutstudy.com/cut/english/grading1/sphere1.htm
|Shallow 1.16ct GIA VG||‘Ideal’ 1.19ct GIA Ex||Deep GIA 1.28ct Ex|
|Potential to see the Diamond Dock lights in a diamond from a 45° viewing angle|
|Potential to see the Diamond Dock lights in a diamond from a 23° viewing angle|
|Fig. 5. Effective Total Angular Size (ETAS) images modeled in DiamCalc based on 3D scans from three actual diamonds. The flares of light show the potential and direction that light in the environment around each diamond could come from to cause a sparkle in the face up position. The lines mark 10° increments on a virtual dome, with the zenith in the centre facing the observer. The diamond is shown as a small blue dot in the centre. The 6 arcs in the upper rows represent the position of the diffuser over two fluorescent tubes in Diamond Dock.|
The steeper lighting-viewing angle makes most round diamonds look brighter, which probably explains why Diamond Dock® users seem to naturally prefer to view diamonds from higher angles than the original designers intended. We believe GIA would have arrived at shallower Excellent proportion parameter sets had the observers in their observational tests been directed to use a higher viewing angle, but then these proportions may have been too shallow.. However, as we have already shown, the light background and very close proximity and brightness lights in the CVE lighting box its narrow source of angular size of lighting and lack of consistency in viewing angles make this method of cut grading useless. Basing a cut grading system on observation and survey results from this method is, we believe, flawed. In particular this lighting environment would have even less validity for brightness grading of fancy shaped diamonds, many of which tend to gather very little light from the illuminated area in the CVE.
In summary we consider that the round brilliant cut grading system established by the GIA Diamond Cut Team is below the standard expected from The World’s Foremost Authority in Gemology™. In our article and this exchange of letters we have outlined a number of shortcomings: when the same diamond can be graded with very different cut grades by the AGS and the GIA, two of the worlds most respected diamond cut grading labs, then neither the Internet savvy diamond buying public, nor the trade can feel confident. Observers using the common lighting environments had only a 58% with light return and fire metrics from GIA’s previous virtual studies. The implied use of an unpublished metric for scintillation, and the inadequacy of light return and fire metrics bodes poorly for the future GIA’s grading of other shapes of diamond cuts means that round diamonds will continue to dominate the market (one of the main concerns noted in our original article). The issue of ‘green fire’ shows an apparent lack of understanding of the GIA Diamond Cut Team of common illumination sources and of human eye sight. Finally we have traced some major failings in their research work to the Diamond Dock(R) or Common Viewing Environment that was used to assess diamond brightness; the background is too light a shade of gray and the fluorescent lights are too bright or too close, the observers angular orientation relative to the diamond and the light source was not regulated and the illumination area is too narrow an angular source to be useful for grading diamonds with a range of different proportions. The GIA Gem Trade Laboratory began commercially issuing grading reports with cut grades for all D-Z coloured round brilliant cut diamonds in January 2006. The grading reports being issued by the GIA and advice given by its associated Facetware(tm) have already influenced the pricing of tens of billion dollars worth of diamonds in the past year. We urge the Gemological Institute of America to consider our arguments concerning its previous research and withdraw its diamond cut grading system immediately.
Kind regards, The Cut Group
* Members of the Cut Group include Sergey Sivovolenko, Garry Holloway, Yuri Shelementiev and Janak Mistry