GIA''s amazing new patent

[Serg]

Date: 2/2/2007 3:25:30 PM
Author: adamasgem

Date: 1/31/2007 4:42:44 PM
Author: adamasgem
Here is whatI was able to resolve using a 5.7 degree and 10 degree girdle normal criteria.
Neglet the lines from one end to the other as they are artifacts from unfolding the girdle that I have to fix..

I added the capability to vary the criteria used to define (select) the girdle facets, and here is a comparison of three limits used..
Seems the file Garry posted had some interesting variation in the planes defining the girdle profile and you needed at least a 15 degree criteria to adequately define what are girdle facets...

Marty, What is green dash line?

 

[adamasgem]

Date: 2/2/2007 8:07:20 PM
Author: Serg

Date: 2/2/2007 3:25:30 PM
Author: adamasgem

Date: 1/31/2007 4:42:44 PM
Author: adamasgem
Here is whatI was able to resolve using a 5.7 degree and 10 degree girdle normal criteria.
Neglet the lines from one end to the other as they are artifacts from unfolding the girdle that I have to fix..

I added the capability to vary the criteria used to define (select) the girdle facets, and here is a comparison of three limits used..
Seems the file Garry posted had some interesting variation in the planes defining the girdle profile and you needed at least a 15 degree criteria to adequately define what are girdle facets...

Marty, What is green dash line?

One of a series of very POOR attempts, at this time to use the girdle facet information to define a girdle midplane. It is supposed to be a projection of the girdle midplane plane definition to the appropriate radius.

I'm trying to design an optimal Kalman filter to use all the girdle facet info to define the "best " girdle midplane plane parameters.

A lot depends on how much "faith" I put on each measurement which would define the "noise". I've fooled around with adding process noise to prevent convergence that shuts down the effects of subsequant measurements..Tis a pain in the neck problem I'm not satisfied with yet, as to the solutions I get.

Call it an intellectual exercise...and right now I feel a little dumb about it...

It is supposed to be a filter solution to the equation set A*(x+dx) + B(y+dy) + C(z+dz) + D = 0.0, which should work to solve for A, B, C, D and considering the measurement variance to be a function of uncertanties denoted by dx;dy,dz and the covariance matrix of the solution {A,B,C,D}(-), where (-) represent the covariance of {A,B,C,D} prior to the state update..

I've either got a conceptual error or software glitch right now.. or need to add process noise to the problem..

My solution right now is too sensitive to the three facet centroids I use to initialize the probelm (magenta colored facets) which I've ssomewhat arbitrarily chosen about 120 degree from each other..

I haven't tried the vertex solution I described in a previous post..

Any sugestions?

 

[Garry H (Cut Nut)]

Sorry for no ascribing credits to you Marty - but if I did I would firstly have needed your permission.

Re the DiamCalc girdle assesment - the DiamCalc scanned report has managed to identify the girdle facets adequately - even though the scan accuracy is clearly not accurate enough to place the tiny facets - it is part of the old Sarin method - I hope they have changed it - it was very poor at culet measurements also.

 

[adamasgem]

Date: 2/2/2007 11:19:25 PM
Author: Garry H (Cut Nut)
Sorry for no ascribing credits to you Marty - but if I did I would firstly have needed your permission. You owe me a beer in Vegas

Re the DiamCalc girdle assesment - the DiamCalc scanned report has managed to identify the girdle facets adequately - even though the scan accuracy is clearly not accurate enough to place the tiny facets - it is part of the old Sarin method - I hope they have changed it - it was very poor at culet measurements also. But the tiny facets are there, in the generated STL''s. Was this a bruted girdle stone?????

 

[Serg]

Date: 2/3/2007 2:32:39 AM
Author: adamasgem

Date: 2/2/2007 11:19:25 PM
Author: Garry H (Cut Nut)
Sorry for no ascribing credits to you Marty - but if I did I would firstly have needed your permission. You owe me a beer in Vegas

Re the DiamCalc girdle assesment - the DiamCalc scanned report has managed to identify the girdle facets adequately - even though the scan accuracy is clearly not accurate enough to place the tiny facets - it is part of the old Sarin method - I hope they have changed it - it was very poor at culet measurements also. But the tiny facets are there, in the generated STL''s. Was this a bruted girdle stone?????

Marty,
When you just download ANY cut to DC from any format DC do very simple recognition type facets( crown, pavilion, girdle) . We do it for add simple parameterization just( not for reports)
But for reports In DC we use special algorithm for each type important cut( please use button "Scanned reports") Thanks Garry for reminder
And Helium use usually more modern algorithms

 

[adamasgem]

Date: 2/3/2007 3:03:31 AM
Author: Serg

Date: 2/3/2007 2:32:39 AM
Author: adamasgem

Date: 2/2/2007 11:19:25 PM
Author: Garry H (Cut Nut)
Sorry for no ascribing credits to you Marty - but if I did I would firstly have needed your permission. You owe me a beer in Vegas

Re the DiamCalc girdle assesment - the DiamCalc scanned report has managed to identify the girdle facets adequately - even though the scan accuracy is clearly not accurate enough to place the tiny facets - it is part of the old Sarin method - I hope they have changed it - it was very poor at culet measurements also. But the tiny facets are there, in the generated STL''s. Was this a bruted girdle stone?????

Marty,
When you just download ANY cut to DC from any format DC do very simple recognition type facets( crown, pavilion, girdle) . We do it for add simple parameterization just( not for reports)
But for reports In DC we use special algorithm for each type important cut( please use button ''Scanned reports'') Thanks Garry for reminder

And Helium use usually more modern algorithms

Thanks Serg.. I''ll try it out. I just looked at the color coded wire mesh, which showed a broken girdle outline, as above.

Any comment on my girlde scallop plane definition algorithm above???

 

[strmrdr]

While I think finding the girdle center-line is a kewl exercise I see nothing to be gained by using it when a absolute reference is available vs a calculated one for anything serious.
Building a virtual model should be good enough to identify most defects using the table reference.

 

[adamasgem]

Date: 2/3/2007 3:03:31 AM
Author: Serg




Marty,
When you just download ANY cut to DC from any format DC do very simple recognition type facets( crown, pavilion, girdle) . We do it for add simple parameterization just( not for reports)
But for reports In DC we use special algorithm for each type important cut( please use button 'Scanned reports') Thanks Garry for reminder



And Helium use usually more modern algorithms

Thanks Serg and Gary for pointing out the DC feature that I hadn't played with.
DC is a great sanity check for what I am doing and I love the rendering capability.

It even verifies the girdle facet components that we both seem to to miss with my 15 degree girdle normal criteria, not exactly the same points, I'll have to check that, but we are using different algorithms. I'll have to try to phase things up better to see where the difference lie. What do you use for a criteia relative to girdle plane normals. I put a hook in to make it a user input. This is the first STL file that I've had this problem with, but it won't be the last.

I guess you call them extra facets in DC, but they are there in your color coded diagram above also, so I am happy having verified my work. Whether they are EF or part of the girdle is a non issue.

I'm still going to look atsome "better" plane definitions, that might make sense, based on the nominal way a stone is cut.

 

[adamasgem]

Date: 2/2/2007 11:19:25 PM
Author: Garry H (Cut Nut)
Sorry for no ascribing credits to you Marty - but if I did I would firstly have needed your permission.

Re the DiamCalc girdle assesment - the DiamCalc scanned report has managed to identify the girdle facets adequately - even though the scan accuracy is clearly not accurate enough to place the tiny facets - it is part of the old Sarin method - I hope they have changed it - it was very poor at culet measurements also.

Garry It is not necessarily scan accuracy, it is the lack of an industry standard as to what constitutes a girdle facet..I note that you didn''t show the same relative azimuth position that I did, in your color coded diagram overlay. It appears that the standard DC wire mesh color coding and that used for reports is different, as it appeared that the standard wire mesh had a broken girdle plane..

 

[adamasgem]

Date: 2/3/2007 4:03:28 AM
Author: strmrdr
While I think finding the girdle center-line is a kewl exercise I see nothing to be gained by using it when a absolute reference is available vs a calculated one for anything serious.
Building a virtual model should be good enough to identify most defects using the table reference.

Can''t say that i disagree with you conceptually, but we techies like to try to figure out what is going on in the most explicit manner possible.

The girdle plane definition is, however, I think, something to define consistently.
I have to think about what Serg is doing using the mains.
I sort of think using the vertex definition is more explicit, but I have to try it out to prove the point. I think my conceptual definition is not bad, and it is consistent with what I know about cutting procedures, based on somewhat limited observations, and a little experiance in colored stone faceting, probably before you were born, or at least a toddler..

Appreciate your comments though..

 

[diagem]

What are the two purple facets???

 

[Garry H (Cut Nut)]

Marty it was 4 years ago - I can not remeer if it was bruted or polished, but the cutter I bought it from is more likely to have had polished G''s.

But I also know Sarins process at the time - it tried to plot 2 spots on a shadow line and draw a line until it hit the line from crown and one from pavilion. So the metod was flawed for very small shadow edges

 

[strmrdr]

Date: 2/3/2007 4:52:04 AM
Author: adamasgem

Date: 2/3/2007 4:03:28 AM
Author: strmrdr
While I think finding the girdle center-line is a kewl exercise I see nothing to be gained by using it when a absolute reference is available vs a calculated one for anything serious.
Building a virtual model should be good enough to identify most defects using the table reference.

Can''t say that i disagree with you conceptually, but we techies like to try to figure out what is going on in the most explicit manner possible.

The girdle plane definition is, however, I think, something to define consistently.
I have to think about what Serg is doing using the mains.
I sort of think using the vertex definition is more explicit, but I have to try it out to prove the point. I think my conceptual definition is not bad, and it is consistent with what I know about cutting procedures, based on somewhat limited observations, and a little experiance in colored stone faceting, probably before you were born, or at least a toddler..

Appreciate your comments though..

I hear ya...
Here is the problem as I see it if you put a dot at each crown pavilion girdle junction half way between them then connect it to the one opposite with a line you will get multiple planes of which none will match perfectly.
Which one is right?


btw I''m older than you think I am lol

 

[adamasgem]

Date: 2/3/2007 5:08:50 AM
Author: DiaGem
What are the two purple facets???

Just my way of flagging the facets used to initialize a girdle plane solution filter... There is actually a third one at the beginning.. The centroid of those three arbitrary facets ( randomly picked with a minumim angular separation) was used to initialize a solution to a plane through the girdle. (Any three points define a plane. I then used the rest of the girdle facets to further refine the definition of the plane.

 

[adamasgem]

Date: 2/3/2007 5:11:21 AM
Author: Garry H (Cut Nut)
Marty it was 4 years ago - I can not remeer if it was bruted or polished, but the cutter I bought it from is more likely to have had polished G's.

But I also know Sarins process at the time - it tried to plot 2 spots on a shadow line and draw a line until it hit the line from crown and one from pavilion. So the metod was flawed for very small shadow edges

Huh????? Is there a language barrier here Matey? OK , maybe what you are describing is how they defined a girdle edge??

But you explanation seemingly doesn't fit the STLs from this file.. The girdle seems pretty uniform and thick enough..

 

[diagem]

Date: 2/3/2007 2:01:24 PM
Author: adamasgem

Date: 2/3/2007 5:08:50 AM
Author: DiaGem
What are the two purple facets???

Just my way of flagging the facets used to initialize a girdle plane solution filter... There is actually a third one at the beginning.. The centroid of those three arbitrary facets ( randomly picked with a minumim angular separation) was used to initialize a solution to a plane through the girdle. (Any three points define a plane. I then used the rest of the girdle facets to further refine the definition of the plane.

Am i understanding correct? are you averaging the girdle facets degrees?

 

[Garry H (Cut Nut)]

Date: 2/3/2007 2:03:39 PM
Author: adamasgem

Date: 2/3/2007 5:11:21 AM
Author: Garry H (Cut Nut)
Marty it was 4 years ago - I can not remeer if it was bruted or polished, but the cutter I bought it from is more likely to have had polished G''s.

But I also know Sarins process at the time - it tried to plot 2 spots on a shadow line and draw a line until it hit the line from crown and one from pavilion. So the metod was flawed for very small shadow edges

Huh????? Is there a language barrier here Matey? OK , maybe what you are describing is how they defined a girdle edge??

But you explanation seemingly doesn''t fit the STLs from this file.. The girdle seems pretty uniform and thick enough..

Lets try sign language? Read the image and its title

 

[adamasgem]

Date: 2/3/2007 2:37:40 PM
Author: Garry H (Cut Nut)

Date: 2/3/2007 2:03:39 PM
Author: adamasgem

Date: 2/3/2007 5:11:21 AM
Author: Garry H (Cut Nut)
Marty it was 4 years ago - I can not remeer if it was bruted or polished, but the cutter I bought it from is more likely to have had polished G's.

But I also know Sarins process at the time - it tried to plot 2 spots on a shadow line and draw a line until it hit the line from crown and one from pavilion. So the metod was flawed for very small shadow edges

Huh????? Is there a language barrier here Matey? OK , maybe what you are describing is how they defined a girdle edge??

But you explanation seemingly doesn't fit the STLs from this file.. The girdle seems pretty uniform and thick enough..

Lets try sign language? Read the image and its title

Yes, sign language works better

I thought that is what you ment.

I can't believe they used just two points and not curve fit all the data from the edge detection to a best fit straight line ..

Thanks for clarifying Aussie speak..

 

[adamasgem]

Date: 2/3/2007 2:25:29 PM
Author: DiaGem

Date: 2/3/2007 2:01:24 PM
Author: adamasgem

Date: 2/3/2007 5:08:50 AM
Author: DiaGem
What are the two purple facets???

Just my way of flagging the facets used to initialize a girdle plane solution filter... There is actually a third one at the beginning.. The centroid of those three arbitrary facets ( randomly picked with a minumim angular separation) was used to initialize a solution to a plane through the girdle. (Any three points define a plane. I then used the rest of the girdle facets to further refine the definition of the plane.

Am i understanding correct? are you averaging the girdle facets degrees?

No I initialized a filter based on three widely separated vertices defined by girdle facet centroids.
Three points uniquely define a plane.
I then use the rest of the girdle facet centroids to find a "best fit" plane.
In the ideal case (for a normal round briliant) , the Z axis component of each of the rest of the girdle facet centroids will all lie on the initial plane from the first three centroids, but we are not in an ideal world, and no girdle is "perfect" .

If the girdle plane is tilted, the unrolled projections of the girdle on the 2D girdle profile I show, the best fit plane will appear as a sine wave.

 

[adamasgem]

Date: 1/29/2007 7:29:09 AM
Author: Serg

Date: 1/28/2007 3:07:33 PM
Author: adamasgem
Here is an example of phasing up girdle profiles, for identification purposes (sorry GIA, you can''t patent it, it is public domain now), or establishing a zero azimuth position for run to run consistency checks... identifying like facets for run to run statistics, or calibration use..

It is a statistically powerfull tool for identification. These were done with 40 slice RSL girdle profiles..

Any comments Serg on the validity of the approach?

Marty, We use combination rules for establishing a zero azimuth position.
ANY one rule is not enough usually for round diamond with perfect symmetry

I was a little intrigued regarding Serg''s comments about rules for a zero azimuth position.

Rhino kindly ran the same doggy stone (from his estate stuff, not typical of his goods) on the Helium and sent me 5 DMC files on the same stone, run on the same Helium.. Thanks Rhino..

When you change the orientation in DiamondCalc, the next file you open will retain the same orientation, which is really neat, so you get the same presentation.

I wanted to check the azimuth repeatability, and four out of five times the Helium was dead nuts on, as far as I''m concerned. I highlighted the pavilion facet (red outline) next to the extra facet on the left, and had to cut and paste to create the composite because of the 100K limit on uploads and then had to convert to a gif.

Note the very smalll facet azimuth variations on the four scans..

The "natural" on the right side was remarkedly consistent, and I don''t think that any scanner can be more repeatable. As to azimuth reference, I think my suggestion of the correlation analysis for syncing up two girdle profile may improve Serg''s algorithms for an azimuth reference.

Anyway''s, damn good job Serg, given the complexity of the problem, but still would probably need some fixes to be able to automatically use a single stone generically to check on a systems accuracy, without some user intervention..

 

[adamasgem]

Date: 2/3/2007 11:46:36 AM
Author: strmrdr
I hear ya...
Here is the problem as I see it if you put a dot at each crown pavilion girdle junction half way between them then connect it to the one opposite with a line you will get multiple planes of which none will match perfectly.
Which one is right?


btw I''m older than you think I am lol

Centroids of the facets are better to use, in my opinion.
As to which "plane solution" is "correct, that is why you use "optimal" estimation techniques, but you have to know the proper measurement noise model to get the "optimal" global solution to the problem, and that can be a problem in itself.

I''ll bet I might be old enough to be your daddy though ( G-D forbid), goin on 63, I am, I am

 

[strmrdr]

lol Marty your 10 years older than I was thinking...
Im 37....

It is an interesting problem, Iv been thinking about it.

One thing that sticks in my mind is that the crown and pavilion are actually independently cut as different processes. Therefore you would technically have pavilion and crown girdle planes and how well they match would be a huge measure of craftsmanship

 

[adamasgem]

Date: 2/3/2007 5:57:01 PM
Author: strmrdr
lol Marty your 10 years older than I was thinking...
Im 37....

It is an interesting problem, Iv been thinking about it.

One thing that sticks in my mind is that the crown and pavilion are actually independently cut as different processes. Therefore you would technically have pavilion and crown girdle planes and how well they match would be a huge measure of craftsmanship

Ok Son...

That is exactly why I suggested the crown-scallop and pavilion-scallop plane concepts and methodology for defining them above..

 

[strmrdr]

lol ok pops.

I see that now.
I must have missed it with all the noise about who gets credit for what LOL.

 

[strmrdr]

hmmmm
How about an alternative for finding tilt.
How about scribing a line from the table center to the culet center then connecting those lines at going around the outside of the diamond then compare the lines.
Wash rinse repeat.

By the shape and angle of the lines it will tell you where and how much the tilt is.
In a perfect diamond all the lines that are covering the same junctions but on the other sides should be identical.

 

[adamasgem]

Date: 2/3/2007 6:26:30 PM
Author: strmrdr
hmmmm
How about an alternative for finding tilt.
How about scribing a line from the table center to the culet center then connecting those lines at going around the outside of the diamond then compare the lines.
Wash rinse repeat.

By the shape and angle of the lines it will tell you where and how much the tilt is.
In a perfect diamond all the lines that are covering the same junctions but on the other sides should be identical.

Problems I see with that
1) Culet center ill defined
2) Cullet not parallel to table

 

[Garry H (Cut Nut)]

really from an optical view point the average pavilion axis between pavilion main facets is by far the most useful starting point - from that table tilt and crown angle variance etc should be calcuated.

I do not think there should be such a thing as an off centered culet.

This axis could then be used for average radius deviation (rather than out of round which is another useless construct as shown in our article - re triangular deviation not being out of round when measured)

 

[Serg]

Date: 2/3/2007 4:49:57 PM
Author: adamasgem

Date: 1/29/2007 7:29:09 AM
Author: Serg

Date: 1/28/2007 3:07:33 PM
Author: adamasgem
Here is an example of phasing up girdle profiles, for identification purposes (sorry GIA, you can''t patent it, it is public domain now), or establishing a zero azimuth position for run to run consistency checks... identifying like facets for run to run statistics, or calibration use..

It is a statistically powerfull tool for identification. These were done with 40 slice RSL girdle profiles..

Any comments Serg on the validity of the approach?

Marty, We use combination rules for establishing a zero azimuth position.
ANY one rule is not enough usually for round diamond with perfect symmetry

I was a little intrigued regarding Serg''s comments about rules for a zero azimuth position.

Rhino kindly ran the same doggy stone (from his estate stuff, not typical of his goods) on the Helium and sent me 5 DMC files on the same stone, run on the same Helium.. Thanks Rhino..

When you change the orientation in DiamondCalc, the next file you open will retain the same orientation, which is really neat, so you get the same presentation.

I wanted to check the azimuth repeatability, and four out of five times the Helium was dead nuts on, as far as I''m concerned. I highlighted the pavilion facet (red outline) next to the extra facet on the left, and had to cut and paste to create the composite because of the 100K limit on uploads and then had to convert to a gif.

Note the very smalll facet azimuth variations on the four scans..

The ''natural'' on the right side was remarkedly consistent, and I don''t think that any scanner can be more repeatable. As to azimuth reference, I think my suggestion of the correlation analysis for syncing up two girdle profile may improve Serg''s algorithms for an azimuth reference.

Anyway''s, damn good job Serg, given the complexity of the problem, but still would probably need some fixes to be able to automatically use a single stone generically to check on a systems accuracy, without some user intervention..

re:I was a little intrigued regarding Serg''s comments about rules for a zero azimuth position.

Rhino kindly ran the same doggy stone (from his estate stuff, not typical of his goods) on the Helium and sent me 5 DMC files on the same stone, run on the same Helium.. Thanks Rhino..

Rhino,
Was it diamond or CZ?
Model accuracy is too bad for clean diamond. It looks you had scan CZ or dirty diamond for Marty tests.
Could you repeat it for clean diamond?

 

[adamasgem]

Date: 2/4/2007 12:48:52 AM
Author: Serg

re:I was a little intrigued regarding Serg''s comments about rules for a zero azimuth position.

Rhino kindly ran the same doggy stone (from his estate stuff, not typical of his goods) on the Helium and sent me 5 DMC files on the same stone, run on the same Helium.. Thanks Rhino..

Rhino,
Was it diamond or CZ? Rhino said it was a dog diamond bought off the street, EF and large naturals..
Model accuracy is too bad for clean diamond. It looks you had scan CZ or dirty diamond for Marty tests.
Could you repeat it for clean diamond? I thoght the results were pretty damn good..

 

[Garry H (Cut Nut)]

It would help to know which are the naturals - they may not be flat and measurements could vary.