Measurement accuracy choices???? A poll

[oldminer]

Okay, so the response is: The model is good enough to use as far as the market demand side goes. If this is what consumers truly want, I'd be surprised. Some folks won't care at all, but those who demand correctness may not be so pleased. People who want information on buying a diamond, want the best, accurate information. At least they now know that a substantial element of this information is known to be approximate. It is approximate BECAUSE it is PREDICTED. Direct measure of light performance is not predictive and that's been my point.

The other point is that you could measure with increased accuracy for a higher cost for the tool. If one did not wish to make predictions, your present tool is totally adequate for excellent measurements on its own. Even you are content with how well Helium measures and I don't doubt it is sufficiently accurate to make good round model predictions. Do you believe this will be the case for fancy shapes of all different kinds, too?

 

[Serg]

Date: 7/15/2006 4:00:23 PM
Author: oldminer
Serg: I don''t have to stop it just because you say so. If you can''t answer what was asked, then maybe you already know you can''t answer without threatening me. Say you measure everything possible that might have an effect on predicitve outcome or say you don''t measure everything that may have an effect. If I am incorrect, I wait for you to correct me. This isn''t some dictatorship and I can ask questions. You can refuse to answer and we can make our own conclusions. You can answer and set the record straight if I am wrong...... Even if you respond that you are not certain about the answer, people will think you are still just the honest person we believe you are. I am not making this insulting, you are.

Dave ,

I gave correct answer to your Technically incorrect question.
Your question is not correct, see example:
1) Sarin and Helium measure everything possible( Build full real 3d model ) that might have an effect on predictive outcome
2) Sarin and Helium 3D models can give different predictive outcome Light metrics.

re:I don''t have to stop it just because you say so

You should not do it(just because you say so), but you need do it ( or proof your statements).

 

[Serg]

Date: 7/15/2006 4:00:23 PM
Author: oldminer
Okay, so the response is: The model is good enough to use as far as the market demand side goes. If this is what consumers truly want, I''d be surprised. Some folks won''t care at all, but those who demand correctness may not be so pleased. People who want information on buying a diamond, want the best, accurate information. At least they now know that a substantial element of this information is known to be approximate. It is approximate BECAUSE it is PREDICTED. Direct measure of light performance is not predictive and that''s been my point.

The other point is that you could measure with increased accuracy for a higher cost for the tool. If one did not wish to make predictions, your present tool is totally adequate for excellent measurements on its own. Even you are content with how well Helium measures and I don''t doubt it is sufficiently accurate to make good round model predictions. Do you believe this will be the case for fancy shapes of all different kinds, too?

reirect measure of light performance is not predictive and that''s been my point.

Correct. But it does not mean what DIRECT measure of light performance CAN give better accuracy then INDIRECT method.

reo you believe this will be the case for fancy shapes of all different kinds, too?

I belive in:
1) It is impossible do for ANY fancy cut by shadow method only.
2) But it can be done by combination 2-3 methods at least. When it will be necessary for big part market we( or other company) will do it.

 

[adamasgem]

David, I''m a little confused with the intent of the “poll”, and I think your questions might be a little misleading..

The industry wants fast, cheap and accurate, three mutually exclusive objectives…

The “consumer” wants accurate, maybe not realizing that it adds to the cost of their “paper”

There are multiple factors that enter into a linear (or angular) measurement error using a single camera as a measurement device without external information inputs

The basic ones are:

1) The basic pixeliization (quantization) of your system, how finely can you divide up a given "screen"
2) The total virtual screen size you use (X by Y pixels)
3) The magnification, if you will, lens that you use to fill up the "screen" with the stone image, either a zoom lens or multiple lens
4) The calibration procedure you use, i.e. are you correcting for any non linear optic effects, such as pin-cushion distortion or how accurate are the fiducials used to generate a baseline linear calibration (common approach like the multiple pedestals on a Sarin)
5) The base-length you are measuring ie: are you measuring a short star facet, or a pavilion main length (each measurement has a different relative accuracy)
6) The relative angle of the edge you are reading with respect to your 2 dimensional measurement..
7) The angular resolver accuracy as the stone is turned
The number of measurements made (elementary statistics) and the number of angles from which they are made
9) The algorithms used to gather and process the data, i.e. edge detection, optimality of the data processing and correlations between measurements errors and results which can IMPROVE the basic linear measurement error because of constraints based on physical reality (2D versus 3D estimation, two planes intersect in a straight line).

Replacing the same stone in the same machine and taking multiple scans (Bayesian approach) utilizing the data from the previous scans to improve the “estimation” accuracy, after randomizing the initial azimuth orientation.

Just counting the number of pixels doesn’t give you the “measurement” accuracy or more correctly the “estimation accuracy”, as you imply. It is not as simplistic as you make it seem.

I for one would like to see less fluff and more hard information, and those that know my background know that I am certainly qualified to separate the wheat from the chaff on this subject.

Just for a start, I would like to see a covariance matrix of their measurement errors from ANY producer of a device, no one has done that yet.

 

[Garry H (Cut Nut)]

Date: 7/15/2006 4:00:23 PM
Author: oldminer
Okay, so the response is: The model is good enough to use as far as the market demand side goes. If this is what consumers truly want, I''d be surprised. Some folks won''t care at all, but those who demand correctness may not be so pleased. People who want information on buying a diamond, want the best, accurate information. At least they now know that a substantial element of this information is known to be approximate. It is approximate BECAUSE it is PREDICTED. Direct measure of light performance is not predictive and that''s been my point.

The other point is that you could measure with increased accuracy for a higher cost for the tool. If one did not wish to make predictions, your present tool is totally adequate for excellent measurements on its own. Even you are content with how well Helium measures and I don''t doubt it is sufficiently accurate to make good round model predictions. Do you believe this will be the case for fancy shapes of all different kinds, too?

I am trying ot follow this thread and help with languauge and other missunderstandings.
Dave it seems you edited your last thread - the 30th one at 4.00 on 15/7/2006.

Sergey copied this part:
Date: 7/15/2006 4:00:23 PM
Author: oldminer
Serg: I don''t have to stop it just because you say so. If you can''t answer what was asked, then maybe you already know you can''t answer without threatening me. Say you measure everything possible that might have an effect on predicitve outcome or say you don''t measure everything that may have an effect. If I am incorrect, I wait for you to correct me. This isn''t some dictatorship and I can ask questions. You can refuse to answer and we can make our own conclusions. You can answer and set the record straight if I am wrong...... Even if you respond that you are not certain about the answer, people will think you are still just the honest person we believe you are. I am not making this insulting, you are.


So thank you for moderating your question.
Sergey has also answered this edited question - the point about accuracy is that there are other ways to improve the accuracy of 3D models that Sergey can employ - but they are far and away beyond anyones requirements just now.
We plan to develop a cut grading (and new cut generation) system first. Then there will be a need to prove those cuts. In some cases there might be only a few degrees difference between facets - or even curved facets - and no shadow system can currently measure such things.
But as far as proportion accuracy - compared to what GIA think is required - Helium is in a feild of its own.


Can you scan the brochure that has led Dr. A to think Helium becomes less accurate as size goes up? I suspect it is simply a conservative estimate of what a smaller stone on the 2:1 scannner might yeild as an error.

 

[oldminer]

Polls always have some intent, intended or accidental. I want to know what people think, but I used some careful wording in the questions to help elicit responses. Also, the thread has sort of veered a little off course, but any and all discussions are useful to all concerned.

Many of the questions you ask are well beyond me, but I will see that someone qualified takes a look. I will make some effort in getting direct responses next week. If all of this furthers the state of the art, we can be content.

 

[Cehrabehra]

Date: 7/15/2006 7:03:22 PM
Author: oldminer
Polls always have some intent, intended or accidental. I want to know what people think, but I used some careful wording in the questions to help elicit responses. Also, the thread has sort of veered a little off course, but any and all discussions are useful to all concerned.

Many of the questions you ask are well beyond me, but I will see that someone qualified takes a look. I will make some effort in getting direct responses next week. If all of this furthers the state of the art, we can be content.

And speaking from the perspective of a little person, an ignorant person, a consumer, I am grateful that you all continue this debate. It''s interesting to watch and read. ::backing away to let the big boys talk::

 

[oldminer]

I have asked Dr. Aggarwal for some responses and maybe next week we will have some new facts to look over.

I want everyone to understand that the edited message which I replaced was removed because I realized as soon as I had posted it that Sergey did not red highlight all of his questions and comments and I had misread much of the message. It was not approriate to leave it and I completely re-wrote a new message that was more fitting. We do everything here to be friendly and mutually cooperative. Even competitors can be polite, have a dinner together, understand eachother and assist the business we all are deeply dedicated to. We may have different approaches, different opinions and different levels of education, but we all can be gentlemen and friends while we compete. At the same time we seemed opposite to eachother, we are learning from eachother, too.

While Dr. Aggarwal and I work on responses to questions raised, I have not gotten anyone to answer my one, simple, uneducated question which I feel is critical.

"What single thing is Graded by a Predictive measure when a tool extists to take a Direct measure?

It is not an accepted methodology to grade a physcial object that one can grasp and examine by creating a likeness and predicting a grade. It is an accepted methodology to create a likeness to assist in further fabrication or experiment with modification for future production.

 

[Serg]

Dave,

You receive answer from me.
Please now give your answer.
Particular:

1)Why are Helium cameras and lens WRONG?

2)re:These vendors want the trade and consumers to put full faith in their devices, but when you go from providing reasonably accurate measurements into predicting Light Behavior with those same approximate measurements, you have a large or significant problem.
Please show such example for 3D Helium round diamond model.

Your are very often do wrong and incorrect statements about Imagem competitors without any proof and facts.
And your ignore questions about proof such statements.
Such policy is not good for ' friendly and mutually cooperative"

 

[oldminer]

Like I said, I will get the best technical answers for you during next week. Keep cool in the meantime......

 

[oldminer]

Dr. Aggarwal has given me a response to the question he considered relevant. Questions that he feels are simply argumenative, but not valid have not gone unnoticed, but there is little reason to create needless postings. He is always ready to directly involve himself with interested parties that are investigating the technology.
xxxxxxxxxxxxxxxxxx

ImaGem is interested in presenting the current status of technical advances applicable to the gem and jewelry industry which we hope will help the consumers, and other interested technically savvy parties. We wish not to target a particular product or individual. For this reason, technical discussions are presented at considerable length in ImaGem’s monthly newsletter. There are gauging systems with different levels of measurement errors and yet have a broad installed base. Clearly, accuracy or efficiency or even value creation are not always primary reasons for commercial success of a technology.

The errors reported by Octonus are in their literature are larger for larger stones. The back of the “Helium Polish” literature under specifications reports: “3D-model accuracy: 3.7mm – 8.2mm polish size ‘better than 10 microns’; 7.4mm – 16.6mm polish ‘better than 20 microns’; 14.8mm – 33.4mm polish size ‘better than 40 microns’. This indeed is a correct statement given the primary magnifications used in the machines, other things being equal. If subpixel analysis is done at the same fractional level in the systems, the errors will be similar to those reported. Primary magnification and size of pixels are important factors in determining the resolution capability of an image processing system, to state otherwise would be incorrect. Instead of giving the readers fragmented answers and assertions, an excellent primer on image processing can be found in “The image processing handbook” by John C. Russ or Edmund Optics Catalog which will support the explanations of gauging repeatability provided by ImaGem.
Providers of gauging systems must distinguish between accuracy and repeatability. These are important and different concepts which are often used interchangeably. For example, a system may be accurate but may not be highly repeatable. To assess gauging system accuracy, method of calibration and repeatability with a probability statement should be provided.

It can be mathematically shown that errors in angular measurements are correlated with errors in linear measurements as well as the size of a stone. Smaller is the size of a stone, larger will be errors in angular measurements, quid pro quo. Proper reporting of system errors should therefore state errors in linear measurements accompanied by a statement that errors in angular measurements will vary with the size of a stone. An alternative would be to report angular errors for stones of different sizes. The current practice of reporting angular error as a single number does not represent the actual performance of a system, except in those cases in which data are dummied down to calculate angular measurements.

There are really two ways to talk about the capability of an image processing system. One is based on our knowledge of hardware and software and the other is based on actual performance. ImaGem has outlined some of the technical explanations and provided sources which can further illuminate on the subject without using technical jargon that serves no other purpose but confuse the reader.

As we have offered before, we will be happy to process stones sent to ImaGem via AGA for grading to find for yourself the accuracy and repeatability of the systems developed and now in use in U.S.A. and in India.

 

[Serg]

Dave,

Is it all answer?

Imagem and you again do not like to give answer on my questions( to proof Imagem incorrect statements)

 

[Serg]

Dave,

Because answer to below question from Dr. Aggarwal is absent. Please do such tests in your Lab.
I sure it is very important information for market. ( And it can clarify problem with Imagem report #211995CeF)

re:"The Imagem scenario wants weight and depth entered for each diamond. These entries can be skipped, but it verifies the other data accuracy in certain respects. "

If operator will misprint total height . For example instead correct 4.09 he will entry 4.03. What height will in Imagem report? ( 4.09, 4.03 or something between?)

 

[adamasgem]

Date: 7/18/2006 11:48:34 AM
Author: oldminer
Dr. Aggarwal has given me a response to the question he considered relevant. Questions that he feels are simply argumenative, but not valid have not gone unnoticed, but there is little reason to create needless postings. He is always ready to directly involve himself with interested parties that are investigating the technology.
xxxxxxxxxxxxxxxxxx

ImaGem is interested in presenting the current status of technical advances applicable to the gem and jewelry industry which we hope will help the consumers, and other interested technically savvy parties.

We wish not to target a particular product or individual. For this reason, technical discussions are presented at considerable length in ImaGem’s monthly newsletter. There are gauging systems with different levels of measurement errors and yet have a broad installed base. Clearly, accuracy or efficiency or even value creation are not always primary reasons for commercial success of a technology.

The errors reported by Octonus are in their literature are larger for larger stones. The back of the “Helium Polish” literature under specifications reports: “3D-model accuracy: 3.7mm – 8.2mm polish size ‘better than 10 microns’; 7.4mm – 16.6mm polish ‘better than 20 microns’; 14.8mm – 33.4mm polish size ‘better than 40 microns’. This indeed is a correct statement given the primary magnifications used in the machines, other things being equal. If subpixel analysis is done at the same fractional level in the systems, the errors will be similar to those reported. Primary magnification and size of pixels are important factors in determining the resolution capability of an image processing system, to state otherwise would be incorrect. Instead of giving the readers fragmented answers and assertions, an excellent primer on image processing can be found in “The image processing handbook” by John C. Russ or Edmund Optics Catalog which will support the explanations of gauging repeatability provided by ImaGem.
Providers of gauging systems must distinguish between accuracy and repeatability. These are important and different concepts which are often used interchangeably. For example, a system may be accurate but may not be highly repeatable. To assess gauging system accuracy, method of calibration and repeatability with a probability statement should be provided.

It can be mathematically shown that errors in angular measurements are correlated with errors in linear measurements as well as the size of a stone. Smaller is the size of a stone, larger will be errors in angular measurements, quid pro quo. Proper reporting of system errors should therefore state errors in linear measurements accompanied by a statement that errors in angular measurements will vary with the size of a stone. An alternative would be to report angular errors for stones of different sizes. The current practice of reporting angular error as a single number does not represent the actual performance of a system, except in those cases in which data are dummied down to calculate angular measurements.

There are really two ways to talk about the capability of an image processing system. One is based on our knowledge of hardware and software and the other is based on actual performance. ImaGem has outlined some of the technical explanations and provided sources which can further illuminate on the subject without using technical jargon that serves no other purpose but confuse the reader.

As we have offered before, we will be happy to process stones sent to ImaGem via AGA for grading to find for yourself the accuracy and repeatability of the systems developed and now in use in U.S.A. and in India.

This is an interesting restatement of my post outlining factors contributing to inherent accuracy. There are a few interesting statements here..

1) "Clearly, accuracy or efficiency or even value creation are not always primary reasons for commercial success of a technology." That is so true, BS sometimes prevails.

2) "“3D-model accuracy: 3.7mm – 8.2mm polish size ‘better than 10 microns’; 7.4mm – 16.6mm polish ‘better than 20 microns’; 14.8mm – 33.4mm polish size ‘better than 40 microns’. This indeed is a correct statement given the primary magnifications used in the machines, other things being equal. " . Given a stated accuracy of better than 40 microns, in the worst case, that is much better than Sarin stated accuracy levels, i.e. that in use by GIA..

3) "To assess gauging system accuracy, method of calibration and repeatability with a probability statement should be provided." OK, would IMAGEM provide this information.

4) "Providers of gauging systems must distinguish between accuracy and repeatability.These are important and different concepts which are often used interchangeably." I agree. "For example, a system may be accurate but may not be highly repeatable." That is why the probability statement must be given. A system may be highly repeatable, but highly inaccurate. Most specs say things about "accuracy" like +/- 0.02mm, but don''t state what they mean by it (i.e. 1 sigma, 2 sigma, max ????). A statement like "better than" implies, to me, a worst case scenario, at minimum, a 3 sigma bound. An unqualified number, like 34.5 implies a worst case accuracy to one half the number of decimal points used. Are you listening GIA?..

Now I have a question, I seached the imagem web site http://www.imageminc.com, and no where could I find any specifics on measurement accuracy, other than stating that measurements "are reported to three decimal places" and "highly accurate", and as you point out a probability statement is need. Would you please point me to a specific statement about accuracy.

I appologize if I couldn''t find it, it certainly isn''t in the technical FAQ''s. Since you are seemingly complaining about the subject, shouldn''t this be corrected?

 

[Serg]

Marty,

(systematical + 3 sigma(accidental))="better than"
I do not know what is Sarin and Imagem accuracy. I think it could be sigma for Imagem and 3sigma for Sarin.
But Imagem use weight ( from scale) and total height( from micrometer) for control( and may be for correction results)
We need separate accuracy for
1) Total height
2) Min and Max diameter
3) Average diameter
4) Other
a) Diameter Round diamond( or fancy round)
b)Diameter rectangular cuts
c) Other

 

[adamasgem]

Date: 7/18/2006 2:03:37 PM
Author: Serg

Marty,

(systematical + 3 sigma(accidental))=''better than''

I do not know what is Sarin and Imagem accuracy. I think it could be sigma for Imagem and 3sigma for Sarin.
But Imagem use weight ( from scale) and total height( from micrometer) for control( and may be for correction results)

We need separate accuracy for
1) Total height
2) Min and Max diameter
3) Average diameter
4) Other

a) Diameter Round diamond( or fancy round)
b)Diameter rectangular cuts
c) Other

Thanks Sergey for your definition of "better than".. AT worst case then, we are talking one sigma of 14 microns for the one sigma uncertainty, and given that I understand cutters may polish using 1 to 5 micron grit, although some may use sub micron grit.

Sarin used to quote +/-0.02mm and +/- 0.4 degrees but that was from a spec sheet on the older professional system before their camera and software upgrade and they didn''t state what they meant by it.

Imagem''s accuracy is entirely unknown or unstated, other than their statement that they report to three decimal places, that is all i could find.

The total height issue question you raised is interesting, although if done right, could help "improve" overall accuracy, but it could work to the detriment, GIGO.

 

[Serg]

re:The total height issue question you raised is interesting, although if done right, could help "improve" overall accuracy, but it could work to the detriment, GIGO.

in each unique case you do not know reason of error.

It could be wrong vertical scale, dust on table, bad accuracy for culet zone, MISPRINT, wrong data from micrometer.

Most probable Imagem are thinking "It is dust"( if height from micrometer less than optical height) and are changing crown height. It will change Table too much. And change mass. For save mass imagem should change diameter.

Very bad method and too risky.
Think about misprint 60 microns( when you will read Indian example with Imagem height 4.023 and Sarin and Helium height 4.09), or break culet( you need use micrometer very carefully).
Point diamond culet will do dints in micrometer plate.
But if you want show "accuracy" _+ 0.002 and height should be same like in micrometer , you have not choice . Is it good goal?

 

[adamasgem]

Date: 7/18/2006 3:09:21 PM
Author: Serg

re:The total height issue question you raised is interesting, although if done right, could help ''improve'' overall accuracy, but it could work to the detriment, GIGO.

in each unique case you do not know reason of error. You are 100% correct

It could be wrong vertical scale, dust on table, bad accuracy for culet zone, MISPRINT, wrong data from micrometer. Yup

Most probable Imagem are thinking ''It is dust''( if height from micrometer less than optical height) and are changing crown height. It will change Table too much. And change mass. For save mass imagem should change diameter.

Very bad method and too risky. Yup Gabage in garbage out...

Think about misprint 60 microns( when you will read Indian example with Imagem height 4.023 and Sarin and Helium height 4.09), or break culet( you need use micrometer very carefully). Very VERY carefully. Sort of same effect when idiots pick up Princess cuts and break the tips or snap a spring loaded leverage guage on a culet. BTW, that was an very interesting height example...

Point diamond culet will do dints in micrometer plate. And a leveridge guage

But if you want show ''accuracy'' _+ 0.002 and height should be same like in micrometer , you have not choice . Is it good goal? Depends on whether you are talking mm or inches

, but if mm then very decent goal.

 

[He Scores]

Gee.....and some of us cutters use a 7X eye loupe and template guage to get accurate measurements.



I'm one of the cutters that Marty said used sub-micron diamond powder. I buy 0-1/4 micron. I told the guy..."hey....and don't give me a whole jar of zero micron either!"

Bill Bray
Diamond Cutter

 

[Serg]

re:Inaccurate and not repeatable Accurate, but not repeatable Inaccurate but repeatable Accurate and repeatable

(a) (b) (c) (d)

(Farnum, Nicholas R. 1994. Modern Statistical Quality Control and Improvement, Duxbury).

Imagem article


Legend "Accurate , but not repeatable" in picture (b)is misleading for case when we use one measurement( one number in report. Point in map is one value in report).

In our field I recommend use below Legends
1) Bad Accuracy.( Big random error+ big systematical error)
2) Bad Accuracy. ( Big random error)
3) Bad Accuracy( Big systematical error. Small random error)
4) Good Accuracy( Small random error. Systematical error is absent)

 

[oldminer]

Imagem employ a three decimal place Mititoyo screw type digital micrometer for depth measures. It is a lot more precise and instrument than the typical slide micrometer we generally employ in the business with two decimal places that most jewelers and labs use. The depth measure provided from that instrument does calaibrate the device within a certain limited range. One must be careful to use such a precision tool with care. There is a torque limiting handle on it that is to be used with each measure. It does not vary from a zero reading if the right care is taken. I suppose a careless lab tech might get to a non-zero situation and install a gauging error. This may have been the case in Surat. However, such an error is a human one, not a machine error. In this degree of precision, one must protect devices from dust, too. Care is crucial. Vigilant lab staff is crucial to any lab, digital or manual.

The weight of a diamond, while fed directly from a scale into the ImaGem system is not used for calibration, but for comparision to the calculated result by the chief gemologist at the end of the process. It is used as a warning flag and is a part of the re-recognition system. If there is a discrepancy, the chief gemologist looks over the situation to be sure things have been correctly processed, both human and device issues being then addressed.

My way of thinking is that one never creates a model of an existing physical item in order to GRADE it. One grades the actual item.

One does create models of physically available, but valuable items, in order to experiment or modify their configuration. This would allow cutters to experiment without any real cost of development. The idea of modelling in order to prototype samples is worthy enough. The idea of making a model in order to come to a grade is bad science and bad lab work. Every error in measurement leads to errors in the model. If a model is then made perfect by dummying down the measures, then the model is faulty. How can the results be reliable if as suggested by others it is a GIGO situation?

Has anyone thought of even one thing GRADED by creation of a model and not DIRECTLY graded from the physical item? Still no answer for that one? (I have none myself to offer, and was looking for an example that could be used by proponents of modelling in their own defense) One shoud not go into battle unarmed.....

 

[Serg]

Date: 7/19/2006 7:27:07 AM
Author: oldminer
Imagem employ a three decimal place Mititoyo screw type digital micrometer for depth measures. It is a lot more precise and instrument than the typical slide micrometer we generally employ in the business with two decimal places that most jewelers and labs use. The depth measure provided from that instrument does calaibrate the device within a certain limited range. One must be careful to use such a precision tool with care. There is a torque limiting handle on it that is to be used with each measure. It does not vary from a zero reading if the right care is taken. I suppose a careless lab tech might get to a non-zero situation and install a gauging error. This may have been the case in Surat. However, such an error is a human one, not a machine error. In this degree of precision, one must protect devices from dust, too. Care is crucial. Vigilant lab staff is crucial to any lab, digital or manual.

The weight of a diamond, while fed directly from a scale into the ImaGem system is not used for calibration, but for comparision to the calculated result by the chief gemologist at the end of the process. It is used as a warning flag and is a part of the re-recognition system. If there is a discrepancy, the chief gemologist looks over the situation to be sure things have been correctly processed, both human and device issues being then addressed.

My way of thinking is that one never creates a model of an existing physical item in order to GRADE it. One grades the actual item.

One does create models of physically available, but valuable items, in order to experiment or modify their configuration. This would allow cutters to experiment without any real cost of development. The idea of modelling in order to prototype samples is worthy enough. The idea of making a model in order to come to a grade is bad science and bad lab work. Every error in measurement leads to errors in the model. If a model is then made perfect by dummying down the measures, then the model is faulty. How can the results be reliable if as suggested by others it is a GIGO situation?

Has anyone thought of even one thing GRADED by creation of a model and not DIRECTLY graded from the physical item? Still no answer for that one? (I have none myself to offer, and was looking for an example that could be used by proponents of modelling in their own defense) One shoud not go into battle unarmed.....

Re: Has anyone thought of even one thing GRADED by creation of a model and not DIRECTLY graded from the physical item? How can the results be reliable if as suggested by others it is a GIGO situation?

Dave if you want receive HUMAN grade without Human, You need at least model Human.

Imagem are modeling Human reaction.
But because Imagem has not Model diamond and Light environment, Imagem are sending GI to Imagem Human model and are receiving GO. Because Imagem are using DIRECT input to Human model, Imagem can not receive adequate input to Human .( Stereo, resolution, Dynamical range,…..)


Re: However, such an error is a human one, not a machine error.

Simple Human error is too critical for Imagem technology. It is mean Technology is not good at least.

BTW. When we showed this example to you 1 year ago, you did not tell “ it is Human error”.

You Told “.....
The depth was measured by a high tech digital device, by direct measure.... So, the response from me is that Sarin and Helium have some problems with non-contact measures.... His manager does not quite know how to tell him that thousands of diamonds already in inventory have bad measurements. It is a situation that calls for discretion....”

In real world:
Imagem report is wrong( May be due combination human error and not smart enough Imagem algorithm) , Sarin and Helim reports are right

Re: The idea of making a model in order to come to a grade is bad science and bad lab work.

It is looks like slander. Or could you proof it ?

re:If a model is then made perfect by dummying down the measures, then the model is faulty

What do you means ? Please clarify where is dummying down the measures?

 

[oldminer]

I don't want to slander you or your device. I said it was perfect for certain uses and not perfect for others. If you don't know what dummying down means, I suggest you ask Marty Haske for an good and probably lengthy explanation. He is well acquanited, as am i, with many dummies.

Marty may just not make you as annoyed as I have done.

Do you have an example of the item graded by modelling for us?

 

[Serg]

Re: The idea of making a model in order to come to a grade is bad science and bad lab work. .. I don''t want to slander you or your device.



Dave,

The IDEA to use 3D model polish diamond for grade Beauty had come to me 1o years ago.( May be somebody( for example Shanon from DTI) had find same idea yearly, I do not know)
I am working 10 years for realize with IDEA to technology.

It could be done good or bad.

But this IDEA is NOT BAD SCIENCE.

 

[Serg]

re: If you don''t know what dummying down means

I did not ask "What is it?"
I asked "where is it?"

 

[oldminer]

"Where is it?"

The machine errors inherent in measuring the diamond are fed into your modelling software. When you get the model created, any errors in end points or angles are filled in by software fixes, dummying down, to make a complete model with no overlapping wires or gaps in end points. That''s where it is hiding. This no problem for a prototype model for cutters and sigth holders to use, but it may be that the model does not exactly look or act like the actual diamind you have in your hand. Why not just measure the light behovior of a diamond directly as we suggest? Any reason you object to that strategy?

12 years ago or more, ImaGem had the direct measure IDEA in its mind. Some years later we patented some of these ideas. We think that these are good ideas and also good science.

 

[Serg]

Date: 7/19/2006 11:22:54 AM
Author: oldminer
''Where is it?''

The machine errors inherent in measuring the diamond are fed into your modelling software. When you get the model created, any errors in end points or angles are filled in by software fixes, dummying down, to make a complete model with no overlapping wires or gaps in end points. That''s where it is hiding. This no problem for a prototype model for cutters and sigth holders to use, but it may be that the model does not exactly look or act like the actual diamind you have in your hand. Why not just measure the light behovior of a diamond directly as we suggest? Any reason you object to that strategy?

12 years ago or more, ImaGem had the direct measure IDEA in its mind. Some years later we patented some of these ideas. We think that these are good ideas and also good science.

re:to make a complete model with no overlapping wires or gaps in end points

We do not do such correction( or any other correction) in OctoNus products. Our clients and distributors very often asked do it earlier. But I rejected all such suggestion.
Do you know it?
I see You do not know what is Helium and 3D model from Helium.

re:We think that these are good ideas and also good science.

Is it enough reason to think What other idea is bad?

 

[Serg]

Date: 7/19/2006 11:22:54 AM
Author: oldminer
''Where is it?''

The machine errors inherent in measuring the diamond are fed into your modelling software. When you get the model created, any errors in end points or angles are filled in by software fixes, dummying down, to make a complete model with no overlapping wires or gaps in end points. That''s where it is hiding. This no problem for a prototype model for cutters and sigth holders to use, but it may be that the model does not exactly look or act like the actual diamind you have in your hand. Why not just measure the light behovior of a diamond directly as we suggest? Any reason you object to that strategy?

12 years ago or more, ImaGem had the direct measure IDEA in its mind. Some years later we patented some of these ideas. We think that these are good ideas and also good science.

just see this link

 

[oldminer]

Your link is excellent proof of just how great a device you make. The models look superb,as I have often said. What we measure is beyond the human eye to distinguish, much like what differentiaes VVS1 from VVS2 in certain stones. It is darned near invisible, but there are large financial consequences. A model that is "exactly" like a physical diamond cannot be made from measurements have inherent, though small machine error. The literature you produce, which I previously have named as you requested, tells of as much as 40 microns of error on larger diamonds. The model produced from such measures will simply not be identical to the physical diamond.

I think all who have read this thread from beginning to end will now be pretty well convinced of the fairness of the criticism. It is not personal at all. When one has a diamond in their possession for GRADING, it makes more sense to directly measure its performance than to create a model and then calculate light performance. Models are best used for prototyping and discovery than for laboratory grading.

Different people have different ideas. The path of discovery is filled with good ideas that don't have universal application. Your idea works well in certain situations which you or I have defined. Our idea works well in certain situations which we have defined. It is not up to us, but up to the informed market which strategy to adopt for each instance of use. This is a growing market and there is room for all of us to make customers and money. My goal is to present information that is not so much of a commercial nature as it is educational. I realize how this thread is of huge interest to those who lurk, learn and form opinions. Many are not qualified to participate ,and I make a big effort to make my postings entertaining, informative and generally meaningful. I also want to be seen as open and fair to anyone with another opinion. This wonderful forum, Pricescope, is of benefit to all of us and I value it tremendously.

The last thing I wish is to discourage anyone from particpating on Pricescope or in this thread. It is an open forum. You can make yourself look smart, wicked, informed, idiotic, sarcastic, mellow, or anything else you prefer. Its part of the reason we are reading the commentary.

Thanks to everyone who has posted a response to the POLL. The data is appreciated.

 

[adamasgem]

Date: 7/19/2006 11:22:54 AM
Author: oldminer
''Where is it?''

The machine errors inherent in measuring the diamond are fed into your modelling software. When you get the model created, any errors in end points or angles are filled in by software fixes, dummying down, to make a complete model with no overlapping wires or gaps in end points. That''s where it is hiding. This no problem for a prototype model for cutters and sigth holders to use, but it may be that the model does not exactly look or act like the actual diamind you have in your hand. Why not just measure the light behovior of a diamond directly as we suggest? Any reason you object to that strategy?

12 years ago or more, ImaGem had the direct measure IDEA in its mind. Some years later we patented some of these ideas. We think that these are good ideas and also good science.

Dave: Let me dummy it down. If you IMAGEM, say you "measure light performance", BUT you are measuring only one aspect of it, and in ONLY one(?) lighting condition, up until this point, a "mysterious" and undefined environment(s). Is that "envirionment" flawed in itself, and not representative of what a "consumer" or even the trade, might or might not perceive as THE envirionment to be used to DISCRIMINATE between diamonds, sort of like the example above of repeatable but not accurate. And, it appears, based on the tenor of this discussion, that you are, impling that your "measurements" are error free, which I can guarantee you, with 100% certainty, that they are not.

Sergey''s example of accuracy versus repeatability should be illustrative enough for you, yet the only statement I can find on Imagem''s web site deals with neither aspect, only that you "report" a number to three decimal places.

How about some MEAT and less marketing and hyperbole, so that we can intelligently, hopefully, embrace or critcize the chest beating.