PR# 264872 The analysis view command shows a different hit pattern in v2010 t

huangyhg

<< Don Ruggieri  --  10/14/09  12:33:13>>
data is in 10142009_01a.zip (V35MR2), 10142009_01b.zip (v37mr3), 10142009_01c.zip (V40r), 10142009_01d.zip (V43MR2) and 10142009_01e.zip (V2010QA).  The analysis view command shows a different hit pattern in v2010 than all previous verisons.

Steps to reproduce:
I was researching a question on the histogram as part of the analysis view.  I measured a plane and dimensioned its flatness, and then displayed the histogram and the corresponding color bar.

In V35Mr2, V37MR3, V40r, V43MR2 they all look the same as far as the hit pattern.  There is one hit in every band except for the band which is just to the right of the upper tolerance, which has a quantity of 2 hits associated with it.

The problem is that (only) in v2010QA there are also 2 hits in the grouping just to the left side of the upper tolerance bad.  This is not true in any other version.
<<END>>


<< Changes made by Don Turcotte (Review Notes) -- 11/11/09  09:17:13>>
The reason for the different hit pattern in the histograms between V43 and 2010 is a result of the fix for #245093 (Dimensions are calculated incorrectly because the calculation is based on the selected decimal precision.)  This fix was put in V44B so both 2009 and 2010 will behave the same.  In these versions of PC-DMIS the histogram bins are determined using the full machine precision (for MM this is 4 places).  In earlier versions of PC-DMIS, the histogram bins were determined using the DISPLAYPRECISION setting, which for this program is defaulted to 3.  If you set the DISPLAYPRECISION to 4, then you will get the same histogram distribution in the earlier versions as for 2009 and 2010.  In PC-DMIS 2009 and 2010, the DISPLAYPRECISION will not have any effect on the histogram distribution since these versions are always using the full machine precision.
<<END>>
<< Don Turcotte (Development Notes)  --  11/10/09  17:27:29>>
I am working on this.  Comparing code between V432 and 2010, I don't see why the histograms should be different, but they are.  Still looking...
<<END>>
<< Don Ruggieri (Test Notes)  --  10/14/09  15:32:57>>
There is more evidence of this in the "02" series of attachments as well.  In a through c there are 9 bars in the histogram between tolerance bands, this is consistent and explainable.  In the "d" zip file, from v43MR2, this doubles to 18, and it is my understanding that this quantity correlates to changes that were made in the software.  In v2010 QA, theree are also 18 bars, but the distribution is different, as shown in the respective bitmap screen captuures.
In 10142009_02d.zip the bitmap shows a hit distribution of -
6 3 5 3 5 6 12 6 9 16 3 4 4 2 2 2 1 3
for a total of 92 hits betwen limits
In 10142009_02e.zip the bitmap shows a hit distribution of
4 3 4 3 7 8 10 4 16 10 3 3 4 2 3 2 0 3
for a total of 89 hits between limits.
Both the placement and the quantity of hits (between limits) has changed for this roundness calculation from v43MR2 to v2010QA version.
<<END>>