Suggest to add PV data below Color scale diagram

[VictorXiao22]

To calculate PV value we need to click a icon and drag two line to show the PV value. Is it possible to show PV and Rms values in Color scale diagram directly?

[atsju]

Would it make sense to just add "PV" in left pane instead ? In order not to bloat the diagram.

Let's see what Dale and George think about it.

[githubdoe]

The RMS is already shown several places including the main screen. The PV computed from the pixel histogram is displayed in the report if that option is selected. Yes there are some web sites that show a conversion but that is only valid if computed from zernike values which can filter real PV values down to false values. That is why the pixel histogram was developed.

Having said all of that PV is a very poor quality metric compared to the others that interferometry creates. It use should be avoided.

Since deriving PV from rms is not accurate at all it can not be displayed with any accuracy except by looking at the profile plot or getting a more accurate account using the pixel histogram. Since that is a manual process then no a PV value should not be made up and placed on the main display.

[githubdoe]

Concerning the PV shown as computed from the manual adjustment of the histogram the low and max bars have not been set to what I would consider reasonable values. One need to move them until the min and max areas display colors shows up in an area significant enough to make a difference and not just at the tails of the histogram. So you can see how that is a manual and judgment error prone process that can frequently be wrong.

[githubdoe]

In general all metric like numbers should go over on the left under "Metrics" That is what the display is all about.

[gr5]

I agree with Dale. PV is very subjective. And arbitrary. One can position those bars in many places that seem valid but get different values for PV. I don't think there is an industry standard e.g. place peak and valley at 1/10 standard deviation on histogram.

[gr5]

@VictorXiao22 - Are you trying to create a report to give to customers that includes the PV? If so then you should include the whole histogram.

If you do menu "files" "save pdf report" you can check the box for histogram.

@githubdoe - but in both cases, the PV value is very very small compared to other numbers so if he is taking a screen shot and sending to customers with each mirror, it would be nice for him if the PV value was much larger.

Maybe the PV value should be very large and bold in the histogram? Is that a good solution @VictorXiao22 ?

I don't want the PV value visible unless the histogram is also visible as I don't want to make it too easy for people to falsify the PV value. So a large, bold PV value should be over the histogram.

[gr5]

The "file" "save PDF report" is scaled wrong for me. The pdf comes out with some pictures cut off and blank pages and it's kind of a mess regarding scaling. And regardless, making the PV value larger and bold makes sense to me and would be larger in the PDF as well.

This wouldn't just be for people who sell mirrors for a living but also for individuals selling a single mirror on ebay or cloudy nights.

[VictorXiao22]

@gr5 I want to record my measurement result including fringe pattern, PV and RMS. If the PV value could be calculated by software but not manually, I assume the result would be more consistent because the calcualtion by software is used same algorithm.

[gr5]

The PV code is somewhat arbitrary where it picks the high and low point. Currently it looks at the absolute highest and lowest single pixel, then it takes the 10% and 90% from that. This is silly. There could be a single pixel that is 5 waves high. So a single pixel can mess up the PV value and change your 1/10 wave mirror to be a 2 wave mirror when it's actually a strehl .9 (fantastic) mirror.

rms and strehl are better measurements.

I could improve the PV by using say a "6 sigma" boundary where I calculate the standard deviation of the histogram distribution and pick some arbitrary number of sigmas (say 6) between the high and low cutoff. Say 3 sigma above the median and 3 sigma below. Or one could find say the 1% cutoff where 1% of the area under the histogram is outside the top and 1% below the bottom PV values (this would be the same as "4.6 sigma").

This would be a MUCH better way to find PV. But there is no standard. There is no industry standard for where to pick the peak and valley.

In the past people just did it by eye. And still do. They would look at the highest point and lowest point and could see how many waves apart they were but this was a rough estimate.

You can improve the PV yourself by making sure you have gaussian blur turned on. This will smooth out things so an individual pixel won't skew the PV value so much. But how much blur to use? The more blur you do the better (smaller) will be the PV value.

So let's keep it the way it is. It's subjective. You can put the high and low values anywhere you want.

[gr5]

Are you selling a mirror? Or mirrors? Why do you want these numbers? I think I can improve the display by making the text much larger. Would that be helpful?

[atsju]

@VictorXiao22 I will explain this with other words so you have 3 different wording from 3 people.

PV value distance in wavelength between highest point of mirror VS lowest. First of all lambda/10 doesn't mean same depending on wavelength used. Secondly there is indeed no industry standard to measure or express PV value. DFTFringe analyses whole surface of the mirror and you could have some measurement noise or some hole somewhere leading to a high PV while in reality the mirror has an excellent shape everywhere so it's in fact excellent. Other methods like Foucault where only measuring "zones" and you typically end with like 5 to 10 points on a curve and depending on the distance to ideal you have a theoretical PV value. This is NOT how DFTFringe measures PV as we are measuring whole surface and real peak real valley. Thus the PV expressed by DFTFringe is not a good indicator of mirror quality AND can vary between measurements.

Are you selling a mirror? Or mirrors? Why do you want these numbers?

This would indeed be helpful for us to understand.

If you want a "marketing" PV value here is a converter https://www.bbastrodesigns.com/Strehl%20RMS%20PV%20calc.html
Strehl 0.8 is lambda/4 and Strehl 0.92 is lambda/8 Strehl 0.99 is lambda/16

[githubdoe]

As others have said the PV value is not a good metric. One example is what if the extreme is hidden by the center of the mirror. For it to have any meaning you need to know how large the area is that is the peak and where is it. Software can not answer that easily. That is why the rms which takes area into consideration was developed.

It is also why I developed the histogram method to let you adjust those sliders till the area of the peaks and valleys shows up and become significant. So I recommend you do that work of adjusting them correctly if you want to display a meaningful value.

Georges suggestion of using a sigma value still ignores where the area of deviation is. That is very important. Yes one could mask the area covered by the diagonal and then it would be more meaningful. But perhaps it is all on the edge. Then it is very meaning full if it's area is large. RMS takes all of that into account.

There are different ways to estimate PV from RMS or strehl. But all depend upon a set of conditions that you can not guarantee for mirrors in general. They assume mirrors are smooth enough and symmetrical enough to do the conversion for example.

[VictorXiao22]

@gr5 @githubdoe
Thank you very much for your illustration.
I do/sell plastic injection molding of optical component and would like to know the "transmitted wavefront error" for specific area.
The specification of "transmitted wavefront error" include PV, RMS and I would like show the measurement result of our product to customer.

I assume if I have the fringe pattern and then the PV and RMS could be calculated by software. Like a commercial equipment ZYGO does.

[githubdoe]

I suggest you use the histogram correctly and then include it in the report.

[gr5]

The PV value is subjective. It's almost meaningless. Please look up "subjective" in a dictionary.

I don't know how zygo calculates PV. I'm sure they document this somewhere. If you can find the documentation we can think about duplicating their method. I'm guessing they take the highest and lowest point after a blurring function but which blurring function? If it's a gaussian blur then how large is the size of the blur?

Although I have to admit, the RMS is also calculated after a blurring function in DFTFringe. So maybe we could just display the highest and lowest point?

[VictorXiao22]

Here is the reference book of ZYGO. I am trying to understand if the PV calculation method is described in this guide.
https://www.seas.upenn.edu/~nanosop/documents/MetroProReferenceGuide0347_M.pdf

[githubdoe]

Zygo's interferometer generates very low noise unlike a lot of other interferometers. They can afford to show the PV knowing that it is from the device being measured. I'm sorry I don't have time to research that document. For the general case the PV is not a valuable metric for reasons we already stated.

[gr5]

Zygo created something called PVr (Peak-valley plus residuals). That is well defined mathematically. But the PV in the image above is likely something different and doesn't include the residuals. The pdf above doesn't mention PVr.