no convergence

[gingersoll]

Why am I getting a warning in purple in the Integration process of Pixinsight saying "no convergence in MRS noise evaluation" have been getting it a lot

[bitli]

[EDIT: Wrong answer, confused message ImageIntegration and ImageCalibration, see next replies]

Hi,
This has been discussed a few times.  See for example http://pixinsight.com/forum/index.php?topic=3320.0 or search for 'no convergence' in the forum.

Basically your process cannot make to use of your darks, as they do not seem relevant for your images.. You can start by looking at them (check that they have the hot pixels at the same place than your image, that they do not have a truncated histograms, stuff like that).  Even if you do not get this message you should check in the log if the factor is reasonable. Ask again if you cannot find the answer in the other replies or by looking up the master dark/images. 

-- bitli

[gingersoll]

I get the same warnings ,when I first got PI I didn't get these warnings now I get them all the time.Also I did not use darks

[sctall]

I am seeing the same problem.
I am just starting to learn , (so experimenting) how to make the masters and do the calibrating of the frames manually using IC and II and other tools.
I have 5 Bias images (just experimenting here), and when I try to run it with the attached screenshot parameters, I keep getting the  'no convergence'  message.
Tried the same with Darks and same thing occurred.
Since I have no idea how to read the console output numbers, or even what to expect. I can't tell whether this is a bug or not. I would think not.
But if someone could look at my attached and see if I am setting something up wrong.
I keep hearing that there is more control when using the individual tools, so I'd like to try to understand them better.
Great write up from Bitli. I got a lot out of that. Got most of it.

I can use the same frames with BPP and make the Masters with no prob.

Note: Since I only have 5 frames I also tried Percentile Clipping, but got the same messages.

At the bottom of the .doc, is the output for the BPP script. It worked

Scott

[bitli]

Sorry, my previous answer was misleading.

The message is in ImageIntegration, not ImageCalibration (they are little twsited messages, all looking alike).

Code: [Select]

MRS noise evaluation: |
** Warning: No convergence in MRS noise evaluation routine - using K-sigma noise estimate.—
** Warning: No convergence in MRS noise evaluation routine - using K-sigma noise estimate./
** Warning: No convergence in MRS noise evaluation routine - using K-sigma noise estimate.done
Scale factors : 1.00000 1.00000 1.00000
Zero offset : +0.000000e+000 +0.000000e+000 +0.000000e+000
Noise estimates : 0.0000e+000 0.0000e+000 0.0000e+000
Weight : 1.00000 1.00000 1.00000
In this case it tries to evaluate the noise of an image with one method (MRS) and as this does not work it tries with K-SIGMA.  This means that you noise or image is very atypical, it could be that you use a CFA image with non CFA settings, or that it is very noisy at all level of images, or just so beautiful that here is no noise, or that you integrated darks/bias on a camera that already truncate most low pixels to zero (this is not the case of the Canon 350D).  I do not know the math well enough to guess the original cause.

In this case you do not use noise evaluation (all weights = 1), so you do not really need the noise for integration.  I suspect that you should also uncheck the 'Evaluate noise' to avoid the warnings.  Although they should not impact the result at all.

The next things to check is that your input parameters are ok.  As you integrate CFA (DSLR raw) images, according to the log, you should make sure that your DSLR settings are correct.  The simplest way is to add 'raw cfa' to the field 'input hints'.  Then it should go through.

Unless you are integrating bias/dark/flats, you should see that  your images have very long stars or set of stars because obviously the raw were not aligned.  This is a cause of panic for ImageIntegration, as it can clearly not make any sensible rejection decision :-)

I think you are right to try to understand the step by step operations,  It is not only a question of more control, it is a question of better understanding so even if you use scripts later you can estimate if your got the best of your image or what did go wrong.

I hope this answer was more helpful than the previous one.

--bitli

[sctall]

Well, I found that whenever I click "Noise Evaluation", I will get the warnings.  So I uncheck those.

So I went thru the whole process following below steps.
Every time I fail the final II,   on the reference frame I used for the SA. Message is "empty image"  it is always on the SA reference frame.
see attached.

Step 1: Generate master Bias with II. Simply integrate them in Image Integration.
Step 2: Generate master Dark with II. Simply integrate them in Image Integration.
Step3:  With IC Generate load all flat frames and calibrate by subtracting master bias and master dark. Click calibrate for both  This is step 1 for Flat.
Step 4: Using the calibrated Flat frames created in step 3, load all flat frames into II and make a Master flat
Step 5: Now with all 3 Masters, use IC to calibrate the light frame. Load all light frames into IC, load all Masters, and uncheck calibrate.
Step 6: Use Batch debayer script to Debayer all _ c calibrated lights.
Step7:  Now use Star Alignment to register all the debayered_c light frames. Pick one of the debayered_c  for a reference frame.
Step8: Now use II to make the final image for processing.  Load all debayer_r light frames into II.

Note: I can load all same frames into BPP script and just works.
What am I doing wrong?

Scott

[astroedo]

Following your worflow, in my opinion, your mistake in step 3 is to check "calibrate" on bias frame if you don't have a overscan region.

In step 5 again you must check the calibrate checkbox only on dark, because if you don't calibrate your dark you will subtract bias twice.

Ciao

[bitli]

I would replace

Step3:  With IC Generate load all flat frames and calibrate by subtracting master bias and master dark. Click calibrate for both  This is step 1 for Flat.

by

Step3:  With IC Generate load all flat frames and calibrate by subtracting master bias. This is step 1 for Flat (no calibrate)

Calibrate bias does not harm - if does nothing if there is no overscan region, but it may be confusing to the humans.

There is usually no need to use darks with flats and anyhow your dark exposures are usually way longer than you flats (the optimize could solve this, but let it keep simple first).

Step 5 should not calibrate the flats again. You should calibrate the darks (to subtract the bias as it is safer not to subtract it in the master). The lights are calibrated anyhow with whatever bias/dark you specify at this step

-- bitli

[sctall]

Thank you all

I am still confused on how this calibration works between types. But I did get it to work
I have a lot more reading to do.

Here is what I did!!!
Following you instructions
Step 1: Generate master bias with II. Simply integrate them in Image Integration.
Step 2: Generate master Dark with II. Simply integrate them in Image Integration.
Step 3: with IC  load all flat frames and calibrate by subtracting master bias only. Do not calibrate for Bias  This is step 1 for Flat.
Step 4: Using the calibrated Flat frames created in step 3, load all flat frames into II and make a Master flat.
Step5: Now with all 3 Masters, use IC to calibrate the light frame. Load all light frames into IC, load all Masters, and uncheck calibrate for Bias, check calibrate for dark and uncheck calibrate for Flat.
Step6: Debayer all calibrated lights.
Step7: Now use Star Alignment to register all the debayered_c light frames.
Step8: Now use II to make the final image for processing.  Load all _r light frames into II.

It seems to work now, and I compared a final light integration between a BPP processed and a manual using the individual processes above.

There is no doubt that the manual handled the Hot pixels much better. 
And I usually have a very green biased Linked Autostretch. And for some reason the BPP is not a bright a green. ( no idea what that means really ), but the brighter green manual image appears much cleaner.

Please what is the difference in adding the masters with calibrate and adding them without calibrate?

Since I use a un-modded DSLR with LP I fight all kinds of issues.

gingersoll, I apologize for hijacking your thread.

Thanks
Scott