More testing, Mesu2/ODK14

[vdb]

Just blew one star up and did a layer of a circle in PS,

not round

[michael.h.f.wilkinson]

Looks very close to round to me.

The mathematical way of finding out whether something is round is computing the central moments of the distribution of grey levels. Best known central moment is the variance, or in 2D the covariance matrix. Higher order (normalized) central moments are skewness and kurtosis (or the equivalent matrices in 2D). In particular, for a fully rotationally symmetric distribution, all odd order moments must be zero. In the presence of noise, these values are approximate. Zernike moments or circular harmonics could also be used.

We also use a number of measures of circularity in our attribute filters, which are one for a perfect circle and deviate for anything not circular. I could give some of these filters a shot on the above image.

[perfrej]

Since I am already booked for end of February, I will feast my eyes on this beast when I get there

I just hope that I get my Losmandy mount before I go. Apparently, the man himself is difficult to communicate with...

/per

[ollypenrice]

The first sentence of Michael's post was more accessible to me than the rest but the test would be interesting! We are now getting far, far better guiding values than for the Trapezium shot and can improve collimation a little as well.

It looks round to me too...

Per, one G11 comes here regularly and performs very well indeed.

Olly

[michael.h.f.wilkinson]

The first sentence of Michael's post was more accessible to me than the rest but the test would be interesting! We are now getting far, far better guiding values than for the Trapezium shot and can improve collimation a little as well.

It looks round to me too...

Per, one G11 comes here regularly and performs very well indeed.

Olly

And I didn't even include the equations

I could make a bachelor project out of assessment of roundness of stars. Maybe a student of ours can make an automatic tool for that.

[Darth Takahashi]

Looks very close to round to me.

The mathematical way of finding out whether something is round is computing the central moments of the distribution of grey levels. Best known central moment is the variance, or in 2D the covariance matrix. Higher order (normalized) central moments are skewness and kurtosis (or the equivalent matrices in 2D). In particular, for a fully rotationally symmetric distribution, all odd order moments must be zero. In the presence of noise, these values are approximate. Zernike moments or circular harmonics could also be used.

We also use a number of measures of circularity in our attribute filters, which are one for a perfect circle and deviate for anything not circular. I could give some of these filters a shot on the above image.

At last someone who gets it:)

Although I do myself hate some of the terms such as "Kurtosis" which means "peakness" of the distribution in question.

central moment = centroid or centre of mass etc...

Should be pretty simple to do this with an image in FIT format.

Finally, I try to stay clear of "Qualitative" statements because unfortunately people are unreliable and get over enthusiastic concerning their purchases, and I'm just as guilty as everyone else. I prefer "Quantitative " information.

As far as making friends, first and foremost I'm here to share my hobby and learn, not make friends, I'm quite happy with myself thank you.

Email and forums are a difficult media to communicate in and I can understand that some of my posts might seem too critical, well that's to bad and if some idiot wants to judge me by electronic posts and email, well, they know where they can go and what to do when they get there, right!

I'll be at SGL7 again this year; you can judge me if and when you meet me there, in person?

For all those people who like pretty images and roundness this is a great forum; but understand some of us are here to try and learn as well. This is still a great opportunity to get some independent "Quantitative " information concerning this new mount.

[michael.h.f.wilkinson]

At last someone who gets it:)

Although I do myself hate some of the terms such as "Kurtosis" which means "peakness" of the distribution in question.

central moment = centroid or centre of mass etc...

Should be pretty simple to do this with an image in FIT format.

Finally, I try to stay clear of "Qualitative" statements because unfortunately people are unreliable and get over enthusiastic concerning their purchases, and I'm just as guilty as everyone else. I prefer "Quantitative " information.

As far as making friends, first and foremost I'm here to share my hobby and learn, not make friends, I'm quite happy with myself thank you.

Email and forums are a difficult media to communicate in and I can understand that some of my posts might seem too critical, well that's to bad and if some idiot wants to judge me by electronic posts and email, well, they know where they can go and what to do when they get there, right!

I'll be at SGL7 again this year; you can judge me if and when you meet me there, in person?

For all those people who like pretty images and roundness this is a great forum; but understand some of us are here to try and learn as well. This is still a great opportunity to get some independent "Quantitative " information concerning this new mount.

I very much agree with the quantitative approach (as an image analysis scientist).

One trick of our so called attribute filters, is that they can simultaneously measure many parameters of all objects in the image at a very high speed (12 Mpixel 16 bit per pixel grey level in 2-3s). If you have a multi-core processor, you can even run it in parallel (current record: 1.2 Gpixel in 60 s on a 24 core beast).

My idea would be to compute a map of deviation from true circles as a function of position in the image, taking into account asymmetries, to separate translational guiding errors from rotational guiding errors, coma, etc. There are also measures of triangularity, elongation that can be included.

[Darth Takahashi]

I very much agree with the quantitative approach (as an image analysis scientist).

One trick of our so called attribute filters, is that they can simultaneously measure many parameters of all objects in the image at a very high speed (12 Mpixel 16 bit per pixel grey level in 2-3s). If you have a multi-core processor, you can even run it in parallel (current record: 1.2 Gpixel in 60 s on a 24 core beast).

My idea would be to compute a map of deviation from true circles as a function of position in the image, taking into account asymmetries, to separate translational guiding errors from rotational guiding errors, coma, etc. There are also measures of triangularity, elongation that can be included.

We need to take this as another subject, because what you propose is extremely interesting. Because essentially you will have mapped the optic system using this technique.

I have often wondered whether or not it is possible to correct an optical system mathematically rather than optically with field flatteners etc...

Simply put, if you know the field curvature and its repeatable, why not apply and inverse function to the image to correct it?

Very very interesting... topic for discussion, please start a new thread.

[rfdesigner]

Simply put, if you know the field curvature and its repeatable, why not apply and inverse function to the image to correct it?

Very very interesting... topic for discussion, please start a new thread.

My concern would be noise.

In my experience with radios, if you try and correct for bad hardware with clever software you suffer noise issues, because you tend to end up with some kind of signal attenuation (spreading etc.) prior to detection which hits the SNR. It's fair enough to use software for gentle tweaks, but I wouldn't want to use a purely software coma corrector for instance.

What you're proposing would be a fantastic tool to add on after a coma corrector to correct for a minor miscollimation that can happen, but then the SNR isn't being damaged much by the errors.

Derek

[michael.h.f.wilkinson]

My concern would be noise.

In my experience with radios, if you try and correct for bad hardware with clever software you suffer noise issues, because you tend to end up with some kind of signal attenuation (spreading etc.) prior to detection which hits the SNR. It's fair enough to use software for gentle tweaks, but I wouldn't want to use a purely software coma corrector for instance.

What you're proposing would be a fantastic tool to add on after a coma corrector to correct for a minor miscollimation that can happen, but then the SNR isn't being damaged much by the errors.

Derek

Agreed up to a point. Deconvolution (the method of deblurring an image by removing the effects of the point-spread-function) is sensitive to noise. I would certainly prefer well-corrected optics over a purely software solution. My proposal as it stood was different: measure what is going on to tune your system optimally. Nothing replaces good optics.

There are however advanced methods that would in theory replace a slightly faulty PSF (e.g. at the edge) with a more symmetrical one without undue sharpening and the related noise issues. Linear guiding errors in particular can be removed without noise exploding.

I could have students look into this aspect as well.

[michael.h.f.wilkinson]

I have just started a thread on this topic:

http://stargazerslounge.com/discussions-software/170214-mapping-point-spread-function-over-image.html

[dave_galera]

Cor blimey, this setup is a magnet for the hypercritical!!!!

You can take it, your back is broad enough and I know you will get it all sorted to your usual VERY high standard!!

Deneb, it is well worth the trip to Les Granges and can recommend it. The problem is it will cost you money when you get home to get some new kit.....it cost me 4.5K

[Darth Takahashi]

You can take it, your back is broad enough and I know you will get it all sorted to your usual VERY high standard!!

Deneb, it is well worth the trip to Les Granges and can recommend it. The problem is it will cost you money when you get home to get some new kit.....it cost me 4.5K

From your signature it looks like the 4.5k was well invested assuming its in the TAK gear. The FSQ85EDX is meant to be something special indeed!!! I don't have one yet!!!

A TAK's for life, not just for Christmas!!!

[dave_galera]

From your signature it looks like the 4.5k was well invested assuming its in the TAK gear. The FSQ85EDX is meant to be something special indeed!!! I don't have one yet!!!

A TAK's for life, not just for Christmas!!!

You are right Neil, it was for the FSQ85, new guide scope, Atik Filter wheel plus a few more bits, I already had the EM200 (bought after exchanging a few e-mails with Olly lol). All this was as a result of spending a week with Olly.........learning his ways of working, and I juuuuuuust fell in love with his FSQ85!!! Up to this point I had been struggling for the best part of a year, but thanks to Olly I am now off and running.

[Darth Takahashi]

Collimation is not 100% that we know ...

and some of those actually quite a lot are close doubles ...

...that is why they are not 100% round.

And maybe why not bit of flex on the system ...

The fact that a 10 minute and 30 minute sub deliver the same error excludes for me mount and guiding error for 99%,

but hey I'm just a starter in this hobby.

Or it could be black matter it is known to deform stars :-)

Yves.

***Collimation for a Dall Kirkham****

Astronomy at 52 degrees!

[ollypenrice]

Neil, be aware that the phrase 'pretty pictures' is like a red rag to a bull in my case. Of course, use it by all means but I'm just informing you of how I respond to it. I have no desire whatsoever to take pretty pictures. My desire (not my acheivement) is to take beautiful and informative pictures. In order to be that they will need a certain level of technical exactitude but that exactitude is a means to an end for me, not an end in itself. I've said this before and doubtless I'll say it again.

Like others on here I regard what I do as being more art than science, though obtaining high resolution is a key part of the craftsmanship side of the art and, yes, I want round stars. We all 'get it' perfectly well on here, I'm sure. Roundness can be defined and measured. The idea, unlike Michael's equations, is not difficult. However, after a certain point it really doesn't matter much - to me.

I come from an arts background so I think and work in a particular way and one unlike your own, I don't doubt.

Is the Internet forum really such a difficult medium in which to communicate? I think it's quite like a pub or the lawyer's archetypal 'Clapham Omnibus.' Describe people as idiots and they'll not be too pleased! Like all forms of good communication I think it requires one to put one's self in the shoes of the reader, that's all. I'm guilty of failing to do that sometimes, sure, and for that I apologize.

I'm off to do my collimation homework. Many thanks for that link. It looks right up our street.

Olly

[michael.h.f.wilkinson]

Like Olly, I tend to look at both sides of the equation: astrophotos have both an artistic appeal and an informative, exact part. In most cases, measurements of roundness should reflect perception of roundness and vice versa. Measurements should quantify visual perception, not be at odds with it, especially given that the human visual system beats any computer vision system at most tasks. Furthermore, most astrophotos are for human consumption, not automated analysis.

In my opinion, the idea of measurement here is just a means to make it easier to judge what is going on. I have often been gobsmacked by images here that some of you still criticize on points like roundness of stars. Once pointed out, I can see what people mean. A tool which facilitates detecting such faults should not be construed to be an "auto-criticism" program, but just a tool to help improving both the scientific exactitude of the image and the visual appeal. This could help both experienced and inexperienced photographers alike.

A good program should be calibrated in the way it reports deviations from the ideal shape in such a way that it gives the user an idea of how bad the defect is, in a way which indicates intuitively what the perceptual impact is, e.g. which percentage of observers notices it, or an average grade given by experienced astrophotographers (maybe we can call it the Penrice-scale ). Qualitatively, it should show what kind of defect it is.

[Darth Takahashi]

***Putting the record straight***

I'm not here to try to offend anyone; however, I realise that my messages may be too direct and to the point for some people. Nevertheless, if I have have upset you I apologise for that.

I'm an amateur astronomy like everyone else here and I'm keen to know how good this mount is because after seeing it at Mijdrecht I have nagging doubts.:(

1) Mechanics made in the Netherlands

2) Servos from America

3) Command system from Oz

4) All put together with zip ties or tie-wraps etc...

It was to say the least not a good first impression but it is a friction drive system that claims a lot, even 100kg payload.

Now you have one at your place I was hoped that you could put some of those concerns to bed!!! Then the roundness campaign started!!!

I have worked 25 years in the process automation industry and the golden rules is as follows; Measure -> Control -> Improve (6sigma DMAIC - I'll let people workout the 'D' and the 'A' for themselves).

To say roundness was not the measure I was looking for would be an understatement. Since I already know true randomness will produce a Normal Distribution, so we weren't going to get any closer to the truth with roundness! :(

As will a number of other factors already mentioned in the posts.

I now work in Marketing and know very well how manufacturer plainly lie or simple overstate their products capabilities - FACT. Unfortunately, my Japanese company is still living in the 1960's and behaves honourably. Not that I'm for the plain lying camp. but we could be more positive etc... Look what happened to MEADE for example. I hope we all know what I'm referring to; the RC's that weren't RC's, right.

Anyway, back to the point, a guide log can be 'A' analysed and it get us nearer to the truth regarding performance! An image tells us nothing.

Where or not they are pretty pictures or artistic images is for others to decide because frankly I don't care. I'm happy when I can achieve the same results that was in the astronomy books during my childhood, if I can do better great.

But lets not kid ourselves; once we start the image manipulation process they are no longer of any scientific use. I have an ST10 and might in the future do some photometry with it.

I suppose I'm a perfectionist and will take any reasonable path or advantage to get closer to the end goal and in that respect this mount was a contender, I will look else where now.

As far as the idiot comment goes, I standby it. Anyone who wants to pass judgement on someone else based upon email and posts is one. I don't use offensive language and I have not made a direct insult towards anyone. In one reply I have made a indirect insult and myself and the other party have shared a PM. Unfortunately, two posts crossed within 2 minutes of each other etc... Otherwise the whole thing could have been completely avoided. Anyway it water of a ducks back as far as I'm concerned. I have been called worse by professionals.:eek:

I'm even using the stupid smily things that I hate because apparently it makes my posts more friendly!!!:icon_salut: go figure!!!

Lets hope that the collimation info helps, although you must be very close so only tweak it.:icon_scratch::icon_scratch:a little bit at a time!

[vdb]

Neil,

in the first light section there was a 100% crop of a 30 minute sub, only stretched .... no hiding of anything ...

A guiding log has it's own caveats, we do dithering for example and we touch the setup as this is still a setup in progress, but believe me the company did not tell any marketing fluff ... the owner rally took the best components which he already used on the mesu 1 ... No doubt you have read the review of Rob Kantelberg ... this mount is more of the same ...

And look at the 30 minute crop in the first light thread, it's on page 2 now ...

Yves

PS he is not interested in a beauty contest, but nuts and bolts are changed, he should update that on his website ...

[vdb]

Here some pictures Olly posted to prove that:

http://stargazerslounge.com/member-equipment-reviews/169671-meet-beast.html

[Darth Takahashi]

;

Neil,

in the first light section there was a 100% crop of a 30 minute sub, only stretched .... no hiding of anything ...

A guiding log has it's own caveats, we do dithering for example and we touch the setup as this is still a setup in progress, but believe me the company did not tell any marketing fluff ... the owner rally took the best components which he already used on the mesu 1 ... No doubt you have read the review of Rob Kantelberg ... this mount is more of the same ...

And look at the 30 minute crop in the first light thread, it's on page 2 now ...

Yves

PS he is not interested in a beauty contest, but nuts and bolts are changed, he should update that on his website ...

Its a little bit a pity that you put Rob Kantelberg name here because being honest his images are nothing special, good of course but in my humble opinion, not special;). I can do better myself but admittely I'm using TAK's and most of his images are with Newts.

But that said there are at least 2 dozen other imagers here that regularly produce better stuff.

I'm excluding those in more beneficial areas of the world in this statement; Olly included. If you want to see what can be done with a Newt look at Mikes images.

Olly can point you in the right -> Mike direction:):)

When last I checked Mike had a Skywatcher 250mm Newt and an EQ6 with an even older HX916 ccd camera. When I last saw him I think his Russian Mink fur hat was more expensive than his kit (this is a joke) but he does own such a hat:rolleyes:

[Tom OD]

Hi Olly,

The round stars look great as does the pic itself. How will seeing influence the pic. Will it be a wait and see (pardon the pun) as each night imaging begins.

Tom.

[ollypenrice]

Hi Tom,

It's going to be seeing dependent so in bad seeing I guess we'll do colour and not Lum. And there will be nights when it maybe won't play at all, or maybe not till the land has cooled. But so far so good. It seems less sensitive than I'd expected.

Olly

[vdb]

Neil,

you mis the point, he did a review of the same manufactures mount with 80% same components ...

You mentioned you where not interested in pretty pictures and now it's the point that you bring up, sorry you lost me ...

Yves.

PS: I think the 30 minute must be really good no remark ;-)

[Darth Takahashi]

Neil,

you mis the point, he did a review of the same manufactures mount with 80% same components ...

You mentioned you where not interested in pretty pictures and now it's the point that you bring up, sorry you lost me ...

Yves.

PS: I think the 30 minute must be really good no remark ;-)

Even I need to eat every now and again.

Its pretty much a mute point for the following reasons;

1) I have only his images to judge the mount by - no guiding log and they are not special, but maybe our standards are too high:rolleyes:

2) The mount is completely different and aimed at those people with Newts (2 of them to be precise!!!)

3) They made a silly platform to stand the Mesu1 higher for those of us with Cassesgains / refractors! (Not a good solution in anyone engineering book)

4) Finally, they realised by themselves that they needed another type of mount with more universal appeal, hence the Mesu2!!!

If I had done the marketing for them they could have skip the Mesu1 completely, too niche!!! VOC = Voice of Customer!!!

I hope I have answered you satisfactory enough?

How did the collimation go?