Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Detailed Lodestar vs. QHY5L-ii study

This topic has been archived. This means that you cannot reply to this topic.
44 replies to this topic

#26 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 14 July 2013 - 06:38 PM

Thanks Korborh. The main thing helpful from the "PTC" plot is the fact that it is log-log, and has many samples along the x-axis. I will write things up in more detail, but a linear plot just doesn't show the transition from read-dominated to shot-dominated. In terms of fitting to the model, the plot doesn't matter - but it does show it well. And the pulse counting is helpful for comparing one camera to another - but it also isn't essential for a single camera. But I do see read/gain values posted sometimes and I don't know how they were derived - and a plot like this makes it more self explanatory.

Thanks,
Frank

#27 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 14 July 2013 - 06:45 PM

Hi Doug-

I think the ssag uses the earlier cmos chip, the mtm9001, and that is the one that didn't impress me much and made me wonder if cmos would ever really take over ccd. But this new cmos chip is much more sensitive.

For OAG I like being in video mode so that you can see immediately when stars are on the (usually small) guide chip. As long as it is well focused and the gain is high, it should be good for finding guide stars - but I still recommend pre-planning them and dialing them in ahead of time - which isn't too hard once you get used to it. But the qhy can run with very long exposure "video" and let you see faint stars anyway.

I will see if I can get similar shots with the two cameras. Note that I am mainly using it in video mode with MetaGuide - so you need to make sure it works with whatever software you intend to use. Katie reports it working with Maxim, so that's good info.

Frank

#28 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 14 July 2013 - 06:48 PM

Thanks Katie - glad it's working for you. Good to know it works with Maxim. I assume it works with other guide software as a normal guide camera with long exposure. I think the increased sensitivity will be a win, and the smaller pixels will help especially with small guidescopes - and they aren't detrimental for OAG work.

Thanks,
Frank

#29 northwolfwu

northwolfwu

    Vostok 1

  • -----
  • Posts: 163
  • Joined: 19 Nov 2008

Posted 14 July 2013 - 09:17 PM

Thank you very much for this detail analysis.
I successfully guided C8 HD with OAG and QHY5 II L by metaguide. Camera is 8300M. The seeing is not good, so I used bin2. The following image is stacked of 10min*3, only with slight curve and denoise:

Attached Thumbnails

  • 5971297-M27_PS-Mean FW Scaled-Crop.jpg


#30 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 15 July 2013 - 01:48 AM

Thanks for posting - that looks like a good result. I have different asterisms I use with different object to judge guiding and in this case the small triangle of stars near the center looks good.

So - video and oag are in fact possible. It helps that you are using EdgeHD because the guidestar won't be comatic and should be small and round - which makes the star spot smaller and brighter for guiding. And since focus is important to make sure the guidestar is small, it helps to have video for feedback as you focus the guide camera.

Thanks,
Frank

#31 northwolfwu

northwolfwu

    Vostok 1

  • -----
  • Posts: 163
  • Joined: 19 Nov 2008

Posted 15 July 2013 - 03:13 AM

Thanks for posting - that looks like a good result. I have different asterisms I use with different object to judge guiding and in this case the small triangle of stars near the center looks good.

So - video and oag are in fact possible. It helps that you are using EdgeHD because the guidestar won't be comatic and should be small and round - which makes the star spot smaller and brighter for guiding. And since focus is important to make sure the guidestar is small, it helps to have video for feedback as you focus the guide camera.

Thanks,
Frank


Because of bad seeing, C8 HD axis is not in perfect collimation. But in the MG guiding, the seeing is good, and the guiding star is very small due to the flat image plane of HD.

Attached Thumbnails

  • 5971648-Metaguide_AG.jpg


#32 astrovienna

astrovienna

    Mercury-Atlas

  • *****
  • Posts: 2772
  • Joined: 04 Dec 2006

Posted 15 July 2013 - 09:58 AM

Frank, I take it your results should be pretty similar for the ASI120MM, which uses the same chip? A lot of planetary imagers have moved to that cam, and many of us use our planetary cams to guide.

I got the ASI120 a few days ago to replace my Flea3, and while I haven't guided yet with it my tests show it runs fine with Metaguide. The ASI is a bit more sensitive than the Flea3 in green and blue, but the big advantage for guiding is that its chip is about double the size of the Flea3. I'm also curious about its 16 bit mode, and whether that might yield better results than 8bit.

Kevin

#33 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 15 July 2013 - 12:18 PM

Hi Kevin-

Yes - the chip is the same as the ASA, but a big advantage of the qhy is that it is 1.25" diam like the lodestar, and very compact. I imagine the ASA isn't too big, but for OAG it helps to have it be very small - especially if you want to insert it into the focuser tube to some depth.

I had to write special code so the st4 port on the qhy would work with MG - so I'm interested to hear if it works on the ASA version. I expect that MG would work fine with the ASA as a videocamera, but it may not be able to use the st4 port. A lot depends on if you can use the camera via directshow, while guiding via ascom.

Although I tested based on a qhy, I expect the sensitivity results to apply for other cameras that use the same cmos chip. I intend to write this up in a separate document and post on the planetary group since I assume there may be interest there.

Thanks for the report on using the ASA with MG - if you can try to connect with the st4 port via ascom I'd appreciate the info. If they have an sdk I could modify MG to talk to the guide port directly.

Thanks,
Frank

#34 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 15 July 2013 - 12:20 PM

Thanks for posting that image of the guide star. That's what I want to see - a small and well defined spot measured with small pixels for a good centroid to guide on.

Frank

#35 AndreyYa

AndreyYa

    Vostok 1

  • *****
  • Posts: 138
  • Joined: 05 Dec 2006

Posted 15 July 2013 - 03:25 PM

Hi Frank,

thank you for cameras comparison, QHY-5 II looks really attractive comparing with pricey and aged Lodestar. Frankly speaking, slow guiding mounts would hardly suffer for Lodestar slow readout. But sensitivity and low noise is a must for OAG guided scope with blown up off-axis stars images.

IMHO, quantum efficiency comparison results could require double checking. It's known that SONY's HAD, SuperHAD etc. CCDs had QE not less then 50%. So, with all my respect to Aptina/Micron, it's hard to believe that their sensors have about 100% QE, especially without backside illuminated sensors technology.

Did you have a chance (or maybe planning) to compare the level of background non-uniformities for these sensors, especially for high gain values? QHY-5 has a lot of issues with interference-like non-stable distortions. It was problematic as it was impossible to get rid of it using usual calibration approaches. Does QHY-5 II have these issues solved?
Additionally, do you have plans to compare readout noise levels of these cameras - for the sake of star centroid measurability?

#36 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 15 July 2013 - 04:57 PM

A high end mount in an observatory may be able to do well with no guiding at all - so I'm just talking about the inherent centroid accuracy and low latency of the guide signal - whether it is needed or not. I find it important for making mid-range equipment work well in long exposures. For OAG, I use video and aim for small and round, well focused star spots. If the scope has coma, I would aim to correct it - as with my c11 and f/6.3 reducer corrector - which gives good oag guide stars.

I'm surprised by the large difference that I measured also - which is why I encouraged checks by others. But note that I am doing things in a very controlled way in terms of identical photon counts in narrow passbands - whereas other sensitivity measures may be based on power or something not directly related to QE.

Also - I have mainly been using Sony sensors - both for imaging (SX) and guiding (lumenera) - so I know they are sensitive - but I never measured it in an absolute way. When I compared the lodestar to lumenera my impression was it was a sony ccd with huge, oblong pixels that operated in binned mode. Others appear to be struck by how sensitive it is - but it didn't seem to stand out from my 7.4um lumenera. I intend to measure it also.

The qhy originally showed banding based on the way I initialized it with the sdk, but once I got it fed right it worked well. It's hard to compare the overall quality since there is dust on both cameras, but the lodestar has several obvious column defects and I don't recall any in the qhy. Any prnu would show in the lateral spread of the plots, and as you can see they are very narrow once out of the read noise regime. The 20x20 patch was selected to avoid dust patches and in the center of the bright diffuse patch of light. "Usual calibration approaches" to me involve two bias and two flats, and selecting a good patch in the frame - and this would pose no problem there either.

My main concern about guidestar centroid accuracy is optimal handling of non-Gaussian star spots blurred by atmosphere in short exposures. The read noise I already stated above, and is less per pixel for the qhy.

Frank

#37 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 15 July 2013 - 05:18 PM

I guess I didn't post plots of the corresponding qhy results, but here is the raw data from a 20x20 pixel region.

Frank

Attached Thumbnails

  • 5972836-qhy_gprimeGainCurve1s.png


#38 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 15 July 2013 - 05:21 PM

The above plot is more noisy looking than for the lodestar because the adu values are shifted left 4-bits - giving it a very small (inverse) gain value. This is also consistent with the pixels and well being smaller. But the linear trend, on the log-log plot, is clear and allows a good gain calculation.

Here is the fit to the plot for the deduced gain and read noise.

Frank

Attached Thumbnails

  • 5972839-qhy_gprimeGainCurve6s.png


#39 PiotrM

PiotrM

    Soyuz

  • *****
  • Posts: 3844
  • Joined: 03 Jan 2010

Posted 15 July 2013 - 05:27 PM

Also - I have mainly been using Sony sensors - both for imaging (SX) and guiding (lumenera) - so I know they are sensitive - but I never measured it in an absolute way. When I compared the lodestar to lumenera my impression was it was a sony ccd with huge, oblong pixels that operated in binned mode. Others appear to be struck by how sensitive it is - but it didn't seem to stand out from my 7.4um lumenera. I intend to measure it also.

In terms of absolute quantum efficiency the ICX429 used in Lodestar is very sensitive - QE max around 68%. Old "planetary" cameras (sensors ICX098, 204, 205, 274) had much lower QE, and vastly lower in IR (that's why ICX618 was a boom). The difference is that ICX429 is interlaced - which is useless in general for planetary imaging. SX made some software/firmware/electronics deinterlacing so that guiding works nicely - although you never can tell if it adds some error to the final image or does it help (deinterlacing adds some blur).

#40 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 16 July 2013 - 01:35 AM

I see numbers thrown around for the QE of these various sensors, but I haven't seen empirical studies like this one to do a comparison. One set of ads for these sensors says the qhy qe is 74% and the qe of the lodestar has a peak of 65% at 620, while it is around 35% at 400 and 720. Right there is the factor of two I am measuring - so the qe does vary quite a bit across the spectrum and much depends on the actual photon count across the spectrum. Even at the peak, the qe of the qhy is greater - and across the spectrum it's not clear how they compare.

So there may be a bias in my study due to the non-uniform spectrum of the LED - but the light from a guidestar isn't uniform either and what you really need is the total fraction of photons from a guidestar spectrum that get converted into electrons - which involves the product of the qe spectral response with the photon spectrum of the star. But since I see the qhy win in each section of the spectrum, it seems like it would win regardless of the star spectrum.

Anyway - I'm doing an actual comparison of two actual cameras in three distinct bands, based on electrons generated. I'd like to see other data for comparison, but so far I don't think my result is very out of line given the sharp peak of the lodestar spectral response.

Frank

#41 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 16 July 2013 - 01:55 AM

I'm also curious about its 16 bit mode, and whether that might yield better results than 8bit.



I neglected to comment on that question and it is something I wanted to talk about. For short exposure applications like guiding and planetary video I don't see much benefit at all from going beyond 8-bit. If you are photon starved in short exposure video, then the frames will be limited by read noise and added bits won't help. If you stack a bunch of frames then you can arbitrarily add more bits of dynamic range with a large number of frames - and the extra bits would be in the noise. If you are measuring the centroid of a faint guidestar it will depend heavily on how the background is treated - and that background will be read noise limited and again the added bits won't help.

The one place I can see bits helping is if you are doing video photometry - in each frame - of stars with a range of brightnesses. Even then, if the exposures are short you will be affected heavily by scintillation and - again - the extra bits won't help.

The qhy is like my lumenera in that it has a 16-bit output mode that is actually 12-bits shifted left by 4, to fill the full range of 16-bits. With MetaGuide and directshow all I see from either camera is 8-bits - and that is fine for finding faint guidestars and calculating the centroid accurately.

Frank

#42 Wembley2000

Wembley2000

    Messenger

  • *****
  • Posts: 416
  • Joined: 07 Jan 2012

Posted 16 July 2013 - 07:03 AM

Frank,

Nice study and nice software I have been using metaguide for collimation for both my sct and refractors and it has worked great. Currently I have been using PhD for guiding with SGP because it allows me to image unattended with meridian flips.

Now my question, with the small pixels of the qhy and the large pixels (binned) of the lodestar do you think there would be much difference in the centroid calculation? I have both of these guiders and use an oag, but I have not used the loadstar yet and last night I was imaging m20 and the seeing that low was just killing the guiding so I was wondering if using the larger pixels, hence lower resolution of the loadstar would help in this instance? My respective % of guided arc/sec per pixel to imager arc sec per pixel is 33% for the lodestar and 144% for the qhy5l-ii, so in effect I am oversampling my guider image by quite a bit .

Thanks for your input.

Wem

#43 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 16 July 2013 - 01:28 PM

Hi Wem-

I have fairly strong opinions about centroid calculations and they differ from pretty much everything you will read in the amateur astro community. A key point is that people tend to use simulations and simplistic descriptions to describe how "modern" centroid calculations are "good to 1/30 pixel" or something - but there is no empirical evidence to support it - largely because it is very hard to demonstrate, empirically, the accuracy of a centroid.

In short - I completely dismiss any write up that uses simulation to "find" centroid accuracy. It usually assumes Gaussian stars and read noise as the only sources of error - and ignores the roles of: atmospheric distortion, and centroid heuristics to determine the "star" region and its background.

To me, for good guiding, you want a very "fresh" measure of the centroid with little delay, so you can make a correction before it "rots" in time. The centroid should be based on the hot spot of the star spot - and exclude the undulating "wings" that contribute no information and bias the calculation. This means you want a well-resolved view of the center of the star - and you want to act quickly. Which points to video with small pixels - in arc-seconds.

As long as there is decent signal from the guidestar you should have no penalty using small pixels. In order to get decent signal with OAG, you should use a FOVI to select the best star available, rather than just expose longer and take what's there.

In summary - you want to resolve the star spot well, and the star spot should be small, in arc-seconds, and well defined. And you want to act on the centroid value promptly with a correction to the mount - before it is old and out of date.

Frank

#44 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • topic starter
  • Posts: 8837
  • Joined: 12 Oct 2007

Posted 17 July 2013 - 02:42 AM

guiding with SGP because it allows me to image unattended with meridian flips.


Hi-

I'm not sure how SGP handles meridian flips, but I instrumented MG years ago with remote message controls that allow different types of meridian flips without recalibrating the camera, under the control of a separate app or script. As long as the telescope is pointing in the right place after the flip and a guidestar is present, you should be able to do an automated meridian flip with MG - either with a guidescope or OAG. It is documented in MG in the section on Remote Messages, and the section on QuickCal, Meridian Flip, etc.

The html documentation online is here.

The sequence would be: StopGuiding, UnLock star, [do meridian flip, then send MG the MeridianFlip notification], LockStar, StartGuiding.

Frank

#45 Wembley2000

Wembley2000

    Messenger

  • *****
  • Posts: 416
  • Joined: 07 Jan 2012

Posted 17 July 2013 - 09:57 AM

Frank,

Thanks for the answers, hopefully SGP will integrate the functions of MG beacuse I have had good success in the past guiding with metaguide.

Wem


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics