Creating Tone Maps

A number of people are using Tone Maps as part of their image processing. The Photoshop Master J-P Metsaivainio may have invented the technique, or perhaps he just popularized it with his spectacular images. Subsequent imagers have developed processes to implement Tone Mapping.

I spent many hours trying to remove the stars from an image. My process is mostly performed within PixInsight, although the last step requires the Spot Healing brush in Photoshop.

Target Image Preparation

First, create the usual S-II, HAlpha, and O-III masters from calibrated subframes. Mine are drizzled from an SBig STF8300M camera on a Takahashi FSQ106ED. Each master is cropped and has DBE applied.

I then use PixelMath to combine S-II, HAlpha, and O-III masters using the following mixing:

  • Red = (0.8 * Sii) + (0.2 * Ha)
  • Green = (0.8 * Ha) + (0.2 * Oiii)
  • Blue = Oiii

Apply Background Neutralization, TGVDenoise, and a standard Histogram stretch to the RGB image.

Image ready for star removal

Star Removal Process

Create several previews on the image. Include areas with bright stars, background areas, strong nebulosity, and so on.

When each star mask is being created, test the mask thoroughly on the previews by applying Morphological Transformation. It takes me many iterations to get the star mask correct. Issues to watch for:

  • If the mask stars are too small you will get halos where the star is not removed far enough out.
  • Too many iterations of Morphological Transformation through the mask will create dark holes where the star was.
  • Is the Tone1Mask including the very small stars? If not, play with Noise threshold or Small-scale Structure Growth. Tone1Mask will likely NOT pick up larger stars.
  • Tone2Mask should ignore smaller stars. Check the biggest stars to verify they are being picked up.
  • Tone2Mask needs Large-scale Structure Growth to ensure the mask is much bigger than the stars.
  • Watch for bright nebulous areas being interpreted as stars, especially along sharper ridges or filaments.
  • Watch for strange artifacts in the vicinity of dark dust lanes, blobs, or “holes” in the nebula.

Test the Morphological Transformation through the masks on your previews on the image (the mask should show clear stars). You may need to either

  1. Run Morphological Transformation / Dilation to expand the mask. Usually 1 or 2 iterations are enough.
  2. Increase / decrease the number of iterations in the erosion.

Create Masks

Create two masks. Tone1Mask is intended to cover smaller stars. Tone2Mask is intended to cover larger stars.

Clone the RGB image to make StarMaskSource. This image serves as the source for the two masks. It is already stretched, so StarMask should work well.

Tone1Mask

StarMask with crop of resulting mask

This StarMask setting generated a star mask which picks up the smaller stars. Note that large stars may not be picked up at all. The dimmer ones need to be brighter so the smallest stars are appropriately affected by the removal process. Run the Range Mask step to brighten the dim star areas.

Crop of Tone1Mask after Range Selection

Tone2Mask

Run HDRMultiscaleTransform on StarMaskSource. This reduces the effect of nebulous areas on the star mask.

StarMaskSource after HDRMultiscaleTransform

Run StarMask below to generate the Tone2Mask. Verify that all of the big stars are masked; sometimes a big star is not actually as bright as others and can be missed. Try to only get the big stars in the mask, not intermediate stars that have already been removed by Tone1Mask. It might be a good idea to use the image after Tone1Mask has been applied; I have not tried this.

 

Crop of Tone2Mask for bright stars

Remove Stars

Now back to the image ready for star removal.

When the previews look OK, apply Morphological Transformation through Tone1Mask to the main image.

The small and medium stars should be gone.

Change the mask to Tone2Mask and apply Morphological Transformation. Again, test on the previews first to make sure it works OK. You may need to change the number of iterations in Morphological Transformation to avoid putting dark holes where the stars were.

Many people combine the masks into a single mask. I use two masks because each mask can require different numbers of iterations. In this case I ended up using 7 iterations. This removed most of the stars without producing black holes.

The image after PixInsight star removal

The image still has some large star remnants, which we will remove in Photoshop.

Save the file as a Tiff for Photoshop processing.

Specify 16 bit unsigned integers! My version of Photoshop does not allow tools to work on 32 bit integers, so I have to convert the image to 16 bit in Photoshop. When this 16 bit file is brought back into PixInsight the colors are different. Saving from PixInsight in 16 bit integers in the first place preserves the colors across Photoshop.

Photoshop Processing

The goal here is to remove the remaining very large stars.

Open the tiff image in Photoshop. Select the Spot Healing brush. Set the brush to be bigger than the target star by perhaps 30%; the circle wants to be significantly larger than the star.

Zoom in on the image. Find the bright stars and click the brush on the star to remove it. Hopefully there are not too many stars remaining. You may want to change brush size for larger / smaller stars.

Note:  J-P Maisvainio has a prior step where he applies Dust & Scratches filter many times to remove the smaller stars. This appears to be the equivalent to my Morphological Transformation steps.

Save the tiff file back to disk (still 16 bit unsigned integer) and reopen in PixInsight.

Image after Photoshop star removal

 

 

Yet Another Flat Frame Panel

Edge Lit LED panel is installed in a wood frame for mounting on the observatory wall. The dark gray plastic sheets slide in covering the white panel.

I have tried a variety of devices in an effort to have flat panel flats in the dome. All were failures:

  • White cardboard square mounted on dome, lit with lamps below. The flats were slightly tilted since the screen / light / OTA are not aligned well.
  • Light boxes. Tried two versions; both failed to provide an even illumination across the surface.
  • Electroluminescent panel. It is not white, kind of a green/blue. Hard to get enough intensity in the Red, HAlpha, and SII filters. Cannot control from the computer. Too small for the C11.
  • Light table. LitEnergy A2 model, although all the different vendors look identical and are likely produced by the same Chinese factory. Unstable electrically, will change intensity for no reason. Cannot control from computer. LEDs do not stay on, but seem to be “panning” in strips across the surface, requiring longer exposures.

So, here is the next hare-brained approach. The lighting industry has developed LED lighting panels intended to be used as replacements for the typical fluorescent light fixtures. They typically come in 2’x2′ and 2’x4′ sizes so they have the same footprint as the fluorescent fixtures. In addition, some models allow dimming by a wall switch slider.

I bought a 2’x2′ model for $45 on Amazon. It is an ASD Edge-Lit Flat Panel, Standard Series, model ASD-ELP22D4040. It is a 40W dimmable panel, Color temperature 4000K, 3400 Lumens. It is nicely solid, unlike the Light Table monstrosity.

Dimming Mechanism

The panels use “standard 0-10V dimming”. There are two sets of wires in the “controller” (looks like the typical ballast box on fluorescent lights). One set is the usual three 110V wires (black, white, and ground). The second set of two wires is for the 0-10V dimming, marked as 0 and +10V.

My original thought was that I should provide a 0-10 V signal on the wires. Wrong! The controller is sending the 10 volt signal – I just need to put a resistor across the wires. If I short the wires together the panel is off. Put a small (100 Ohm) resistor across the wires and the display is dim. A large resistance (leave the wires separated) and the display is on full brightness.

Even at the dimmest settings the panel is too bright, especially for binned flats. I inserted two sheets of smokey gray acrylic plastic over the panel. Each sheet is about 3/16 inch thick, retrieved from the local plastic company’s scrap bin. The sheets attenuate the light nicely, so even 4×4 binned Luminous flats work well.

Solution

I set up a simple Arduino Uno ($10 knockoff) which adjusts 2 digital 10K potentiometers connected in series. The light controller wires are connected to the 2 pots. The Arduino can set each pot in steps of 0-255. So, for a dim light I set the first pot to something like 10, and the second pot to zero. For more light I need to set a value into pot 2 as well. The brightest value of 20KOhms is probably not the max available from the unit, I could add a third pot if needed.

Control Program

I wrote a little control program to send commands to the Arduino. Features are:

  • The COM port is the one assigned by the Arduino IDE when it communicates to the Arduino.
  • The power cord to the light panel is controlled through a Digilogger IP power switch.Two special values are used to control power: a value of 0 turns the power off, and a value of 255 turns the power on. These values are hard-coded in the ACP AutoFlat script. The DigiLogger Program, IP address, and Port are used to perform these power operations.
  • The program stores its settings in a config file. Establish the settings by running FlatPanel manually, setting the values, and exiting the program.
  • Subsequently FlatPanel can be run like a console program. If the runstring has an argument (for example, 255) then the Arduino is set to the argument value and the program exits without displaying an interface. This is the mode used by the AutoFlat script. If the argument happens to be 255 the power is turned on; a value of zero turns the power off.
  • The two panel values are sent to the Arduino, with the two bytes combined into a single integer. The value of that integer is displayed in the label at the lower left of the screen. These integers are used in the settings placed in the AutoFlatConfig.txt file.

AutoFlatConfig.txt file Settings

I manually determined the values for each filter/bin combination so that a flat image requires about a 1 second exposure. These settings are then entered into the AutoFlatConfig.txt file. My settings for this portion of the AutoFlatConfig file are:

; Starting with ACP 8.1 you can specify separate panel brightness values
; for higher binning, accounting for the higher sensitivity at those
; binning levels.
;
;                        Lum R   G   B   Ha     Oiii   Sii
PerFilterBrightness      22, 70, 65, 75, 65530, 65530, 65530
PerFilterBrightnessBin2   6, 20, 20, 20, 41210, 65335, 65530
PerFilterBrightnessBin4   1,  5,  4,  4,    50,    50, 50
;
LightCtrlProgram D:\Dropbox\BrewSky\Programs\FlatPanel\FlatPanel\bin\Debug\FlatPanel.exe
LightOnCommand #BRT#
LightOffCommand 0
LightOnDelay 2 ; Time needed (sec) for brightness to stabilize

 

 

Davis Weather System Problem

Davis VantagePro2 system

The Davis VantagePro 2 weather system has been running well now for perhaps 5 years.

We got actual rainfall the other day, a real soaker probably 0.5-0.75 inch. We last had rain back in January; even when it rains in the area it typically misses us.

Well, the data was pretty strange. It showed we got 68 inches that night…. Probably not.

I went out and removed the rain shield; it had a small dead bug, but wasn’t particularly blocked. The buckets were clean, I don’t see anything wrong.

Toggled the bucket back and forth a few times, but didn’t see any response. It should register 0.01 inches per tip.

Called Davis support. As we were talking, I started seeing a steady amount of rain being registered. It was a constant 0.12 inches every 30 seconds. It was certainly not raining! Conclusion: something is wrong with the transmitter device.

They mailed a replacement device ($60). I got it installed OK, seems to be reading OK. Sprayed some water into the rain cone, saw it register. It takes a minute or two to register due to the update frequency of the console and of weather display. Sent back the old one.

Light Table as Flat Panel

Light table is a thin (3/8″?) plastic assembly. The single button is in the upper left, along with power connection.

A few weeks ago I started using a light table to produce flat fields. These units use LEDs to illuminate a large screen. Apparently artists use them to do tracing, and some people use them to entertain small children.

I first bought a cheaper version, but it failed after the first power on. I then bought a LitEnergy table for about $110 on Amazon. The lighted surface is about 24″ by 33″ (size A1), so both OTAs can point at it at the same time. It has a 110 Volt power connector and a single capacitive button for control. Touch the button once and the unit powers on. Hold the button and the unit changes intensity down to a fairly dim level, or up to the brightest level. I never know if it will go up or down.

The basic unit is too bright for flats, even at a dim setting. I collected three 1/4″ smoky acrylic sheets (Filter Sheets) to put on top of the table to filter the intensity further.

Overall, it kind of works but I am not happy with it.

Issue 1: Power On

For running remotely, devices are generally left in the “On” state with their power switch. Device are turned on and off  through an Internet Power Switch (Digilogger). In this case there is no On switch I can leave on all the time. I have to physically touch the On switch to power on the unit. This makes it unable to run remotely.

There is also no way to remotely adjust the light intensity.

Issue 2: Static

The unit seems to be poorly wired or grounded or something. When turned on I feel static; a couple of times I got small zaps when touching the button or the frame of the unit. The first unit displayed the same problem. I think it died as a result of one of the zaps.

When I place a filter sheet on the unit, static immediately charges the sheet. Placing the filter sheet can cause the intensity level to change as though I touched the capacitive button. Touching the sheet or the frame can also change the dimness setting on the unit, without touching the button.

In addition, the static can cause the unit to change intensity settings even without touching it. In the middle of running a set of 25 flats the intensity can suddenly change a couple of levels, causing the flat process to either fail or take bad flats. The static does not play well with the capacitive button.

Issue 3: Setting Intensity Level

It is difficult to set a particular light level on the screen. There is no indication of the currently selected level. I try to adjust the setting to the maximum level, then hold the button and count the number of dimmings, but

  1. the dimmings are subtle so I can be off, and
  2. it is hard to know when the unit is on maximum. Even if I count correctly I might have started off by 1 or 2 from maximum when I started. There really needs to be an indication of the current level setting.

Adjustment is finicky and not very reproducible.

Issue 4: LED Flicker

It turns out that the LEDs are not on constantly. At dim levels, it appears that the LEDs perhaps are turned on a few rows at a time. The resulting image shows “bars” rather than an even image. Visually the intensity looks even, but exposures less than 2.5-3 seconds are very different.

Thus, I need to select filter sheets and a light level so I get more than 3 second exposures. In practice, exposures tended to be 7-15 seconds depending on the filter. This doesn’t seem like a big deal, but if I take sets of 25 flats per filter at 15 seconds per flat, over 7 filters, at 3 binning levels, it takes over 2 hours. It is much more convenient to take flats at less than a second. If the overall process could be automated this problem would not be as serious.

To take flats over several filters I need to physically be in the observatory. For each filter/bin combination I need to select the necessary plastic sheet, adjust the dimness setting, and run the ACP Autoflat script. I might be able to run 2 or 3 filters at this sheet/setting, then I need to select a new dimness and sheet combination. It takes a couple of hours to run all of the combinations. I have to be in the observatory the entire time to continually change settings. As noted in Issue 2, I also have to watch for the light intensity shifting on its own, requiring me to restart the process.

I am looking for another alternative.

Radiant Barrier

Squares attached to dome. Covering the inside roof areas is under way.

I have been running about a year now with the two air conditioners. In the beginning this worked very well, but now the air conditioners are struggling. It seems that after a year they do not cool as well. I hate to think I have to buy a new AC every year:(

Previously we had insulated the walls of the observatory with fiberglass batting and drywall. The dome itself continues to get very hot, perhaps 116 degrees as measured by the IR gun.

So, I am implementing two more temperature control measures.

Phase 1: Radiant Barrier

In this first phase I am attaching radiant barrier material to the inside of the dome. I am using simple perforated barrier, not the kind with “bubble wrap” in between. The company says it needs a 3/4 inch gap between the barrier and the roof to work correctly.

Notice the shutter; lower shutter is covered, one piece on the upper shutter.

The material comes in a roll, 24 inches wide. I cut a 24″x24″ rough square from the roll and attach it to the dome with aluminum tape. Pieces are attached to the dome when possible, to other squares otherwise. I overlapped the pieces several inches at least; I am putting square pieces on a round dome, so things don’t line up neatly. The tape is kind of dangerous – I got a number of “paper cuts” on my finger before I realized what was happening. I figured the tape would attach well to the barrier, but I was concerned that it might not stick well to the dome. It turns out the reverse seems to be true;. It sticks very well to the dome, but may develop a problem with un-sticking from the barrier. Time will tell.

I covered the dome, the lower shutter, and under the flat portion of the roof around the dome. I tried one piece on the upper shutter, and it seems to be OK. I worry that it will catch on something when the shutter opens. In addition, it is very difficult to get to the zenith of the dome, so I did not put barrier on the rest of the upper shutter.

The observatory is noticeably cooler now. The IT gun shows temperatures around 90-93 degrees. The AC is still struggling, but not as much as before. We will see if the barrier will stay attached.

Phase 2: Reflective Coating

There are a number of reflective coatings out there that are supposed to reflect the sunlight, or prevent “heat loading”, using nano particles of some sort. Some are Sunshield (Home Depot), Tex-cote, Kool-seal, ThermoSeal, and SureCoat.

I have selected a product from APOC. They make a standard product APOC 247 White for this purpose. In addition, they have a special version APOC 248 – Arizona White which is supposed to be better formulated for the Arizona heat and UV. It is certainly true that the Arizona sun is very high in UV and is extremely destructive, so I elected to try this one.

I have a guy coming in later this week to clean the dome and apply the coating. It looks straightforward, but at my age I am too nervous to climb around up there.

 

Dust Prevention

Image result for haboob in phoenix

My observatory is in the Arizona desert. As such we get a tremendous amount of dust in the air.

  1. I’m sure there is a large amount of dust suspended in the air in general.
  2. We get occasional high winds which pick up a lot of dust as they blow across the desert.
  3. We get occasional dust storms which are amazingly dense dust – you can’t see 3 feet in front of your face. Apparently “dust storm” is politically incorrect, so the media has taken to calling them “haboobs”. The image above shows a storm coming across the desert. Notice the mountains in the background; that wall of dust is BIG.

I spent 4 days cleaning the observatory. The dust is tough to get up; it is thick and deposits like mud (probably electrostatically charged). You can’t use a Swiffer or dust cloth – at best that just smears the dust around. You have to use rags and water to pick it up.

I have implemented some anti-dust strategies which I hope will help:

  • I made a cloth skirt which goes around the outside of the dome. The skirt is about 6 inches long, hanging off the rotating dome and dragging on the roof. I’m hoping this will provide another barrier to dust coming in that way. The cloth has two ropes sewed into the edges. The top rope allows the cloth to pull tightly against the plastic band holding it against the dome, making a tight seal. The lower rope weights the bottom edge of the skirt to make sure to drapes down onto the roof.
  • I plan to install some weatherproof sealing against the edge of the shutter. The upper shutter in particular does not seal tightly when closed.
  • I put cloth drapes over the computer table and various equipment cabinets. I also put a drape around the base of the mount. When I need something I just lift the drape (to get at a tool chest, for example). I put Velcro on the computer drape to hold it on the monitor – it kept sliding off to the back.  Now, every so often I will just need to wash the drapes.
  • My electronic cabinets were designed for upstate New York, where I encountered serious insect invasion issues. The cabinets have open front and back doors for air flow, with two layers of screen keeping out bugs. I added a cloth layer to the doors to hopefully keep out dust while still allowing air flow.
  • I changed the scope Park position back to pointing the scope downward. I will need to change this in the summer; the heat causes grease on the Celestron Edge baffling to drip onto the corrector plate. However, the rest of the year I hope that less dust will settle on the telescope glass if it is pointing down.

New FSQ106, maintenance

FSQ

Have obtained a new-to-me FSQ106EDIII. After some false starts I have the adapters from OPT and Texas Nautical. Built another bracket to attach the EZFocus motor.

While testing the focuser for slippage the wall wart died. Very hot, no voltage. Replaced it with a cable to the old 12V power supply, seems to be working now.

Odd issue – while running the VCurves and ACP Focus Offset script, I notice that the focus position seems to be consistently changing. If I run multiple focus runs in a row, as quickly as possible, the resulting position seems to shift by about 35-40 focuser tics. This shift is probably roughly the size of the crtical focus zone.

So, when Focus Offsets runs, it seems like the focus position of each filter is off by n*35 tics. When HAlpha comes up with an offset of 189 tics, it may actually be off by 140 tics!

I don’t know what is causing this shift. The scope certainly doesn’t shift temperature that fast. I thought maybe the focuser was slipping, but measuring the position of the camera with a dial caliper shows to slippage. I ran the focuser back and forth 5 times (1000 pulses each way) and it ended up back at the correct position each time.

I did find that the backlash was off; I changed from 200 to 500 pulses. I will re-run the Focus Offset script again.

Maintenance

Updated ACP to 8.1, including updating the web system.Cleaned up the web menu system a bit, added a couple of items (AstroCalc).

Updated FM to 4.1.0.44.

Drizzle Kernel Function Note

Playing with Drizzle again.

Saw a mention on the PixInsight forums about Drizzling with factor 1. Apparently this can help reduce the noise in your stacked image? Odd idea, I will have to try that sometime.

Tried varying the drizzle Kernel function. There are several possibilities, primarily Square (default), circular, and Gaussian. Potentially Gaussian can provide better results, but needs a larger number of subs to work well.

So, I tried doing 16 subs (Blue filter of M100, binned x2, 15 minutes each), using each of the 3 functions. I used a 0.9 drop size for each, drizzling by a factor of 2. The hardware was the STF8300M / Edge 11 combination. Measured the noise of the integrated image using the Noise script.

Kernel Noise (e-4)
No Drizzling 2.590
Square 1.511
Circle 2.015
Gaussian 2.580

Clearly in this case the default Square function performed the best. The Gaussian took significantly longer to calculate and was no better than not drizzling at all..

Bias Frame Stability

In the course of chasing down a weird problem, a couple of issues came up in discussions with other people.

Does the Bias Master frame change at different temperatures?

I.e., do I need separate Bias Masters at each temperature, or does one master work at all temperatures? Maxim is clearly set up to use different temperatures.

I generated Bias Masters at several temperatures from 0C down to -13.5C. This was as low as I could go, since Arizona is still too warm for the camera to get any cooler. Forty subs were taken directly in Maxim and converted to the Bias Master in Maxim using the SD Mask combine of 40 subs. The subs were taken after the new temperature set point appeared to be stable (roughly 20-30 seconds after the set point was reached). I ran a couple of them twice to see how much variation might be seen from run to run. Also, I have a couple of older masters (1-2 years old) at -15C and -20C.

For each master I measured the median intensity at the center of the image using Maxim’s Information tool with the biggest aperture available (20 pixels).

biasvstemp

It seems clear that the recently built masters are the same out to -13.5C. The colder ones are likely too old, which is why they are different. I would not expect some dramatic shift from -13.5C to -15C.

Is it necessary to wait some period of time for the CCD cooling to stabilize?

In discussions with Joe Mize (he has had a number of interesting points!), he pointed out that he waits a couple of hours for the CCD to stabilize before taking Bias frames (he calls this “cold soaking”).

As a typical impatient person, I generally set the cooling target, set up the AutoSave sequence to take the desired Bias and Dark frames, then start the imaging. The Bias subs are taken first, for no particular reason. As the temperature approaches the set point the temperature typically oscillates a bit around the set point as the system control settles. This usually happens for 15-30 seconds. Once the cooling target appears to be stable I start the AutoSave acquisition.

Now Joe indicates I need to wait a couple of hours:(

So, I did a couple of runs. Run 1: I set the cooling point to -10C, waited as usual for the temperature to settle, and took a set of 40 Bias subs. These were combined into a master in Maxim’s Set Calibration tool using SD Mask median combination. I then used the Graph Information tool to draw a horizontal line in the exact center of the image.

After 30 minutes I took another set of 40 subs and created another master. I repeated this process at 1 hour and 1.5 hours. The following animated gif flips among the resultant graphs. The red text at the upper right indicates the elapsed time before creating the master.

30minbiascool

It appears that the initial master is in fact somewhat lower than the subsequent graphs by 32 counts (about 1.5%). After 30 minutes the graphs seem to be relatively consistent. So, it looks like I do in fact need to wait perhaps 30 minutes.

Run 2: I turned the system off for about three hours, then repeated the test. This time I created a master every 5 minutes in order to get a better estimate of when it has stabilized.

5minbiascool

I should have changed the label at the top; 1 is the first master, 2 is after 5 minutes, 3 is after 10 minutes, etc. It looks like by 10-15 minutes the masters have equilibrated.There is less of a shift (15 counts, or 0.75%); perhaps the system didn’t completely warm up from the previous test.

So, it looks like I should let the system cool for 15-30 minutes before starting the Bias frames.

I might expect the same effect when taking darks. However I expect the issue to be less noticeable since the subs are typically 15 minutes or so; the first sub might be off but subsequent subs should be equilibrated. There might be an effect if doing short subs like 1 minute.

Another Measure

I was reading an interesting article Signal to Noise: Part 3 – Measuring your Camera by Craig Stark. I thought it would be fun to go through the exercise of measuring the various characteristics of my STF8300M.

One of the checks is this very topic – how long does it take your camera to cool to a stable temperature? His approach is to turn on the cooler, then start taking 1 minute dark frames. He expects to see a pattern in the beginning followed by a steady state dark median after some time.

In my case, the camera reached the target within 1 minute! Only the first dark showed any significant difference while the camera cooled.

darkstability

Single Star versus Multi-star guiding in Maxim

I have always used single star guiding in Maxim. However, in Maxim version 6 they implemented multi-star guiding. I wanted to test the multi-star compared to the usual single star.

Basic conclusion: multi-star guiding does not show an appreciable gain in guiding precision. There may be other reasons for using it; perhaps it is more robust somehow, perhaps it is better in poor seeing conditions? I was unable to test these issues; I just looked at guiding precision.

The clear difference in guiding was better seeing as the night went on. The initial guiding started after astronomical twilight. Both guiding techniques improved significantly as the evening went on. I often see this trend when using the Subframe Selection script in PixInsight; as the evening wears on the FWHM moves steadily downward.

Data Collection

I imaged several 5 minute exposures, binned 2×2, using the Edge 11/STF8300M setup. The guider is the integrated STi in the 8300 package. This setup has an image scale of 0.80 arcseconds per pixel.The STi is binned 1×1. The same calibration settings were used for Single star and Multi-star guiding.

The images were well guided; I happened to have better seeing this night. In poorer seeing I typically see larger fluctuations in the guide errors. It may be that multi-star guiding would help more in poor seeing.

I used three targets: one in the Milky way to ensure a good number of guide stars for multi-star guiding, and two targets with few guide stars. The images are not calibrated or stacked, although the Maxim Filter tool was used to remove hot and cold pixels.

I measured the guiding accuracy using the graph tool in Maxim. After guiding has been started I clear the graph and let it run for 30-50 points. I then read the RMS X and Y errors from the graph, clear the graph, and repeat. In each situation (single or multi star guiding, a particular target) I repeat this process 5 times and average the results.

I also measured the FWHM of the resulting images using both CCDInspector and PixInsight. Their results were essentially the same, which is reassuring.

Results

rmsguideerror

This shows the RMS guiding error as the evening went on. The green points are using single star guiding, the blue use multi-star.

In the first half of the graph the RMS errors are more variable. In the second set of three readings the guiding has settled into a more consistent pattern.

fwhm

These are the FWHM measurements from PixInsight (in pixels). The first image in particular was poorer (3.3 px * 0.8 as/px = 2.64 as. The last six images are around 2.4 arcseconds.