1189 replies to this topic

[corsiar]

Jon what was your average tracking taking these images?

[Jon Rista]

Jon what was your average tracking taking these images?

It varied a lot. I had some mount problems which resulted in some tacking issues there, but more significantly I had very widely ranging seeing. It ranged from around or even under 1.5" seeing to as much as 4-5" seeing, across the last couple months. Some nights were fairly consistent, some nights I could go from sub 2" seeing to 5" seeing within a matter of an hour or two. It was very inconsistent. I was never actually able to get consistently good results from an FWHM standpoint because the seeing was so variable. I have several hours to maybe a nights worth of very high resolution subs on most targets, with FWHMs at or below 2". But to get enough SNR for a good image, I had to mix those high resolution subs with subs with FWHMs as large as around 3".

 

It's been a very strange year, and really a bad couple of years for imaging here. Since January/February in 2016, I had mostly clouds, with only a night or two clear every couple of months or so. The entire summer cloudy or smokey last year. I had a couple sets of clear nights in late fall, but they were very hazy nights, so SNR suffered due to higher LP as well as poorer object signal (due to being reflected or scattered by the haze). I had a couple clear nights in late winter and early spring this year, then months, a small handful of clear nights during summer this year, then months. Most of those nights were either hazy or had smoke...and some of them the smoke was so bad I could barely see any stars, so didn't bother imaging. I finally got enough clear nights to get enough data for some full multi-channel narrow band images starting in late September through now...however as mentioned, the seeing has been wildly variable and inconsistent.

 

I tell ya, what I would give for consistent 2" or better seeing. I'll take some haze, now that we have LocalNormalization in PI I think I could normalize a lot of the haze issues out...but the wildly varying seeing, especially once it tops 3", is basically sub killing, even night killing. And that's the case either with the ASI183 or the ASI1600...both cameras have pixels small enough that beyond 3" seeing I lose too much resolution for me to achieve my goals. I've tossed entire nights worth of data, sometimes multiple nights, from both cameras, on many targets over the last year and a half or so. Hoping the rest of this year and next year end up better.

[Henry from NZ]

Very nice images of the Cave Nebula, Jon.

 

I wonder if you (or anyone else) have images taken with this camera (or its colour version) using a focal length of around 800-1000 mm? 

[Jon Rista]

Very nice images of the Cave Nebula, Jon.

 

I wonder if you (or anyone else) have images taken with this camera (or its colour version) using a focal length of around 800-1000 mm? 

I'll be trying at 840mm. I have a 1.4x TC that I can attach to my 600mm lens, and I intend to give that a try once some galaxy season targets roll past my trees (probably about a month before that happens). I don't know how the optical quality will be....I guess I'll find out.

 

And thanks. I am actually a bit disappointed in Cave myself. I definitely lost SNR there. I put in a ton of integration time, and I feel I only got about half what the effort should have delivered. The undithered subs only varied about half a pixel to a pixel, in each set, and the different sets only varied by about a pixel and a half from each other. I am kind of surprised my tracking was that good, but I think that was also when I was getting around 0.45-0.55" RMS guiding, and my PA must be pretty good right now. Anyway, the lack of dithering definitely bit me in the rear on that one.

Edited by Jon Rista, 24 November 2017 - 10:55 PM.

[Jon Rista]

IC1871 - Soul Nebula Detail (#3)

 

Third image with the ASI183MM Pro, I finally started to get a proper feel for the camera, and had ironed out the issues from the previous two images. One set of data for this object, which I was imaging concurrently with Cave nebula (early morning object after cave had set below the housing and treeline to the west), was also undithered. I managed to get enough OIII to do a basic blend, however OIII acquisition for this object is ongoing as the channel is currently very shallow and doesn't quite have the necessary contrast. Hopefully I'll get a chance to finish this object up before it's gone.

 

The Ha channel is quite strong in this object, and as some of it was imaged during the moon, I acquired more Ha than OIII:

 

 

This channel ended up quite clean at 115x180s subs at gain 200, which is about 5h45m integration. The OIII channel did pick up some of the oxygen gas structure, which as usual was much more cloudy and tenuous than the Ha structures:

 

 

Only 48x180s subs here also at gain 200, which is 2h24m integration. I am hoping to get another 6 hours of data in OIII before the winter is out. I am hoping some more data will at the very least clean up the signal, as the OIII channels is the primary source of noise in the following HOO combination. I am not sure if getting more OIII will help much with the contrast...the linear alignment approach I use to combining my images is about the least biased and most natural way to combine NB channels, and when the star colors come out fairly natural (vs. heavily blue cast and bloated), it usually means the two channels are naturally related. The linear alignment could be biased a bit by an overly weak signal in one channel, though, which would affect the median or mean measurements and can result in the alignment being shifted a tiny bit to the right. Hopefully with a stronger OIII signal, the linear alignment will produce a more accurate alignment and improve the OIII contrast a bit relative to Ha.

 

 

This is a custom HOO blend. The original pure HOO blend without any cross-blending was furiously blood red, and the OIII signal was hardly visible at all. I adjusted the blend as follows to balance it just a slight bit:

 

R: Ha*1.0

G: OIII*.85 + Ha*.135

B: OIII*.89 + Ha*.1

 

The color should, as with Elephant Trunk, be a pinkish-magenta red, rather than just a pure red. More pinkish where the oxygen is. It may show up a bit more of an orangish red on sRGB screens, especially if they are not well calibrated. Color may be a bit more muted on sRGB screens. Those with true wide gamut AdobeRGB or better screens should see the image as it was intended with the proper natural color.

[Jon Rista]

Here is a crop of the 2x2 binned (downsampled 2x) version of the image:

 

 

Noise is not too bad here, although it is still definitely worse in the blue and green channels. Below is a full 100% crop of the core detail region, which is noisier due to the weak OIII channel:

 

 

Detail in this image is fairly good. As with the previous images, I had a mix of nights with good seeing and nights with poor seeing. I culled out the worst (usually the stuff over ~3" FWHM). These images have not had an deconvolution, however due to the fact that the data is well sampled, deconvolution should provide considerable benefits for improving resolution. This is in contrast to undersampled data, which does not deconvolve well.

Edited by Jon Rista, 25 November 2017 - 12:00 PM.

[Jon Rista]

A couple more detail crops, showing some of the many isolated little globs of dust and gas in the region:

 

 

 

With 8-9 hours of OIII, I am hoping these tiny isolated little details will clean up considerably, as they are much cleaner in the Ha channel. That said, longer exposures with both channels would still be better. A three minute sub is just not quite enough to swamp the read noise by the minimal 3xRN^2, which is the lower limit I aim for these days. You might notice a slight bit of horizontal banding here. This is in part due to the weaker exposures, and longer subs at a lower gain should help bury the banding. It is also in part due to weaker dithering. I recently purchased a new 60mm f/3.6 guidescope, and it seems my dithering config in PHD2 needs to be tweaked so that my dithers are aggressive enough. At the moment they do not seem to be changing the star positions quite enough to fully average out the remnant FPN.

Edited by Jon Rista, 25 November 2017 - 12:02 PM.

[Jon Rista]

These are with a QHY183C OSC on a William Optics 98mm with a f0.8x reducer/flattener using 600 sec subs. Knowing what I do now, if I re-image them it will be with more subs of a shorter length.

Scope is f/6.3 @ fl 618mm with the reducer it is f/5 @ fl 494mm.

 

 

Hey Paul, I think I missed these before (AstroBin thumbs often don't show up for me). The M33 looks quite nice. Slight greenish tint to both images. I am curious, have you ever had any trouble calibrating out the amp glows? That seemed to be a problem for many early adopters of the QHY183C. I tested a few data sets, and following the no-scaling/no-optimization rule for darks, it always seemed to work for me, but I was always curious if there was a deeper issue.

[Pauls72]

Jon, I did initially have problems with amp glow and then I got some help from you.   Stacking the CMOS camera subs with Nebulosity3 or 4 just did not get rid of all of the amp glow. But using DSS and selecting 'Median Kappa-Sigma clipping' does cleanup the amp glow.

 

Once I got things straightened out with the amp glow on the QHY183C I really enjoyed it. Mostly used it on my William Optics 98mm APO with a 0.8x reducer flattener. Scope is f/6.3 @ 617mm which is .8 arcsec/pix, with the reducer f/5 @ 494mm and 1 arcsec/pix.

 

I did buy a QHY163M and QHY filter wheel to move to narrow band mono and did image a few things in B/W. Then got a Baader Filter Set LRGBC + Narrowband H-alpha/O-III/S-II, 36 mm. About that time I ran into a number of medical problems and other life events. So I haven't done any imaging this year. But I have been able to keep up with most of my visual outreach events. However I am in the process of getting things set to image again.

 

Yea, my processing skills still need some work. I still have the original subs, so I can go back and reprocess them.

[Jon Rista]

Yeah, I remember now. I actually emailed Craig Stark about the issue. From what I can tell, Nebulosity is performing dark optimization, however there did not appear to be any UI options for it, and Craig assured me that Nebulosity doesn't do dark optimization. I wonder, however, if it may be doing something else internally that is missmatching the darks to the lights. I don't have the program myself, so I was never able to fully figure out the issue.

 

At least DSS works. It's kind of a bummer that there aren't more alternative calibration programs out there. DSS is getting a bit dated these days... I've wondered what it would really take to create a simple command line calibration tool. You could still use DCRAW (everything does) for DSLR support, and there are libraries out there for FITS support. Something bare bones, though. Something that, maybe, you could even run and have it "watch" a directory for new files, and calibrate on the fly as you acquire your data. I could probably write something like that myself...however I honestly don't know when I'd find the time. Anyway...I wonder when the day will come that DSS ceases to function with modern data files...

[poobie]

Jon - you might want to give the windows port of Siril a try.  You'll need more disk, as siril explicitly preserves all the intermediate files, and the workflow isn't as automated as DSS, but it's up to date, and seems to produce tighter stacks for me.  I've been using it with an ASI178MCC, and it calibrates those subs just fine.

[Jon Rista]

Jon - you might want to give the windows port of Siril a try.  You'll need more disk, as siril explicitly preserves all the intermediate files, and the workflow isn't as automated as DSS, but it's up to date, and seems to produce tighter stacks for me.  I've been using it with an ASI178MCC, and it calibrates those subs just fine.

I use PI myself (which also uses lots of disk as it preserves the intermediates), but it is good to know that there is something else out there besides DSS. I'll give it a try and see how it goes, though. Good to know for recommendation purposes. Thanks!

[xiga]

For calibrating and stacking, i give APP a big thumbs up! For whatever reason, having used it i now seem to know what a lot of the terminology really means (calibration, registration, integration, i always got them mixed up but now i know the difference). Also, the ability to see individual lights before and after calibration on the fly is amazing. I was always a DSS guy, and never had a single problem with it, so i had no real need to buy APP, but once i used it and got used to it, i bought it as soon as the trial ended. I can see it being my pre-processing program of choice for a long time to come. 

[Jon Rista]

Alrighty! Time for some read noise stuff. I tend to throw a the math around, but since this is the beginners forum, I'm going to try and demonstrate read noise visually for everyone as well, with corresponding numbers. I am also going to try and demonstrate the difference between observing and measuring noise in terms of ADU, vs. observing and measuring noise in terms of e- (electrons). Hopefully that will clear something up that I think is a constant sticking point for many people.

 

Bias Signal

 

The bias signal of the Sony IMX183 is pretty stable all the way down to 0s exposure times. This differs from the Panasonic MN34230ALJ, which seems to have some slight instability below 0.2s exposures, resulting in variable (frame to frame) bias shading that can throw off calibrations. The recommendation with the QHY163 and ASI1600 (and probably the Atik Horizon once it gets here) has been to use 0.2s as the minimum exposure for calibration frames, which ultimately leads to the recommendation of using flat darks rather than biases to calibrate all flat frames, even those only a fraction of a second long.

 

With the IMX183, the stability of the bias in terms of shading means you could use 0s bias frames and calibrate your flat frames like any normal calibration workflow. That said, as with most Sony IMX sensors, there is some random horizontal banding noise. I am honestly not sure where this comes from, however I've observed it with data I have acquired myself as well as data acquired by other individuals with Sony IMX sensors. This horizontal banding can be fairly strong, and in the case of the IMX183 it is definitely not the worst I have seen, however it can pose a bit of a problem if you are getting shallower signals (i.e. narrow band). Stacking lots of frames combined with fairly aggressive dithering is the ultimate solution, since random banding will not fully calibrate out with normal calibration routines.

 

 

In addition to the random horizontal banding, the Sony IMX183 has fixed vertical banding, both small scale and large scale. The vertical banding structure appears to be quite consistent (I still need to test across gain settings and power off/on cycles), and so far has been calibrating out of dark and light frames just fine. The only potential issue with biases is the horizontal banding. As you will see shortly, the magnitude of the bands fades with increasing gain, and even by gain 53 (~2e-/ADU, much akin to gain 76 on MN34230ALJ based cameras) the banding is diminished a good deal.

 

Since this camera has amp glow (more on this soon once I get into dark signal and dark frames), I still recommend skipping biases for light calibration, and just using a single master dark from well matched dark frames. Well matched means same exposure, temp, gain and offset (offset should be set automatically by the driver now).

 

Read Noise

 

With this particular sensor, the normal gain range is from 0-270 (up to 27dB gain), however the effective range in my testing seems to top out at around gain 200. I have not noted any meaningful improvement in read noise beyond Gain 200 (199 to be explicit, as you'll see shortly), even though the maximum viable is 270. Actual read noise ranges from just shy of 3e- @ gain 0 down to 1.5e- exactly at gain 199. At unity, gain 111, read noise is 2.04e-, and at a stop slower than unity, tested as gain 53, read noise is 2.43e-.

 

In terms of the most effective setting for this sensor. I have found that gain 53 seems to be the universal ideal, although for narrow band unity gain may do just as well for a small loss in DR. I have been using 10 minute NB subs, 30 second L subs and 90 second RGB subs all at gain 53 lately. I'm liking the results, however there is a slight bit of horizontal banding that requires a good number of frames with fairly aggressive dithering to clean up (I recommend at least 50-80 NB subs, which at 10m NB subs is 8-14 hours of integration...which really is just fine!!) For LRGB, stacking up to 200 frames or so offers plenty of benefits, and better opportunity to average out any remnant banding.

 

Below is a comparison of read noise in two scales. The top row is read noise at the specified gain settings, but displayed and measured in terms of ADU. This is how most people will normally SEE their read noise. As gain increases, the apparent noise in terms of ADU also appears to increase. I think this often leads to the assumption that increasing gain increases noise. This is, however, a misconception, one which I hope will be remedied by the bottom row:

 

 

The bottom row is read noise at the specified gain settings, but displayed and measured in terms of electrons (e-). This is not how most people normally see their read noise, however this is the TRUE noise, in terms of actual analog signal. Note how the minimum gain appears to be higher noise, and also note the increased prevalence of horizontal banding. In contrast, unity, gain 178 and gain 199 have little to no banding and a much cleaner, gaussian noise distribution. This is the benefit of increasing gain! This was achieved through the use of a very simple little PixelMath formula that is effectively a DAC, digital-to-analog converter. It takes the ADU counts, and converts them back to electron counts using the gain in terms of e-/ADU for each gain setting. The formula is as follows:

 

K: $T*g - (median($T)*g- median($T)); g=3.6|2|1|.5|.25

 

There are a few things going on here. First off, the symbol g represents the gain. This is 3.6 for gain 0, 2 for gain 53, 1 for gain 111 (unity), 0.5 for gain 178 and 0.25 for gain 199. When converting from electrons to ADU, you divide the electron count by the gain. So if you have 10e-, and your gain is 3.6e-/ADU, then you have 10/3.6 = 2.78 ADU. This is effectively ADC, analog-to-digital conversion. The first term of the above formula does just the opposite...it multiplies the ADU count by the gain, so if you have 2.78 ADU and you multiply by 3.6e-/ADU you end up with 10e-. This is effectively DAC, digital-to-analog conversion.

 

The second term is a normalization term. All this does is make sure that after converting ADU back to e-, the original median level of the image is restored. The original median for all images was also originally initialized to that of the unity gain sample via another formula, one which I've shared here before, what I call linear alignment:

 

L: $T + (median(g111) - median($T))

 

This eliminated any discrepancies in offsets between the various gain settings. This is another interesting characteristic of note about this sensor. There seems to be a slight bit of internal offset control by the camera. At gain 0, the offset settles around 41 ADU, and as you increase gain, it can shift upwards, settling at around 58 ADU by gain 300. I am not sure if this offset tuning will be consistent from camera to camera. In my case, I have found that most gain settings seem to have around 40-45 ADU offset until I get above gain 200, then they shift quickly to over 50 ADU. I had some conversations with ZWO about this, and their measurements were slightly different, so we may expect some variation here. In my testing, it does appear that for any given gain, the offset is quite consistent, it only seems to vary across gain settings (which is good, as if the bias level varied for a given gain, then calibration would be near impossible, and I'd be recommending against this camera!)

[Jon Rista]

Quantization Error - ADU

 

I took a vertical center slice from each of the bias frame crops in the above comparison image. I then plotted all the pixel values to a normalized floating point range of 0 to 0.5 (this is for the stretched data, BTW). This is a nice way to demonstrate another value to increasing gain: minimization of quantization error. With these 12-bit readouts, quantization at lower gain settings is higher. It can be high enough to cause obvious quantization of the output, especially when you plot the data this way:

 

 

While the standard deviation of the noise in terms of ADU is lower at gain 0, you can easily see the quantization. You can still see quantization at gain 53 as well, although it is definitely less. By gain 178 and 199 the quantization error has all but been buried by other noise. This is the benefit of "sampling" the signal at a higher resolution. At gain 178 and 199, we are sampling the signal at a sub-electron scale...0.5e-/ADU and 0.25e-/ADU respectively. You can see that we are getting a much wider variation in terms of possible values for each pixel at these gain settings. These measurements are a much smaller scale of noise in the voltage applied to the ADC, as well as variations in the ADC circuitry itself, in the sensor during readout. At these gain settings the IMX183 easily samples the signal as well as a CCD. At unity, you can still see some quantization, but not nearly as much as at gain 0.

 

It should be noted that this quantization error is only an issue with very small signals. In this case, these are bias frames, and otherwise completely devoid of any image signal. I'll be sharing some similar plots in the future of some LRGB and NB signals to demonstrate how large signals can "swamp" the quantization error, and once again give you a clean well-sampled signal.

 

Quantization Error - Electrons

 

The same kinds of plots as above, only this time from the bottom row of the comparison image, the data in terms of electrons. This better demonstrates both the improvement in quantization error with higher gain, as well as the reduction in read noise:

 

 

Again, at gain 0 and gain 53 you can see the quantization. At unity there is still some quantization but it is much more reasonable, and at gain 178 and 199 quantization error is basically gone.

Edited by Jon Rista, 27 November 2017 - 10:44 PM.

[andysea]

This is awesome! 

Thank you Jon for putting so much effort into this very detailed review.

I was wondering if you could elaborate on the QE of this sensor. I think QHY published a value of 84% which is quite high.

 

Thank you!

[Jon Rista]

This is awesome! 

Thank you Jon for putting so much effort into this very detailed review.

I was wondering if you could elaborate on the QE of this sensor. I think QHY published a value of 84% which is quite high.

 

Thank you!

Welcome, Andy.

 

Regarding Q.E. I honestly don't have any good way to measure that. I am not even sure of the proper procedures, but one thing I am sure about, you need a very well controlled environment with a very well understood illuminant so you can properly evaluate the sensor's response to each wavelength of light.

 

In my searching, I did find one absolute Q.E. chart that indicated peak was well over 70%. I'll see if I can dig that up again.

[akulapanam]

 

This is awesome! 

Thank you Jon for putting so much effort into this very detailed review.

I was wondering if you could elaborate on the QE of this sensor. I think QHY published a value of 84% which is quite high.

 

Thank you!

Welcome, Andy.

 

Regarding Q.E. I honestly don't have any good way to measure that. I am not even sure of the proper procedures, but one thing I am sure about, you need a very well controlled environment with a very well understood illuminant so you can properly evaluate the sensor's response to each wavelength of light.

 

In my searching, I did find one absolute Q.E. chart that indicated peak was well over 70%. I'll see if I can dig that up again.

 

 

The 84% is close to what PtGrey is reporting at 79%.  The QE for this guy and the 178 is quite high.  QHY has invested a ton of time in reducing amp glow too so I expect we will see better results in this area vs. ASI.

[Jon Rista]

 

 

This is awesome! 

Thank you Jon for putting so much effort into this very detailed review.

I was wondering if you could elaborate on the QE of this sensor. I think QHY published a value of 84% which is quite high.

 

Thank you!

Welcome, Andy.

 

Regarding Q.E. I honestly don't have any good way to measure that. I am not even sure of the proper procedures, but one thing I am sure about, you need a very well controlled environment with a very well understood illuminant so you can properly evaluate the sensor's response to each wavelength of light.

 

In my searching, I did find one absolute Q.E. chart that indicated peak was well over 70%. I'll see if I can dig that up again.

 

 

The 84% is close to what PtGrey is reporting at 79%.  The QE for this guy and the 178 is quite high.  QHY has invested a ton of time in reducing amp glow too so I expect we will see better results in this area vs. ASI.

 

Does anyone have one of the QHY183M cameras...? (Not sure if those ever found their way out into the wild for beta testing...)

[Jon Rista]

 

This is awesome! 

Thank you Jon for putting so much effort into this very detailed review.

I was wondering if you could elaborate on the QE of this sensor. I think QHY published a value of 84% which is quite high.

 

Thank you!

Welcome, Andy.

 

Regarding Q.E. I honestly don't have any good way to measure that. I am not even sure of the proper procedures, but one thing I am sure about, you need a very well controlled environment with a very well understood illuminant so you can properly evaluate the sensor's response to each wavelength of light.

 

In my searching, I did find one absolute Q.E. chart that indicated peak was well over 70%. I'll see if I can dig that up again.

 

Here is the Q.E. chart:

 

https://www.ptgrey.c...ion-sony-imx183

 

It looks like it is indicating 84% Q.E. At the Ha line it still looks to be around 50%, OIII line close to 80%.  

[Jon Rista]

Dark Signal

 

The Sony IMX183, like any other sensor, has dark signals that grow, in the absence of light, over time. There is an underlying dark current, like any sensor, and there are also glows. Whether they are true amplifier glow (unlikely, amplifiers are built into each pixel now, however there may be other heat-generating units on the sensor die), or due to heat or even other signal (IR?) generated by SoC (system on chip, the integration of all the necessary readout and image processing logic into a composite package paired directly with the sensor, something Sony does as a matter of course these days), I cannot say for sure. Regardless, the impact is the same, and the patterns and practices to manage it are the same as with any other camera that exhibits amp glow.

 

Dark Current

 

The true dark current for this sensor appears to be very low. The official rating is < 0.002e-/s @ -20C. With exposures up to 10 minutes long, I have had very few problems with dark current, the most notable exhibition of which is hot pixels. The hot pixel count of this sensor is very low, lower even than the Panasonic MN34230ALJ. In part, the MN34230ALJ also appears to have RTS (random telegraph signal) which leads to semi-hot pixels, pixels that appear hot but only for a short time before returning to normal, which limits the effectiveness of dark frames for correcting hot pixels. Only true hot pixels will be corrected by a dark, and in these terms the MN34230ALJ seems to dark current almost as low as the IMX183, at around -0.006e-/s @ -20C.

 

As temperatures rise, the IMX183 has better dark current characteristics than the MN34230ALJ. Where the MN34230ALJ seems to double fairly quickly, around every 4.5C or around there, the IMX183 seems to double closer to every 10C. Even as hot as 25C, dark current on the IMX183 is only 0.06e-/s, which is actually incredible. This is an area where Sony seems to do significantly better than anyone, managing dark current.

 

Amp Glow

 

Given that dark current is basically a non-issue with the IMX183, that means the main dark signal concern is amp glow. The glows on this sensor are definitely time-dependent, they grow brighter with longer exposures. The general characteristic of the glows seems to be about the same as with any other Sony CMOS sensor, which all (based on my testing of IMX178, IMX174, IMX294, and IMX183 sensors) seem to exhibit the same general pattern: A primary starburst from the right edge of the frame, two tighter radial glows to the lower left and right corners, and a very faint glow to the upper left corner. Under very deep integrations, some slight glow along the top and bottom edges, and a very faint bubble about half the height of the sensor at the middle left edge.

 

Below is a worst case example of the glows (1 hour of 900s dark frames), stretched to an extreme so that the full characteristic of the glow can be seen:

 

 

I say the above is a worst case scenario, because it really is. The strange thing about the amp glows here is they seem to grow faster with longer exposures. I tested this by integrating the same total integrated dark frame time, 1 hour, using a range of different exposures at unity (gain 111). Comparing 60x60s, 30x120s, 20x180s, 15x240s, 12x300s, 8x450s, 6x600s and 4x900s, a clear pattern emerges:

 

 

I am honestly not sure how to account for this, however it does indicate that stacking more shorter subs will result in less total amp glow than stacking fewer long subs. This lends itself better to LRGB imaging than NB imaging, as well as imaging at shorter focal lengths rather than long. LRGB galaxy imaging at 800mm should be quite ideal, with subs in the 2-3 minute range. That said, I have found that 10 minute narrow band subs at gain 53 seem to fair ok with amp glow, and the sky signal buries the glows enough that there is no readily observable increase in noise from them. This may warrant further testing of the glow growth rate at other gain settings to see if the behavior differs with gain.

 

In the area of amp glow aesthetics (), the Panasonic MN34230ALJ is the clear winner. It's glows are far less intrusive, fainter, and don't seem to grow quite as much or the same way as the Sony glows do. If you are chasing faint details with narrow band, I think the MN34230ALJ based cameras like the QHY163 or ASI1600 are better options.

Edited by Jon Rista, 28 November 2017 - 02:45 PM.

[Pauls72]

Jon,  I tool some fresh bias with the latest QHY ASCOM drivers and I'm not seeing the banding like you have.

I sent you a PM with a link to the bias I just took.

[Pauls72]

I also went back and reprocessed my M33.

 

 

[Jon Rista]

Jon,  I tool some fresh bias with the latest QHY ASCOM drivers and I'm not seeing the banding like you have.

I sent you a PM with a link to the bias I just took.

000s.jpg

Is that a single bias frame, or a master? My master is 100 frames. I can actually see some banding in yours, it's just that your stretch isn't aggressive enough to readily show it, and in a single bias frame the read noise will largely overpower the pattern.

[Jon Rista]

I also went back and reprocessed my M33.

 

 

M33-reprocessed2s.jpg

Looking better. It still seems to have a slight green hue, which can be corrected with SCNR in PI, Hasta La Vista Green in PS, etc.