Another Full Frame camera is released

[billdan]

Both QHY and ZWO have announced another Colour Full Frame camera that is released this month. These are much better suited to long focal length telescopes.

The QHY-410C and the ZWO ASI2400C uses the Sony IMX410 sensor (aka Sony A7iii DSLR).

The camera is a 14bit, CMOS BSI, full frame format, 5.94um pixel size, 24m pixels colour camera, 120ke fullwell.

More specs and graphs here from QHY.

https://www.qhyccd.com/bbs/index.php?topic=8233.msg39203#msg39 203

[gregbradley]

Why 14 bit when the ASI2600 and 6200 are 16bit?

Seems like a step backwards. At this time only OSC as well which limits its appeal. Its more of an "upgrade" to the QHY367C.

I wonder how QHY get 120K full well when ZWO get 100,000? Probably the speed of digitisation making for lower read noise but slower downloads.

Greg

[Atmos]

It's an older Sony sensor, probably from before they were mass producing the 16-bit CMOS sensors.
It's a very intriguing one though, if it was mono it would really interest me definitely over the newer ones. I'm a fan of the slightly larger pixels :)

[gregbradley]

US $3695 seems high for an older sensor not 16bit.

I don't think this is an old sensor. Its only recently been released along with 5 other full frame sensors with various features. The 14bit seems to be aimed at increasing the frames per second.

https://www.sony-semicon.co.jp/products/common/pdf/IMX410CQK_Flyer.pdf

6nm pixels is a good average size for many setups under average seeing conditions. if you get access to better seeing you could go smaller pixels.

I am finding the 183mm with its 2.4nm pixels can be quite sharp at times but it does show up the seeing pretty quickly.

So this camera would be competing with the KAF16200 sensor 16 bit. I must say I am not 100% sure how important 16 bit is. Per the promo on ZWO it shows 16 bit as having smoother gradations of grey rather than distinct steps. What that does to your final astro image I am not sure.

Per Cloudynights thread it was hinted at by QHY that a mono version may be possible. Not sure how often these rumours come true though.

Greg.

[billdan]

Speaking of rumours - This what Dr Qiu wrote recently (23rd May) on the QHY forum.

Hello,

This year we will release a new Sony 4/3 inch monochrome sensor and maybe APS-C monochrome sensor. Maybe consider this.

Best regards,
QiuHongyun

[Sayfog]

The mythical APS-C mono.... this is gonna be interesting if it's true.

[billdan]

+1 - I'd love an APS-C Mono, as long as it doesn't have tiny pixels.

In my perfect world, 6 micron pixels would be nice. A low noise CMOS version of the QHY10, 10Mp, 3900 x 2600, I could live with that.

[Camelopardalis]

Maybe I’m losing it after lockdown, but I’m struggling to see the advantage of the 2400 vs the 6200 :shrug:

Maybe on a larger scope where sub-4 micron pixels are less desirable?

[gregbradley]

You can bin the 6200 and get 7.5nm pixels and a bit of extra sensitivity and still be at 15mp (the famous full frame KAI11000 sensor is 11mp).

Then again the 2400 could also be binned into a 6mp full frame nearly 12nm pixel sensor. That sounds like it would cost a fair bit of resolution though.

File size would be one large benefit. 24mp files are probably around 48mb whereas the 6200 they are 122mb. That would also slow down processing time a lot as well.

I'd love to see some comparable images highlighting the difference between 14 and 16 bit.

I remember seeing some comparisons between the 14bit Nikon D800 and 16bit medium format cameras. It was subtle. Better graduations between colours in the image.

Also I am wondering just how important this matching pixel size to optics and seeing conditions really is. I am not seeing much of an issue with the 183mm on my Honders at .43 arc secs per pixel. A bit harder to focus and more susceptible to seeing are the main problems. On the plus side the 20mp small sensor gives rounder stars as there are more pixels in each star than normal. It makes them look very round, rounder than normal.

Greg.

[The_bluester]

That is the pity of it isn't it, there will never be a comparison between 14 and 16 bits. I think the full well potential is a little meaningless. Unless you don't mind the quantisation uncertainty of 7e/ADC you cant make full use of it. I went back with my ASI294 to zero gain recently (Which stuffs the entire 64K full well capacity of the sensor pixel wells into the 14 bit ADC range) and after one night went straight back to unity gain and deleted the resulting subs, the image quality difference was starkly visible. Even just trying out shooting flats at zero gain (So that the ADC of the image directly correlated to the sensor electron count) made a yucky looking result when those flats were applied to light frames shot at either gain setting.

All that said, it will probably make quite a nice camera when paired to the right telescope, but I think it is likely to suffer from some of the same compromises as my ASI294 where you have to decide between capturing faint detail with longer exposures or keeping good star colours on the brighter stars. Once the ADC output is saturated for any colour channel that is it for faithful star colour reproduction. All you can hope for is to saturate all three so you get a white star, though there will be a bit of a fringe around the white core until all three channels are back in a working range again.

[gregbradley]

Hi Paul,

This is the major difference I am seeing in images between CCD and CMOS. CMOS images often have only white stars or weak star colours versus CCDs where star colours are usually vibrant.

So that to me means full well depth is a more important factor to consider when comparing models than perhaps some other aspects.

The high QE seems to mean the wells fill up more quickly.

Greg.

[Camelopardalis]

I get the maths part :D

If you compare a single image, there is that potential for differences in gradations, but when that is coupled with higher read noise (in a CCD with deep wells, for example), then the uncertainty of the obtained value of a pixel is increased.

But...the ADC width becomes largely irrelevant when you start stacking a bunch of subs, as you're averaging and increasing the precision along the way. You should end up with the same resulting numbers over the same period of time with either 14-bit or 16-bit. (let's imagine we had the same sensor but could change the ADC width...)

I'm sure either of them would make a great cooled, calibration-enabled, un-messed-with full frame camera with any camera lens or scope. I can't help thinking that for the extra US$300 I'd rather have the finer resolution, seeing as all the other stats seem to be otherwise comparable.

I'm loving the 183 on my f/4 newt...although living close to the ocean can show up a fair bit of variance within a night and from one night to another. Such is life :lol:

[gregbradley]

Yeah I get that too. That is the small well playing their part in imaging vagaries.

I remember threads about digital cameras and 14 bit and that lot of the capability of 16bit is not even used. You see that in astro images where the bulk of the data is in one small part of the histogram. All the rest is empty space not even being used.

Greg.

[The_bluester]

The bulk of the data might be in the middle of the histogram, but the brighter stars are not. If you use Astro Pixel Processor, the histogram display relates to the current preview window. If you display a typical image and use the linear histogram display, the histo will probably be somewhere around the middle, but if you zoom in towards an even moderately bright star the histogram will shaft significantly to the right.

I have wondered more than once if the vibrant star colours you more often see out of a CCD is related to the anti blooming gate sensors that many astro cams seem to use (The typical KAF8300 being the workhorse) The anti blooming gate by my reading will essentially make the sensor less linear at the top end of the pixels range, basically the more accumulated charge, the less sensitive the pixel is so stars would be less inclined to saturate than in a CMOS camera of the same full well capacity.

[Camelopardalis]

Anti blooming prevents charge spilling over into adjacent pixels by bleeding off the excess charge, but it doesn’t change that a pixel at its maximum value is always going to look saturated (because it is). Loss of star colour is exactly because of this, regardless of chip design. Balancing exposure time is crucial for optimal results.

Of course, this is exacerbated when we start compressing a 16-bit image into something we can actually display...most computer displays are only 8-bit (although 10-bit displays are available, at a price) and our venerable jpeg is (unsurprisingly) only 8-bit also.

[The_bluester]

Ones that I have read about the working of don't exactly work that way (I would have to dig deeper into the KAF8300) where the pixel accumulates electrons until saturation and then the anti blooming stops it from accumulating more and spilling over to adjacent pixels, the ones I was reading about, progressively bleed electrons at an increasing rate as the pixel approaches saturation.

A telling factor having finally looked up a KAF8300 camera that I know a lot of people use (QSI683) is that it has 25Ke wells, and 16 bit conversion. In fact when you look at the QSI spec page, only 3 out of 17 models listed have pixel wells which cant be quantified at least 1 to 1.

[Camelopardalis]

Look at the gain factor.

The value of electrons in the well is merely multiplied by the gain to produce a 16-bit number.

[The_bluester]

While that is true, to make an easy example of it, if you had a 32K pixel well and 16 bit ADC, you can increase gain to 0.5e/ADC to stretch the data into a 64K count, and then divide by 2 again later to have the original number with good accuracy (I say good as a most of the conversion ratios are not in whole electrons)

But if you use gain of 7e/ADC (100KE into 14 bits in the cam mentioned in this thread) multiply that by four to make it a 16 bit number you can not then divide by 4 and multiply by seven to determine the original number, the original information is lost. Does that matter? Eye of the beholder, but I did decide months ago that for astro use I would not buy another camera which had an ADC that could not reproduce the entire range of the sensor, or at least a really good chunk of it. 16 bits and 80KE, I would probably look at good and hard.

I just see this as a sensor which may have the same colour reproduction issues as my ASI294 where it produces the most asthetically pleasing images (In terms of smooth backgrounds and dark areas) at a gain which means the effective full well capacity is a quarter of it's stated one, so it may be best suited to shorter exposures, like my 294 is. I can do 10 and even 20 minute subs if I really want faint detail, but even reasonably faint stars are pretty "Meh" if I do. Time will tell if I am right or not. I can get some pretty decent images out of my ASI294 with some work and I would assume the same would be the case with this camera. But if we are talking a marginal cost, I would jump to the 6200 for my application.

I still really want an APS-C mono CMOS cam for a bit more FOV with less hassle and cheaper filters, I can dream..

[Camelopardalis]

I’m sorry Paul, I’m always up for healthy discussion, but why must you challenge every comment I make?

[The_bluester]

Well, I could say much the same really. I am not offended if you are not. I am always interested in what others have got to say, just in case they put something out there that makes me think I am wrong! It has happened before.

This is all fairly esoteric anyway given no matter what I might think of this particular camera (Rightly or wrongly) the pixel scale is quite unsuitable for my current gear or anything I am likely to have in the foreseeable future.

[Camelopardalis]

Ah well, in that case I'll have to try harder :P

Just keep in mind that there is well established maths and physics involved, and "alternative facts" like in the realms of faith and power don't cut it.

Regarding the gain and ADC situation, it doesn't matter all that much (generalising hugely). What's important is total exposure time. Run the calculations and you'll see... so long as the subs are sky-limited, then 6x10 minute subs will average out to the same as 60x1 minute subs.

[The_bluester]

I will have to reshoot my demo to show how the ASI294 behaves in the two situations by way of gain (To either allow the highest full well at about 4e/ADC at "zero" gain versus unity gain of just over 1e/ADC) I actually made the experiment earlier this year as when I first bought the camera I used to use zero gain and then changed to unity a few months later, but I had changed so many things in a short time it was hard to compare the data. The difference between images using the same total exposure time at different gain settings was pretty stark. Basically I shot M42 for three hours at each setting using 5 minute subs for both, with flats and darks shot at the same settings as each set of lights and integrated in APP using the same settings. Unfortunately I built myself a new processing PC last week (Talk about timing) and decided when I was moving data over that I didn't need to keep that!

I am in the middle of a mosaic that I will be lucky to finish this season unless the weather gods are really nice to me, once I knock that over I will pick a suitable target and do it again for the point of the exercise and post the results here, happy to prove myself wrong if that is what happens.

Maybe it is just that the ASI294 behaves in a peculiar way that is leading me to my assertions? Given it is the only camera I have, I don't have any way to test that, but they are certainly known for being tricky to calibrate by way of flats.

[Camelopardalis]

Hmmm that’s interesting...which was better? The zero gain or the unity gain?

The 294 does seem to be a bit different than those that came before it. The 1600 is very predictable, a solid performer. The 183 seems to work well too, but it has some frightening amp glow that calibrates out well most of the time.

The newer sensors in the 2600 and 6200 feature the High Conversion Gain mode as does the 294. I guess we’ll have to see how they perform when more of them are out in the wild :shrug:

[The_bluester]

Unity gain produces the nicer looking subs by a long way with the 294. I am just trying to negotiate with the minister for finance on a 2600 at the moment. I would love a 6200 mono but the finance just won't go there for now. I would hope to use the 2600 on zero gain, which conversion wise equals about the same as the 294 at unity gain (120 gain in ZWO speak for that camera) but at the lowest gain the 2600 has half the read noise of the 294 at lowest gain, I reckon it has the possibility of being a very useful gain level with loads of full well capacity to hopefully bring out star colours and still quite low noise.