Back to forums

ZWO vs QHY slightly confusing information

Takahashi FSQ-106N ZWO ASI6200MM Pro Equipment CMOS CCD
19 replies•1.2k views
Bill McLaughlin avatar
Bill McLaughlin avatar
I preface this by saying that I have 30 years or so experience in imaging, mostly with CCD (started in 1993 with an ST4 guider/"imager"). I am shopping for a full frame CMOS camera to replace one of my CCD cameras. I have a ZWO 2600 already and have ordered the Moravian equivalent for a remote site (it has a shutter). The next thing will be a full frame Sony IMX455 based CMOS camera (not medium format, those are too unwieldy to process and display) to go on my FSQ106N.

Looking at the ZWO it is clear they only offer the consumer grade sensor, whereas QHY offers both that and the industrial version. Just to be clear, I am talking about the Sony IMX455 chip in both cases.  The confusion comes when the subject is bit depth/full well:

From the ZWO 6200 page:

* True 16-bit ADC – giving a high dynamic range of up to 14 stops
* in one place it says "Full well: 50000e" and in another place on the same page it says "Full Well 51 ke"

From the QHY 600 page:

* Native 16 bit A/D: The new Sony sensor has native 16-bit A/D on-chip. The output is real 16-bits with 65536 levels. 

I assume that this is due to what they call (on the same QHY page) "Extended full well Capacity": 

"With a pixel size of 3.76um, these sensors already have an impressive full well capacity of 51ke.  Nevertheless, QHYCCD has implemented a unique approach to achieve a full well capacity higher than 51ke- through innovative user controllable read mode settings.  In extended full well readout mode, the QHY600 can achieve an extremely large full-well charge value of nearly 80ke- and the QHY268C can achieve nearly 75ke-.  Greater full-well capacity provides greater dynamic range and large variations in magnitude of brightness are less likely to saturate.  The QHY600 / 268C have three readout modes with different characteristics."

So I am guessing this is something QHY has done in the camera circuitry?  Note the graph below from their page:

Clearly this "well advantage" dwindles to zero by the time the gain gets to 100 or so but does seem to offer some advantage at lower gains and this could be even more important with a large chip on an FSQ where the large FOV makes many bright stars  likely.

Any thoughts or information would be appreciated. Thanks! 

Original page with graph can be found at: QHY 600 Product Page

Tim Hawkes avatar
Tim Hawkes avatar
Bill McLaughlin:
With a pixel size of 3.76um, these sensors already have an impressive full well capacity of 51ke.  Nevertheless, QHYCCD has implemented a unique approach to achieve a full well capacity higher than 51ke- through innovative user controllable read mode settings.  In extended full well readout mode, the QHY600 can achieve an extremely large full-well charge value of nearly 80ke- and the QHY268C can achieve nearly 75ke-.  Greater full-well capacity provides greater dynamic range and large variations in magnitude of brightness are less likely to saturate


Just wondering under what specific circumstances this faciility would matter most in practice?   I suppose that given the relatively low read out noise of CMOS --- far lower than sky shot noise for us Bortle 6 types  - most times just halving the exposure time and stacking  more subs would do as well?  

I am guessing that this might come into its own under really dark skies and for fields where there is a bright star adjacent the object of interest  ..e.g the Sadre nebula?

Tim 

sorry a very tangential point to your question but you did welcome any comments :-)
Bill McLaughlin avatar
Bill McLaughlin avatar
Tim Hawkes:
I am guessing that this might come into its own under really dark skies...


Exactly what we have.
andrea tasselli avatar
andrea tasselli avatar
As measured the FWC+ of the ASI6200 is 65,384 ADU @ gain 100. Or 17,000 e-.
Tim Hawkes avatar
Tim Hawkes avatar
andrea tasselli:
As measured the FWC+ of the ASI6200 is 65,384 ADU @ gain 100. Or 17,000 e-.

Yes it is interesting when you look closely at the figures.  I was contemplating upgrading the ASI 294 but - as far as I can see -  the main driver for getting the 2600 or 6200 has to be sensor size and perhaps the slightly smaller pixels which you might also bin depending on set up.  But everything else looks very similar to me   ~90%  peak QE, ~  17000 e- at unity gain and fairly similar read noise (with HCG).

I think that one of the things that is slightly confusing in comparing the data is that for the 2600 and 6200 unity gain is zero while for the 294  unity gain is 120.  

Also - something that I don't really understand -  what is the practical advantage of having a FWC of 64 k if only at a gain of 0.25 e/ ADU --  there must be limited circumstances under which gain less than unity is useful?  ... avoiding overexposure might equally as well be achieved through combining shorter exposures at a gain above where HCG cuts in?

Tim


POSTSCRIPT .   Apologies my comment above that the  6200  is similar to the 294 in having 17k e- at unity gain is just wrong - actually it is 51k e- so that is indeed a pretty fair advantage and a reason to buy although that is at a cost of about twice the read noise v the 294 at unity gain with HCG.
Helpful
Torben van Hees avatar
Torben van Hees avatar
The higher full well of lower gain is useful if you don‘t want to increase the number of exposures you take. With fast optics/large apertures this can be a problem with bright stars or it can become a problem under bright skies. If you have both, a large full well capacity or reducing gain is extremely helpful. With the extended full well mode I do 60-120s L exposures with my Epsilon in Bortle 6. Without it, I need 2-3x shorter exposures. My workstation is running hot processing 1000 frames from the IMX571 - I shudder to think about the processing time needed for 3000 frames from the IMX455.

One thing to consider: The extended FWC-mode from QHY introduces significant nonlinearity at the middle and upper ADU ranges. So much so, that you can‘t really saturate the pixels. That means you‘re paying with a decrease in dynamic range and you can‘t really use the data for photometry any more. For me, that‘s a small price to pay: The DR is easily recovered during stacking and I only need photometric data for star color calibration with PixInsight which is accurate enough for the dim stars.
Well Written Helpful Insightful Engaging Supportive
Tim Hawkes avatar
Tim Hawkes avatar
Torben van Hees:
The higher full well of lower gain is useful if you don‘t want to increase the number of exposures you take. With fast optics/large apertures this can be a problem with bright stars or it can become a problem under bright skies. If you have both, a large full well capacity or reducing gain is extremely helpful. With the extended full well mode I do 60-120s L exposures with my Epsilon in Bortle 6. Without it, I need 2-3x shorter exposures. My workstation is running hot processing 1000 frames from the IMX571 - I shudder to think about the processing time needed for 3000 frames from the IMX455.

One thing to consider: The extended FWC-mode from QHY introduces significant nonlinearity at the middle and upper ADU ranges. So much so, that you can‘t really saturate the pixels. That means you‘re paying with a decrease in dynamic range and you can‘t really use the data for photometry any more. For me, that‘s a small price to pay: The DR is easily recovered during stacking and I only need photometric data for star color calibration with PixInsight which is accurate enough for the dim stars.

Thanks.  That is a good point about reducing the number of frames to process - I have not so far  worked with frames that big but it is a tip that might come in handy in future.   

On your second point I suppose that you would also have to be careful then about not making flats at low gain and letting the saturation get above 50% or so?
Torben van Hees avatar
Torben van Hees avatar
Tim Hawkes:
andrea tasselli:
As measured the FWC+ of the ASI6200 is 65,384 ADU @ gain 100. Or 17,000 e-.

Yes it is interesting when you look closely at the figures.  I was contemplating upgrading the ASI 294 but - as far as I can see -  the main driver for getting the 2600 or 6200 has to be sensor size and perhaps the slightly smaller pixels which you might also bin depending on set up.  But everything else looks very similar to me   ~90%  peak QE, ~  17000 e- at unity gain and fairly similar read noise (with HCG).

I think that one of the things that is slightly confusing in comparing the data is that for the 2600 and 6200 unity gain is zero while for the 294  unity gain is 120.  

Also - something that I don't really understand -  what is the practical advantage of having a FWC of 64 k if only at a gain of 0.25 e/ ADU --  there must be limited circumstances under which gain less than unity is useful?  ... avoiding overexposure might equally as well be achieved through combining shorter exposures at a gain above where HCG cuts in?

Tim

The newer chips do not have any amp glow. This increases the resilience against small calibration errors in the data. It also allows longer exposures: In very long exposures, the noise from the amp glow becomes visible. 

Also, the IMX294C uses a quad-bayer array. That means its pixels are actually smaller than those of the IMX 455/571. An individual subpixel has a FWC of 4.2ke- at unity gain (which only matters if you are still undersampled with these small pixels - I don‘t know any telescope for which that would be true). 

The peak QE of the chips is approximately the same, but the IMX455 has slightly better Ha-response (see here: https://www.cloudynights.com/topic/798063-quantum-efficiency-of-the-sony-imx455-in-qhy600-and-imx492-in-qhy294/). Note that some CCD cameras still best the CMOS in Ha-sensitivity but that is bought with a lot less average QE in the visible spectrum and far larger read noise.

From my practical experience, the IMX571/455 are leagues easier to use than the IMX294/492: Calibration does not need to be quite as perfectly matched (the 492 is best calibrated with darks matching ambient and sensor temperature, for example). Also, the noise pattern of the IMX294/492 shows a very noticeable banding. This can be averaged out but you need plenty of long enough exposures and heavy dithering. And, of course, 4x the sensor area means 4x more light grasp.
Helpful Engaging
andrea tasselli avatar
andrea tasselli avatar
Torben van Hees:
The newer chips do not have any amp glow. This increases the resilience against small calibration errors in the data. It also allows longer exposures: In very long exposures, the noise from the amp glow becomes visible. 

Also, the IMX294C uses a quad-bayer array. That means its pixels are actually smaller than those of the IMX 455/571. An individual subpixel has a FWC of 4.2ke- at unity gain (which only matters if you are still undersampled with these small pixels - I don‘t know any telescope for which that would be true). 

The peak QE of the chips is approximately the same, but the IMX455 has slightly better Ha-response (see here: https://www.cloudynights.com/topic/798063-quantum-efficiency-of-the-sony-imx455-in-qhy600-and-imx492-in-qhy294/). Note that some CCD cameras still best the CMOS in Ha-sensitivity but that is bought with a lot less average QE in the visible spectrum and far larger read noise.

From my practical experience, the IMX571/455 are leagues easier to use than the IMX294/492: Calibration does not need to be quite as perfectly matched (the 492 is best calibrated with darks matching ambient and sensor temperature, for example). Also, the noise pattern of the IMX294/492 shows a very noticeable banding. This can be averaged out but you need plenty of long enough exposures and heavy dithering. And, of course, 4x the sensor area means 4x more light grasp.


I'm not sure the ASI294MC has sub-matrix pixels at all, unlike its monochrome sibling. I disagree that the colour camera needs matched darks and I have yet to see banding in the final stacked up images. My integrations are typically of 180s but even with 60s I can't find anything worth noting. And I never ever ever dither. The last assertion isn't true in the only meaningful sense, that of specific illumination. Whether that covers an area 4x as wide makes no difference at the only relevant parameter, that is pixel illumination.
Tim Hawkes avatar
Tim Hawkes avatar
andrea tasselli:
From my practical experience, the IMX571/455 are leagues easier to use than the IMX294/492


Astroimaging is a broad church and I guess that the advantages/ disadvantages just depend on exactly what your interests and activities are..

Being at Bortle6  I  am on fairly short exposure (90s or less)  except for narrow band ( 180s) pretty much according to Robin Glover's Sharpcap analysis  - i.e. that basically says where skyglow shot noise so far dominates  read noise barely registers.    

For me,  the ability to unlock the individual pixels of the 294MM 492 chip has actually  been very useful because it allowed sampling and oversampling better attuned to lucky imaging and deconvolution without need to change the physical set up  of anything.

Although unlike Andrea, I do always dither I do agree with his observations that if there are problems with banding and dark matching then they have never been apparent to me  in final images.   I do have and use libraries of  darks that are well matched in terms of sensor temperature, offset, gain  etc.- so that part of the work is already done.  But  had not before heard of ambient temperature being a concern ?

At the moment I am doing generally small-field long focus stuff seeking to improve detail in small angular size objects under rather bright skies.  I can however well see  that the  6200 would have real advantages  for wider vista shorter focus imaging and especially under darker skies.   So I if ever achieve my dream of  getting long term access to a dark sky site that I could visit for weeks at a time at short notice - then I would like to move in that direction and would then probably trade up.

Tim

PS .  Apologies Andrea and Torben.  I realise that I sent this as a reply to you Andrea when actually it was to Torben.
Helpful Insightful Respectful
Torben van Hees avatar
Torben van Hees avatar
andrea tasselli:
Torben van Hees:
The newer chips do not have any amp glow. This increases the resilience against small calibration errors in the data. It also allows longer exposures: In very long exposures, the noise from the amp glow becomes visible. 

Also, the IMX294C uses a quad-bayer array. That means its pixels are actually smaller than those of the IMX 455/571. An individual subpixel has a FWC of 4.2ke- at unity gain (which only matters if you are still undersampled with these small pixels - I don‘t know any telescope for which that would be true). 

The peak QE of the chips is approximately the same, but the IMX455 has slightly better Ha-response (see here: https://www.cloudynights.com/topic/798063-quantum-efficiency-of-the-sony-imx455-in-qhy600-and-imx492-in-qhy294/). Note that some CCD cameras still best the CMOS in Ha-sensitivity but that is bought with a lot less average QE in the visible spectrum and far larger read noise.

From my practical experience, the IMX571/455 are leagues easier to use than the IMX294/492: Calibration does not need to be quite as perfectly matched (the 492 is best calibrated with darks matching ambient and sensor temperature, for example). Also, the noise pattern of the IMX294/492 shows a very noticeable banding. This can be averaged out but you need plenty of long enough exposures and heavy dithering. And, of course, 4x the sensor area means 4x more light grasp.


I'm not sure the ASI294MC has sub-matrix pixels at all, unlike its monochrome sibling. I disagree that the colour camera needs matched darks and I have yet to see banding in the final stacked up images. My integrations are typically of 180s but even with 60s I can't find anything worth noting. And I never ever ever dither. The last assertion isn't true in the only meaningful sense, that of specific illumination. Whether that covers an area 4x as wide makes no difference at the only relevant parameter, that is pixel illumination.

Actually the IMX294 chip does have a quad-Bayer architecture which outputs bin 2x2 (https://www.sony-semicon.com/files/62/pdf/p-12_IMX294CJK_Flyer.pdf). Note that the  spreadsheet refers to a 4.6 um „unit“.

I‘m not saying the IMX294 isn‘t a good sensor. Just that it needs more care to employ than the IMX571/455 and the venerable Panasonic that it succeeded. With very narrow filters/under darks skies the banding in the noise does become visible in the integration. Exposing long enough to swamp it is sometimes impossible. At Bortle 6 (where I am), this only becomes a problem with 3nm filters or with very fast, fairly large optics (the RASA 11 blows out stars very quickly - due to the large aperture this is actually quicker than the signal increase of extended objects from the fast f-ratio).

If you look at any bias frame, I‘m sure you will see the banding. Now create two master biases and compare them: The banding will be different between the two and between any two. That‘s the pattern in the noise that you need to swamp with shot noise and is difficult to overcome with dithering alone. I believe this is RTN - but am not sure.

Actually my last sentence is very true: If you keep the FOV (and aperture, of course) the same, and pick the corresponding telescope FL, a larger sensor will gather more light. That‘s why large sensors are desirable for any low-light application. I think planning a capture from the sensor to the target feels backwards. Due to the very low read noise of the modern CMOS, pixel illumination has become quite a lot less relevant. Unless, of course, you need to swamp some banding.
Helpful Insightful
Torben van Hees avatar
Torben van Hees avatar
@Tim Hawkes: Even for long-FL work the ASI6200 would be advantageous: You could increase aperture and FL and so keep the FOV the same and bin in post. That increases your light grasp. An expensive proposition, of course, leading not only to aperture- but also to sensor-size- and therefore filter-size fever.
andrea tasselli avatar
andrea tasselli avatar
Torben van Hees:
Actually my last sentence is very true: If you keep the FOV (and aperture, of course) the same, and pick the corresponding telescope FL, a larger sensor will gather more light. That‘s why large sensors are desirable for any low-light application. I think planning a capture from the sensor to the target feels backwards. Due to the very low read noise of the modern CMOS, pixel illumination has become quite a lot less relevant. Unless, of course, you need to swamp some banding.


In the context we're interested here it makes no sense at all, if the field requirements are matched. Unless you're in survey mode and then the more sky area you cover the better it is. If that is you application, then yes, the larger the sensor the better it is. The amount of light falling on each and every pixels is no longer relevant in a detector? I beg to differ...

At the moment I cannot test your assertions about banding but so far I haven't see any banding at all in any conditions and with any filter, once dark-bias has been removed.
Bill McLaughlin avatar
Bill McLaughlin avatar
As the OP and using the information (which has been interesting) to make a purchase decision, the most critical point is this:

It appears that use of these "modes" in the QHY 600 is optional. In other words, if you stick to the standard mode the system will behave much as the equivalent ZWO unit.

As such, the main  question one needs to ask as a buyer whether the additional modes are worth the additional cost given the usage intended.

Does that sound right?
Well Written Insightful Respectful Concise Engaging
John Hayes avatar
John Hayes avatar
The OP didn’t show the notes for that chart and didn’t show the curves for the QHY mode 3 operation.  In mode 3, unity gain occurs at gain 25 with a FWD of about 65k.   This is an ideal mode for remote operation since it maximizes response along with well depth.  The QHY 600M operated in mode 3, gain 25 has better QE, similar well depth at 1x1 binning, and about half the RN of a 16803 CCD sensor.  If the data is binned 2x2 (for ~7.5 pixels), the FWD is 4x that of a 16803.  That mode of operation is an excellent way to run the camera for long exposure imaging that avoids the bandwidth problems created by using high gain (with lower RN) values.  It’s not best for lucky imaging but lucky imaging is generally not workable in remote operations simply due to the amount of data it generates (unless you can run a windowed acquisition system.)

John
Helpful Insightful
Bill McLaughlin avatar
Bill McLaughlin avatar
John Hayes:
The OP didn’t show the notes for that chart and didn’t show the curves for the QHY mode 3 operation.  In mode 3, unity gain occurs at gain 25 with a FWD of about 65k.   This is an ideal mode for remote operation since it maximizes response along with well depth.  The QHY 600M operated in mode 3, gain 25 has better QE, similar well depth at 1x1 binning, and about half the RN of a 16803 CCD sensor.  If the data is binned 2x2 (for ~7.5 pixels), the FWD is 4x that of a 16803.  That mode of operation is an excellent way to run the camera for long exposure imaging that avoids the bandwidth problems created by using high gain (with lower RN) values


Thanks John, very useful information. I have also updated the OP with a link to the QHY page for the 600 that gives the details.  FYI, if you are still in Central Oregon, I live about an hour from you.  
Jay Grevell avatar
Jay Grevell avatar
I use the QHY600m myself and having the different read modes gives great flexibility in how you use your camera. I was initially using mode 0 photographic mode but after reading johns post and a bit of trialing i now use mode 3 for its larger full well depth allowing longer subs without saturating.

These cameras spit out 130mb subs so depending on your computer and storage setup can be a problem and longer subs can be  a benefit.

For narrowband mode 1 has a lower read noise which is more suited to the lower signal of narrowband.

It was a no brainer for me as i didnt have any existing ZWO gear to try and tie in with either.
Helpful
John Hayes avatar
John Hayes avatar
Bill McLaughlin:
John Hayes:
The OP didn’t show the notes for that chart and didn’t show the curves for the QHY mode 3 operation.  In mode 3, unity gain occurs at gain 25 with a FWD of about 65k.   This is an ideal mode for remote operation since it maximizes response along with well depth.  The QHY 600M operated in mode 3, gain 25 has better QE, similar well depth at 1x1 binning, and about half the RN of a 16803 CCD sensor.  If the data is binned 2x2 (for ~7.5 pixels), the FWD is 4x that of a 16803.  That mode of operation is an excellent way to run the camera for long exposure imaging that avoids the bandwidth problems created by using high gain (with lower RN) values


Thanks John, very useful information. I have also updated the OP with a link to the QHY page for the 600 that gives the details.  FYI, if you are still in Central Oregon, I live about an hour from you.  

Bill,
I think that there's a pretty good argument to be made the the QHY 600M is a superior camera to the ASI 6200.  In fact, QHY just came out with their "Lite" version of the QHY 600 to more directly compete with the ZWO camera.  It has a consumer grade chip, half the memory, it's slightly smaller, the price is more comparable the the ZWO camera.  In addition the QHY 600 offers more operating modes than the ZWO camera.

The one problem that I've had with QHY is service.  One of the two cameras that I have is dying in a strange way after only 18 months in the field.  For comparison, I've never had a FLI camera die.  They are bullet proof but if you had service questions, FLI answered their emails and phones within a day.  QHY, well...not so much.  It took over 6 weeks of emails to finally get their service guy to contact me.  When he did, I give him high marks for spending 3 hours remotely running my camera to diagnose the problem.  Along the way, he tried valiantly to convince me that there was nothing wrong, but I was able to show him that the camera was not operating correctly every time.  The symptom is that sometimes the camera comes up in a mode were it appears to only transmit the upper 8 bits of the full 16 bit data word.  The lower byte appears to be filled with zeros.  If you fool with it by cycling power and by turning the camera off for a while, after maybe 50 tries, it will randomly connect correctly.  (No...it's not a cable issue.  We swapped numerous cables with no change).  It's like having a car that only starts at random after 50-70 tries.  Like the car, if I leave it running, it will continue to work, but every once in a while the observatory loses power and then I'm back to spending hours (or days) trying to get the camera working again.  Cha finally conceded that "Yes...it is broken and needs to be returned for repair."  That's not so easy from Chile.  I've heard from other folks that the QHY drivers have memory leaks but I personally haven't experienced that problem.

I am told (yes...this is hearsay) that Moravian cameras have similar reliability to FLI equipment.  So, I have one on order.  The big problem there is that when I ordered it back in December I was given a 6 week delivery time.  Now Moravian tells me that it will be another 6 weeks.  DOH!  My calendar is full so that pushes my ability to get the scope repaired out into May, which is unacceptable!  This thing broke sometime early last November!   I've got to get my scope fixed, so today I had to order yet another QHY 600M to take down there when I go to install my refractor some time in mid February.   So, it looks like I'll be having a big camera sale come about June.  I've got 3 FLI cameras, 3 QHY cameras, a Starlight Press guide camera, a ZWO ASI 1200 Cool, and a Moravian on the way to support just two telescopes!  It's crazy.  Remote imaging is not for the financially faint of heart.

John



PS. Yes I'm in Bend (when I'm not in Tucson during the winter.)
Helpful Insightful Respectful Engaging
Bill McLaughlin avatar
Bill McLaughlin avatar
John Hayes:
I am told (yes...this is hearsay) that Moravian cameras have similar reliability to FLI equipment.  So, I have one on order.  The big problem there is that when I ordered it back in December I was given a 6 week delivery time.  Now Moravian tells me that it will be another 6 weeks.  DOH!  My calendar is full so that pushes my ability to get the scope repaired out into May, which is unacceptable!  This thing broke sometime early last November!   I've got to get my scope fixed, so today I had to order yet another QHY 600M to take down there when I go to install my refractor some time in mid February.   So, it looks like I'll be having a big camera sale come about June.  I've got 3 FLI cameras, 3 QHY cameras, a Starlight Press guide camera, a ZWO ASI 1200 Cool, and a Moravian on the way to support just two telescopes!  It's crazy.  Remote imaging is not for the financially faint of heart.


I feel your pain.   Just ordered the smaller Moravian C3-26000 PRO myself for a remote. Hopefully it will not be 12 weeks! The 455 based camera will be for local use, which helps some. I also will have several cameras for sale soon - a Moravian 16200 and a really vintage but still quite functional STT 8300 (my last SBIG camera after almost 30 years owning at least one). They are very yesterday and I often wonder how they are still around.
Bill McLaughlin avatar
Bill McLaughlin avatar
John Hayes:
I am told (yes...this is hearsay) that Moravian cameras have similar reliability to FLI equipment.  So, I have one on order.  The big problem there is that when I ordered it back in December I was given a 6 week delivery time.  Now Moravian tells me that it will be another 6 weeks.  DOH!


An update on this subject. I did order my Moravian (the smaller IMX571 version) about three weeks ago from Tolga and just heard today that it is ready to ship (3 weeks instead of the 6 weeks estimate). It must be some full frame specific part holding yours up rather than the Moravian production schedule.
Well Written Concise