These are characterization measurement results from my QHY268M showing System Gain, Read Noise, Full Well Capacity and Dynamic Range as function of the camera's Gain Parameter for each of its four operating modes.

Recently, I made similar measurements of my QHY600M.  You may refer to this link for more details on how I made the measurements and calculated the results.  The acquisition and processing details applied here are nearly identical to the previous ones and need not be repeated here.

Last year I received a QHY268M and performed similar measurements on Mode 1 only.  Several other users of this camera model posted similar independent results that corroborated my findings.  Unfortunately, that camera had to be returned due to connection issues.  The measurements posted here are from its replacement.  The newest results are nearly identical to the ones I made before.

Comparison with the QHY268M characteristic curves from QHY's website, however, show major differences that I am unable to reconcile.  QHY was contacted last year about these differences.  I have no updates on the matter at this time.  I can only speculate that QHY may have posted results from another sensor or perhaps from an earlier model of the camera that has since been modified.

What I find the most interesting is just how well the results from my QHY268M match those from its "big brother" QHY600M.  (My results for the latter closely match those of QHY's published characteristic curves as I note in the aforementioned posting.)  It is my intention to use both cameras simultaneously on separate imaging rigs and interchange them as I wish.  Given how well their operating characteristics compare, it ought to be quite easy to swap the two cameras between systems and utilize the same acquisition settings.

An example of this similarity in performance could be seen when taking flats with each camera.  For both the QHY268M and QHY600M tests, I used the same Artesky flat panel in front of my TS RC10.  I was able to vary and use the same panel intensity levels for both cameras for all four of their respective modes.

I configured each camera system with its respective filter wheel and OAG such that they have almost exactly the same back focus.  This allows me, for example, to connect either system to the same flat field or focal reducer on several of my scopes and come to the same required spacing without having to maintain two sets of adapters.  I am also able to power each system in the same manner with interchangeable cables.

I hope that these results may be of some use to those who are contemplating purchasing or may already own the QHY268M.  I have imaged with my earlier camera and found it to be an impressive imager.  Hopefully, the weather will clear soon and I can get my new camera out under the stars.  Thank you for your attention!


ADDENDUM 2022-01-17

I took some additional test data to produce Photon Transfer Curves (PTC's) for the following camera parameters:

Rev G - Mode #1, Gain Parameter 0
Rev H - Mode #1, Gain Parameter 56
Rev I - Mode #3, Gain Parameter 0

PTC's offer a wealth of information, and for this exercise I focused mainly on verifying my earlier quick calculations of Offset, System Gain and Read Noise for these three conditions.  I was also interested in examining the behavior of the camera when illuminated towards the top of its bit range in these modes.

I won't go into too much detail here on how I made these PTC's.  I follow the procedures detailed in James R. Janesick's Photon Transfer book as well as materials provided over the years by Richard Crisp.

I used the same scope and flat panel described above and took flat frames of varying exposure time such that I illuminated across most of the range of the camera.  I gathered the statistics from a small 500 x 500 pixel region in the middle of my frames.  Shot noise is obtained through "frame differencing", in which two frames of equal illumination and exposure are subtracted one from the other.  The resulting standard deviation of this difference is divided by the sqrt(2), and this is the combination of Read Noise and Shot Noise, which ought to be devoid of Fixed Pattern Noise (FPN).  For those interested in learning more, I refer you Janesick's book and/or Crisp's online materials.

By careful and precise adjustment of the Offset and Read Noise, the PTCs' Shot Noise curves can be derived with precise slopes of 0.5 on the log-log scale as shown.  (Note that "Signal" here refers to the average value of the pixels in the 500 x 500 region with properly adjusted Offset subtracted.)  The Read Noise values I came up with closely match what I show in Rev B.

One can get the System Gain from the Shot Noise curves in units of [e-/ADU] either by examining their intercepts on the y = 1 [ADU] gridline or using the coefficient of the power fitting curve.  Solve for x with y = 1, and one can get the System Gain value.  These values corroborate what I got before in the Original plot image.

Effective Full Well Capacity and Dynamic Range

The Fixed Pattern Noise (FPN) curves, however, did not come out with slopes of 1.0 as expected.  There is a nuance with this camera that escapes me for the present, although my power law fit is close to 1.  Given this, I did not use these curves to make any derivation or interpretation of, for example, Photon Response Non-Uniformity (PRNU).

Combining the Total Noise results from all three cases  in Rev J, one can see that there is something different happening towards the high end of the bit range in Mode #3.  When my sensor integration trends towards filling the full well in Mode #3, the average Signal level actually hits an earlier saturation value and the noise increases dramatically.  One could falsely think that, say, a star has not yet saturated at the highest allowed 16-bit value and be tempted to push an exposure further, but these curves demonstrate that there is actually an effective cutoff in Mode #3 that is lower than the normal cutoff seen in Mode #1.

In Rev K, I show this result again, this time in units of linear steps, in which I simply took the Signal levels in [ADU] and divided by the respective Read Noise values in [ADU].  This shows how many effective levels are available for the different modes.  I show these same results in terms of stops or bits on the log scale that demonstrates that, at least in terms of bit-sizes, these differences are not as dramatic as they appear on the linear scale.

I've also added Rev L in which I show the same results, but this time I have multiplied both the Offset-Subtracted Signal and Total Noise by the respective System Gain values.  This puts the curves on a more equal footing, and in this way we can see why Mode #3 is called extended range in that it does indeed have a larger Full Well Capacity as measured in [e-].  However, the previous plots demonstrate that this is done at the cost of slightly lower Dynamic Range and reduced range of effective signal.

Conclusions and Future Planning

What does this imply?  My big takeaway from all this is that one needs to be careful in interpreting the effectiveness and range of Mode #3.  I can't say what the camera is doing with bright signal towards the top of the range, but as I see it, this demands a departure from my usual way of planning and using exposures such as with Mode #1.  Also, given the hit taken on the upper range, Mode #3 is no longer the winner when it comes to effective Dynamic Range, and so I will probably not be incorporating it into my imaging.

By the way, it is my assumption that these same results also correspond to the QHY600M-PH as these two cameras operate and have already been measured so similarly.  I may form PTC's from my QHY600M in the future, but not anytime soon.  These results take a lot of time to acquire and present!

For those interested, there is an ongoing discussion on Cloudy Nights concerning these points.  (So far it has remained a profitable discussion and not devolved into outright warfare or crimes against humanity.)

I look forward to any comments and insight others may have about these results, particularly if my results or conclusions need to be called into question.  I'd appreciate the correction if needed.  Thanks for sticking with me on all this!