QHY268M review

I have several cameras in my testing queue, and a couple of remaining posts regarding the QHY600.  For now, I'd like to give some preliminary results for the QHY268M.  As I have more information, and get plots posted on some website, I'll expand this post.

QHY has donated one of these cameras (thank you, Dr. Qiu!), and it was going to be installed on BSM_TX.  Because that camera had issues, I've actually tested my personal QHY268M that will be sent out to TX at least temporarily.

The QHY268M uses the Sony IMX571 sensor, a smaller APS-C sized version of the IMX455.  It has 6280x4210 pixels, though only 6244x4168 are active pixels; the remainder are overscan and black-level pixels that do not see light.  Each pixel is 3.76micron in size, and is digitized to 16 bits.  This gives a physical active size of 23.5x15.7mm, with a diagonal of 28.3mm.  Note that the diagonal is approximately the same size as the clear aperture of a 1.25" filter, so with any converging beam, you will have vignetting in the sensor corners with the smaller filters.  You can live with this (and not have to buy new filters!), or go with 36mm or larger filters to fill the entire sensor.  With USB3, you can read approximately 2.5 frames per second.  Like all CMOS cameras, the shutter is electronic, so you need to use a dark filter for darks, or cover the front of the camera or telescope in some manner.

When you connect to the ASCOM driver, QHY brings up a window for setting parameters.  I select the readout mode and gain as below.  I set the offset to 10, but this is not critical.  What you want is for the offset to yield a bias level >100 so that all of the noise is digitized.  Setting it too high just reduces the dynamic range that can be digitized.  So a value between 100 and 200 for the bias seems reasonable, and that yields that "10" offset value that I use.  The selection window also has a clickbox for including/removing "overscan".  I'll have a separate post on this feature.  Just beware that if you include overscan, there will be rows and columns that aren't detecting sky photons, and in fact can yield counts close to 0 no matter what gain setting is chosen.  Just leave the binning parameter at 1x1, as you will do the binning selection inside of your acquisition program (like MaximDL).  QHY offers four readout modes for the 268:

0 (photographic mode)  amplifier switch to lower noise at gain=26

1 (high gain mode).  amplifier switch to lower noise at gain = 56

2 (extended full well mode). Higher noise, but full well around 80Ke-

3 (extended full well mode 2CMSIT) Full well around 80Ke-, but readnoise 1.3x lower than mode 2

I've tested most of these modes, but will report only on mode 1 and mode 3 as they are the most likely modes that someone would use.

Regarding mode 3:  if you set gain=0, the read noise is about 5.57e-, with a system gain of 1.28e-/ADU.  Doing a linearity test, this mode is linear (<1%) to about 49K counts, which converts to a well depth of about 63K electrons.  I did not see an amplifier change with gain setting, which matches the plots that QHY provides.  The readnoise is substantially higher than for mode#1, my preferred mode, and the well depth is not that much greater, so I don't see any reason to use this mode.  Mode 2 is similar in this respect.  For mode 0, the readnoise is higher than mode1 until the amplifier switch near gain=26, but at that point the well depth is about 27Ke-, not really a great combination.  So again, I avoid mode 0 for photometry.

That leaves mode 1 (high gain mode).  For this mode, I measured 2.85e-/ADU readnoise at 0 gain setting.  This yields a system gain of 0.77e-/ADU, and a linear full well depth around 62K counts or 48Ke-.  I consider this the best combination for photometry, as fairly short exposures will have enough sky signal in broad-band filters that the sky noise will dominate over the readnoise, and you can stack images without a noise penalty.  Using gain setting 58, I get a readnoise of 1.26e- and a system gain of 0.32e-/ADU.  This gives a linear full well of 62K counts or about 20Ke-.  I consider this gain setting to be pretty much ideal for spectroscopy, where the background is always low and readnoise becomes a much bigger noise source.

For all of these tests, I used an offset of 10.

Binning is done in software, as for most CMOS sensors.  This means that the readnoise is basically sqrt(npix)*native_readnoise value.  For mode 1, gain 0, 2x2 binning yields a system gain of 0.77e-/ADU, the same as for 1x1 binning, but the readnoise increased to 6.27e-.  Surprisingly, this is not the 2x theoretical increase from the 1x1 value (which would be 5.7e-).  I'll investigate this further.  Note that the system gain does not change.  This means that the 2x2 pixels are summed.  If all native pixels had the same value, this means that the binned pixel would saturate (65535 counts) when the four raw pixels reached 16383 counts, so software binning via the driver does not increase the dynamic range.  It is actually better to keep the system in 1x1 mode and then do the binning in software like MaximDL after the readout, storing a floating-point image that then can keep the larger dynamic range.

Looking at dark current, with mode=1, gain0, offset=0 and temperature=-5C, the bias level is 166.5 counts and a 300-second dark is 0.8332 counts, for a real value of 0.00214e-/sec/pixel, which is comparable to QHY's measurements.  Again, dark current (on average) and readnoise will be much less than the sky background for wide-band filters like UBVRI.  More important will be other noise sources, such as hot pixels (a good reason to subtract darks) and Random Telegraph Signal (RTS, a subject of an upcoming post).

I do not have any on-sky data yet, but will include this when the QHY268M is mounted on BSM_TX.  As this sensor is in the same family as the full-frame IMX455 that is used in the QHY600, I think it will show the same excellent results.

As with other QHY cameras, the camera is shipped with a power supply, a USB3 cable and lots of adapters.  My only complaint is that the included power supply cable does not have the threaded shell to create a solid connection to the camera, but they do supply a pigtail cable that the power supply plugs into and has the threaded shell on the other end.  You could tape the joint to prevent it pulling open, but it is not quite as good as a single cable with a threaded shell.  One thing that I'd like to try is to see what happens to the system noise when you use a high-quality power supply, and I may do that with the original QHY268M when it gets repaired.

I think this camera is an excellent choice for the majority of amateur use.  It is at a great price-point, and has wonderful specs.

Arne