Planetary camera comparison, QHY5 QHY5L2 DMK21AU618

[Shiraz]

Hi. thanks for looking - this is a follow-up to Jarrod's earlier thread. Grateful for any feedback on the following assessment. Regards Ray

Field of view and resolution: the effects of pixel scale and field of view of the three cameras (DMK21AU618, QHY5L2 and QHY5) are as in the attached images of the same scene taken with the same scope (note: scaled 0.33x and the image quality varied due to lighting changes). For a given scope, the 618 has the smallest field of view, the QHY5L2 a bit bigger and the QHY5 has the largest. All have enough pixels for planetary imaging, but the varying pixel sizes require different focal lengths to give the same resolution. Expressed in Fnumber, the 618 needs f28, the QHY5L2 needs f19 and the QHY5 needs f26 to resolve the best possible detail from any scope.

Signal to noise: I set up a white LED source at a distance that produced broadband pixel illumination levels similar to those that I get from from Saturn through my 12 inch scope at ~f28 on the DMK21AU618. The source distance was then adjusted to provide equivalent pixel illumination on each of the three cameras and the single frame SNR (mean signal / SD) was measured in sub regions of each chip where there were no obvious hot pixels. With a gain setting that produced ~2/3 histogram, with 14.5ms exposure (corresponds to the max rate of the DMK of 60Hz) and for the same pixel illumination, the cameras yielded the following SNRs:

DMK21AU618 required a gain of 1000 (in Firecapture) for ~2/3 histogram and the SNR was ~16
QHY5L2 required a gain of only 300 for ~2/3 histogram, at which the SNR was ~22
QHY5 could not reach 2/3 histogram at 14.5ms – it required 25ms at full gain (in QGvideo with gain boost and no noise reduction) and the SNR was 12.

since the QHY5L5 had significant extra gain available, an additional test was performed where the gain was set to maximum and the exposure time was reduced to the point where 2/3 histogram resulted. Under these conditions, the QHY5L2 result was:

QHY5L2 at a gain of 1000 required only 4.5ms to reach 2/3 histogram at which the SNR was 12.

The QHY5L2 had significantly higher SNR than the DMK21AU618 under identical and realistic test conditions. The DMK was operating at flat-out gain at its maximum frame rate of 60Hz, but, for the same illumination, the QHY5L2 had plenty of SNR and gain in reserve and could operate at much higher rates on a Region Of Interest (ROI) with usable SNR. Thus, the higher SNR and gain of the QHY5L2 could be traded for higher framerates if required to reduce the effects of atmospheric turbulence. The older QHY5 was not in the same class as the other two, but could still run at about 30Hz on an ROI with usable SNR in gain boost mode – the SNR could also have been improved by flat and dark frame calibration to get rid of fixed pattern noise, since this was a quite obvious feature of this camera. The QHY5 can produce a band across the image at some exposures, which can limit the maximum usable framerate, but I did not delve into this.

So, from this limited test, the QHY5L2 is the best planetary camera in this group, the DMK21AU618 is still very capable and the original QHY5 may only be acceptable on brighter objects (or as a guider where the large field of view is a major advantage). You probably will not go too far wrong if you use this comparo to get some idea of the performance of similar cams - eg the ASI120 should be quite similar to the QHY5L2 and the Orion guider should equate to the QHY5. The software may offer different features, but the underlying technologies should be similar.

Have not touched on achievable frame rates, since that is highly variable depending on computer and software. Others may care to comment.

[JB80]

Thank you Ray for sharing this, very interesting.
:)

[Troy]

Hi Ray,
Another technical but enjoyable post.
I'm not really sure how you measured the SNR mean signal / sd, but if I use the formula of
signal / noise (max histogram value / FireCapture noise estimate) would this work?
For example a histogram of 66 and noise estimate of 5 would give a SNR of 13.2
Also, would it be better to have the same exposure and gain settings and adjust the light source distance so they have the same histogram value?


Troy

[Astro_Bot]

Thanks for this very useful comparison, Ray.

[Shiraz]

Thanks Guys.

Troy, Nebulosity shows the pixel stats for a 21x21 pixel region in an image - the mean signal/SD is the measure of SNR that I used.
Not sure what the Firecapture noise estimate is, but it should give a measure of the variance or SD in some form. The best signal estimate is probably the mean value rather than the maximum histogram value. The ratio of the mean to the noise estimate should give a signal to noise ratio that you could use for comparisons - might not be the same as I get, but would still be valid.

I tried to make the eval as simple as possible and the best test seemed to me to be to set the illumination at levels like those from a dim planet and then see what the cameras could do in both signal and signal/noise. Agree though that there are other equally valid ways to test.

[Troy]

Thanks Ray,
you have given me plenty of ideas :thumbsup:
Yeah after I wrote my last post I then thought that the mean would be the best way to go not sure how much the max histogram value would jump around indoors.
In my last post I wrote
>>Also, would it be better to have the same exposure and gain settings and adjust the light source distance so they have the same histogram value?
I think if it was done this way the only difference would be the noise value without having done this myself it seems like a good way for the SNR test.

For another signal test I thought about screwing different filters onto the camera's, such as red, green and blue while pointing it at a white light source. It would be interesting to compare the different channels particularly the blue because this wavelength struggles with signal on some of the planets and as a result a lower frame rate is used.

Thanks again Ray I like your real world testing.

[Shiraz]

It sure is an interesting task and I would be keen to know what you find on the colour evaluation. I went into it thinking that it would be relatively easy to just do a simple test and compare cameras, but it turned out to quite difficult to ensure that the comparison was fair and that the results were realistic enough actually mean something to someone trying to image Saturn.

A few things that I found and that you might consider:
1. the cameras I tested have quite different internal gains (adu/photon), so using the same nominal gain setting in Firecap will set the cameras up differently and you won't be comparing apples with apples - in particular, the histogram values for different cameras will not mean the same thing in your test method. To get around this, I used a constant and realistic illumination to ensure that the light input per pixel was constant, and then compared the outputs at various gain and exposure settings - that way at least the input was fixed in an absolute sense.
2. the cameras have different pixel sizes. In use the Fno of the optics would be varied to get the same pixel scale for each camera and you need to adjust the source distance so that the pixel illumination reflects the effect of the real-world variation in the Fno of the optics - for an unfocused source, the distance must be adjusted in proportion to the pixel linear dimensions in order to get an apples to apples comparison.
3. both gain and SNR are important, but they are largely independent. Insufficient gain and you end up with quantisation noise and onion rings - too high a gain and you don't change the SNR, but can get saturation. So what is needed is enough gain to get above the onion rings and below saturation at high framerates and enough SNR to operate at high framerates without excessive noise. Different cameras can have different optimum operating points, so (again) comparing apples with apples is not so easy.

Have fun! Regards Ray