Evaluation page concerning ZWO ASI294MM camera

[Christian Buil]

An evaluation page of rhe very new camera ASI294MM, in French AND in English

http://www.astrosurf.com/buil/asi294mm/

Christian Buil

[Olivier GARDE]

Merci Christian pour ce test très intéressant et fort complet.
Bon je vais conserver mes ATIK 460, 414 et mon ASI 183MM encore un peu...

[Ken Hudson]

Merci Christian! Your evaluation was very helpful to me. I am trying to decide which camera to buy for the UVEX and after reading your evaluation I think the ASI183MM Pro would be my best choice. Again, merci!

[Olivier GARDE]

Hi Ken,

For the UVEX, I think that an ASI 183 MM pro is the best choice with this spectrograph at this day and the CMED process (ISIS 6.0).

[Ken Hudson]

Thank you Olivier. I think you are right.

Cheers, Ken

[Terry Bohlsen]

Thank you Christian. I now own both the ASA183MM and a ASA294MM and it is interesting to see the comparison.
I use the ASA183MM on my LISA with a 23um slit and potentially on my UVEX with probably a 35um slit on a C11. For the LISA I use a focal reducer but not the other spectrographs. I bin it 2x2 anyway due to file size but the tests tell me that I still have room to use The CMED processing as well.
The ASA294MM is now on my LHIRES also with a 35um slit. The CMED processing certainly makes a very big difference in the noise of the final spectra. Clearly there are differences between the cameras but I think that both will work well on my system. It may not be the same with a slit less than 23um though.

[Christian Buil]

This ASI294MM is not a so bad product. One quality for me is the size of the sensor for the price.

Concerning the noise, the interpretation of the values is nether an easy task. For a correct analysis it is important to compare cameras with the same equivalent pixel size, an ASI294MM with these pixels of 4.63 microns on one side, and an ASI183MM camera operated in 2x2 binning on the other, i.e. with 2 x 2.4 = 4.8 microns, roughly the same value as for the ASI294MM. Also, consider that a fine performance criteria is the the Signal to Noise Ratio (SNR) in function of the situation.

For simplicity, it is assumed here that the pixel size is ultimately the same on ASI183MM (after 2x2 binning) and on ASI294MM. It is also assumed that for a given exposure time, a ASI294MM pixel collects 100 photo-electrons (arbitrary, only for the demo). In the elementary pixels of the ASI183MM camera this gives 100/4 = 25 photo-electrons. By putting noise the measured values, the result in term of signal to noise ratio (SNR) for some processing hypothesis is :



And here we see that pixels binning of the 183 in a traditional way give a much lower signal-to-noise ratio compared to the 294. Basically, the 183 is twice worse if it is acceptable to make it work in 2x2 binning (if the observation absolutely requires small pixels, say 2.5 microns, then there is no more battle, the ASI294MM does not offer this possibility and the ASI183MM wins hands down!).

This is also why I insist quite a bit on CMOS oversampling and the use of noise reduction algorithms. If we applied a median filtering before binning, the RSB rises to 50 for the 183, the difference is only 1.4, but still in favor of the 294.

It is necessary to apply all the possibilities of an algorithm like CMED (last row of the table) so that the ASI183MM finally equals the ASI294MM (and again, a consequent oversampling for the 183 is necessary).

Note that doing statistics (standard deviation, average) on data whose noise distribution is not Gaussian can generate interpretation difficulties, but the trend is there.

The morality of this thing:

(1) for an application which is satisfied with large pixels of 4.6 to 4.8 microns (for example UVEX on a relatively large telescope and a 35 microns slit, see also Therry Bohlsen comment concerning LHIRES + 35 microns configuration for an appropriate usage of the ASI294MM. ), the ASI294MM is probably the most efficient, but with a nuance for me which poses difficulty, this famous RTS noise which gives the impression that the image contains artefacts (1/f noise), which also gives this mixed result (we hesitate to filter the RTS noise here because there is a great risk of breaking the images sharpness).

(2) for use with small pixels, only the ASI183MM can be used while waiting for a status of the possible "super resolved" mode of the QHY294M.

Christian

[Ken Hudson]

Once again, thank you Christian. The included table and the additional information is very useful and very helpful to all of us!! So, now I have a decision to make!

Cheers, Ken

[Christian Buil]

I have added more info about the computation of RSB and also I have introduced a simple and nearly global criteria for evaluate all the imaging detector, CCD or CMOS, the detectivity (D*), given by the formula:

D* = (f x QE x p^2) / RON

with, f, the binning factor, QE, the quantum efficiency at a given wavelength, p, the physical width of the pixel, RON, the read noise in the raw image. See my webpage for some results.

Christian

[Michel Verlinden]

Christian,
Merci infiniment pour ce test très instructif et extrêmement documenté.
Michel Verlinden