Canon EOS Ra Sensor Measurements at PhotonsToPhotos

Thanks for running the numbers. Is the change in PDR/read noise strictly due to the way the red channel is operating or is there something else like binned sensor/CPUs?  The 60Da has the same firmware as the 60D, but the Ra has specific firmware.

The 60D and 60Da didn't have the same degree of difference in PDR or read noise.

ASTRO PERSPECTIVE
It's interesting because a lot of dedicated astrophotographers have found the Ra to be really impressive in use for a simple one-shot camera.

https://www.youtube.com/watch?v=x8bYAc0cGM4
Richard Wright from Software Bisque was gushing over it in this video. Software Bisque makes one of the flagship "tripod heads" (mounts) for astrophotography

https://www.bisque.com/product-category/mounts/

These are the types of mounts used by universities and other researchers to do things like exoplanet discovery like KELT and MEarth.
https://keltsurvey.org/
https://www.cfa.harvard.edu/MEarth/Telescopes.html

DAYLIGHT PERSPECTIVE
Both the 60Da and EOS Ra are less aggressive with their Ha filters than aftermarket modification. Daylight performance has increased IR leakage (like the old Leica M8) but it's somewhat fixable if your standard for quality is lowered.

https://www.youtube.com/watch?v=uZA-vuy0uG8

https://skyandtelescope.org/astronomy-equipment/first-look-canons-eos-ra-mirrorless-camera/

https://astrobackyard.com/canon-60da/

sharkmelley wrote:

Very interesting.

I particularly note that the Ra appears to have lower read noise than the R:

https://photonstophotos.net/Charts/RN_e.htm#Canon%20EOS%20R_14,Canon%20EOS%20Ra_14

For both the R and the Ra, do you know of the reason for the ripples in the PTC at the low end? For some cameras it's caused by digital scaling of the raw data but this isn't something normally associated with Canon.

I'm showing averages across channels, which isn't normally an issue, except the Ra channels are very unbalanced. Very low red read noise is dragging the average down. If it were green I think it would be unchanged. And strangely blue EOS Ra is noisier than EOS R.

The EOS R also has ripples indicative of scaling.

There are some other oddities with the Ra histograms.

I'm showing averages across channels, which isn't normally an issue, except the Ra channels are very unbalanced. Very low red read noise is dragging the average down. If it were green I think it would be unchanged. And strangely blue EOS Ra is noisier than EOS R.

I was using an OLED TV for the test patterns.  (for both the 60Da and Ra). Does that affect the analysis?

Is there something with the timing for read noise that Canon could somehow optimize for red Ha nebula at the expense of blue Reflection nebula?

GXAlan wrote:

I'm showing averages across channels, which isn't normally an issue, except the Ra channels are very unbalanced. Very low red read noise is dragging the average down. If it were green I think it would be unchanged. And strangely blue EOS Ra is noisier than EOS R.

I was using an OLED TV for the test patterns. (for both the 60Da and Ra). Does that affect the analysis?

For my tests the spectral characteristics of the light source don't affect the results.

Is there something with the timing for read noise that Canon could somehow optimize for red Ha nebula at the expense of blue Reflection nebula?

Timing would be the same for all channels. They are scaled differently.

sharkmelley wrote:

bclaff wrote:

Timing would be the same for all channels. They are scaled differently.

So the gain is different in each channel?

And what kind of histogram oddities are you seeing? Gaps?

Probably digital scaling.

Columns are DN, R, Gr, Gb, and B

Note that only the red channel appears to be unscaled.

bclaff wrote:

Note that only the red channel appears to be unscaled.

There's definitely scaling on the red at some ISOs. Take a look at

in the 1/250 FPN tests
114A8184.CR3 ISO 125
114A8187.CR3 ISO 250

But yes, the blue and greens are more consistently scaled than others.

This must be the digital white balance pre-scaling (which explains why the white balance on EOS Ra is always easy to fix in post).  I recall the D810 did this and the D810A was considered a very good astro camera.

GXAlan wrote:

bclaff wrote:

Note that only the red channel appears to be unscaled.

There's definitely scaling on the red at some ISOs. Take a look at

in the 1/250 FPN tests
114A8184.CR3 ISO 125
114A8187.CR3 ISO 250

But yes, the blue and greens are more consistently scaled than others.

You seeing how Canon routines does intermediate ISO settings by scaling.
An exception being the 5DS/5DSR models.

bclaff wrote:

You seeing how Canon routines does intermediate ISO settings by scaling.
An exception being the 5DS/5DSR models.

Ah!  My other Canon is a 5Ds.

Very interesting, thanks for running the numbers, and thanks also to GXAlan for running the tests

I am intrigued by the lower read noise of the EOS Ra w.r.t. to EOS R / 5DIV at base ISO.

Assuming only the optical low-pass filter was changed, this should not change the read noise (in e-) in dark frames in my understanding. So here my questions:

• Have you spotted some kind of 'NR' in the dark frames of the EOS Ra, like for the R5?
• Or is it because the camera can use a lower conversion gain (e- to DN) due to increased sensitivity in the reds? (I would think this is the more likely explanation)

Cheers,

Ethan

EthanCam wrote:

Very interesting, thanks for running the numbers, and thanks also to GXAlan for running the tests

I am intrigued by the lower read noise of the EOS Ra w.r.t. to EOS R / 5DIV at base ISO.

Assuming only the optical low-pass filter was changed, this should not change the read noise (in e-) in dark frames in my understanding. So here my questions:

• Have you spotted some kind of 'NR' in the dark frames of the EOS Ra, like for the R5?
• Or is it because the camera can use a lower conversion gain (e- to DN) due to increased sensitivity in the reds? (I would think this is the more likely explanation)

No NR spotted but since more photons are captured you would need to lower the conversion gain to avoid clipping the ADC and that would result in slightly lower noise.
So that's a good theory.

Okay, thanks!

What’s also interesting is that long exposures are different between the R and 5D4

https://www.brendandaveyphotography.com/canon-consistent-with-eos-r-rp/

I have sent Brendan some files from the Ra.  At 5 min, I was getting 54-ish with SD of 18-ish Which is also better than the regular R.

The Sony sensors are very good for terrestrial but they still have filtering for astro.  The GFX100 is spectacular but it’s not cost effective to convert it for Ha sensitivity.  A variant of the A7R4 sensor is available in a cooled Astro camera, but you lose the flexibility of the DSLR.

GXAlan wrote:

What’s also interesting is that long exposures are different between the R and 5D4

https://www.brendandaveyphotography.com/canon-consistent-with-eos-r-rp/

I have sent Brendan some files from the Ra. At 5 min, I was getting 54-ish with SD of 18-ish Which is also better than the regular R.

The Sony sensors are very good for terrestrial but they still have filtering for astro. The GFX100 is spectacular but it’s not cost effective to convert it for Ha sensitivity. A variant of the A7R4 sensor is available in a cooled Astro camera, but you lose the flexibility of the DSLR.

That's interesting, here my theory:

5D4: older DIGIC6+ -> higher thermal noise (affecting long exposures) than EOS R
R: DIGIC8
Ra: DIGIC8,  but probably lower conversion gain -> similar thermals as EOS R but better noise statistics because of lower conversion gain

Yes, Sony are known for doing some sort of 'star eater thing', Mark Shelley's website is informative on that. But I am not an astro-photography expert