592 replies to this topic

[A. Viegas]

Hi Alberic

I am not sure but yea it could be 57dB!

[Kaikul]

Al, how's the weather looking there? I hope it's a fortuitous one for your 294 test. I hope to catch it tonight. What time will you begin your broadcast?

[A. Viegas]

weather looking good.  but having issues with nightskiesnetwork!

[ccs_hello]

Never seen IMX294's full datasheet.

I think its HCG mode is 4x, which is 12dB.  You can see it in sensor measurement chart @ gain value=120.

 

Gut feeling: its max analog gain 27dB (22.624x) and

max. digital gain 18 dB (8x)

 

12 + 27 = 39 (you'll notice the chart stops at gain value = 390), and

39 + 18 = 57

 

Worthwhile to mention that IMX294, is a BSI image sensor with a pixel pitch of 4.63um.  This is rarely seen in (fairly) large pitch image sensors.
Also it's LCG gain is fairly low (due to large pixel area and BSI) and HCG is 4x.
The ZWO software design sets HCG to kick in at gain=120 (12dB), i.e., analog gain back to 1x.
Prior to that setting, say gain=119, the analog gain is almost 4x.

 

Clear Skies!

 

ccs_hello

[Dragon Man]

Never seen IMX294's full datasheet.

I think its HCG mode is 4x, which is 12dB.  You can see it in sensor measurement chart @ gain value=120.

 

Gut feeling: its max analog gain 27dB (22.624x) and

max. digital gain 18 dB (8x)

 

12 + 27 = 39 (you'll notice the chart stops at gain value = 390), and

39 + 18 = 57

 

Worthwhile to mention that IMX294, is a BSI image sensor with a pixel pitch of 4.63um.  This is rarely seen in (fairly) large pitch image sensors.
Also it's LCG gain is fairly low (due to large pixel area and BSI) and HCG is 4x.
The ZWO software design sets HCG to kick in at gain=120 (12dB), i.e., analog gain back to 1x.
Prior to that setting, say gain=119, the analog gain is almost 4x.

 

Clear Skies!

 

ccs_hello

ccs, 

Thank you for the work you put into your post, 

but I don't understand your technical explanation. For people like me who don't understand all the maths and terminology, are you saying it's a good sensor or maybe not so good.

 

Sorry, not picking on you. I appreciate your technical know-how. I'm just trying to decipher what your figures mean  

[XS_Man]

,ccs,

Thank you for the work you put into your post

but I don't understand your technical explanation. For people like me who don't understand all the maths and terminology, are you saying it's a good sensor or maybe not so good.

 

Sorry, not picking on you. I appreciate your technical know-how. I'm just trying to decipher what your figures mean  

 

 

Don't worry Ken,  CCS is only saying that this IMX294 is a great new color sensor !

 

Albéric

Edited by XS_Man, 17 September 2017 - 07:34 AM.

[XS_Man]

And now QHYCCD has just announced a camera with IMX294 :

 

http://qhyccd.com/bb...hp?topic=6069.0

 

Rich week-end, for sure !

 

Albéric

[mclewis1]

It will be interesting to see the whole sale price of the sensor and/or get some idea of where the two vendors are thinking about pricing the new camera. Right now I think it's a bit of chicken and egg (existing interest vs. aggressive pricing to generate interest) and seeing who will blink first and release a number.

 

I just hope it's not something like "well it's twice the size of a 224 sensor so ... " 

[A. Viegas]

So I hope anyone who tuned in last night to my broadcast got something out of it. I was plagued with some technical issues from no sound to really poor tracking from my CPC11...  

Nevertheless I put the new ASI294 Pro through some basic EAA paces using Sharpcap.   I think there are still some bugs in the driver as the pixel array was in 17:9 format (4144x2802 or something like that..) and binning really did not do anything, maybe shrink the image size by 10%... at least visually binning was not helping.   I ran the camera mostly at 75% gain I am using percent here because I am not sure how the sharpcap max gain of 570 maps onto the spec graph of 400 max.   In any event, the images are pretty clean.   I mostly used live stack, little tweak of histogram and thats it.    Workflow was convert from .FITS to .TIFF using Fitswork and then convert again to .JPG for posting to the gallery which does not allow .TIF.

I mostly used the C11.   I had the C0.63x reducer but my spacing was a bit short so effective focal length was about 2000mm or F7.3  which is pretty high and again my tracking sucked (I did not balance the scope )   So I kept the average exposure time short and I stacked on average 5-20 frames.  Later in the evening I switched to my AstroTech 65mm APO and its 420mm FL really opened up the FOV to a whopping 2.6 x 1.7 degrees from the C11's 32'x22'   - initially my focus was a bit off, but as I dialed in Focus  M31 was just cresting over some trees and as usual the clouds came in, so turned it off at about 1am...   so about 3 hrs all in.   Nice thing about www.nightskiesnetwork.com  and broadcasting is users get to see all the warts and problems.  No editing! This was first light under the sky so there were a number of technical issues...  Anyhow here are some of the images:

M15 with AT 65 - 8s exposures 10 stack

 

M31 AT65 -15s exposures 15 stack   (it was in the trees)

 

Pelican - AT 65 - 120s exposure  single frame  (note my focus was off for this)

 

NGC7000 (North American Nebula)  - 120s single exposure  (again focus was off for this)

 

 

 

NGC7331 group - C11 @F7.3    20s exposures  stack of 15

 

 

NGC6888 - C11 also  - 10s exposures  17 frames

 

M57  - C11  -  10s   28 frames

 

M27 - C11 - 8s  stack of 5 frames

 

 

 

Al

Edited by A. Viegas, 17 September 2017 - 09:23 AM.

[Astrojedi]

 

Never seen IMX294's full datasheet.

I think its HCG mode is 4x, which is 12dB.  You can see it in sensor measurement chart @ gain value=120.

 

Gut feeling: its max analog gain 27dB (22.624x) and

max. digital gain 18 dB (8x)

 

12 + 27 = 39 (you'll notice the chart stops at gain value = 390), and

39 + 18 = 57

 

Worthwhile to mention that IMX294, is a BSI image sensor with a pixel pitch of 4.63um.  This is rarely seen in (fairly) large pitch image sensors.
Also it's LCG gain is fairly low (due to large pixel area and BSI) and HCG is 4x.
The ZWO software design sets HCG to kick in at gain=120 (12dB), i.e., analog gain back to 1x.
Prior to that setting, say gain=119, the analog gain is almost 4x.

 

Clear Skies!

 

ccs_hello

ccs, 

Thank you for the work you put into your post, 

but I don't understand your technical explanation. For people like me who don't understand all the maths and terminology, are you saying it's a good sensor or maybe not so good.

 

Sorry, not picking on you. I appreciate your technical know-how. I'm just trying to decipher what your figures mean  

 

What css_hello is saying is that BSI based sensors usually don't have large pixels as the SNR is already pretty high. The pixel size is quite large relatively speaking. SNR should be good assuming QE is high (which seems like it is given the low light ratings)

[jimthompson]

Hi Al,

 

Despite your technical difficulties it looks like you had a successful first test.  For 75% gain your images look quite clean and no visible (amp) glow around the edges.  I am wondering if you are able to access the High Dynamic Range (HDR) feature of the IMX294 sensor in Sharpcap?  This is a very interesting feature to me for objects like M31, M42, or globular clusters.  I am especially wondering about it for lunar imaging, which is probably my #1 interest followed by deepsky. 

 

Best Regards,

 

Jim T.

[Alien Observatory]

Al, TY for taking the time to broadcast (warts and all), looking forward to your next bcast...  Pat Utah  

[Relativist]

Got to see a bit of the show after the change in scopes, looks very promising so far.

[neaptide]

Al, last night was the first broadcast I tuned into on NSN. It was great! Thanks for letting us watch the testing process for the new camera. I am used to using an asi224, so when you hooked up the AT65mm to the 294 the FOV seemed huge to me. Looking forward to your next broadcast. 

[Ptarmigan]

SONY Launches Highly Sensitive 4/3" CMOS Sensor for 4K Surveillance
https://www.framos.c...4k-surveillance

 

It is a 4/3 sensor, so the diagonal is 21.60 mm.

[mclewis1]

Interesting to see the technical marketing descriptions. We're reaping the benefit of a bunch of folks willing to pay money for the ability to see real details in the shadows of security camera images. It's also nice that the full frame is 4:3 format in addition to being a type 4/3" size sensor, that provides more area coverage using the same focal reducer (vs. a "true" 16:9 format sensor). We also have to remember that this sensor like all type 4/3" sensors will generally require a larger diameter focal reducer to prevent or reduce the amount of vignetting.

[XS_Man]

 

Very important to point out IMX294's part number is IMX294CJK.
J is a unusual suffix and can give an important clue.

http://www.sony-semi...mg/imx294_1.png

As you can see in the above pictures, the Bayer RGGB pattern actually is a
4x4 arrangement.

RRGG
RRGG
GGBB
GGBB

While two different readout groups can be read separately/independently with
different exposure time.

RRGG
RRGG
GGBB
GGBB

BTW, for full resolution without pixel value summing (software binning), the spatial
resolving power will not be that good (has to use 4x4 for de-mosaic algorithm as
oppose to the common 2x2 demosaic.)

Clear Skies!

ccs_hello

 

I add to this post one ccs_hello's answer which is very important.

The arrangement is strange. For normal use, it's a kind of mix between a bin2 monochrome

sensor which is covered with a bayer matrix.

 

So this sensor seems to me closer to a 9.3 micron pixels sensor than a 4.63 microns sensor.

But as far as I understand it's not equal to 9.3 micron pixels as each pixel is read independently.

 

So as CCS said, debayering algorithm should not be 2x2 but 4x4.

 

That's why I'm suprised to see that Sharpcap gives directly clear images ?

As far as I know 4x4 is not supported in SharpCap. Or does it has been recently added ?

 

Please let me know ?

 

Albéric

[A. Viegas]

Hi Alberic

I was wondering about that as well after CCS posted his comment.   In particular I was wondering which debayer pattern to use when I opened up my .fits files from the broadcast.   In fact when I open the file using Fitswork (http://www.fitswork....re/softw_en.php)  I can save the debayered file in color with no additional setup...   So while there may be something else here in terms of the summing I think the traditional RGGB pattern works just fine.   As people saw on my broadcast, using RAW16 I was able to show images in color on the screen in Sharpcap without any issues.    My one oddity was that I did not notice any particular improvement when I selected binning from the sharpcap drop down menu...  but I think Hiten has gotten it to work, so maybe it was just user error on my part about binning.

 

 

Al

[XS_Man]

Hi Al,

 

Many thanks for these first images. Not perfect but gives a first idea of what is possible at F 7.

Considering a classical Newton between F/4 and F/5, the image will be more rich. Cool !

 

 

I can save the debayered file in color with no additional setup...   So while there may be something else here in terms of the summing I think the traditional RGGB pattern works just fine.   As people saw on my broadcast, using RAW16 I was able to show images in color on the screen in Sharpcap without any issues.

 

 

I don't understand why it works ?

Do we loose informations working in this way ?

 

Albéric

Edited by XS_Man, 18 September 2017 - 12:57 PM.

[Relativist]

RRGGRRGG

RRGGRRGG

GGBBGGBB

GGBBGGBB

 

Compare to:

 

RGRGRGRG

GBGBGBGB

RGRGRGRG

GBGBGBGB

 

It seems there is no loss of information, it's just a different pattern. 

                       

R|RG|GR|RG|G
R|RG|GR|RG|G
G|GB|BG|GB|B
G|GB|BG|GB|B

 

Compared to:

 

|RG|RG|RG|RG|
|GB|GB|GB|GB|
|RG|RG|RG|RG|
|GB|GB|GB|GB|

[ccs_hello]

SONY Exmor image sensor has internal digital circuit and a small amount of memory in it.

What it can do in one of the readout mode is to

(as opposed to read out image array pixel matrix value as it is), it can reshuffle the image matrix readout sequence such that

the order is manipulated back to the classic

RGRGRGRG

GBGBGBGB

RGRGRGRG
GBGBGBGB

sequence such that the downstream image processing/demosaic algorithm will not see surprises.

 

Also, this special arrangement actually opens the door for charge-domain binning (same as the binning as used in CCD image sensors) since 

the like-color pixels are neighbors to each other.  If they do, the net pixel pitch will be huge.

 

P.S. I have no full datasheet so I am just doing an educated guess here

 

Clear Skies!

 

ccs_hello

[XS_Man]

Hi Relativist,

Hi CCS,

 

SONY Exmor image sensor has internal digital circuit and a small amount of memory in it.

What it can do in one of the readout mode is to

(as opposed to read out image array pixel matrix value as it is), it can reshuffle the image matrix readout sequence such that

the order is manipulated back to the classic

RGRGRGRG

GBGBGBGB

RGRGRGRG
GBGBGBGB

sequence such that the downstream image processing/demosaic algorithm will not see surprises.

 

 

OK, I agree, it could be a logical explanation.

 

Also, this special arrangement actually opens the door for charge-domain binning (same as the binning as used in CCD image sensors) since

the like-color pixels are neighbors to each other.  If they do, the net pixel pitch will be huge.

 

Yes, that would great, 9.3 microns associated pixels using a real binning like CCD  should give performances near A7s sensor ! That would be useful in both surveillance and astronomy. But Sony don't speak about

this point. And no communication about that in flyer is not a good "omen" I guess !

 

Albéric

[Astrojedi]

Instead of starting a new thread I will contribute to Al's thread with some captures and thoughts on the ASI294 Pro.

 

I have had a grand total of just 30 mins of time under the stars with this camera due to the cloudy weather so these are very preliminary thoughts.

 

Pros:

- Very sensitive camera. Quick characterization of the sensor shows read noise <2e. Subjectively QE seems high - most Sony BSI sensors are close to 80% so observation is consistent.

- No amp glow at all,even at 500 gain - I don't do 600s exposures like Al

- Driver is very stable and camera has a buffer so you are not losing frames even at high frame rates.

 

Cons:

- 4/3 size is hard on my EAA setup. Significant vignetting even with my Hyperstar - Next light will be using my newly acquired 102mm ED refractor

- Needs dark frame calibration - Even with cooling on I am getting a lot of hot pixels - Al's camera does not seem to have this problem. This is bugging me as I prefer to use cameras with no calibration.

 

Overall very promising but still too early for a conclusion. The big unknown is price. I will keep posting results in this thread.

 

Had to use heavy jpg compression to fit captures to CN file size - Originally saved full size PNGs available on my Flickr account: https://www.flickr.c...157686110792301

Edited by Astrojedi, 19 September 2017 - 03:08 PM.

[Ain Soph Aur]

Hiten--

 

Great report. Very pleased to see you re beta testing this camera also! Now if SX would just announce they have a version coming out I would be  Would love to see how my MN190 would handle this sensor.

Edited by Ain Soph Aur, 19 September 2017 - 01:49 PM.

[Astrojedi]

M31 - A single unstacked 15s exposure using the C8 Hyperstar and Astronomik CLS filter. x2 bin. SharpCap Pro v3 running on Compute Stick. No calibration or post processing. Click to see full size.

 

Edited by Astrojedi, 19 September 2017 - 02:05 PM.