ASI 294 read noise curve explanation ?

xs_man

Hi Sean,

Hi Sam,

Hi all,

AS you know IMX294 and new announced IMX299 are Quad Bayer sensors.

Please, could you tell me if the readout noise curve for ASI 294 is given for one

basic pixel of 2.4 microns, or is it the read noise for each cluster of 4 pixels ?

It's not clearly indicated even if it's always an important question.

Many thanks.

Best Regards.

Albéric

chad

Hi Alberic,

It is 1 pixel.

These basic parameters are related only to the circuit design, and have nothing to do with their readout values.

Thanks

Chad

xs_man

Hi Chad,

Sorry I'm not sure it's clear for me.

Sony says :

"The IMX294CJK uses a Quad Bayer structure, and outputs data binned in 2 × 2

pixel units in normal mode"

https://www.sony-semicon.co.jp/products_en/new_pro/may_2017/imx294cjk_e.html

The output use binning 2x2 pixel clusters.

It's a CMOS sensor (not a CCD) so each 2.4 microns pixel should be read separatly.

On the curve, in HCG mode, medium readout noise is nearly 1.5 e-. For me it's

related to each 2.4 microns pixels.

So for each 4.63 micron cluster of 4 pixels, readout noise is 1.5+1.5+1.5+1.5 = 6 e-

Do you agree with me ?

Or is it wrong ? And in this case, why ?

Many thanks.

Albéric

chad

Hi Alberic,

I think you made a mistake.

In that dialog, I think they means the normal bin, not the normal mode. In their datasheet, usually, they will call it as full size mode, not the normal mode.

Beside, you also could made a small calculation: we know it's pixel size is 4.63µm, and it has 4168 pixels on the width. So (4.63*4168) = 19297.84µm = 19.29mm, If it use 2*2 in normal mode, it means we have to multiply by 2, means at least, it should be 19.29*2 = 38mm, about 4cm, you can measure it. then you will fine the sensor is so small.

Thanks

Chad

xs_man

Hi Chad,

For me there is only 3 ways to understand Sony's datas :

1- A classical CMOS used as a "wired" bin 2x2 with 6 e- readout noise :

-----------------------------------------------------------------------------------------------------

Each basic pixel is 2.4 microns pixel. The readout noise curve is related

to this basic pixel. Nearly 1.5 e- in HCG mode for 2.4 microns pixel.

So for each 4.63 microns cluster of 4 pixels, readout noise is 1.5+1.5+1.5+1.5 = 6 e-

if we consider that each pixel is read separatly.

2- A revolutionnary CMOS used as a "hard wired" bin 2x2 with 1.5 e- readout noise :

------------------------------------------------------------------------------------------------------------------------

The readout noise curve is related to a cluster of 4 pixels.

Nearly 1.5 e- in HCG mode for 4.63 microns cluster.

So for each basic 2.4 microns pixel, readout noise is 1.5 / 4 = 0.375 e- if we

consider that each pixel is read separatly.

Only 0.375 e-, the absolute jewel of the year 2018 !!!!

Sorry, I don't believe in miracles...

3- An hybrid "hard wired" CMOS as a CCD in bin 2x2 mode with 1.5 e- readout noise :

----------------------------------------------------------------------------------------------------------------------------

Each basic 2.4 microns pixel are read in a 4 pixels cluster at the

same time. So readout noise is applied one time to the cluster.

And the readout noise curve is related to a cluster of 4 pixels.

I need to know how it works before considering to buy a camera with this sensor...

Or not...

Many thanks Chad.

Best.

Albéric

chad

Hi Alberic,

I did not understand you clearly.

It should be your method 3. I find a description/in datasheet that about the read.

Thanks

Chad

xs_man

Hi Chad,

"When readout the image, 4 adjacent pixels with same color are added and regarded as one pixel"

Yes, it's exactly this point which is crucial !

And Sony gives only a "fuzzy" answer.

We don't know if the 4 adjacent pixels are firstly added (as done for a CCD in bin2x) and read as one pixel ???

Or if if the 4 adjacent pixels are firstly read (as done for CMOS sensor) and then added (software binning)

and used as one pixel ???

The readout noise can't be the same for these 2 possible cases...

Many thanks.

Albéric

chad

Hi Alberic，

It should be added first.

First, it is much easier to implement an adder on the circuit than ADC. So there are usually more adders in the chip.

Second, Usually, when we talk about reading data from a pixel, we're talking about the value we get from the ADC. It means that if they did adder after ADC, they will say reading data from 4 pixles and add them. Usually, they describe the Bin mode like this.

Third, maybe it is the cultural difference between the east and the west, we think this sentence has no other meaning, what it expresses is to add first and then adc.

Fourth, I asked sony about this problem, they give me the same answer, do add first.

Thanks

Chad

xs_man

Hi Chad,

Hi Sam,

Hi ZWO team,

Hi all,

Firstly, I wish you an happy (european) new year 2019 !!

A lot of success for all !

Chad,

Fourth, I asked sony about this problem, they give me the same answer, do add first.

So OK, if Sony confirm that the 4 pixel are firstly added and then read after (when processed by ADC), that's

a real good new !

Many thanks.

Best.

Albéric

Support@ZWO

Thank you! Happy new year!