Why don't we get the GOOD data anymore?

[rnclark]

And by today's standards, those specs are not very good.

15 electron read noise, but today's CMOS are under 2.

The 7D Mark II is on the order of 0.01 electron/second!

I have to say that I doubt this number if it is at RT, and I think it is at least 100x higher.

Do you have some sort of reference for this data point?

(I saw your data point of 0.016e-/s at 10C so I mean corroborating measurements)

http://www.clarkvision.com/reviews/evaluation-canon-7dii/

This is 15fA/cm^2 and is unlikely, even at 10C.

I am not sure what evidence you would like. As you say, the read noise derived, in the 1.6 electron range, is reasonable. Long exposure are then a simple derivation.

This image was made with the internal camera temperature at 16 C, and is 18 minutes total exposure (18 one-minute exposures) so 1080 seconds, and the faintest nebulae in the image are less than 1 photon per minute per pixel--on the order of 10 photons. Apparent read noise at ISO 3200 is 1.9 electrons, so noise from read noise alone is 8 electrons. With dark current on the order of 0.025 e/sec, dark current total would be 27 electrons, so noise from dark current would be about 5 electrons, with combined noise under 10 electrons, consistent with the noise level for 10 photons (yes, that less than S/N of 1). If dark current were much higher, images like this would not be possible. In fact before I did it, astrophotographers were saying it was impossible. So the observational evidence is showing that the 7D2 has impressively low dark current.


Spiral Perseid meteor spears the Heart Nebula

More info on this image is here:

ClarkVision Photograph - A Spiral Meteor Spears the Heart

It was also a NASA APOD: http://apod.nasa.gov/apod/ap160905.html

And this is not the only image. I am routinely making images with sub-photon per minute exposure. That is only possible with very low dark current.

Roger

 

[Eric Fossum]

And by today's standards, those specs are not very good.

15 electron read noise, but today's CMOS are under 2.

The 7D Mark II is on the order of 0.01 electron/second!

I have to say that I doubt this number if it is at RT, and I think it is at least 100x higher.

Do you have some sort of reference for this data point?

(I saw your data point of 0.016e-/s at 10C so I mean corroborating measurements)

http://www.clarkvision.com/reviews/evaluation-canon-7dii/

This is 15fA/cm^2 and is unlikely, even at 10C.

I am not sure what evidence you would like. As you say, the read noise derived, in the 1.6 electron range, is reasonable. Long exposure are then a simple derivation.

This image was made with the internal camera temperature at 16 C, and is 18 minutes total exposure (18 one-minute exposures) so 1080 seconds, and the faintest nebulae in the image are less than 1 photon per minute per pixel--on the order of 10 photons. Apparent read noise at ISO 3200 is 1.9 electrons, so noise from read noise alone is 8 electrons. With dark current on the order of 0.025 e/sec, dark current total would be 27 electrons, so noise from dark current would be about 5 electrons, with combined noise under 10 electrons, consistent with the noise level for 10 photons (yes, that less than S/N of 1). If dark current were much higher, images like this would not be possible. In fact before I did it, astrophotographers were saying it was impossible. So the observational evidence is showing that the 7D2 has impressively low dark current.


Spiral Perseid meteor spears the Heart Nebula

More info on this image is here:

http://www.clarkvision.com/gallerie...ark.av18.c08.12.2016.0J6A0119-64.g-1400s.html

It was also a NASA APOD: http://apod.nasa.gov/apod/ap160905.html

And this is not the only image. I am routinely making images with sub-photon per minute exposure. That is only possible with very low dark current.

Roger

So I have spent at least a couple of hours trying to find something to support your numbers. I did find a relatively recent Canon paper that reports a very low dark signal of 50e-/s at 60C for a pixel of about twice the area. Taking a half bandgap activation energy and scaling the temp and area suggests about 1.3e-/sec at 16C - close to what I had been thinking. But, using a nearly full bandgap activation energy in order to get your doubling temperature of 4.8C, suggests about 0.065e-/s which is still too high compared to what you report, but much lower than I expected. Makes it hard to say you are way off base and I retract my doubts!

BTW, this dark current is also much lower than I find being reported for scientific image sensors, including p-channel CMOS pixels. I am going to have to investigate further.

Thanks for bringing this to my attention and sorry for doubting you!

 

[bclaff]

BTW, this dark current is also much lower than I find being reported for scientific image sensors, including p-channel CMOS pixels. I am going to have to investigate further.

FWIW, I don't study dark current much but I also suspect that on some cameras the raw data has had some sort of signal processing applied, perhaps with information gleaned from the optical black area; very confusing to "researchers" but perhaps nice for the photographer.

Regards,

 

[SmilerGrogan]

Back up there, buddy. We need to take a quick detour. How does one measure the internal camera temperature? Or is this only something available on astro cameras.....This is a level of granularity that I didn't know was possible.

This image was made with the internal camera temperature at 16 C, and is 18 minutes total exposure (18 one-minute exposures) so 1080 seconds, and the faintest nebulae i

 

[JimKasson]

Back up there, buddy. We need to take a quick detour. How does one measure the internal camera temperature? Or is this only something available on astro cameras.....This is a level of granularity that I didn't know was possible.

This image was made with the internal camera temperature at 16 C, and is 18 minutes total exposure (18 one-minute exposures) so 1080 seconds, and the faintest nebulae i

With Sony a7x cameras, it's part of the EXIF:



Jim

--
http://blog.kasson.com

 

[Iliah Borg]

BTW, this dark current is also much lower than I find being reported for scientific image sensors, including p-channel CMOS pixels. I am going to have to investigate further.

FWIW, I don't study dark current much but I also suspect that on some cameras the raw data has had some sort of signal processing applied, perhaps with information gleaned from the optical black area; very confusing to "researchers" but perhaps nice for the photographer.

Raw data is not an accurate indicator of sensor noise performance. CDS is only the first barrier.

 

[bclaff]

BTW, this dark current is also much lower than I find being reported for scientific image sensors, including p-channel CMOS pixels. I am going to have to investigate further.

FWIW, I don't study dark current much but I also suspect that on some cameras the raw data has had some sort of signal processing applied, perhaps with information gleaned from the optical black area; very confusing to "researchers" but perhaps nice for the photographer.

Raw data is not an accurate indicator of sensor noise performance. CDS is only the first barrier.

Agreed, but raw data is what we get to observe, so we must make do. :-(

Regards,

 

[Xasan]

Output sensitivity is 25µV/e-, not 25V.

 

[bclaff]

Output sensitivity is 25µV/e-, not 25V.

That typo has already been pointed out by multiple people.

 

[Xasan]

Output sensitivity is 25µV/e-, not 25V.

That typo has already been pointed out by multiple people.

And now, by multiple + 1.

My point is it is not good data, and not only because of this typo. It is half-a$$ed data.

 

[bclaff]

Output sensitivity is 25µV/e-, not 25V.

That typo has already been pointed out by multiple people.

And now, by multiple + 1.

My point is it is not good data, and not only because of this typo. It is half-a$$ed data.

The typographical error doesn't change the validity of the underlying data.

 

[Erik Kaffehr]

Much noise no signal…

Output sensitivity is 25µV/e-, not 25V.

That typo has already been pointed out by multiple people.

And now, by multiple + 1.

My point is it is not good data, and not only because of this typo. It is half-a$$ed data.

 

[Xasan]

Output sensitivity is 25µV/e-, not 25V.

That typo has already been pointed out by multiple people.

And now, by multiple + 1.

My point is it is not good data, and not only because of this typo. It is half-a$$ed data.

The typographical error doesn't change the validity of the underlying data.

Gosh. Don't even try to derive sensor performance, less a camera based on that sensor, from that data. If you think otherwise, think again.

And no, it is not "just a typo". It is the attitude. They simply do not care.

 

[Xasan]

Much noise no signal…

You are so right, but not in the way you meant it. That data tells a photographer nothing about a camera, and very little about a sensor.

It is better then nothing data, but it is not good data, by a long shot.

 

[alanr0]

My point is it is not good data, and not only because of this typo. It is half-a$$ed data.

The typographical error doesn't change the validity of the underlying data.

Gosh. Don't even try to derive sensor performance, less a camera based on that sensor, from that data. If you think otherwise, think again.

And no, it is not "just a typo". It is the attitude. They simply do not care.

Who is "they" that are lacking in care?

Are you complaining about the cut-and-paste which lost the Greek "mu" symbols from the pixel size and output sensitivity, or do you have a more fundamental objection to the original Kodak specification?

http://www.kodak.com/ek/uploadedFiles/Content/Small_Business/Images_Sensor_Solutions/Datasheets(pdfs)/KAF-10500LongSpec.pdf

Is your concern that this specification is erroneous, or merely that the summary table does not include the parameters you prefer to see?

 

[Xasan]

My point is it is not good data, and not only because of this typo. It is half-a$$ed data.

The typographical error doesn't change the validity of the underlying data.

Gosh. Don't even try to derive sensor performance, less a camera based on that sensor, from that data. If you think otherwise, think again.

And no, it is not "just a typo". It is the attitude. They simply do not care.

Who is "they" that are lacking in care?

Those who lost µ.

do you have a more fundamental objection to the original Kodak specification?

Yes. Try using that specification in practice, you will see immediately.

 

[DSPographer]

But, to get to the OP's point, these are specs for a sensor, not for a camera. I beleive that sensors today are spec'd in at least this much detail.

I think the underlying problem alluded to by the OP is that these specifications are generally not publicly available.

I don't think the specs for many sensors are a secret. There are a host of exceptions.

...

Here are specs for many Sony sensors:

http://www.sony-semicon.co.jp/products_en/IS/sensor0/products/cmos.html

Well, take Sony for example; those are all indusrial and surveillance sensors; not those used in our digital cameras.

Regards,

--
Bill ( Your trusted source for independent sensor data at http://www.photonstophotos.net )

Also notice that the Sony documents are just flyers without critical specs like QE, read noise, or full well capacity.

For some of the industrial sensors Point Grey has characterized them and they provide comprehensive specs:

Loading…

www.ptgrey.com