Production K-7 Dynamic Range vs. K20D...

[GordonBGood]

I thank you for your answer, 300 dpi may be the delivery standard for supplying images, but photo books and quality magazines (with quality prints) seldom are more than 200 ppi. (quality magazines are often 150 ppi, compared to typical newspapers magazines which are less than 90 ppi).

Roland, so as per my post step back away only two times normal viewing distance instead of three times for 200 dpi rather than 300, one and a half times for 150 dpi, and don't step back at all for newspaper/magazine quality at 90 ppi, although for the smaller dpi, one would normally apply some downsampling, which you can play with as well.

I just answered your question of how the K-7 images would look printed at A4 in my earlier post, and most printers work with a ppi of about 300 dpi even though the dpi is much higher.

Regards, GordonBGood

 

[rgmwa]

You obviously have in-depth knowledge of sensor design, computing and electronics, and your posts reflect that. Are you involved in the industry. Just curious?

--
Robert
rgmwa

 

[philipfla]

Have any of you guys deteremined yet how many angels can dance on the K-7 sensor? My concern is that it may not be enough and the camera will not be able to make a photograph up to my exacting standards.

 

[jamesm007]

Do you think your hypothesis of the OP amps or what ever they are called being placed on the Samsung CMOS chip is still the likely cause of the noise? I would think if Samsung could remove it and have a nice decrease in dark read noise this was their big chance. But its still about the same. I am not a sensor scientist I am only thinking through logic, but have you re-thought the idea? Perhaps something else in the design of the chip. They claimed to have done some radical things such as making the distance from the micro lens to the PD (photo diode) much shorter and capture more light not coming perfectly perpendicular to the PD compared to other designs. Who knows what else they did. But the big question is do they even know why they have highish dark read noise; at least compared to other sensors. They apparently did not address it at all, which given the opportunity seems odd.

Now that being said, I had another look at K-7 samples and as you said if your happy with the K20D (as I am) you will be happy with the K-7, no problem if your willing to do some work such as me with Noise Ninja or another program getting decent iso3200 pics about the equal of the K20D. So I am not knocking it. In fact parts of the pic looks better than the K20D at high ISO. The blacks seem blacker?
--
jamesm007,
http://s195.photobucket.com/albums/z77/jamesm700/
WSSA member 266PX

 

[jamesm007]

BTW nothing personal about your hypo, I don't ask in a sarcastic or ? demeaning way. I know you know your stuff and respect it, so I am just curious as to your thoughts now.
--
jamesm007,
http://s195.photobucket.com/albums/z77/jamesm700/
WSSA member 266PX

 

[tomtor]

BTW nothing personal about your hypo, I don't ask in a sarcastic or ? demeaning way. I know you know your stuff and respect it, so I am just curious as to your thoughts now.

And when Gordon is answering sensor questions, I have a question of my own

I understand that some sensors (Nikon D5000/90 etc) have a native ISO sensitivity of 200 and that 100 is a software? emulation. I do not understand how that would work. It is clear that you can raise the ISO in software by simply scaling the sensor values by eg a factor 2, but I do not understand how you can lower the ISO. Why don't you get blown highlights when lowering the ISO with a trick? What is the method and what are the drawbacks involved?

Thanks for any info.

--
Tom - http://www.pentaxphotogallery.com/tomvijlbrief

 

[dotnik]

I also posted it here:
http://forums.dpreview.com/forums/read.asp?forum=1036&message=32400605

With the K7 I tested the other day I detected a huge difference between ISO 1600 and ISO 3200. At least I now fully understand why Pentax did not want beta testers to post images at the highest ISOs.

Is it possible to control noise in a way so that it lowers in some areas at the cost of higher noise in other areas in order to push the worst noist above ISO 3200 so that ISO 1600 in teh K7 could be better than the K20D but the K20D performs better at ISO 3200 and ISO 6400? A kind of noise envelope so to speak.

Or does noise always behave in a linear way?

--
.......
Have a nice day (a picture is worth a thousand words)
Jim

dotnik:
'Don't overestimate technology - nothing is knowledgefree'

 

[GordonBGood]

You obviously have in-depth knowledge of sensor design, computing and electronics, and your posts reflect that. Are you involved in the industry. Just curious?

Robert, I'm just an electronics engineer with a background in specialized data acquisition systems with an intense interest in digital photography and experience with Pentax digital cameras since the Optio 330 in 2001.

Regards, GordonBGood

 

[GordonBGood]

Do you think your hypothesis of the OP amps or what ever they are called being placed on the Samsung CMOS chip is still the likely cause of the noise? I would think if Samsung could remove it and have a nice decrease in dark read noise this was their big chance. But its still about the same. I am not a sensor scientist I am only thinking through logic, but have you re-thought the idea? Perhaps something else in the design of the chip. They claimed to have done some radical things such as making the distance from the micro lens to the PD (photo diode) much shorter and capture more light not coming perfectly perpendicular to the PD compared to other designs. Who knows what else they did. But the big question is do they even know why they have highish dark read noise; at least compared to other sensors. They apparently did not address it at all, which given the opportunity seems odd.

I think there are two sources of too high noise in the K20D/K-7, as follows:

1) the black read noise of the sensor itself is somewhat higher than the competition being about twice as high as that of Nikon D300 and about three times as high as the Nikon D90/D5000 and the Canon 50D/500D etc.

2) the "OP amps or what ever" are the charge buffer and variable gain amplifiers that are limiting the low ISO noise by adding a more-or-less constant source of noise to the black read noise that is relatively large as compared to the sensor black read noise as ISO decreases. This is similar to the problem that the Canon designs have but a little bit worse. The Nikon/Sony sensor designs don't have this problem and have very little of this constant noise source injected relative to the size of the black read noise of the photocell (sencel) itself.

I think that had Pentax not divorced Sony sensor in favour of Samsung sensors, we would now have image quality of the order of the new Nikon cameras. As to whether Pentax engineers know, how could they not know when the results of raw testing are published on the DxOMark Sensor Score website? I think that this is not completely in Pentax's control as the sensor technology belongs to Samsung, and the priority for this generation has been to add video capability to the sensor as well as tune the feature set. I would think that the next generation will be the one that intense research is done on image quality as well as furthering the video capabilities.

Now that being said, I had another look at K-7 samples and as you said if your happy with the K20D (as I am) you will be happy with the K-7, no problem if your willing to do some work such as me with Noise Ninja or another program getting decent iso3200 pics about the equal of the K20D. So I am not knocking it. In fact parts of the pic looks better than the K20D at high ISO. The blacks seem blacker?

I still have a slight problem in comparing the K-7 to the K20D as to noise in that there seems to be less noise reduction applied to the K-7 data right across the ISO range including low ISO's or there is a small extra source of noise being injected into the K-7 raw data, but it isn't anything that just a small touch of noise reduction and post processing of ones favourite flavour can't mostly cure and make the end results pretty much the same.

Regards, GordonBGood

 

[GordonBGood]

I understand that some sensors (Nikon D5000/90 etc) have a native ISO sensitivity of 200 and that 100 is a software? emulation. I do not understand how that would work. It is clear that you can raise the ISO in software by simply scaling the sensor values by eg a factor 2, but I do not understand how you can lower the ISO. Why don't you get blown highlights when lowering the ISO with a trick? What is the method and what are the drawbacks involved?

Lower ISO's than the "base" sensitivity are achieve by a combination of reducing the available "highlight headroom" and allowing the brightest tones to clip for some images; this means that in most cases the so called ISO 100 isn't really ISO 100 as compared to ISO 200, but is actually about ISO 150 but is metered as if it were ISO 100 with a half stop reduction in headroom and/or clipping.

Regards, GordonBGood

 

[GordonBGood]

I also posted it here:
http://forums.dpreview.com/forums/read.asp?forum=1036&message=32400605

With the K7 I tested the other day I detected a huge difference between ISO 1600 and ISO 3200. At least I now fully understand why Pentax did not want beta testers to post images at the highest ISOs.

The difference between raw data for ISO 1600 and ISO 3200 on the K-7 is that it appears that absolutely no Noise Reduction (NR) is applied to the ISO 1600 readings where there is a moderate amount applied at ISO 3200 and above.

Is it possible to control noise in a way so that it lowers in some areas at the cost of higher noise in other areas in order to push the worst noist above ISO 3200 so that ISO 1600 in teh K7 could be better than the K20D but the K20D performs better at ISO 3200 and ISO 6400? A kind of noise envelope so to speak.

No, there is no way to accomplish this of which I am aware. I believe that what you are seeing is just the results of different types of noise NR applied by the two models, as per my comment above.

Regards, GordonBGood

 

[dotnik]

--
.......
Have a nice day (a picture is worth a thousand words)
Jim

dotnik:
'Don't overestimate technology - nothing is knowledgefree'

 

[Leo]

Lightroom 2 has the Exposure adjustment + - 4.00EV I think.
Leo

 

[exdeejjjaaaa]

some update from Oleg_V :

http://www.penta-club.ru/forum/index.php?showtopic=62912&hl=

PS: text is mostly in Russian

--

 

[copejorg1]

They claimed to have done some radical things such as making the distance from the micro lens to the PD (photo diode) much shorter and capture more light not coming perfectly perpendicular to the PD compared to other designs.

Hi James,

I would be very surprised to find that Samsung had really made the change you mentioned above in the K-7's sensor design (relative to their previous K20D/GX-20 sensor, that is), as this benefit was already being claimed in their advertising for the K20D/GX-20 DSLRs. See the link below, from Samsung's GX-20 web page:

http://www.samsung.com/us/consumer/detail/detail.do?group=camerascamcorders&type=digitalcameras&subtype=dslr&model_cd=ER-GX20ZBBA/US

In order for this claim to be true for the K-7 sensor (vis-a-vis its predecessor from the K20D/GX20 products), it would mean that they did it again, putting the photodiode even closer to the surface of the silicon than they did in their earlier sensor. I doubt this is the case.

I personally think that in Pentax's K-7 advertising, they were just recycling this bit of information from earlier advertising, hoping it wouldn't be noticed that this benefit (versus competing sensor manufacturers) had already been incorporated into their previous-generation product.
--
Regards,
Greg