DXOMark Pentax K-5 vs Nikon D7000 vs Canon 7D

Reportage · 11-08-2010, 07:38 AM

Originally posted by kittykat46

Don't tell me some bright engineer in Pentax was given an assignment - go study how dXOMark does its measurement and come back with some recommendations on how to tweak the sensor performance give us that little bit of an edge.

Its not unheard of in the corporate world

, if the index is a well publicised and respected one, but I'm just toying around here...

well, has Nvidia or ATI in times past ever optimized their drivers for optimum 3dmark scores rather then actual real world performance?

falconeye · 11-08-2010, 07:40 AM

Originally posted by kittykat46

Don't tell me some bright engineer in Pentax was given an assignment - go study how dXOMark does its measurement and come back with some recommendations on how to tweak the sensor performance give us that little bit of an edge.

An engineer (Pentax) on assignment to beat a physicist (DxO)? Poor engineer ...

OTOH, they tried with their raw-level NR and fooled about everybody else

glanglois · 11-08-2010, 08:43 AM

Originally posted by falconeye

An engineer (Pentax) on assignment to beat a physicist (DxO)? Poor engineer ...

OTOH, they tried with their raw-level NR and fooled about everybody else

Spoken as a true physicist!

macTak · 11-08-2010, 09:02 AM

I think (considering how idential the results are save for a little greater ISO cheating by Nikon) this shows that it was a really clever move by Pentax to add an ISO 80 option and thus move their camera's ranking ahead of the D7000 and any other cameras going to use that sensor. Just wait, next generation all the compaines (instead of compeating to have the highest ISO as they have been), will be competing to have the lowest ISO

stormcloud · 11-08-2010, 09:32 AM

an iso 50 would be nice addition - even if it was just overexposed then brought down in processing - come on pentax firmware update for iso 50

falconeye · 11-08-2010, 09:39 AM

Originally posted by macTak

add an ISO 80 option and thus move their camera's ranking ahead of the D7000 and any other cameras going to use that sensor. Just wait, next generation all the compaines (instead of compeating to have the highest ISO as they have been), will be competing to have the lowest ISO

Again.

You can't just lower the ISO setting.
The way DxO tests Dynamic Range, it would make no difference at all. The translucent mirror in the SLT-A55 makes no difference either.

The only way to increase DR (for a given read-out noise) is by a higher full well capacity which allows for a true lower iso. Cheating on iso doesn't give you a single extra electron in the well.

I don't fully understand what happened in the ISO 70/83 region to the K-5/D7000 sensors. But it's not about "adding an option". Not at all.

macTak · 11-08-2010, 09:56 AM

Originally posted by falconeye

Again.

You can't just lower the ISO setting.
The way DxO tests Dynamic Range, it would make no difference at all. The translucent mirror in the SLT-A55 makes no difference either.

The only way to increase DR (for a given read-out noise) is by a higher full well capacity which allows for a true lower iso. Cheating on iso doesn't give you a single extra electron in the well.

I don't fully understand what happened in the ISO 70/83 region to the K-5/D7000 sensors. But it's not about "adding an option". Not at all.

Falk, I take it that the sensor (for both cameras) had that full well capacty allowing for a true lower ISO, but for whatever reason, Nikon decided not to go that low (unlike Pentax). That's what I meant by adding the option.

falconeye · 11-08-2010, 11:10 AM

Originally posted by macTak

Falk, I take it that the sensor (for both cameras) had that full well capacty allowing for a true lower ISO, but for whatever reason, Nikon decided not to go that low (unlike Pentax). That's what I meant by adding the option.

I know what you meant. Taken two sensors with the same full well capacity, the same quantum efficiency and the same analog gain setting (e.g., the same sensor actually).

Then both sensors have the same true base iso value, assuming the same microlens fill factor and bayer filter transparencies (which seems to be the case from the other curves).

Now, to go below this true base iso value doesn't increase the DR anymore. It just stays constant (with a risk of overexposure). Even if such a thing as negative gain existed, if the well is full it is full. And Dxo measures ISO by observing pixel saturation. At negative gain, the pixel will never saturate and the so-called output dynamic range is damaged. That's not the case here. DxO observes 100% luminosity at K-5 vendor ISO 80.

The point is that your statement implies that Nikon wouldn't use the sensor at base ISO at all. But always with a gain of 91/70=83/70 or +19%.

That's a different statement from just cheating on iso and adding an option. E.g., Pentax could call their extended ISO "50". 50 isn't unheard of for a real value of 70. But it wouldn't change the result.

So, why shouldn't Nikon use the sensor at base gain already at ISO 100, or add an extended ISO 50 step? Not fully exploiting the sensor just doesn't make sense.

The only other idea I could come up with is that the Sony sensor documentation mentions some strange setting lowering the chip's quantum efficiency, e.g., by lowering the photon-electron conversion rate, maybe by applying some reduced voltage somewhere. That would be a strange thing possibly explaining why it is an extended setting for Pentax. But I don't believe this. Just enumerating the possibilities.

Also possible is that the sensor documentation lists gain factors which Pentax choose to underrun after finding out that there is some headroom, i.e., the lowest listed gain isn't the lowest preventing clipping the output DR.

This is a document of Sony's 12MP Exmor chip which may be used in the K-x: http://www.sony.net/Products/SC-HP/cx_news/vol50/pdf/imx021.pdf

It talks about a programmable 24dB gain amplifier. 24dB are 4 stops spanning a region e.g., from ISO 160 to ISO 2500. Any gain above must be done in software, any "gain" below by overexposure. As the gain amplifier must be on chip, Nikon and Pentax have to use the same one. I can't see why Nikon wouldn't use the 0dB setting...

However, I've not found a corresponding document for the sensor in the K-5.

Tom in Delaware · 11-08-2010, 05:45 PM

D7000 comes close. Close but no bananas. The 7D has suddenly become an expensive doorstop. Heh, heh.

It's good to be the king.

zxaar · 11-08-2010, 05:54 PM

@falconeye , the reason for nikon for not going to lower iso than iso 100 might be related to the way they implement their hardware with sensor. Both nikon and pentax after-tune the base sensor so i take it that it is the hardware than connects sensor that could hamper. If it were a software issue, they might have worked around it.

Just a guess though, most probably i am wrong on this.

falconeye · 11-08-2010, 06:27 PM

Originally posted by zxaar

@falconeye , the reason for nikon for not going to lower iso than iso 100 might be related to the way they implement their hardware with sensor. Both nikon and pentax after-tune the base sensor so i take it that it is the hardware than connects sensor that could hamper. If it were a software issue, they might have worked around it.

Just a guess though, most probably i am wrong on this.

I would normally agree. But the charm with that sensor is that it is used almost like memory. You read out digital values. Of course, the quality of power supply, cooling and protection from electromagnetic perturbations increases IQ. But by decreasing noise, not by increasing the signal. So, that's no explaination. Maybe, you can force the programmable gain amplifier to a lower gain at 0dB by providing undervoltage to the image sensor. And observe that the ADC did't utilize the entire full well capacity before overrun. But that's an only remote possibility.

OTOH, I know of no other logical explaination except to protect a forthcoming D400.

11-08-2010, 11:10 AM	#23
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by macTak Falk, I take it that the sensor (for both cameras) had that full well capacty allowing for a true lower ISO, but for whatever reason, Nikon decided not to go that low (unlike Pentax). That's what I meant by adding the option. I know what you meant. Taken two sensors with the same full well capacity, the same quantum efficiency and the same analog gain setting (e.g., the same sensor actually). Then both sensors have the same true base iso value, assuming the same microlens fill factor and bayer filter transparencies (which seems to be the case from the other curves). Now, to go below this true base iso value doesn't increase the DR anymore. It just stays constant (with a risk of overexposure). Even if such a thing as negative gain existed, if the well is full it is full. And Dxo measures ISO by observing pixel saturation. At negative gain, the pixel will never saturate and the so-called output dynamic range is damaged. That's not the case here. DxO observes 100% luminosity at K-5 vendor ISO 80. The point is that your statement implies that Nikon wouldn't use the sensor at base ISO at all. But always with a gain of 91/70=83/70 or +19%. That's a different statement from just cheating on iso and adding an option. E.g., Pentax could call their extended ISO "50". 50 isn't unheard of for a real value of 70. But it wouldn't change the result. So, why shouldn't Nikon use the sensor at base gain already at ISO 100, or add an extended ISO 50 step? Not fully exploiting the sensor just doesn't make sense. The only other idea I could come up with is that the Sony sensor documentation mentions some strange setting lowering the chip's quantum efficiency, e.g., by lowering the photon-electron conversion rate, maybe by applying some reduced voltage somewhere. That would be a strange thing possibly explaining why it is an extended setting for Pentax. But I don't believe this. Just enumerating the possibilities. Also possible is that the sensor documentation lists gain factors which Pentax choose to underrun after finding out that there is some headroom, i.e., the lowest listed gain isn't the lowest preventing clipping the output DR. This is a document of Sony's 12MP Exmor chip which may be used in the K-x: http://www.sony.net/Products/SC-HP/cx_news/vol50/pdf/imx021.pdf It talks about a programmable 24dB gain amplifier. 24dB are 4 stops spanning a region e.g., from ISO 160 to ISO 2500. Any gain above must be done in software, any "gain" below by overexposure. As the gain amplifier must be on chip, Nikon and Pentax have to use the same one. I can't see why Nikon wouldn't use the 0dB setting... However, I've not found a corresponding document for the sensor in the K-5. Last edited by falconeye; 11-08-2010 at 11:39 AM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Nikon D7000 seems like a better deal than K-5	crossover37	Canon, Nikon, Sony, and Other Camera Brands	60	11-07-2010 01:22 AM
Pentax K-5 vs Nikon D7000 image quality	seliscan	Pentax K-5 & K-5 II	5	11-02-2010 01:54 AM
Nikon D7000 First Unboxing videos	jogiba	Canon, Nikon, Sony, and Other Camera Brands	3	10-18-2010 08:23 PM
Nikon D7000 pictured	deadwolfbones	Canon, Nikon, Sony, and Other Camera Brands	4	10-18-2010 04:19 AM
Nikon D7000 officially announced!	JohnBee	Canon, Nikon, Sony, and Other Camera Brands	41	10-04-2010 02:30 PM

11-08-2010, 07:38 AM	#16
Reportage Veteran Member Join Date: Feb 2009 Photos: Gallery Posts: 739	Originally posted by kittykat46 Don't tell me some bright engineer in Pentax was given an assignment - go study how dXOMark does its measurement and come back with some recommendations on how to tweak the sensor performance give us that little bit of an edge. Its not unheard of in the corporate world , if the index is a well publicised and respected one, but I'm just toying around here... well, has Nvidia or ATI in times past ever optimized their drivers for optimum 3dmark scores rather then actual real world performance?

11-08-2010, 07:40 AM	#17
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by kittykat46 Don't tell me some bright engineer in Pentax was given an assignment - go study how dXOMark does its measurement and come back with some recommendations on how to tweak the sensor performance give us that little bit of an edge. An engineer (Pentax) on assignment to beat a physicist (DxO)? Poor engineer ... OTOH, they tried with their raw-level NR and fooled about everybody else

11-08-2010, 08:43 AM	#18
glanglois Veteran Member Join Date: Aug 2007 Location: Madison, Wis., USA Posts: 1,506	Originally posted by falconeye An engineer (Pentax) on assignment to beat a physicist (DxO)? Poor engineer ... OTOH, they tried with their raw-level NR and fooled about everybody else Spoken as a true physicist!

11-08-2010, 09:02 AM	#19
macTak Veteran Member Join Date: Apr 2010 Photos: Gallery Posts: 759	I think (considering how idential the results are save for a little greater ISO cheating by Nikon) this shows that it was a really clever move by Pentax to add an ISO 80 option and thus move their camera's ranking ahead of the D7000 and any other cameras going to use that sensor. Just wait, next generation all the compaines (instead of compeating to have the highest ISO as they have been), will be competing to have the lowest ISO

11-08-2010, 09:32 AM	#20
stormcloud Senior Member Join Date: Oct 2009 Posts: 232	an iso 50 would be nice addition - even if it was just overexposed then brought down in processing - come on pentax firmware update for iso 50

11-08-2010, 09:39 AM	#21
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by macTak add an ISO 80 option and thus move their camera's ranking ahead of the D7000 and any other cameras going to use that sensor. Just wait, next generation all the compaines (instead of compeating to have the highest ISO as they have been), will be competing to have the lowest ISO Again. You can't just lower the ISO setting. The way DxO tests Dynamic Range, it would make no difference at all. The translucent mirror in the SLT-A55 makes no difference either. The only way to increase DR (for a given read-out noise) is by a higher full well capacity which allows for a true lower iso. Cheating on iso doesn't give you a single extra electron in the well. I don't fully understand what happened in the ISO 70/83 region to the K-5/D7000 sensors. But it's not about "adding an option". Not at all.

11-08-2010, 09:56 AM	#22
macTak Veteran Member Join Date: Apr 2010 Photos: Gallery Posts: 759	Originally posted by falconeye Again. You can't just lower the ISO setting. The way DxO tests Dynamic Range, it would make no difference at all. The translucent mirror in the SLT-A55 makes no difference either. The only way to increase DR (for a given read-out noise) is by a higher full well capacity which allows for a true lower iso. Cheating on iso doesn't give you a single extra electron in the well. I don't fully understand what happened in the ISO 70/83 region to the K-5/D7000 sensors. But it's not about "adding an option". Not at all. Falk, I take it that the sensor (for both cameras) had that full well capacty allowing for a true lower ISO, but for whatever reason, Nikon decided not to go that low (unlike Pentax). That's what I meant by adding the option.

11-08-2010, 05:45 PM	#24
Tom in Delaware Senior Member Join Date: May 2009 Location: Delaware Posts: 177	D7000 comes close. Close but no bananas. The 7D has suddenly become an expensive doorstop. Heh, heh. It's good to be the king.

11-08-2010, 05:54 PM	#25
zxaar Veteran Member Join Date: Jul 2009 Location: kobe/japan Posts: 510	@falconeye , the reason for nikon for not going to lower iso than iso 100 might be related to the way they implement their hardware with sensor. Both nikon and pentax after-tune the base sensor so i take it that it is the hardware than connects sensor that could hamper. If it were a software issue, they might have worked around it. Just a guess though, most probably i am wrong on this.

11-08-2010, 06:27 PM	#26
falconeye Veteran Member Join Date: Jan 2008 Location: Munich, Alps, Germany Photos: Gallery Posts: 6,871	Originally posted by zxaar @falconeye , the reason for nikon for not going to lower iso than iso 100 might be related to the way they implement their hardware with sensor. Both nikon and pentax after-tune the base sensor so i take it that it is the hardware than connects sensor that could hamper. If it were a software issue, they might have worked around it. Just a guess though, most probably i am wrong on this. I would normally agree. But the charm with that sensor is that it is used almost like memory. You read out digital values. Of course, the quality of power supply, cooling and protection from electromagnetic perturbations increases IQ. But by decreasing noise, not by increasing the signal. So, that's no explaination. Maybe, you can force the programmable gain amplifier to a lower gain at 0dB by providing undervoltage to the image sensor. And observe that the ADC did't utilize the entire full well capacity before overrun. But that's an only remote possibility. OTOH, I know of no other logical explaination except to protect a forthcoming D400.