DXOMark rates GH4 at 13 stops dynamic range - 1 stop better than 5D Mark III

[hmcindie]

you confused Ev with Stop. 14.4 Evs on D800 is equal to 13.5 stops. so 12.8 on GH4 is much lower. however, DxO measures the DR down to 1:1 signal to noise situation, that is not practical (an image which has same amount of noise as it has signal, is useless for you, no matter how much DR is there). I usually ignore 3 stops of the score, so 10.5 stops for D800 (still) image could be a safe bet.  

 

EV = stop. Same thing. EV 2 is one stop away from EV 1. Etc.

 

Straight from DXOMark: "Maximum dynamic range is the greatest possible amplitude between light and dark details a given sensor can record, and is expressed in EVs (exposure values) or f-stops, with each increase of 1 EV (or one stop) corresponding to twice the amount of light."

 

I've shot with the D800 and it definitely has more DR than 12 stops just by eyeballing it. It has DR to spare (atleast in ISO 100). It makes no sense to just "ignore" measurements.

[Quirky]

I surely read it. Reason why someone needs to introduce some common sense and real scope over such geek questions. 

 

That's the whole point. When we lose it, we lose any basis for our wondering. A discussion about cameras serves if balanced. If not, if we distort the blanket only for one side of the bed, we will lose the whole purpose for.

 

Actually, you are overlooking one little tidbit, namely the fact that ever since cameras became digital, not all cameras are being bought for making great still or motion pictures. Sometimes the digital gadget itself is the core of the hobby. Or sometimes the camera buyers are just a bit geekier than you. In such a case, the scores matter to the intended target audience. You quite apparently aren't in that target audience, so you might just want to ignore the scores, as well as the discussions around them.

 

You are right, the DXOMark scores don't tell much about the devices as a whole, and most people do realise that, but reading and comparing them is a popular form of socialising for the right target audience. Nothing wrong with that per se. 

Besides, the nature of this thread, and thus the main target audience, is clearly marked right in the header. Not much point in delving into such a thread, if that's not your cup of tea. You might as well let the intended audience just have their geeky fun without disrupting it. Even though you did have a point in your comment. Whatever floats your/their boat.

Just sayin'.

[Emanuel]

This is the real problem in this industry, in any artistic industry tbh. Injecting your ideals of how anyone should go about learning their craft, with such brazen arrogance and reverence for your experience, aren't the kind of conversations we should be having. 

Again, if you want to talk craft, go to a forum post about craft. If you want to provide meaningful conversation, to the point at hand, then chime on in. I for one appreciate these kind of tests. The bulk of my knowledge is in computing. As I transition to film, namely digital film, I've found dynamic range information incredibly useful.

 

 

No one here intends to be arrogant, I believe. And I am giving the benefit of doubt for both on the spot. Perspective, you see? That's my purpose with my entries, to put everything in perspective. If not, all this might sound rather useless if not taken in perspective and altogether. Technical aspects can be passionate, for sure. If not, why would we be here discussing it? I just think my own POV and experience out there can be useful to others who come here to read this.

 

That said, the activity is not so diversified as it seems from the complexity itself which defines this business. I also appreciate these tests and no one is putting such in doubt. On the contrary. Again, trying to give accurate perspective in order to make them more useful, if well contextualized.

 

E :-)

[Emanuel]

Actually, you are overlooking one little tidbit, namely the fact that ever since cameras became digital, not all cameras are being bought for making great still or motion pictures. Sometimes the digital gadget itself is the core of the hobby. Or sometimes the camera buyers are just a bit geekier than you. In such a case, the scores matter to the intended target audience. You quite apparently aren't in that target audience, so you might just want to ignore the scores, as well as the discussions around them.

 

You are right, the DXOMark scores don't tell much about the devices as a whole, and most people do realise that, but reading and comparing them is a popular form of socialising for the right target audience. Nothing wrong with that per se. 

Besides, the nature of this thread, and thus the main target audience, is clearly marked right in the header. Not much point in delving into such a thread, if that's not your cup of tea. You might as well let the intended audience just have their geeky fun without disrupting it. Even though you did have a point in your comment. Whatever floats your/their boat.

Just sayin'.

 

You've 51% nailed it : ) On the leftover, I confess I can respond to your point much focused in a filmmaking forum I believe Andrew is intending to, as the filmmaker I see on his eye. And if I've always defended this corner as one of most valuable for indie filmmaking on web nowadays, it is exactly because of that. Where symbiosis between technical and aesthetical sides of the game are closer than separate realms.

 

I salute your vibe to put the things on perspective, anyway. Cinema is made of that ;-)

[Noah Yuan-Vogel]

The Alexa's dual-gain sensor structure remains unique though AFAIK.

 

Actually my understanding is that BMCC and BMPCC both also use similar dual-gain pathways to overcome ADC noise limitations like Alexa.

 

Also it seems magic lantern has a dual-gain mode for the 5DMK3 in RAW mode that allows different gain for alternating rows.

 

 

What I found interesting to see is that they measured the greatest dynamic range and best Signal to Noise at ISO 100 (measured at ISO 88), you have only access to ISO 100 when you turn on the Extended ISO range, something I never did before now :-)

 

I have just published a blogpost comparing the usable dynamic range of the BMPCC, Nikon D800 and GH4. I also show how I conducted these tests. In my opinion the usable dynamic range is around 7 stops. If I did something wrong in these tests I would really like to know how I could improve it.

 

 

Interesting test but it seems to be more of an exposure latitude test than a dynamic range test.  The exposure steps seem very coarse, to the extent that I would think there is at least a 2-stop margin of error.  Also, in such a test, judging images from different cameras, it might be better to grade the images back to "normal" using more than just offset and contrast adjustments, since to a great degree, the subjective appearance of noise and of "usability" can vary greatly depending on how the image is graded.  For example, how black level and shadow contrast is handled has a huge effect on the visibility of noise.  All of the cameras create images with very different starting points so unless you match them all at every step, its a bit hard to judge much.

[Simon Bailey]

No need to state the obvious!

There is a need because if you read most of the comments here people think that its 13 stops in video mode. It was the first thing the article should have mentioned. 

 

Never let the truth get in the way of a good story when a GH4 is concerned...

[Emanuel]

There is a need because if you read most of the comments here people think that its 13 stops in video mode.

 

hahaha True. But I think people will easily conclude by themselves to read between the lines. And a number is only a number. Like Fahrenheit or Celsius to be read from different angles and in opposite readers/sides of Atlantic. There are no magical numbers, that's part of my try to elaborate here. A GH4 can equal a Dragon (under control of certain variables), except on price :D

 

A comparative approach based on comparative terms is what can bring some light from there.

[marienho]

Interesting test but it seems to be more of an exposure latitude test than a dynamic range test.  The exposure steps seem very coarse, to the extent that I would think there is at least a 2-stop margin of error.  Also, in such a test, judging images from different cameras, it might be better to grade the images back to "normal" using more than just offset and contrast adjustments, since to a great degree, the subjective appearance of noise and of "usability" can vary greatly depending on how the image is graded.  For example, how black level and shadow contrast is handled has a huge effect on the visibility of noise.  All of the cameras create images with very different starting points so unless you match them all at every step, its a bit hard to judge much.

 

Hi Noah, great feedback, thanks. Now that you point it out, it is indeed exposure latitude test, I should change the description. It indeed is very subjective and highly dependable on grading, I wasn't sure if I was going to post it exactly because of this, but I learn from the feedback I get, so that's at least something positive for me :-)

[sanveer]

Its amusing how everyone has an opinion on how DXO Labs seems to have gotten everything wrong, and yet, none can substantiate their theories about how to calculate actual dynamic range, in anything even remotely scientific. 

[ND64]

EV = stop. Same thing. EV 2 is one stop away from EV 1. Etc.

Straight from DXOMark: "Maximum dynamic range is the greatest possible amplitude between light and dark details a given sensor can record, and is expressed in EVs (exposure values) or f-stops, with each increase of 1 EV (or one stop) corresponding to twice the amount of light."

I've shot with the D800 and it definitely has more DR than 12 stops just by eyeballing it. It has DR to spare (atleast in ISO 100). It makes no sense to just "ignore" measurements.

Check this chart from Imaging Resources:
' alt='' class='ipsImage' >

Highest detected range: 13.3 f-stop
Where is that 14.4 number?

and this one is even more realistic (also closer to my guess):

http://home.comcast.net/~NikonD70/Charts/PDR.htm#D800E

[Quirky]

You've 51% nailed it : ) On the leftover, I confess I can respond to your point much focused in a filmmaking forum I believe Andrew is intending to, as the filmmaker I see on his eye. And if I've always defended this corner as one of most valuable for indie filmmaking on web nowadays, it is exactly because of that. Where symbiosis between technical and aesthetical sides of the game are closer than separate realms.

 

I salute your vibe to put the things on perspective, anyway. Cinema is made of that ;-)

 

Well, the point was, the EOSHD forum is not much different from a bunch of other online forums, be it a stills or video oriented one, they're all filled with the same gadget geekery and nerdy-nam-nam these days, and it's normal these days. Going on a crusade in defence of the pure craft of filmmaking is pretty futile, and waste of your time.

 

You could assume the role of the noble king anointed by some watery tart distributing swords from a lake and go on a quest for the Holy Reel, along with the noble Knights Who Shoot Cinema, but in threads like this you might end up being taunted by those outrageous French soldiers throwing down cows wrapped in DXOMark spec sheets over the wall of their castle time and again. So why not letting them be and go looking for the light on your own instead?

 

Those with a genuine interest in the pragmatic side of photography, video and filmmaking will get bored with the DXOMark specs eventually, anyway. Then they'll start following you and knight Andrew on the quest for the Holy Reel. 

 

But meanwhile, let the French soldiers keep their outragous geek accents and their cows wrapped in DXOMark sheets.  After all, this thread is their castle, it's only a model, and they're minding their own business behind it peacefully. Save your Holy Hand Granade for the killer rabbit, which is much more dangerous to all of us, you know.  :P

 

Life online isn't supposed to be too serious, anyway, is it. Carry on. 

/OT

[Clayton Moore]

Someone I know was hired to shoot a video for SONY using the F55.  He used a GH4 for B-Roll and said when the whole thing was down scaled to HD it was freeking amazing.  Said both he and SONY were very pleased.  Said that the GH4 when the in camera 4k gets downscaled to HD 1920x1080, it makes by far the best looking HD he has ever shot using an affordable camera.  Even better he says then the BMCC 2.5k. Since most of us wont actually deliver much of anything to clients in actual 4k.  Sounds like a prretty good deal to me.

[hmcindie]

Check this chart from Imaging Resources:
' alt='' class='ipsImage' >

Highest detected range: 13.3 f-stop
Where is that 14.4 number?

and this one is even more realistic (also closer to my guess):

http://home.comcast.net/~NikonD70/Charts/PDR.htm#D800E

 

That home.comcast (whoever does those) measurements makes no sense. 7d has six stops of latitude at ISO 1600 according to that? Yeah no. They must be measuring completely bonkers.

 

And about that Imaging Resources link. They say this: "The total dynamic range score is about as high as it can get with the Stouffer 4110 stepchart using an actual lens (the lens likely introduces a small loss in dynamic range)"

 

You CANNOT get a higher score than that with the Stouffer 4110 stepchart. DXOMark measures better.

[Emanuel]

Well, the point was, the EOSHD forum is not much different from a bunch of other online forums, be it a stills or video oriented one, they're all filled with the same gadget geekery and nerdy-nam-nam these days, and it's normal these days. Going on a crusade in defence of the pure craft of filmmaking is pretty futile, and waste of your time.

 

You could assume the role of the noble king anointed by some watery tart distributing swords from a lake and go on a quest for the Holy Reel, along with the noble Knights Who Shoot Cinema, but in threads like this you might end up being taunted by those outrageous French soldiers throwing down cows wrapped in DXOMark spec sheets over the wall of their castle time and again. So why not letting them be and go looking for the light on your own instead?

 

Those with a genuine interest in the pragmatic side of photography, video and filmmaking will get bored with the DXOMark specs eventually, anyway. Then they'll start following you and knight Andrew on the quest for the Holy Reel. 

 

But meanwhile, let the French soldiers keep their outragous geek accents and their cows wrapped in DXOMark sheets.  After all, this thread is their castle, it's only a model, and they're minding their own business behind it peacefully. Save your Holy Hand Granade for the killer rabbit, which is much more dangerous to all of us, you know.  :P

 

Life online isn't supposed to be too serious, anyway, is it. Carry on. 

/OT

 

 

I guess life online is only part of the real XXI deal. There's no exactly B&W or there shoudn't be. I see Andrew's work as fair serious enough, though. Or you woudn't see people like me or Noah Yuan-Vogel, as for instance, posting in the same page. There's absolute zero of cruzade over here. Just a complementary attempt to give meaning to certain discussions people entertain themselves to have in such boards ;-)

[jgharding]

 

Actually my understanding is that BMCC and BMPCC both also use similar dual-gain pathways to overcome ADC noise limitations like Alexa.

 

Also it seems magic lantern has a dual-gain mode for the 5DMK3 in RAW mode that allows different gain for alternating rows.

 

 

Yes I forgot about Magic Lantern, it does come with heavy aliasing though becuse it's alternate channels, not every channel having both levels, we aren't talking about dual-gain stages at a hardware level, but it's the same theory.

 

The BMCC etc I didn't know about, I'll have to have a look for more info. It certainly should be cleaner and have a broader range with proper dual channel. Maybe they just use it for noise processing in ProRes mode?

 

I know the sCMOS2 has dual 11-bit ADC, but that's not in their cameras yet.

[Andrew Reid]

I think (but not 100% sure) that the old sCMOS had dual ADC too. The specs are quite similar. The new chip doesn't seem to bring a lot new to the table.

 

Dual-gain on Canon DSLRs is nice for stills, but not usable for video in my opinion.

[squig]

On 17/06/2014 at 1:08 PM, Noah Yuan-Vogel said:

Also it seems magic lantern has a dual-gain mode for the 5DMK3 in RAW mode that allows different gain for alternating rows.

Dual ISO mode. It's great for stills, 2 stops more DR. But there's a lot of aliasing in raw video mode.

You getting the 5K or 6K Terra?

Looking at the specs it doesn't look like the 5K Terra records CDNG. So what I'm wondering is how are the Kineraw files going to talk to the usual suspect in post.

Now their site's down.

[hmcindie]

It's funny reading these old posts. The 5d mark III still does great looking shots while the GH4 has been relegated long ago to the dust bin.

[Inazuma]

24 minutes ago, hmcindie said:

It's funny reading these old posts. The 5d mark III still does great looking shots while the GH4 has been relegated long ago to the dust bin.

Very subjective... to me the 5d3 has been in the dust bin for years

[TheRenaissanceMan]

Lots of people are still loving and doing great work with the GH4, so I'm not sure what you're on about there.