Full frame cameras, worth the upgrade? - Part 1 - ISO and Noise Performance

November 10, 2014  •  2 Comments

I often hear people say that they hope some day to upgrade to full frame.  People talk about this upgrade in the same mystical fashion as finally getting to move up from the kids table on Thanksgiving.  Some feel as though it will really improve the quality of their images.  Possibly even take their photography to "the next level."  Are they right?  Over the next several weeks I'll be sharing my experience with this upgrade.  Where it was magical and where it was just kinda ho hum.  I assure you it was most certainly BOTH!  I'd like to dispel some of the mystical speak and put it into comparative terms that most photographers should understand.  If something is unclear, please drop me a line and I'll do my best to answer your questions.

Mystical quality number 1:  Full frame cameras have way better low light performance.

There are a few factors that play into low light performance.  ISO noise performance is one of the more important things you'll hear bandied about.  If you're not sure what ISO Noise performance is, it's the ability for a camera to get a good clean, noise-free image when using high ISOs.  Noise is something that people are familiar with from trying to take a picture of a darker scene with a cell phone camera.  To be able to distinguish between light and dark areas, the camera sensor has to become more sensitive to light.  This is similar to turning up the volume on a radio station that is fading out as you drive away from it.  You get a lot of static and noise.  When you look at a "noisy" image it looks about like you would expect that radio static situation to look if it was converted into image form.

It is generally true that full frame cameras have better ISO noise performance because the pixels are bigger.  To help visualize this think about placing a 5 gallon bucket outside during a rain storm.  In this example the bucket is like a pixel on your camera's sensor and the rain is light falling on that pixel.  If you now set a kiddie pool next to this 5 gallon bucket which one is going to collect more rain (light) faster?  Granted, they'll both have the same water depth but the container with the bigger opening will be able to collect a larger volume of rain given the same amount of time.  It can also handle a larger volume of rain before overflowing.

When I talk about the pixels being bigger I mean the actual pixel down deep inside of your camera's sensor.  On full frame cameras the bucket for collecting light is actually physically larger just like in the example above.  These buckets are measured in micrometers (aka microns or uM) and their sizes are told to us by manufacturers in the camera specs.  

Here are are the pixel sizes of a few popular cameras on the market today:

Canon 5D Mark III   6.25uM 
Canon 6D                6.5uM

Canon T3i                5.3uM
Canon 1DX              6.95uM
Nikon D7000            4.78uM
Nikon D750              5.9uM
Nikon D800              4.88uM
Nikon D4                  7.3uM
Nikon D3200           3.85uM

Nikon D5300           3.85uM

The Nikon D7000 has a 4.78 micrometer pixel size and the D800's pixel size is 4.88 micrometers.  The difference here is quite small actually.  Compare a D5300 to that of the D7000 and now you're talking about a larger size difference even though both of these are non full frame cameras.  Based on this you'd probably see a larger difference in ISO noise between the D5300 and the D7000 than you would between the D7000 and the D800.  Pixel size is probably one of the more important performance factors of a camera and its one that is rarely heavily advertised.  You have to do a search to find this bit of info but it's well worth it.
A great question would be to ask why the D4's pixel is so much bigger than the D800's?  They're both full frame cameras right?  To explain this think of a pixel as a marble and the sensor as a kitchen pan.  
If you get a smaller (crop sensor) cookie sheet and you put 16 million marbles into they are going to have to be quite small marbles.  However if you get a bigger (full frame) cookie sheet and you still have 16 million marbles now you can use bigger marbles than you did before.  This is what is happening when you compare the 16MP D7000 to the 16MP D4.  
The D800 has a bigger area (its full frame) but someone asked them to fit 36 million marbles into it instead of 16 million.  Thus the marbles weren't able to get much bigger.  Thankfully new technologies have allowed for better low light performance in addition to just physical bucket size.

How important to you is this?  Well lets look at a somewhat real world example.  I have a crop sensor and a full frame camera at my disposal for doing a quick test.  I have a Nikon D7000 and a Nikon D800.  DXOMark specifies an ISO setting up to which a camera can capture "excellent quality images".  

DXOMark Max ISO for Excellent Quality Images:

D7000:  1167
D800:    2853

I took pictures of a brick wall using the same settings on both of these cameras.  The images were taken with light appropriate to the camera settings, provided by a speedlight in TTL mode.  Settings are 1/125sec ISO2500 f/5.  Based on these settings we should be in an area were DXOMark says one camera can generate an "excellent quality image" and the other cannot.  This is the type of situation a wedding shooter may end up in if they were to use small flashes to bring up the illumination in the room yet want to use a higher ISO to keep the the room feeling light in the final images.

Here are the results side by side.  Both files have been sampled to the same size JPG from the camera RAW file.


Above Nikon D800, 1/125, f5, ISO2500, 200mm


Above D7000, 1/125, f5, ISO2500, 130mm

And for you pixel peepers out there here are zoomed in crops from each image:
D800 CropD800 Crop

Above Nikon D800 Zoomed Crop

D7000 CropD7000 Crop

Above Nikon D7000 Zoomed Crop

The texture of the bricks tends to hide the noise.  Much like the interest of a well taken shot will cover up noise.  There are some interesting things going on here but so far I'm not so sure I've seen a justification for the $2000 price difference between these cameras.  Granted these images were both properly lit and no post processing gain was needed.  If you accidentally under exposed an image and had to brighten it in post, the ISO performance would show itself a bit more.  

Next is a more direct comparison of the noise of the two sensors in these cameras.  I took a picture of a door illuminated by a sun tube skylight.  I purposefully underexposed these images and shot them at ISO3200.  We're now above DXOMark's acceptable range for excellent quality images on both cameras.  Camera settings are 1/200, f5, ISO3200.  Then each image was bumped up by 2 stops from the original RAW file and then both images were exported to same size JPGs.

D800 Door @ ISO3200D800 Door @ ISO3200

Nikon D800 - 1/200, f5, ISO3200, 200mm

D7000 Door @ ISO 3200D7000 Door @ ISO 3200

Nikon D7000 - 1/200, f5, ISO3200, 135mm

Now for the pixel peepers:

D800 CropD800 Crop

Nikon D800 Crop

D7000 CropD7000 Crop Nikon D7000 Crop

This is becoming more of a test just to show faults rather than a real world example but it gives you a better ability to see differences in noise performance at a fairly high ISO of 3200.  I also noticed that seeing the differences in these pictures is a lot harder on my phone which I've always regarded to have quite an excellent full HD display. 

All this being said if you REALLY have to push your camera to the limits on a regular basis shooting in low light; especially while shooting moving subjects.  The Larger pixel sizes on most full frame cameras can really make a difference there.  Here is a shot of a very dark corner inside where ISO 6400 was used to get the shutter speed up to 1/50.  A 50mm f/1.4 prime lens was used wide open to get maximum light gathered.  On previous images the shots were taken from the same location with equivalent focal lengths to compensate for the crop factor. Since this lens doesn't have a zoom capability to compensate for the crop factor, the full frame camera is a bit closer to get the same framing.  There is a lot of difference here.  If this is something you have to do frequently, you can look below and determine if it's worth the cost to you.

Note:  Both of these images were shot with auto white balance.

Nikon D800

Nikon D7000

Nikon D800

Nikon D7000

I hope this gives you an idea of some of the subtle (and not so subtle) differences between a full frame and a cropped sensor camera.  Next I'll be looking at one of two topics: Autofocus considerations for low light performance or depth of field equivalence and field of view differences.  Both of these areas show much greater differences than what we explored today.  If you would prefer to see one these topics first, please let me know in the comments below!  


2.amanda myers(non-registered)
fantastic article! I have researched this and had come to the conclusion that I really didn't need a full frame, since I have grown accustomed to the focal length difference. I didn't realize the noise was such a huge factor. Your examples and explanation were spot on and easy to understand. Thank you!
1.Ken Roland(non-registered)
Great work, hope to read some more coming soon.
No comments posted.