2011-02-19

Sony HX9V think 4MP not 16MP and be happy


If you think of the Sony DSC-HX9V as a 16MP camera and you look at the results in 100%, you might be unhappy. But if you think of the HX9V as a 4MP to max 8MP camera you will be much more pleased with the results.

I will try to explain in a very simplified level the why.


The Airy Disk is the smallest dot of light any lens can produce.
The diameter in microns is about 1.35 times the f/number (using the green wavelength our eyes see most brightly). So, for example, the Airy disk at f/ 4.0 is 5.4 microns across.
It is understandable now, why prime lenses have so low f/ numbers, because the lower they are the smaller is the dot of light they can put on the sensor = higher resolution.
Also these lenses have a much higher optical quality.

The 16MP 1/2.3" sensor has 57 MP/cm² pixel density —> 1.3 micron pixels,
Panasonic´s 14 MP has 50 MP/cm² —> 1.4 micron pixels
the Sony HX5 has 35 MP/cm² —> 1.7 micron pixels
the Fujifilm F31fd had 16 MP/cm² —> 2.5 micron pixels
Sony Alpha NEX-5 has 3.9 MP/cm² —> 5 micron pixels
Sony A55 has 4.4 MP/cm² —> 4.8 micron pixels

To better understand the problematic I cropped the sensor from 4608 x 3456 to 15 x 15.
Then I overlayed an Airy Disk at f/4.0  and full tele over the cropped part of the sensor.
With the size of the 16MP sensor and the f number range of the Sony lens the smallest light dot (Airy Disk) will cover already more than 9 pixel at f 4.0.
If we do an honest count we can cover the surface of nearly 16 combined pixel.

Now you understand why every camara looses resolution and detail when you zoom in.

Let´s compare that to the Sony HX5
What we crop from the sensor has the same size but a lesser density. 1.7 micron per sensor
If we do an honest count we can cover nearly 10 pixel - fascinating.


We cover nearly 16 sensors at 16MP and 10 at 10 MP. So if we wanted to have an as sharp coverage as with a 14MP APS-C sensor, then a 1MP 1/2.3" sensor will do the trick.

Here the 14MP APS-C Sony sensor what is in the NEX-3, NEX-5 and Sony A33
5 micron per sensor


After letting sink in what I just fabricate I need a drink.



here is a nice explanation how the sensor gets the pixel color information.


A closer look at the Sony 16MP sensor.
To make things easy to understand we take a 16 pixel sensor.
Above the 16 pixel sensor is the Bayer filter which covers every pixel with one R, G or B filter. Green gets the double amount of pixels to cover.
Our smallest possible light dot covers already pixel 1-3,5-7 and 9-11.
you see there is a lot of overlapping information when we move the Airy Disk just 1 pixel to get our 16MP resolution.
Our smallest possible light dot covers 1-3,5-7 and 9-11 with some overlapping of 8 and 14  you get the picture why we have smudged detail what can not be recovered especially in low contrast and low light situations.


You understand now that when we combine 1,2,5,6 then 3,4,7,8 then 9,10,13,14 and at last 11,12,15,16 we reduced our resolution from 16 to 4 but with a much more pleasing result.

from a forum post:
theranman
Although a vast oversimplification, there is an easy way to get an idea of what the noise issue is all about. Draw a square (our sensor, so to speak). Subdivide it into four. Let's say that each of the four squares represents a pixel. Now, draw another same size square. Divide it into 16. Let's say that each of those squares represents a pixel. On the face of it, you'd think that both boxes would receive the same amount of light, right? Wrong. See the problem? The square with 16 pixels has ALLLL those lines inside of it. Those lines represent the pixel walls. Those pixel walls block the downward shining light, light that would otherwise reach the sensor if they weren't there. That's noise.

VarmintCong
Here's a cross-section of D7000 that helps illustrate this point. I don't know how much it's a factor though. The light passes through those trapezoidal shaped lines before hitting the photo sensors, which you can't really see here but are at the bottom of the image.
You can also see the mirolenses on top which make sure that more photons hitting the surface of the sensor reach their intended destination.

with the Sony BSI cmos sensor the middle part (wires) are at the bottom so the light has to travel less before hitting the photosensitive part.

A look at the full 16MP resolution with ISO 200

and after reducing to 4MP, auto color correction and some sharpening


I did the same with the images of my HX5 when the originals were just not good enough.

With this little quick and dirty explanation you have now a better understanding why I think of the Sony HX9V as a 4MP camera.

17 comments:

  1. Thanks. What are the settings you use to downsize?

    Regards,
    Frank B

    ReplyDelete
  2. I used faststone viewer my quick and easy tool for that
    Resize / Resample 50%
    Lanczos3

    ReplyDelete
  3. I don't know what formula you have used to calculate the airy disk size. With my math the HX9V on widest angle and max aperture will have a disk 1,7 μm in size.

    X = 1.22 x lamba x f/d (lamda for visible light approx. 420 nm)

    ReplyDelete
  4. Airy Disk =1.35 * f --- calculator at:
    http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm --- http://en.wikipedia.org/wiki/Airy_disk ---

    ReplyDelete
  5. read some comments in dpreview - pixel density is pixel density. no matter how you downsize megapixels total pixel density in the sensor remains the same, therefore image quality will not improve. detail/quality will be lost due to downsizing. taking a picture at full resolution and *not* peeping at 100% always produces good/acceptable results. every picture on flickr 1024x768 or smaller looks like it came out from DSLR.

    ReplyDelete
  6. ps - been there done that. used to have a 7mp canon, tried shooting at lower resolution, no improvement.

    ReplyDelete
  7. yes, that is the beauty with downsizing, everything looks better and the trick is to shoot in full resolution and then do some pp on the computer as I have shown in lot of samples.

    ReplyDelete
  8. or, we can shoot with ZS10, HX9V, HX100V, and now some new old canons but pretend they are high end SLR's with expensive lenses. No need to downsize. Just view at whatever resolution image is shown on tv/monitor and don't view "original" size. Thumbnails also look good in windows...lol

    at 640x480 i can't tell a POS kodak from a pro SLR. Downsizing is not always an option for some of us.

    i think all camera manufacturers' slogan should be "if we make it they will buy it" because they WILL buy it and that's the sad part.

    ReplyDelete
  9. yepp you are right there, but the mass of people just go for the numbers. How many MP, how big zoom and how cheap.

    With the NEX series and all the other slim cameras with a big sensor this is where the camera makers want the customer. Make the quality from small sensor P&S so bad and the small large sensor cameras so much tempting and P&S like then there is a whole new market they can open and sell all the nice addon´s like flash, lenses etc.

    ReplyDelete
  10. I found this site while researching the HX9V.

    I recently acquired a HX5V but now learn of the improvements of the HX9V.

    Excuse my camera tech ignorance but if the problem of the HX9V is the high MP, then why can't you simply adjust the image resolution setting to 10MP or lower?

    When saying to adjust it in the computer, is that referring to just reducing the physical size of the photo so it appears to be sharper?

    So why are they even bothering with the HX7V?

    ReplyDelete
  11. I tried that with the HX5 to shoot in a lower resolution but you get a better result shooting full res and then resizing on the computer. (see my HX5 blog for more info)
    the HX7 was made so there was something in the 299 price bracket.

    ReplyDelete
  12. Not important: 100% detail perfection, or should it be looked at as 16MP or 4MP. Important: how good the output is for your personal needs. If you post photos online at 1024x, any camera is fine. If you make small prints, HX9 is fine. If you make 8x10 prints, HX9 looks capable. If you need all of the 16MP resolution, for extreme crops or megaprints, you are shopping the wrong camera (and arguably the wrong segment of cameras altogether). I compared the HX9 shots to shots from a TX1, a T100, an H5, and an F717 - all resized to 5Mp (2560x). The end result? The F717 and H5 looked best, the HX9 just behind, the TX1 slightly better than the T100, but close enough between them all to be a non-issue for web gallery use and medium prints.

    ReplyDelete
  13. agree, for a good 16MP you need to go APS-C and have a capable prime lens. No zoom and definitely no no superzoom and if you want to do macros, you better have a macro lens. Been there done that had the whole bag full. Love the compromise with the much more capable HX5 and soon HX9V have with you every where and shoot whatever comes at you. Picture quality wise the HX whatever is OK with me.

    ReplyDelete
  14. Joe, do you do USM sharpening?

    ReplyDelete
  15. yes, Bernd - see: Sony HX7V / HX9V analyzing the 16MP sensor pictures - Part 5

    ReplyDelete
  16. please mail jo.cherry -at- gmail. id love to chat with you offline about using some of your material. thanks.

    ReplyDelete

If you want to talk to me, use the comments box at the end of the post. No need to register or log in. If possible choose a name for easy answering.