What is ISO and a Deeper Look at Your Image Sensor

(This is Lesson 7 of my “How It’s Done” series)

I was going to make this lesson a quick and breezy discussion of ISO.  From the many conversations I have with new DSLR owners, I’ve found that ISO is one of the most misunderstood concepts in digital photography.  It doesn’t have to be that way.  Understanding ISO is a pretty basic concept.  It’s a lot less confusing than other concepts we’ve already covered like aperture control and depth of field.  I could probably explain ISO in a paragraph or two, but that would be doing you a disservice.  The whole purpose behind this series of lessons is to give you enough photographic knowledge that you feel confident about your photographic decisions.  The last thing this world needs is another photographer that leaves his ISO permanently set at 100, or even worse, sets his ISO to auto and leaves it there for the life of the camera — yuck.  America, you can be smarter than that!  You have a real technological marvel living in the heart of your camera.  By the end of this lesson you will know how to push that Ferrari hidden in your camera to the edges of its performance envelope.  To get to that level we’re going to dip a little into the technicalities, but don’t worry, we won’t go so deep that you fall asleep.  At the end of this lesson you’ll know how ISO effects image quality, when you want to use a high ISO, and when you want to go low ISO.  Also, you’ll know enough basic image sensor theory that you’ll have confidence in your ISO decisions.

Let’s start by delving deeper than we have previously into the workings of your image sensor.

Your image sensor is an array of millions of light sensitive pixels.  When you were shopping for a camera undoubtedly mega pixels were a factor in your buying decision.  One mega pixel equals one million pixels.  Each pixel is a microscopic light sensor.  The more pixels packed onto an image sensor, the more detailed a picture that camera is able to produce.  My first digital camera was a 3.2 mega pixel Canon Powershot S30.  I couldn’t make prints larger than 8×10 inches from that ancient Powershot because of pixilation (Note: pixilation is when a picture starts to show the individual pixels that make up the image because it is printed too large.  It looks like a bunch of jagged squares).  Today’s DSLRs pack in so many pixels that you don’t’ have to worry about pixilation until you print your pictures larger than poster size.  Who really needs more pixel horse power than that?

A more important question than “how many mega pixels?” is “how does an image sensor work?”  Here’s the quick (all you really need to know as a photographer) explanation.

We’ll start with the smallest component of the sensor, the pixel, and work our way up.  Each one of your millions of pixels is a tiny hyper sensitive solar cell.  When light hits a pixel, the pixel converts the light energy into little charge of electricity.  The amount of electricity created is proportional to how bright the light is.  A brighter light means more electricity generated.

To me, the next part of image sensor technology  is even more amazing than the millions of tiny solar cells .  Attached to each pixel on your camera’s image sensor is an itsy-bitsy voltage meter.  Think of how amazing that is!  Millions of perfectly calibrated microscopic voltage meters living and working inside your camera.  (When I took high school electronics class we did a lot of voltage metering.  The voltage meters we used were the size of a shoebox and weighed five pounds.)  The voltage meters attached to your pixels each take a very precise measurement of the charge on each individual pixel. The voltage reading is then converted to a brightness value.  The brightness values are a range from 0 to 255.  For example, a pixel that has no measurable charge is assigned a brightness value zero value, while a pixel that is a midtone would be assigned  value of 128, and the brightest highlights in a photo would have a brightness value of 255.  (Does that sound familiar?  You might want to reread lesson # Understanding exposure one pixel at a time)

All this information generated by the image sensor would be useless if it weren’t organized in a specific way.  The millions and millions of pixels on your sensor are arranged in a very specific pattern called a Bayer array.  Each one of your pixels actually has a red, green, or blue color filter covering it.  In a Bayer array 50% of the pixels have green filters, 25% have red, and the remaining 25% have red filters.  Don’t ask me why green gets twice as much representation.  That’s a doctoral thesis above my level of understanding.  Just know that the array works, and make a note in the back of your mind that in photoshop the green channel has twice as much detail as the red and the blue channels.  Below is a picture of a Bayer Array lifted by permission from Wikipedia.

Each one of your pixels sees only in a single color.  Fortunately, red, green, and blue can be mixed to form all the colors of the rainbow.  Your camera knows the location of each pixel and what color filter is covering the pixel.  It’s the job of your camera’s image processor to place each pixel’s data, mix the colors, and create a recognizable picture.

Now that we have an understanding of all the parts , lets break down what happens in the split second your camera takes a picture.

Step 1 — The shutter opens and the image sensor is blasted with light

Step 2 — Each of the millions of pixels absorb light and transform the light energy into a tiny electrical charge (brighter light equals more electricity  — less light equals less electricity)

Step 3 — The electrical charge at each pixel is measured by the pixel’s own individual voltage meter and the measured voltage is translated into a brightness value on a scale of 0 to 255

Step 4 — The brightness value from each pixel is tagged with the pixel’s location on the Bayer Array and the information is sent to the image processing computer

Step 5 — The image processing computer sorts all this data creating a recognizable photo

Step 5 and ½ — If you didn’t turn off all the automatic junk we discussed in lesson one “Disillusionment is a Positive First Step”, this is when you camera’s imaging computer adds stuff like auto saturation, and auto sharpening.  These things are better dealt with by you, the photographer, in Photoshop, rather than in a hap hazard, uncontrolled way in the camera.  Turn that stuff off!  Give your camera and yourself a break.  Skip step 5 and ½.

Step 6 — The processing computer then sends the photo data to the camera’s memory card for storage.  The file type is either encoded as Jpeg, or RAW, or one of each.

You got it?  Is this all making sense?  Good.  We’re almost back to our main topic of discussion, ISO, but first we have to cover a couple quick topics.  We have to talk about how your image sensor screws up.

The bane of  digital image sensors has always been shadow detail and digital noise.  Digital sensors have, in their short history, never been very good at recording details in the darkest areas of photos.  Things are getting better.  Newer sensors can see better in dark areas than could sensors of just a few years ago.  However, engineers are far from perfecting digital sensor’s shadow performance.  Ironically, film was the opposite.  Color film was great at shadow detail and poor at recording highlights.  Color film photographers were able to work around this limitation in their medium just as digital photographers make do with the limitations of their equipment.  The key is to understand the problem first.

There are two reasons why digital camera’s have a tough time with shadows.  The first is simply the physical limitations of a pixels sensing ability.  Remember, a pixel is basically a solar cell with a volt meter attached.  It’s hard for a pixel to sense light where there isn’t much light to begin with.  The tiny electrical charge created by a pixel that’s pointed at a shadow might be too small to be metered .  Or, the charge could be so small that the camera’s computer has to throw away the data because it’s not above a certain threshold.  For advanced readers, think of the blacks slider on Photoshops RAW converter.  Anything that falls below the black threshold is rendered as black.  You can roll back the black threshold to get more shadow detail, but you also get a lot more noise.  The noise is random pixel readings.  When things are this dark,  a pixel’s voltage meter is reading tiny charges.  At this low range your pixels might be measuring a charge created by the heat of the camera rather than a charge created by light striking the sensor.  You’re at the limits of your sensor’s abilities.  It can’t tell the difference between a good and bad measurement at this range.

Where shadows are concerned, digital cameras have another factor working against them.  Remember the 0 to 255 brightness scale discussed earlier?  Well, I neglected to tell you one important fact about that scale.  The scale is nonlinear.  (Holy crap Batman!  Nonlinear!  That smacks of High school algebra!)  Don’t despair dear readers.  I’m not going to graph some indecipherable curves in my quest to teach you how your sensor works.  Just understand that when your camera records shadow values the brightness levels are a lot wider than in the highlights.  In the 0 to 255 brightness scale a step from a value of 2 up to 3 is a much bigger step than a step between the values 128 and 129.  Because of this method of measurement, your image sensor is simply more sensitive to the highlights than it is the shadows.

To better illustrate the problems recording shadow details I posted below a snapshot of my dogs.  This photo was taken at ISO 640.  For my old Canon 5D that ISO is starting to push into the realm of noise and poor shadow detail.

Notice that this web size (950 pixels wide) photo looks OK.  Jack, my black dog, looks pretty sharp and his fur looks fairly well rendered. Now look below at the two 100% crops of Jack’s face.

The image on the left is Jack’s face as it was displayed in the full sized photo.  It doesn’t look terrible.  However, I want you to take a look at the areas of darkest shadow.  Notice how there isn’t a very smooth transition from dark black to slightly lighter gray.  That’s because of the big steps between lightness levels that we talked about earlier.  Also, there is a touch of noise in the dark spots.  That’s from the camera making incorrect readings at the lower end of its sensing abilities.  Take a look at the right image now.  In this image I opened up the blacks a little in the Photoshop’s RAW converter.   I moved the blacks threshold slider (those super imposed sliders at the bottom of each picture) from 5 to 2.   That move tells photoshop I want to display data that is even deeper in the shadows.  Don’t worry if you don’t understand that photoshop stuff yet.  What I am trying to do here is demonstrate how thin the shadow detail is in the dark areas.  In the right image take a look at how grainy and pixelated the eyes look, and also take a look at the messy shadow in the lower right corner.

To contrast with those messy shadows in Jack’s face, I posted below a 100% crop from the other dog’s neck and head.  Gatsby’s fur is a nice midtone and my camera does a good job capturing it, even at this high ISO setting.

Take a look at that nice sharply rendered red collar.  It looks great.  It’s a midtone.  It should look great.  The fur around it looks good too.  Now, look at the inside of Gatsby’s ear.  Uh Oh, here comes the bad shadow detail again!  It’s the same down near his dark eye.  Granted, those areas are slightly out of focus, but the camera still is putting noise in the dark spots regardless of focus.

So what exactly am I trying to tell you?  You aren’t ever going to be satisfied with pictures of your black dog?  Unless your picture is perfect looking at 100% size it’s a total failure?  No!!  All I am trying to do is point out some limitations at the extreme edge of your sensor’s ability.  Now that you understand the limits of your sensor, you can work to overcome them.  Maybe, when taking pictures of your black lab you can try some fill flash, or maybe you can experiment with a slight overexposure and then push details backwards into the shadows via Photoshop.  Heck, you could even say, “That slight shadow noise doesn’t even look bad.  Only a perfectionist would be bothered by it.  Screw it!  I’m gonna go out and take some pictures.”  All those are correct answers.

OK.  Now, we can finally circle back to ISO.

Your camera’s ISO setting often (sort of incorrectly as we’ll discuss later) defined as how sensitive your image sensor is to light.  An ISO of 100 is less sensitive than ISO 200.  Each 100 doubling of ISO value is equal to one stop of aperture or one stop of shutter speed.  For example, let’s say you are shooting a football game on a cloudy day.  You have a really expensive lens and you are shooting at f/2.8.  Since it’s cloudy you notice that your shutter speed is only 1/60th second.  Your photos aren’t coming out very crisp.  The action on the field is just too fast for that slow shutter speed.  Being a smart photographer you decide you want a faster shutter speed.  Your lens is already dialed down to its fastest aperture.  What else can you do?  You turn up the ISO.  Let’s say you dial up the ISO to 400.  That’s two full stops faster than ISO 100 (going from 100 to 200 is one stop, from 200 to 400 is another stop).  Now, with the same lighting you get camera settings f/2.8 and 1/250th.  That’s fast enough for this action.

Here’s the big catch when it comes to ISO, you are not actually turning up the sensitivity of the image sensor.  You are turning up the gain.  There is a big difference.  With a high ISO your image sensor doesn’t miraculously find a better way to see.  What happens as you raise ISO is the million little volt meters we talked about earlier add some amplification to their signals.  This trick of electronics does a great job brightening midtones and highlights, but it doesn’t do such a stunning job with shadow detail.  Remember, your image sensor already has some problems when dealing with shadows.  Adding amplification doesn’t always solve those problems.  A lot of times it actually amplifies them.  That’s where you can get a lot of digital noise — using a high ISO and taking photos of dark objects or deep shadows.

Don’t let that last paragraph persuade you away from shooting at high ISO settings.  That’s the last thing I want to do.  Too many photographers are afraid to bump up their ISO for fear of a little noise in their images.   Modern DSLRs do a great job suppressing noise at high ISO.  Also, Photoshop keeps coming up with better and better noise suppression software.  The combination of these two technologies makes high ISO photography better today than it ever was.  I think there are a lot of photographers out there who had bad high ISO experiences with older DSLRs and now they never try high ISO with their newer cameras.  It’s a shame because the new cameras do a great job.  My suggestion is to try lots of different ISO settings with your camera and determine where the noise level becomes unacceptably high.  Keep in mind how big you will be displaying your images.  If the biggest you’ll print your picture is 8×10, you can go hog wild with ISO before any noticeable noise appears.  If you’re printing a poster you’ll have to be more conservative.

Just like in previous lessons I’m going to give you a list of ISO settings and possible uses.  Of all my lists, this one needs to be taken with the biggest grain of salt.  These settings are based on my Canon 5D Classic.  That camera design is old, but it’s also a full frame camera with big pixels.   The sensor size might give some of  your newer APS-C sized cameras a run for their money.  Or, your  new camera might have a better ISO response than mine.  Keep that in mind as you read this list.  I also included some photo examples to spruce up this otherwise blah looking article.

ISO 50 — This is the setting for long exposures when shooting on a tripod.  It eliminates noise when the shutter is open more than several seconds

This photo of downtown Seattle is an ISO 50 photo.  My camera was mounted on a tripod and the shutter was open 30 seconds.  Look at the flag in the bottom of the frame to get an idea of how long the exposure was.  I had almost no digital noise in this photo.  ISO 50 does a great job when taking pictures at the edge of dark.

ISO 100 — I use ISO 100 for 80% of my shots.  As long as the shutter isn’t open more than a few seconds, ISO 100 gives virtually no image noise.

This photo of a Wisconsin farm in the dead of winter was taken at ISO 100.  There was plenty of sun and snow so shutter speed wasn’t a concern.  In fact, I was so cold while taking this photo that I probably didn’t even think about ISO.  It was zero degrees and the wind was whipping.  Luckily, my camera lives at ISO 100.  If it’s not set at ISO 100 when I turn it off there’s a good chance that the next time I use my camera I’ll be shooting in the wrong ISO for 10 minutes or so.

ISO 400 — This is my setting for action shots.  If I am shooting something with lots of motion I’ll bump up the ISO to 400.  At this setting I get almost no noise as long as the shutter is open less than 1/250 second.  That makes ISO 400 perfect for chasing action.  I also use this setting if the light is low, and I don’t have a tripod.  At slower shutter speeds I start to notice a bit of noise in the shadows, but this can usually be cleaned up with some creative noise reduction in Photoshop.

This photo of Ella riding the General Lee at the State Fair was taken at ISO 400.  It was sunny and bright, but I wanted to be sure that I had enough shutter speed to stop the action.  I discuss this photo in more detail here.

ISO 3200 — This is my last resort setting.  If I am hand holding the camera and already way down at f/2.8, sometimes ISO 3200 is the only way to go.  The noise is pretty bad, but it can be cleaned up enough to make a decent web sized photo.

Here is a shot taken at ISO 3200, f/2.8, and 1/40th second.  I could have used flash for this and got a crisper picture, but I really wanted the fire to do all the lighting.  I cleaned up the noise in Photoshop.  The picture could have been sharper.  I tried to focus on Matt’s head, but the auto focus grabbed the shoe hanging in the right side of the frame instead.  Night photography is hard.

Well, that’s about all I have to say about ISO and image sensors.  For further reading and a little ISO pep talk let me recommend this article I wrote last winter.

Continue to the next lesson

Return to “How It’s Done“