Camera BASICS

Introduction

All cameras are different (duuh) and there are so many parameters to look for. A very common mistake is believing that higher resolution and more expensive is always better!  Short story, this is not the case and could in many cases actually be the complete opposite. If you really want to know what is best for you, keep on reading. 

Summary:

Expensive is probably not best (for you).

Very High res cameras also have many disadvantages, which over all generally makes them bad.

Pro tip:

Consider a camera in the 3-5 Mpix range, unless you have very specific (low magnification) requirements.  

CAMERA Resolution

The greater resolution the more details you can see!
 

This is all great and the with todays computers and the price of cameras, we strongly recommend to use at least 1.2 Mpix cameras preferably more if you match the rest of your system accordingly.

But don't over do it! Today's high end cameras have higher resolution than most desktop monitors can display. Some cameras even have higher resolution than your microscope can resolve. We think it makes very little sense having a camera with such a high resolution that you can't see it, nor that your microscope can resolve!

 

Be aware, high resolution always comes with a number of drawbacks. So don't fall for for the "high resolution" pit fall. It is likely to give you more drawbacks than advantages.

Summary:
Higher resolution <=> low sensitivity.
Higher resolution <=> more lag and generally slower system.
Higher resolution <=> requires an expensive computer.

Extremely high resolution <=> cameras with big sensor <=> expensive optics <=> extreme computer.

Pro tip:

Make sure your monitor have at least 30% higher resolution than your camera.
For example, if you have a 5 Mpix (~2500 x 2000 pixels) camera, you should have a 3840 x 2160 monitor.
Or if you have a camera with 1280 x 1024 pixel resolution use a monitor with 1600 x 1200 pixels.

Pro tip:

If you have a 4K monitor and a 5 MPix camera (or more), you really need a powerful computer with a dedicated GPU Graphic board.  If you don't, you will just have a slow, lagging and annoying system. If you don't have the money for this hardware, you are actually better of with a cheaper camera.

Optical RESOLUTION

The optical resolution determines the how small parts you can see. This is pure physics and can not be worked around, not even with the best microscope in the world. The optical resolution is determined by the wavelength of light and the Numerical Aperture (NA) of the optics. The wavelength of visible light is roughly between 400-700 nm, but as an average when we do our computations, we use green at 550 nm = 0,55 um.  

The Numerical Aperture (NA) is always printed on your objective, often next to the magnification. For normal optics it can never be higher than 1.0, but for oil immersion optics it can be up to 1.25


 


 

 

The max resolution on the camera sensor is equal to optical resolution times the total magnification.


An example:
On a microscope with a Plan Achromat 20x lens with NA=0.40 and a 0,5x camera adapter, the resolution on the chip is 8,39 um. The sampling theory says the pixel size should be half of that, equals 4,19 um. Today, a common pixel size is 3.45 um, which is smaller than the optics can resolve. 

Given a sensor size of 1/1.8" (equals a chip size of 7,2 x 5,4 mm) this gives us the theoretical maximum resolution of roughly 1700 x 1300. See bottom of this page for tip on how to compute optical resolution and suitable chip resolution for your system.

The above calculations is for a monochrome camera. A color camera have less "true" resolution, since four pixels are combined in a so called Bayer pattern to compute red, green and blue. So this requires to roughly double the resolution in both direction, ending up with a recommended max chip resolution of approximately 2800 x 2000 pixels (~5 Mpix).

For higher magnification (40x, 60x and 100x), the demands on the camera becomes even lower, because the magnification increases more than the NA. The opposite is also true, for lower magnifications (10x, 5x etc) you will need higher chip resolution to match the optical resolution.

The resolution can be further improved by using oil-immersion objectives, with gives you a higher NA, but since this only exist for high magnifications it does not mean any practical difference when choosing camera.

Summary:

So as a rule of thumb, the resolution on your camera probably does not need to go beyond 5 Mpix.
A camera with a resolution higher than this is for most microscopes and applications rather pointless, the exception would be if you have very high requirements on low magnification images, in other words, where you need BOTH a big field of view and very high resolution. 

Objectives_with_NA.png
Spectrum.png

Frame Rate AND LAG

Having a high frame rate (fps) might not sound important, after all, we are looking at non moving samples. But this is utterly wrong, we do move and focus our sample very quickly.

Imaging driving a car where there is a half second delay between your movement of the steering wheel and the actual steering. It would be impossible to steer the car. At high speeds you would most likely overcompensate for the lack of response, resulting in a very uncomfortable or even catastrophic ride!

So to be more precise, the important factor is not the frame rate, but the lag. Since lag and fps are somewhat related, fps is important.

On top of the above, with increased frame rate, all auto-functions such as Auto white and Auto exposure will works much faster. There are simply more images to analyze. Double the frame rate means the cameras auto features operates at double speed.

There are three main factors that determines the lag:

  • The resolution of your camera 

  • The performance of your computer

  • The performance of your camera.

For an in-depth explanation please see the chapter "Understanding lag" bellow.

Pro tip:

In the old days, one use to say that 25 fps is enough. This true for old analog camera with virtually no lag (well, some things where better in the old days), but for digital cameras, refresh rate under 25 fps it gives us an unacceptable lag!
For reference, g
aming enthusiast uses up to 200 fps today (which requires special monitors), this is purely because the want to minimize the lag/time between an enemy shows up "in memory" and this enemy is displayed on screen.


We do not need that. Aim for at least 30-40 fps but even better 50 fps, this will dramatically improve how much you will enjoy your ride. 
 

Make sure you have the proper hardware to accommodate these settings without making your computer scream! For 3 Mpix and above, make sure you have a high quality USB-3 board in your system and invest in a proper GPU based graphics card. The built in USB-3 port and built in standard graphics card on your ordinary desktop computer will not be sufficient.

Also remember to never connect the camera to a USB-hub, nor to connect the monitor via a HUB/Extender.

Check your CPU and GPU consumption, it should not go beyond 70%.

Chip Size/Pixel Size

The size of your chip (or sensor) should match the optics of your microscope. If the chip is to small, you will only see a fraction of the image you see in the eye-pieces and this is quite annoying. If you have a to big chip for your optics, the edges will become dark and unevenly illuminated. Always check with your microscopy supplier for the recommended camera chip size for your particular optics (or vice versa).

Here is a short list that at least give you a hint about right adapter magnification, given you have a 10x eye-piece.

 

 

 

 

If you have the possibility to chose both adapter and camera, and have the money, choose the camera with the biggest sensor (given it gives you bigger pixels) and buy an adapter designed for this chip size.

 

Bigger chip size (with same resolution) means bigger pixels. Strive for the camera with biggest pixels.
Bigger pixels means better light sensitivity => shorter exposure time => better dynamic range (that is, the cameras ability to see both dark and bright spots at the same time). All these factors are normally important in metallurgical applications. For color cameras, bigger pixel size also generally means better color representation.
 

 

Think of a sensor as small glasses placed side by side, where every glass represent a pixel. By measuring falling water (water in this case representing light), we have built a water measure array. With very small glasses (pixels), the spread/resolution of the water is of course better, but on the downside, some of the glasses (or all) are more likely to flood when the water start pouring (bright illumination). This would be equal to over-exposure. 

 

 

 

 

 

 

 

 

 

 


 

Also observe, big chip and matching optics often cost more money.

The above logic also gives you a reason why you should not buy a camera with unnecessary high resolution. High resolution means many pixels cramped into the same sensor size, which means smaller pixels. This is not good for image quality!

Pro-tip
The optimal camera can't be found. Improving one parameter often comes with other drawbacks. So our advice is to avoid the extremes, they are often built for a very special purpose. It's likely you will dislike the disadvantages more than you appreciate he advantages.

ChipSizeToCamaraAdapterConversion.png
glasses.png

TRANSFER TECHNOLOGY

Most cameras support one of the USB standards.
USB 2 must be considered outdated by today's means and should not be used.

USB 3.0 which, by the way, is the same as USB 3.1 Gen 1 and USB SuperSpeed (Yes, someone failed at the marketing department), supports up to theoretical speed of 5 GBits. This allows for very high frame rates. For example a standard 3 Mpix camera often support up to 50 frames per second (fps). 

However, please note that in reality most standard, built in USB 3.0 ports can often not handle speeds higher the 2Gbits. This can simply be explained by the connection between the USB controller and the host computers often only supports 2GBit. This is less than half of the potential.

Some camera manufactures are now staring to support USB 3.1 Gen 2 standard, with transfer rates up to double of the speed of USB 3.0/3.1 Gen 1 (10 Gbit). But please observe, few computer support USB 3.1 Gen 2. You most likely need to invest in a high quality USB 3.1 Gen 2 that uses high speed PCIe ver 4-8/16 connection (again, PCI boards less than this does not support these high speed) to take advantage of this.


Pro tip:
If you purchase a camera with very high frame rate/transfer rate, make sure you have a high end USB port on your computer. Adding a external USB 3.1 card with PCIe ver 4, with 8 or 16 lanes is the only way actually achieve this speed. Low cost USB extension board does NOT support this speed.

If you buy a USB 3.1 Gen 2 camera, you really must make sure you both have a compatible USB-3.1 Gen 2 port AND a computer that can work at this speed.

Computer performance

With the extreme bandwidth of today's cameras, especially the high resolution cameras, enormous amount of computing power is required. For every images processed a lot of computations is performed. Just the raw internal RGB data stream of a 5 Mpix camera is  about 500 million bytes per second. On top of this, every pixel is processed, such as "Bayer" to color computation, flipping, mirroring and color processing. Eventually the image is sent to the Graphics board, where descaling and image presentation/interpolation is computed.  All-in-all at least 10 billion (10 000 000 000) computations is performed every second! And this is just showing the live video!

So, every part of your computer system must be design to allow for this. If not, the image will randomly freeze, you will see missing frames or lag. The auto-white/auto-exposure will also work poorly and you can forget about starting other applications at the same time. 

Don't underestimate the importance of this! Above we have provided you with suitable hardware for a few cameras. 

Pro tip:

Invest in a GPU-based graphics board. This will offload the CPU dramatically and will increase your fps, minimize the lag and just add to the overall feeling. You don't need a top of the line ultra modern gaming graphics board, as long it has a GPU.

Monitor and monitor Resolution

Having a live image with higher resolution than your monitor is generally a bad idea. This either forces you to "downscale" the live image (and the high resolution is lost!) or you will only see a fraction of the image (compared to what you see in the eye-pieces), which is very inconvenient.

So make sure you monitor support the resolution of your camera + another 30% to give space for menus etc.

Also, make sure you always connect your monitor directly to a PC with a HDMI or Display Port adapter. Never use a

Hub (often used with laptops), they simply do not support the refresh rate you want and also puts additional demands on the CPU/Hardware. 

Computing the MAX camera resolution

This is an advanced chapter for the super interested. Here you will learn how to compute the highest theoretical chip resolution for your particular system.

Optical resolution:

The optical resolution determines how close small objects can be each other, and still be seen or detected as separate objects. We think Nikon demonstrates this well in this article.
 

Resolution r = 1,22λ/(2NA)    where λ = 0,55 um         (there are variants of this theoretical equation, not covered here!) 

Here is a list of typical NA (and corresponding resolution) for common objectives

Table_NA.png

Resolution/Pixel size at the chip:

This is simply calculated as:

R = r * M

Where M is the total magnification in the system, including camera adapter.
 

Given the 20x Plan Achromat above (NA=0.40) and a camera adapter of 0,5x the total magnification is 10x.

R =  (1,22*0,55)/(2*0,40)*10 = 8,38 um

The sampling theorem says the resolution on chip should be the double, that is 4,19 um.

In other words, having a pixel size smaller than 4,19 just give you "Empty magnification".
 

Most camera supplier specify the pixel size.
For example, the Alvium 1800 U-319 (3 Mpix) have 3,45 um pixel size, which would be sufficient to resolve all details for this setup.

Given the nature of how color cameras work, the true color resolution is a bit lower. 
Dividing by a value between 1,5 and 2 again will give you a good approximation. 
So, for a color camera, the perfect match might be a pixel size of 1,8 um

Also observe:

      High quality optics <=> requires high quality camera.

      Low magnification <=> requires high quality camera

In other words, if you work mostly in high magnifications (40x and higher), the requirements on the camera is quite moderate, even if you have high quality optics 

Camera Chip Resolution:

The camera resolution is computed by taking the sensor size (width and height) divided by pixel size

A conversion between typically used chip sizes can be found here:

Sensor_siez_in_mm.png

For example: 

With the above example, and a chip of size 1/1,8" the size of the chip in um is 7160 x 5370 um (~7,2 x 5,4 mm). Divided by the pixel size 3,45, it gives us a the max resolution of 1709 x 1282.

Again, for a color camera this can be doubled ending up in ~3200 x 2400 pixels.

Summary:

The above calculation will give you an approximate theoretical maximum resolution. In the reality this resolution is never achieved, affected by a number of  factors, such as cleanliness, the microscope is not perfectly calibrated/aligned according to Köhler, dirt and unevenness on the sample and much more.

Also observe, resolving all the details in low magnifications demands more from the camera compared to high magnification, which might be a little contrary to what you might think.

For example, using the top of the line 100x Plan Apocromat lens, with oil immersion,  give you a pixel size of:

   
w = 1,22*0,55/(2*1,4)*100*0,5 = 11,5 um    => a sensor with a resolution of 1200 x 1024  would be sufficient to resolve the details. 

 

Understanding LAG

A digital camera processes one image at a time, that is, the image is captured in it's entirety and then sent to the computer over a serial connection (USB-3). After this, a number of processing steps takes place. In short the complete process looks like this:

 

 

ImageProcess.png

Some of these steps involve transfer of data between hardware components (Camera->USB->Memory->Graphics card), some quite intensive computation and one step is just waiting for the monitor to be ready for a new image.

Depending on the size of the image and the speed of your computer, these steps takes time. As a base line, each steps takes roughly between 20-30 ms, given a 3 Mpix camera/image on an average computer. In total, this adds up to roughly 0,25 seconds. This is, if you think about it, quite amazing. The number of operations/computations for one 3 Mpix image is at least 300 million. Given 25 fps, you add up in 7.5 billion operations per second, just to watch live video!

Also bare in mind, that the system is able to process "a train" of images at the same time, for instance the transfer of one image from the camera to memory is done simultaneously as the previous image is computed, and yet another image is processed in the graphics card.  

With this being said, it starts to be quite obvious that your computer performance and the image size of your camera can affect this lag quite severely.