How to choose a microscope camera

There are a lot of ways to get images out of your microscope.

But not all are equal to give you good image quality.

Different methods are more or less expensive, but in the end, if you want an optimal image quality, very few choices are available.

Sensor size

One of the most important factor to consider is the size of the sensor.

A small sensor size will result in a cropped field of view.

In most cases, a 1” sensor is recommended.
Depending on the FN of the objectives (field number), a bigger sensor can be used.

Reduction lens

Some USB C-mount cameras come with a 0.5x reduction lens/eyepiece adapter, to mitigate the fact that they have a small sensor.

This may sound appealing, but will sadly lead to poor performances:

  • Blurred corners (that can’t even be fixed with focus stacking)
  • High level of chromatic aberrations, especially on the blue channel, even with high end apochromat objectives.
  • Slight distortions

Keep in mind: any new lens between the objective and the camera sensor will lead to reduced performances. In some cases this is necessary (objectives working only with specific photo-eyepieces), but most of the time you want to avoid that.

Sensor resolution

TLDR:

  • 1” color camera: 20MP is recommended
  • 1” grayscale camera: 5MP is enough
  • 1/2” color camera: 5MP or higher

The higher the magnification of an objective, the lower the resolution is needed.
→ Less details per area

A color camera have a bayer pattern. For this reason, you usually want 4x more pixels than with a grayscale camera.

How to determine the optimal pixel size

Determining microscope resolution

d = λ/(2×NA)

With:

  • NA: Numerical aperture of the objective
  • λ: Wavelength of the light
    • 450nm for blue
    • 550nm for green
    • 650nm for red

Determining pixel size

P = (d×m×Mo)/2

With:

  • P: Pixel size
  • m: Magnification multiplier (usually 1.0)
  • Mo: Objective magnification

Magnification multiplier

Using intermediate optics like a magnification changer or reduction lens will have an influence on the system magnification.

Examples

I will use the 550nm green color for the following examples.
→ Green is corrected for chromatic aberrations and spherical aberrations on achromat objectives.

100x/1.25 oil objective

d = λ/(2×NA) = 550/(2×1.25) = 220nm
P = (d×m×Mo)/2 = (220×1×100)/2 = 11µm

Here we can see any camera with 11µm (or smaller) pixels will be enough to resolve the details.

50x/0.80 objective

d = 550/(2×0.8) = 343.75nm
P = (344×1×50)/2 = ~8.6µm

20x/0.40 objective

d = 550/(2×0.4) = 687.5nm
P = (687.5×1×20)/2 = ~6.9µm

10x/0.25 objective

d = 550/(2×0.25) = 1.1µm
P = (1.1×1×10)/2 = 5.5µm

4x/0.1 objective

d = 550/(2×0.1) = 2.75µm
P = (2.75×1×4)/2 = 5.5µm

High N.A. objectives

We can notice the lower the magnification, the smaller the pixel size we need.

But this is even worse with high N.A. objectives, like apochromats.

Objective Resolution Pixel size
4x/0.20 1.4 µm 2.75 µm
10x/0.40 688 nm 3.44 µm
20x/0.80 344 nm 3.44 µm
60x/1.40 196 nm 5.88 µm

For a Nikon Plan Apo 4x/0.20, with a color camera, you want a pixel size of about 1.4µm to make the most of it.

For example, my E3ISPM20000KPA 1” 20MP color camera have a pixel size of 2.4µm. This is enough for the resolving power of this objective, but because of the debayer, some details might be lost.

On the other hand, my SC1803R 1/2” 18MP color camera have a pixel size of 1.2µm, which is much better for this objective, but with it I would lose a lot of the field of view.

This is something to keep in mind when buying high end apochromat objectives.

Automated captures

DSLR

When doing automated captures or/and focus stacking, avoid using a DSLR camera with a physical shutter.

Some panoramas, like this high magnification capture of an ATMEGA8 may require thousands of images.
This ATMGA8 was made of 43x41x47 = 82861 images.

A common issue with DSLR is the difficulty to automate captures.

For a good pipeline / capture system, you need at least an access to the video stream. This is usually possible with a DSLR but you may end up with much lower resolution, and the need to go through an HDMI capture card.

USB cameras

USB cameras are usually much more easier to work with, at least when the API / compatibility is not an issue.

Toupcam API

I’m mainly using cameras with a toupcam API. There is a high level of control, and I’m working directly with the RAW video stream to get the best of the camera, but I wouldn’t recommend for someone who is not comfortable with C/C++ programming. The software is proprietary, and this may be a barrier for some platforms. But reverse engineering the USB protocol seems feasible (it’s basically a RAW video stream) and the Linux support is good for my 3 toupcam cameras.

UVC

Cameras compatible with UVC are interesting, because they are cheaper and can be used through OpenCV easily, and thus Python scripting. The main issue is that the stream is most of the time compressed, and UVC cameras have small sensors.

USB3 Vision

Common on industrial cameras. Don’t recommended for someone who is not comfortable with C/C++ programming. High level of control through the API, but software is usually proprietary.

My only experience is the GO-5000M-USB Monochrome Camera that I got second hand for 200€. Compatibility under linux was a little painful, but the API is not too hard to use and this is a great monochrome 1” 5MP camera.

Raspberry Pi camera modules

Raspberry Pi HQ camera module

A good tradeoff if you don’t have a lot of money to spend on a camera is the Raspberry Pi HQ camera module. It can be mounted to C-Mount and the API can be used through Python on yout Raspberry Pi. Sadly, it’s only a 1/2” sensor.

OneInchEye v2.0

I didn’t had the opportunity to test one, but it was recommended by Zeptobars.
Probably one of the cheapest options for a 1” camera, especially if you already have a compatible RPi.

Please note that the OneInchEye is designed for Raspberry Pi Compute Module 4 boards with a 22-pin FPC connector with 4-lane MIPI interface and use the same pinout as the Raspberry Pi Compute Module 4 IO Board.

Afocal projection


Image by Paul Tadrous, PUMA Microscope


Not recommended when it can be avoided.

Necessary with old optics that requires special eyepieces and don’t have any dedicated photo-eyepieces (even if it seems, some eyepieces can be converted to photo-eyepieces).

Quality will be reduced because of the addition of glass between the objective and the sensor, but this may be mitigated by using an apochromat lens for the projection. A DSLR is probably a good choice for this kind of corner case.

I still never tried this method myself.

Note: usually, result will be the full image circle.

Honorable mention

For 1050nm IR imaging / high sensitivity, the Sony IMX585 is one of the best solutions.

It can be found in the Touptek Astronomy line with cooling.
The color camera with or without the IR Cut filter (AR glass recommended):
→ ToupTek ATR585C |ATR3CMOS08300KPA
There is also a monochrome version:
→ ToupTek G3CMOS08300KPA IMX585 Fan-cooling Camera

Important: the front glass is heated to prevent condensation, it should not be removed.

This camera was tested and recommended by 4e71 on the siliconprawn discord.