Zephyr: DIY low-cost fluorescence scanner

Zephyr is a low-cost Do It Yourself (DIY) machine which can scan petridishes and 96 well plates for expression of fluorescence at micrometer scale. The Typhoon is the commercial machine that does the same, only it is priced around 120.000 dollars. The main difference is the use of low-cost optics. This allows you to pick exactly which fluorescence you want to detect and not to pay for the ones you do not use. Furthermore, it does not have confocal optics, as this is not that often when scanning bacteria and protein gels. This DIY machine can be built by anyone with one or two days on their hands and the costs are around 1500 dollars.

The machine is built from a plastic frame, machined by laser-cutting. This is a widely available technique and can be done by many companies. The resulting parts can be assembled like a puzzle, clicking the parts together, making it accessible. The petridishes/gels/plates are moved on a 2D table under an optical tube resembling a fluorescent microscope. By taking images one after another and combining them with the supplied stitching software a high resolution image of the entire object is obtained.

Why? Reason d’être

Research is not cheap and synthetic biology is no exception. Much of the lab equipment has a running price of ten thousand dollars. For some teams this is no hurdle, their lab has all the equipment they possibly may need, while other teams may struggle with their characterization because of lack of needed equipment. This may be an explanation why in the iGEM competition certain regions/continents (e.g. Africa and Latin America) have few teams and this over the past recent years. [1][2] In our view, being able to participate in the iGEM competition should be accessible to everyone and the cost of equipment should not come to hinder creativity all over the world.

For most of the mentioned equipment, only the high tech versions are available, which makes it so costly. However the simple versions of these machines would be enough to carry the work an iGEM team has to do. For instance, we would like to draw a parallel: there are only high tech Bentleys available and no Ford Fiestas, while Fiestas would be enough for simple transportation. This is why we decided to build a low-cost Typhoon, which would be easy to make on your own. This machine is of course not as high-tech as the Typhoon, but it measures at the same scale and has roughly the same performance.

Affordable lab tools for everyone is primordial to making synthetic biology open, accessible, and innovative. As part of our Human Practice endeavor, we wanted to try to make one of these essential tools affordable in order to allow more teams to participate in iGEM in the future. We believe the more teams can participate the more we will all be able to share and build together on new ideas.In the next sections we show the working principle, how to build the Zephyr, the explanation of the design, the results, discussion and conclusion.

What? Working principle

The Zephyr is shown in Figure 2, where the optical parts are in the black tube. In Figure 1 this optical set-up is schematically shown. In this figure the fluorescent object (e.g. cell with GFP) is at the bottom and excited with a LED through an excitation filter. The emitted fluorescence passes through the objective, dichroic mirror, emission filter and eyepiece to be detected using a webcam.
In Figure 2, the excitation is seen as the small blue spot on the 2D table. The 2D table contains a petridish that can be moved around to image the entire plate.
This movement is shown in Figure 3, you move in a snake wise manner around the entire plate to make an image of it. Note that this is not a continuous motion, but a step-wise one. So, the plate is given a small displacement, it is stopped allowing the webcam to take a sharp image and then moved again to take the next image.

Figure 1: Schematic of the optical set-up;              Figure 2: Foto of the motion of the Zephyr
Figure 3: Schematic of the scanning motion

How? The Zephyr DIY guide

How to make the Zephyr can be broken down in different modules: first the buying of materials and parts, then the making of several parts, assembling them, wiring the electronic circuit, programming the microprocessor, controlling the set-up from the pc and calibrating the image stitching to make a complete image. The explanation on this is for readability on this separate page.

Explanation of the design

The design of the Zephyr is of course not thoughtless one, many considerations took place. Several of them are listed below, as to explain the design:

  • As a machining technology the laser cutting is chosen, because it is a technique that is widely available and can be done by a company for a reasonable price (around 150 euros). This way the user does not have to have experience in milling or similar techniques. The same goes for the assembly, by using a click-wise assembly the accessibility of it is high.
  • The frame material is chosen as PMMA, because it is one of the plastic materials that yields the highest accuracy with laser cutting.
  • For excitation of the cells a high power LED is chosen over a laser. A laser would give a higher power excitation, but the LED is enough to excite colonies. Since the LED is an order 10 to 100 cheaper it is chosen.
  • In assembly A, the dichroic holder, next to a dichroic mirror a excitation and emission filter are used. This is common practice in fluorescence microscopy [4], since the selectivity and sensitivity of the measurement go up: the cells are excited with a more precise wavelength and a narrower region of wavelength is measured.
  • The maximum dimension of the object to be scanned are 140 cm by 140 cm. This is big enough for many protein/DNA gels, petridishes and can fit a 96 well plate.
  • For actuation stepper motor are chosen as actuation, as they provide relative accurate displacement. The disadvantage is that it uses a lot of power, even when the motor is not moving. Note that 4 motors instead of 2, which would be enough to actuate in two directions. In a first prototype 2 motors were used, however this made the displacement of the 2D table wobbling.
  • No displacement sensor was used, to improve the accuracy an accelerometer was tested. However this accelerometer in combination with Arduino could not provide high frequency measurements. This had as a result that the measurements did not correlate with the discontinuous displacement.


To test the performance of the Zephyr, three experiments were performed. The first experiment is to test the selectivity: how is an E. coli colony with constitutive GFP expression seen with respect to a colony without GFP and a colony with constitutive RFP expression?

The second experiment is to test the sensitivity: what levels of fluorescence can be detected. This is done with a nucleic acid stain at different concentrations. Finally, a part of a plate with E. coli colonies with constitutive GFP expression is imaged to test the scanning capabilities of the Zephyr.


How selective is the imaging, do you see much background at objects other than GFP? To test this we made a plate as in Figure 12, which is divided into three partitions: one with E. coli with constitutive GFP expression, on E. coli no GFP expression and one with E. coli constitutive RFP expression. The resulting images are also shown in Figure 4 (the black boxes). The GFP picture was unfortunately somewhat out of focus, but the bright shot is the GFP being detected. The two dark pictures have no detection at all.

Figure 4: Image of three partitions: E. coli with constitutive GFP expression, E. coli with constitutive no GFP expression, E. coli with constitutive RFP expression and there the images taken by the Zephyr in the black boxes.

Sensitivity of imaging

To test the sensitivity of the Zephyr, the YOYO1 dye is used. This is a nucleic acid stain that shows fluorescence in the presence of DNA. [3]. This stain shows fluorescence at 510nm, very similar to GFP. This way we use different concentrations of this stain to characterize the sensitivity of the Zephyr to detect fluorescence. The dye is recommended to use at 100 nM. Thus a range of dilutions is made from 2µM to 10nM in water. To all these solutions 500ng of DNA was added. As a control, the 500ng of DNA diluted in water is used.

All these solutions were then scanned by both the Typhoon and the Zephyr, leading to the results of Figure 5. In these bright spots are the fluorescence being detected.

Figure 5: Image of different concentrations of YOYO1 dye, the one on the left being most concentrated. The ‘DNA’ is the control without fluorescent dye. The top lane are the results of the Typhoon and the bottom lane are the results of the Zephyr.

Petridish reading

As explained in the ‘How?’ section, for petridish reading first a calibration must be done. This is done using the calibration text of Figure 6. Using this text (and without the assembly A present), 25 rows of 25 images are scanned. The calibration software finds the displacements between them and first stitches the individual rows together as in Figure 7. Pasting all the individual rows together is done in Figure 8.
Now that the pattern of the displacements is found through this text calibration a part of a plate containing E. coli colonies with constitutive GFP expression, Figure 9, is scanned. The resulting image of this scanning is in Figure 10.

Figure 6: Example of calibration text on the 2D table of the Zephyr, with a 5 eurocent coin as reference.
Figure 7: Row of calibration text stitched together (25 individual pictures)

Figure 8: Rows of calibration text stitched together (25 rows of 25 pictures: 625 pictures)

Figure 9: The part of the plate with E. coli colonies with constitutive GFP expression which is scanned

Figure 10: The resulting image of the scanning of Figure 9.

Video 1: Impression of the Zephyr scanning the calibration text.
The Zephyr was displayed on the Discovery Festival to the public, good discussions took place on the mechanism and structure.


The selectivity of the Zephyr is good, you only see the expression of GFP, and other fluorescent proteins like RFP do not seem to influence the retrieved image. This is to be expected, since the chosen filters and dichroic mirror are of high quality and are also used in fluorescent microscopy.

The sensitivity of the system is equally good as the Typhoon. It detects the same concentrations and does not detect the same concentrations. Thus the sensitivity of the imaging of the Zephyr is good.

The petridish reading is a difficult task and this becomes clear from the pictures. The text stitched together from 625 individual images is not very successful: the rows do not fully overlap. This is a result of the relative measure: the pictures only have information of their displacement with respect to their neighbor, thus errors will add up over a large amount of pictures. In Figure 10, some colonies are clearly visible, however the alignment is not perfect which is due to the same reason. Adding a sensor to help get a better alignment would greatly improve the performance on this scanning. For more on this, see the section ‘future aspirations’.


The selectivity of the Zephyr is good; it is seeing one type of fluorescence protein at a time. The sensitivity is alright, it can still see a low amount of fluorescence (10 nM YOYO dye). The petridish scanning is working, however the images are not stitched together very well, which results in a fuzzy image.

Future aspirations

Based on these results some improvements can be thought of. The most important one is to improve the petridish reading. This requires a better detection of the displacement. As discussed in the section ‘Explanation of the design choices’, the relative sensing with an accelerometer did not work, probably due to the discontinuous short movements. An alternative to this is using an absolute measurement using a grid. This is schematically shown in Figure 11, where an extra camera is added under the 2D table. This is capturing the grid, so from these images you can find the absolute positions. In this way the image retrieved is more precise.
The image stitching in Matlab works well, only it is relatively slow. This is probably due to the recurrent use of the function nanmean, which is not numerical optimal. Thus an improvement to speed up the image stitching would be to make optimize this function numerically.

Figure 11: Schematic of the improvement of the absolute displacement measurement using a second camera.


  1. “Teams Registered for iGEM 2012”,[Online]. Available From: viewed on 1 Oct. 2013.
  2. “Teams Registered for iGEM 2013”,[Online]. Available From: viewed on 1 Oct. 2013.
  3. Molecular Probes “Dimeric Cyanine Nucleic Acid Stains” at Life Technologies Manuals, Jan-2000
  4. K.R. Spring, "Introduction to Fluorescence Microscopy", [Online]. Available From: viewed on 2 Oct. 2013