How to read an MTF lens curve

We recently published a TechBrief “What is MTF?” to our Knowledge Base. It provides an overview of the Modulation Transfer Function, also called the Optical Transfer Function, and why MTF provides an important measure of lens performance. That’s particularly useful when comparing lenses from different manufacturers – or even lenses from different product families by the same manufacturer. With that TechBrief as the appetizer course, let’s dig in a little deeper and look at how to read an MTF lens curve. They can look a little intimidating at first glance, but we’ll walk you through it and take the mystery out of it.

Figure A. Both images created with lenses nominally for similar pixel sizes and resolution – which would you rather have in your application?

Test charts cluster alternating black and white strips, or “line pairs”, from coarse to fine gradations, varying “spatial frequency”, measured in lines / mm, in object space. The lens, besides mapping object space onto the much smaller sensor space, must get the geometry right in terms of correlating each x,y point to the corresponding position on the sensor, to the best of the lens’ resolving capacity. Furthermore, one wants at least two pixels, preferably 3 or more, to span any “contrast edge” of a feature that must be identified.

So one has to know the field of view (FOV), the sensor size, the pixel pitch, the feature characteristics, and the imaging goals, to determine optical requirements. For a comprehensive example please see our article “Imaging Basics: How to Calculate Resolution for Machine Vision“.

Figure B. Top to bottom: Test pattern, lens, image from camera sensor, brightness distribution, MTF curve

Unpacking Modulation Transfer Function, let’s recall that “transfer” is about getting photons presented at the front of the lens, coming from some real world object, through glass lens elements and focused onto a sensor consisting of a pixel array inside a camera. In addition to that nifty optical wizardry, we often ask lens designers and manufacturers to provide lens adjustments for aperture and variable distance focus, and to make the product light weight and affordable while keeping performance high. “Any other wishes?” one can practically hear the lens designer asking sarcastically before embarking on product design.

So as with any complex system, when transferring from one medium to another, there’s going to be some inherent lossiness. The lens designer’s goal, while working within the constraints and goals mentioned above, is to achieve the best possible performance across the range of optical and mechanical parameters the user may ask of the lens in the field.

Consider Figure B1 below, taken from comprehensive Figure B. This shows the image generated from the camera sensor, in effect the optical transfer of the real world scene through the lens and projected onto the pixel array of the sensor. The widely-spaced black stripes – and the equally-spaced white gaps – look really crisp with seeming perfect contrast, as desired.

Figure B1: Image of progressively more line pairs per millimeter (lp/mm)

But for the more narrowly-spaced patterns, light from the white zones bleeds into the black zones and substantially lowers the image contrast. Most real world objects, if imaged in black and white, would have shades of gray. But a test chart, at any point position, is either fully black or fully white. So any pixel value recorded that isn’t full black or full white represents some degradation in contrast introduced by the lens.

The MTF graph is a visual representation of the lens’ ability to maintain contrast across a large collection of sampled line pairs of varying widths.

Let’s look at Figure B2, an example MTF curve:

Figure B2: Example of MTF graph
  • the horizontal axis denotes spatial frequency in line pairs per millimeter; so near the origin on the left, the line pairs are widely spaced, and progressively become more narrowly spaced to the right
  • the vertical axis denotes the modulation transfer function (MTF), with high values correlating to high contrast (full black or full white at any point), and low values representing undesirable gray values that deviate from full black or full white

The graph in Figure B2 only shows lens-center MTF, for basic discussions, and does not show performance on edges, nor take in account f# and distance. MTF, and optics more generally, are among the more challenging aspects of machine vision, and this blog is just a primer on the topic.

Click to contact
Give us some brief idea of your application or your questions –
we will contact you to assist

In very general terms, we’d like a lens’ MTF plot to be fairly close to the Diffraction Limit – the theoretical best-case achievable in terms of the physics of diffraction. But lens design being the multivariate optimization challenge that it is, achieving near perfection in performance may mean lots of glass elements, taking up space, adding weight, cost, and engineering complexity. So a real-world lens is typically a compromise on one or more variables, while still aiming to achieve performance that delivers good results.

Visualizing correlation between MTF plot and resultant image – MORITEX North America

How good is good enough? When comparing two lenses, likely in different price tiers that reflect the engineering and manufacturing complexity in the respective products, should one necessarily choose the higher performing lens? Often, yes, if the application is challenging and one needs the best possible sensor, lighting and lensing to achieve success.

But sometimes good enough is good enough. It depends. For example, do you “just” need to detect the presence of a hole, or do you need to accurately measure the size of the hole? The system requirements for the two options are very different, and may impact choice of sensor, camera, lens, lighting, and software – but almost certainly sensor and lensing. Any lens can find the hole, but a lens capable of high contrast is needed for accurate measurement.

Here’s one general rule of thumb: the smaller the pixel size, the better the optics need to be to obtain equivalent resolution. As sensor technology evolves, manufacturers are able to achieve higher pixel density in the same area. Just a few years ago the leap from a VGA sensor to 1 or 5 MegaPixels (MP) was considered remarkable. Now we have 20 and 50 MP sensors. That provides fantastic options to systems-builders, creating single-camera solutions where multiple cameras might have been needed previously. But it means one can’t be careless with the optical planning – in order to achieve optimal outcomes.

Not all lens manufacturers express their MTF charts identically, and testing methods vary somewhat. Also, note that many provide two or even three lens families for each category of lenses, in order to provide customers with performance and pricing tiers that scale to different solutions requirements. To see an MTF chart for a specific lens, click first on a lens manufacturer pages such as Moritex, then on a lens family page, then on a specific lens. Then find the datasheet link, and scroll within the datasheet PDF to find the MTF curves and other performance details.

Contact us for a quote

Besides the theoretical approach to reading specifications prior to ordering a lens, sometimes it can be arranged to send samples to our lab for us to take sample images for you. Or it may be possible to test-drive a demo lens at your facility under your conditions. In any case, let us help you with your component selection – it’s what we do.

Finally, remember that some universities offer entire degree programs or specializations in optics, and that an advanced treatment of MTF graph interpretation could easily fill a day-long workshop or more – assuming attendees met certain prerequisites. So this short blog doesn’t claim to provide the advanced course. But hopefully it boosts the reader’s confidence to look at MTF plots and usefully interpret lens performance characteristics.

1st Vision’s sales engineers have over 100 years of combined experience to assist in your camera and components selection.  With a large portfolio of lensescablesNIC card and industrial computers, we can provide a full vision solution!

Acknowledgement / Credits: Special thanks to MORITEX North America for permission to include selected graphics in this blog. We’re proud to represent their range of lenses in our product offerings.

Effilux LED bar lights for machine vision – adjustable and modular!

Various LED bar configurations

Effective machine vision outcomes depend upon getting a good image. A well-chosen sensor and camera are a good start. So is a suitable lens. Just as important is lighting, since one needs photons coming from the object being imaged to pass through the lens and generate charges in the sensor, in order to create the digital image one can then process in software. Elsewhere we cover the full range of components to consider, but here we’ll focus on lighting.

While some applications are sufficiently well-lit without augmentation, many machine vision solutions are only achieved by using lighting matched to the sensor, lens, and object being imaged. This may be white light – which comes in various “temperatures”; but may also be red, blue, ultra-violet (UV), infra-red (IR), or hyper-spectral, for example.

LED bar lights are a particularly common choice, able to provide bright field or dark field illumination, according to how they are deployed. The illustrations below show several different scenarios.

Example uses of LED bar lights

LED light bars conventionally had to be factory assembled for specific customer requirements, and could not be re-configured in the field. The EFFI-Flex LED bar breaks free from many of those constraints. Available in various lengths, many features can be field-adapted by the user, including, for example:

  • Color of light emitted
  • Emitting angle
  • Optional polarizer
  • Built-in controller – continuous vs. strobed option
  • Diffuser window opacity: Transparent, Semi-diffusive, Opaline
EFFI-Flex user-configurable LED bar
Contact us for a quote

While the EFFI-Flex offers maximum configurability, sister products like the EFFI-Flex-CPT and EFFI-Flex-IP69K offer IP67 and IP69 protection, respectively, ideal for environments requiring more ruggedized or washdown components.

SWIR LED bar, backlight, and ringlight

Do you have an application you need tested with lights? Contact us and we can get your parts in the lab, test them and send images back.   If your materials can’t be shipped because they are spoilable foodstuffs, hazmat items, or such, contact us anyway and we’ll figure out how to source the items or bring lights to your facility.

Test and optimize lighting with customer materials

1st Vision’s sales engineers have over 100 years of combined experience to assist in your camera and components selection.  With a large portfolio of lensescablesNIC card and industrial computers, we can provide a full vision solution!

Do I need cables designed for machine vision?

Do I really need cables designed specifically for machine vision? As a distributor of machine vision cameras, lenses, camera systems, cables, and accessories, we hear this question many times a day. Why does your GigE or USB3 cable cost so much?  I can just buy a cable online from Amazon, Ebay, etc. for $5 when yours costs $25 or more!

The answer is: You can…  sometimes… but it depends upon many things, and how critical those things are to your application. 

Here are 5 key variables you need to consider in a camera cable selection

  1. Distance from camera to computer
  2. Data rate in which the camera is transmitting  
  3. Importance of reliability of application 
  4. Structural integrity of connection at camera and computer
  5. Total cost of your process and / or down time

From many years of diagnosing industrial imaging problems, especially after incorrect software setup, BAD CABLES ARE NEXT ON THE LIST FOR  “MY CAMERA DOESN’T WORK” problems!! (Inadequate lighting or sub-optimal lensing also come up, but those are topics for another day.)

Distance, the killing factor!  If you were to look at a “bode plot” of the signal transmitting from the camera to the computer you would see dramatic attenuation of the signal vs. distance, and also versus the data rate. In fact, at the distance limits, you might wonder if it actually works as the signal is so low! 

GigE is rated at 100 meters, however, the signal does degrade quite a bit, so cable quality and data rate will be the determining factors.  USB3 does not have a real specification and it is difficult to find consumer grade cables greater than 2 meters in length. In fact, we have experienced poor results with consumer cables greater than 1 meter in length!

Click here for all USB3 Industrial camera cables

Click here for all GigE Industrial camera cables

What are the differences between ‘Industrial’ and ‘’consumer’ cables?  

8 differences are listed below: 

Assorted machine vision cables
  1. Industrial cables are tested to a specification for each cable.  There are no batch to batch differences. 
  2. That specification usually meets organization requirements such as IEEE or Automated Imaging Association (AIA) standards
  3. Industrial cables give you consistency from a single manufacturer (when buying online, you are not always sure you are getting the same cable)
  4. Industrial cables have over-molded connectors
  5. Industrial cables have screw locks on the ends
  6. Industrial cables are usually made with larger gauge wire
  7. Industrial cables typically specify bend radius
  8. Industrial cables are made with flex requirements (bend cycles they can meet)

When should we consider using an “Industrial cable”?  Here are a few examples to consider:

Example 1: In a research lab, using a microscope 1 meter from the computer running low data rates, non automated.

Distance is small, data rate is low, chance of someone pulling on the cable is low, and if the data doesn’t get delivered, you can re-acquire the image. There is no big need for a special cable and can buy it off the internet.

Example 1a: Let’s change some of these parameters, now assuming you are not in lab, but the microscope is in an OEM instrument being shipped all over the world

If the system fails because you went with an unspecified cable, what is the cost of sending someone to fix this system 3000 miles away? In this situation, even though the distance is small, and the data rate is low, the consequences of a cable failure are very high!

Example 2: GigE cameras running at close to the full bandwidth. If you don’t need screw lock connectors, and the distance is not too great (< 10 or 20 meters), 

You can probably get by with ‘higher quality’ consumer cables. At distances greater than 20 meters, if you care about system reliability, you will definitely want industrial cables.  

Example 3. Two to Four GigE cameras running at close to full bandwidth in a system.

If you need system repeatability, or anything close to determinism, you will need industrial cables. On the other hand, if you your application is not sensitive to packet re-sends, a consumer cable should work at under 20 meters

Example 4. GigE cameras in an instrument.  Regular GigE cables are just locked into the RJ45 with a plastic tab. 

If your product is being shipped, you can’t rely on this not to break. You want an industrial cable with screw locks.

Example 5. GigE cameras in a lab. 

Save the money and use a consumer cable!

Key takeaways:

  • If you running USB3 cables at distances more than 2 meters, DO NOT use consumer cables.
  • If you are running multiple cameras at high speeds, DO NOT use consumer cables.  
  • Obviously, if you need to make sure your cables stay connected, and need lock downs on the connectors, you cannot use consumer cables. 
  • If you are running low speed , short distance, and you can afford to re-transmit your data, consumer cables might be just fine.
Contact us for a quote

Below are additional remarks provided by our cable manufacturing partner Components Express Inc., that help to support the above conclusions. It’s good reading for those who want to understand the value-added processes used to produce specialized cables.

  • Families of connectors for vision systems include right-angle options to address commonly found space constraints and/or avoid overstressing the cable strain relief. Generic cables are typically “straight on” only.
  • The test process for machine vision cables go beyond routing hi-pot testing to include the electrical testing that ensures conformance with the latest and most stringent machine vision performance standards. Machine vision configurators – using customer application parameters – prevent mis-applying a cable that won’t meet the performance requirements.
Screenshot for GigE cable configurator; also available for USB and other types
  • Machine vision cable families cater to de-facto standards. For example, pin-outs vary by ring lighting makers for the same 5-pin connector. So it’s more labor and cost-intensive to support the permutations of pin-outs across diverse camera and lighting manufacturers.
  • The IP67 versions of standard electrical interfaces can vary by camera body. Machine vision cameras have different part numbers for specific camera bodies. For example, a screw lock Ethernet cable might damage the camera body of another maker if the mold-to-connector nose distance varies.
  • Machine vision y-cables are a unique breed and typically bought in small quantities. Pin-outs are higher and the semi-standard interfaces are different.

1st Vision’s sales engineers have over 100 years of combined experience to assist in your camera and components selection.  With a large portfolio of lensescablesNIC card and industrial computers, we can provide a full vision solution!

SVS-Vistek HR + SHR cameras: high resolution and speed

SVS-Vistek HR and SHR camera series

Some applications demand high resolution from 16MP or up to 151MP. Thanks to dual and 10GigE interfaces, Camera Link, and CoaXPress, getting image data from the camera to the computer can be accomplished at speeds matched to application requirements, using camera series HR and SHR from SVS-Vistek.

What kind of applications require such resolution? Detail-demanding manufacturing inspection, geo mapping, life science, film industry and other applications require, or benefit from, a high resolution image delivered from the camera directly to the PC. Prior to the convergence of high-resolution sensors and high-speed interfaces, one might have needed multiple smaller-resolution cameras to capture the entire field of view – but with complex optical alignment and image-stitching (in software) to piece together the desired image.

The HR series offers resolutions from 16 – 120MP. The SHR series ranges from 47 – 151MP. While every machine vision camera offers various features designed to enhance ease-of-use or applications outcomes, here are some particular features we highlight from one or both of the HR or SHR series:

  • Minimal 128 MB internal image memory, burst mode – capture sequences rapidly on the camera and transfer them to the computer before the next event
  • LED controller for continuous & strobe built into camera – avoids the need to purchase and integrate a separate controller
  • Programmable logic functions , sequencers and timers – critical for certain applications where programmed capture sequences can be pre-loaded on the camera
  • RS-232 serial data to control exposure, lights or lenses
  • Long exposure times up to 60 seconds (camera model dependent) – useful for low-light applications such as those sometimes found in life sciences or astrophysics
  • Camera Link, CoaXPress and 10GigE interface options (varies by model)

For pricing on the HR / HSR Series, follow the family links below to the series table, then click on “Get quote” for a specific model of interest. Or just call us at 978-474-0044 to discuss your application and let us guide you to the best fit.

SVS-Vistek HR series

SVS-Vistek SHR series

The HR series uses a range of CCD and CMOS sensors from CANON, Sony and ON Semi. The SDR series use both CCDs from ON Semi and  CMOS sensors from Sony. The same sensor choices and feature sets are offered across several popular machine vision interfaces, permitting users to tailor their own need for speed to specific application requirements. SVS-VISTEK engineering and manufacturing precision mounts these high-resolution sensors, which allows users to have distortion free, high quality, high content images.

SVS-Vistek shr661 – 127 megapixel camera

At the time of writing, note the newest member of the SHR series, the shr661. At 127 megapixels this CMOS sensor camera has remarkably high resolution with a global shutter. With the IMX661 sensor from the Sony Pregius series, the backlight technology enables very high light sensitivity and above-average noise behavior. This enables an image quality with which even the finest structures can be resolved. The shr661 is one of the most powerful industrial cameras on the market. 

Those familiar with high-resolution sensors may know about dual and quad-tap sensors, whereby higher frame rates may be achieved with electronics permitting two or more sections of the sensor’s pixel values to be read out in parallel. A traditional challenge to that approach has been for camera manufacturers to match or balance the taps so that no discernable boundary line is visible in the composite stitched image. SVS-Vistek is an industry leader with their proprietary algorithm for multi-tap image stitching.

1st Vision’s sales engineers have over 100 years of combined experience to assist in your camera and components selection.  With a large portfolio of lensescablesNIC card and industrial computers, we can provide a full vision solution!