FEATURE
Issue: 
Topic tags: 

A test for metrology

Aspheres and other freeform optics are becoming more commonly used in areas like consumer electronics - but, as these lenses move further away from rotational symmetry, can the metrology equipment to test them cope? Tom Eddershaw finds out

The use of freeform and aspheric lenses are becoming more popular – examples include reflector designs for car headlights, heads up displays (HUD) and optics for consumer electronics such as Google Glass. Advances in the software and machining tools mean more extreme freeform designs are being produced, but the metrology equipment needed to inspect these complex optics is struggling to keep pace.

Mark Middleton, business development manager at optical metrology equipment provider RedLux, commented: ‘The metrology market is lagging behind the production capability. At the moment you are lucky if you can get metrology that can match the reproducibility of the manufacturing machines. And this isn’t good because, ideally, you want to be able to measure a component five or 10 times better than the required tolerances of the part.’ Middleton said that the technology just simply hadn’t caught up yet with the manufacturing process. This often means systems are very expensive and taking a measurement can be very time consuming.

The advantage of an aspheric lens – a lens with a non-spherical curvature – is that one asphere can replace a number of spherical lenses in an optical system and therefore reduce the size, weight, and expense of a device. This is why they are being used more often in areas like consumer products.

Matthew Stairiker, senior manufacturing development engineer at Edmund Optics, explained the advantages of using aspheres: ‘If the designer is not constrained by correction of their system utilising only rotationally symmetric optics, tolerances and performance can be greatly improved. The ability to correct non-rotationally symmetric errors affords a more easily assembled optical train which can then be corrected via a single freeform.’

The lack of rotational symmetry can prove problematic for some measurement techniques; the bizarre angles and points of inflection used to improve the functionality of the lens also mean the system cannot be allowed to make any assumptions. Conventional lens metrology is often left wanting and adaptations of existing systems can be costly, less accurate, or lead to a time consuming process.

Middleton stated that optics manufacturers are looking for metrology systems offering greater accuracy. Stairiker commented that there are four key requirements for a metrology system that Edmund Optics would typically look for: accuracy, repeatability, measurement time, and if it can measure in two and three dimensions.

Stairiker said: ‘What is the accuracy of the machine? Now this question typically comes with a lot of hand waving and “ifs” and “buts”. The reason for this is each asphere is fairly unique. What is the repeatability? A gauge that we look for is the maximum aspheric departure that is provided by the metrology equipment manufacturer. This does not mean that every asphere which has this departure or less can be measured, but that the machine has the ability to measure a lens that has this amount of departure.’ He also said that the type of data output should be questioned to ensure it is compatible with existing systems.

However, a variety of techniques to inspect the part are available and, as is often the case, each has its benefits and limitations. As the complexity of the lenses increases, the inspection equipment has to advance with it. One of the issues with transferring from one step to the next is dealing with aspheric departure – the amount the geometry of the lens deviates from that of a sphere.

Middleton said that, if a company has an existing system that was used for spherical lenses, a computer generated hologram (CGH) could be used to adapt the system and account for the aspheric departure. However he also said: ‘These are custom optics that have to be very well machined and the curvature can only deviate by 63nm (a tenth of the wavelength of a helium neon laser) which is the accuracy we are working to here. The upshot of this is that these CGHs will cost around £10,000 a piece. So if the same lenses are being manufactured many times, this may save the user buying a new machine.’ However for each new type of lens manufactured, a new custom optic must be bought to account for the differing deviation and another £10,000 must be spent.

An alternative approach is to use a stylus profilometer. Middleton said they are found to be more flexible than the CGHs, and very accurate provided that the lens was correctly aligned. He explained the method is to drag either a ball or needle over the surface while measuring the displacement from the original position. This provides a sliced view of the lens, typically over the centre, which is then compared to the design data to provide an error map.

He said: ‘This is commonly in place today. The limitations of this is, in addition to only giving a 2D profile, if the lens contains a rise angle which is too large for the stylus, a false reading can be given due to “flanking”. This is when the side of the stylus or ball is in contact with the surface because the angle is too steep for it to travel over smoothly, typically around 50°. For certain applications which use micro-optics such as the cameras used in the smartphone market, very small aspheres are used and the angles of the lens can get very steep.’

He also warned: ‘What gets overlooked by a lot of companies who are selling these profilometers is that you have to be very conscious of the alignment of the lens with the system. You have to centre and level the lens very carefully. So it can be quite a time-consuming activity to centre and level it properly, even though you are only going to get a two-dimensional profile.’

The data given out is also only a two dimensional reading, which means for non-rotationally symmetric lenses, such as freeforms, multiple scans have to be taken and stitched together, which again can take time.

The process is also contact-based which means it is at risk of causing damage to the part. For mass production, this may not prove to be an issue when measuring the lens, as it can be scrapped if it becomes damaged, but when inspecting the cast for injection moulded lenses a scratch could result in an entire batch of flawed lenses.

Middleton said that RedLux uses a confocal light pen to create a data cloud across the optic. ‘We can then join these dots through interpolation. We have a white light source which is focused to a 4µm spot. This spot experiences chromatic aberration; so the blue wavelengths are focused first, then the other colours are focused afterwards. The sensor is moved so that the lens lies between the blue and red focal points. By collecting the reflected light and analysing it using a spectrometer the dominant wavelength of the reflected light can be given, thus calculating the exact distance from the probe to the particular measured point on the lens. We can do this very quickly – at up to 16,000 times per second – and then move to the next spot and the next.’

‘Another advantage, such as for smaller lenses, is it collects reflected light from both the front and backside of the lens which means the depth of the lens and the shape of the back of the lens can be calculated by analysing the second signal from the same scan.’

He continued: ‘This forms the basis of our optical coordinate measurement system. We have a little dance between the pen moving in and out, keeping itself in focus, while the lens is manipulated to follow the design data. The operator would load in the CAD file, the computer itself would make its own program in order to control all of the axis so that the part is always somewhere between the red and the blue focal points. It then subtracts the CAD model from the observed data to produce a residual error map in 3D. From here you can get all of the surface parameters and even roughness if you take a dense enough data cloud.’

Metrology company, Sensofar, offers an alternative method in confocal tracking. Agustí Pintó, project manager for the company’s Plu Apex, an aspheric optical profiler, explained: ‘The traditional contact measurement is only applying a low force so it wouldn’t damage the lens, but instead the coating. But, also, if too much force is applied, the actual lens can be scratched and this is more of a problem for the mould. Our technology has the ability to cover aspheres and freeform and is completely non-contact and thus does not cause any damage to the part.’

This confocal approach gives a very precise focal sensor. A focused green light is used measuring from a reference point. From then it works the same as a contact profiler but instead of using a mechanical tip, it is being touched with light. Pintó said: ‘We focus on a point, the object is moved laterally at constant speed and the microscope moves toward and away from the surface in order to keep the light in focus with a continuous autofocus technology.’

‘Confocal tracking provides 3D measurements as opposed to the 2D profiles using the stylus. We also offer a live image so that the user can see where they are measuring and this is another advantage. The user doesn’t need any prior knowledge of the lens design and the full alignment procedure is offered. If there is a mathematical curve it can be compared to the result using our software or it can use CAD data to compare between the design and the actual part.’

Pintó continued: ‘There is no limit within the aspheric departure that we can measure; the only limit that we have is the working distance of the microscope objective that we are using.
‘We offer different objectives for customers. One operates from 10mm and can measure up to 30° slope, the other operates at 4.5mm working distance and can measure up to 50°.’

The methods used to measure freeform optics are available, but as aspheric designs advance, there will also have to be advances in the metrology equipment required to test them.