TechArchive

This article was originally written in the period 1995-2000

The different linear encoder technologies have never been as competitive as they are now.

Accuracy and resolution are of great concern in cutting and measurement applications. Machines are expected to uphold tighter tolerances and better repeatability than hitherto in the production of parts. Linear encoder manufacturers now offer measurement devices that are accurate to +/-0.1um and resolutions within the nanometre range (the relationship between accuracy and resolution is explained in the panel on the right).

Although we deal here with three types of linear encoder, there is in fact a host of technologies with various advantages relating to the application, reliability, accuracy, ease of alignment, cross-sectional size and resolution.

The linear encoder in its various guises is becoming an increasingly essential element of machine tool guidance systems. The biggest battle at present in the UK is getting any type of linear encoder accepted on machine tools other than grinding machines, such as milling machines, whereas in Europe, it is now becoming commonplace.

Many traditional machine tools use systems employing the drive screw as the measuring standard. A photoelectric rotary encoder interpolates over one revolution and generates digital signals. This configuration is suitable if a sufficiently accurate leadscrew is being used in conjunction with a backlash-free nut and precision leadscrew bearings. Advantages of this system lie in lower cost and high reliability, since the rotary encoder can be perfectly sealed against contamination and fluids.

But the disadvantages are inferior accuracy and repeatability. Firstly, a leadscrew cannot be as accurate as a precision scale. The length of the screw is also dependent on temperature. With heavy machining or frequent traversing at rapid feeds, the temperature increases. For a steel leadscrew a metre long, a temperature variation of 1Cdeg will cause a 10um measuring error.

In a measuring system comprising rack, pinion and rotary encoder, eccentricities of shaft and pinion are a major cause of measuring error. Measuring deviations below 10um is hardly possible, even for short distances.

So what are the options? The traditional glass scale can be direct reading or reflective. The reading head alignment may depend on the structure that it is mounted on, or the head may be self-aligned within the scale. There are open and enclosed types, the enclosed types offering more protection against the environment, but because of the need for a coupling, also being subject to an additional error term, backlash error.

Most optical types are incremental, but there are also a few absolute types and some which have distance coded reference points – next to the incremental track is a track with several reference marks individually spaced according to a mathematical algorithm, so that the absolute position can always be established after traversing two consecutive reference points. Some give quadrature output direct from the head, others output a raw sine wave and need additional electronics to provide a sensible output signal.

Bendy scales like Renishaw, Siko and the Sony Digiruler are also quite common. Renishaw uses an optical reflective system, while Siko and Sony use magneto-resistive technology.

Renishaw’s ‘flexible steel tape’ linear encoder system is now being made generally available. The RG2 linear encoder consists of a 20m pitch scale on a metal strip and a choice of read heads. The scale comes in stock reels up to 50m long and is cut off to suit, so the possibilities for large structures are intriguing. The advantages over conventional linear encoders are numerous. Firstly, there’s the speed of installation. The scale comes with an adhesive backing tape along its length, and is fixed with epoxy adhesive ‘clamps’ at each end. A built-in set up indicator makes things even easier with quick alignment. Then there’s the thermal stability on a wide range of machine structures. And finally, the almost impossibly low profile will enable the system to be used in some very tight situations. Of course, as the scale is exposed, the system won’t lend itself to dirty environments.

The electronics is all built into the read head, with digital or analogue output on offer. Whilst the data sheet for the 5.0m read head suggests a maximum speed of 1m/s, the company mentioned a figure of 5m/sec to us. There are 1.0m and 0.5m versions too, all with equivalent repeatability quoted. The system works by diffracting infra-red light scattered from the scale to form fringes – these are filtered to reject optical noise and distortion. Four photodetectors then feed the analogue electronics, which have built-in interpolation.

The third group, typified by the Sony Magnescale and Newall Spherosyn, use an inductive measuring system (the older Inductosyn is still around but is rarely used now). Their suppliers say they are more appropriate for dirty environments, a point only grudgingly conceded by Heidenhain, the leading supplier of optical systems. Sony has also developed Laserscale, which is a laser hologram on quartz offering nanometre resolution. A magnetic system called Spiral, pioneered by Scientific Generics, is working towards micron accuracy. A key feature of Spiral is the ability to monitor the position of a number of objects simultaneously, by using a series of resonant coils all communicating via a single reference strip.

Heidenhain UK Managing Director Malcolm Smith says that maintaining the gap between scale and reader head is critical with magnetic systems, because small variations can change the signal level and a few microns can halve it.

This is disputed by Axis’ Adrian Silcox, who says that the gap is not critical. The majority of magnetic scales use a phase modulated output signal, not an amplitude modulated signal, to translate movement. In fact, says Silcox, the scales will work perfectly accurately with 30% (0.7V if memory serves correctly) of the full signal amplitude (2V). It is the phase modulation that is critical, not the amplitude. The majority of Sony scales work with a rod passing through a hole in the reading head with a clearance of approximately 20um. This is the normal specified working clearance of the head gap for scales with resolutions of up to 0.0001mm.

Sony scales not using the phase modulated system operate on a resistive output signal. These scales have even better misalignment properties (head gap), claims Silcox, who asserts that the Sony Digiruler allows up to 1.5mm variation in head clearance and 2mm misalignment axially.

According to Silcox, historically, the optical grating has always been far more vulnerable to gap problems with the head guided along the scale by a delicate sprung-loaded carrier which can wear and jump. Sudden deceleration and vibration are a major problem. This same carrier can cause backlash errors and contains an exposed PCB vulnerable to swarf.

Smith, for Heidenhain, also says that swarf can scrape off the magnetic coating. However, Silcox says that the majority of Sony magnetic scales do not have a magnetic coating. They use a ferromagnetic alloy with a magnetic signal recorded deep into the alloy. They do not deteriorate or suffer due to any coatings being scraped off. One linear encoder in the Sony range (MSS series) uses a magnetic coating. These are used for travels from 3m upwards and they are not commonplace. Silcox suggests that optical scales breaking due to an accidental knock is a more likely occurrence than scratched magnetic scales malfunctioning.

So which system is best for accuracy? One thing to remember is that these suppliers do not just rely on the accuracy of the scale, but they subject the signals to a certain amount of electronic enhancement (interpolation) as well. Interpolation is a process whereby the original (say) sine wave is sub-divided into little pieces by an electronic system, so that the system overall can distinguish individual parts of the sine wave, not just the peaks and troughs in the original wave. It’s clever mathematics, but while it increases the resolution, it also introduces some error terms of its own. One area of debate is whether it is best to have a finer pitch scale, which is expensive, linked to less electronic interpolation, or should you have a coarser pitch scale with higher interpolation?

Heidenhain’s Smith says that magnetic systems have more backlash because of the magnetic backlash (reluctance) component. Therefore backlash in optical systems is typically under 1um whilst it is typically 1.8um in a magnetic system.

But the major advantage claimed by Heidenhain for optical encoders is the scale: the case is that inherent slide accuracy more than outweighs any environmental negative and prices are comparable. It is desirable, says Smith, to have as much resolution as possible in the scale, to minimise the reliance on electronic enhancement. In a magnetic system, the best scale resolution will be 200um, whereas 20um is possible in an optical system. Since signal linearity is a significant source of error and as a rule of thumb is about 1% of the signal period, about 2um error is introduced into a magnetic system. Even with optical systems, considerable work is being done to improve the quality of the sine wave.

Silcox agrees that the accuracy of the measuring scale is determined by the accuracy of the output signal (sine wave). But the better the accuracy, the more he says you can use electronic interpolation to increase resolution without errors. He says Sony can interpolate to a much higher resolution because the basic signal is a lot better, so 20um or 200um pitch makes no difference to the end result. Users can ask to see laser calibration reports for comparative optical and magnetic scales. They should show identical accuracy.

On pricing, Silcox claims that “We knock large chunks off HDH prices with the latest Sony 0.01mm/0.005mm resolution scales and counters. Likewise with the Sony Digiruler measuring systems. Most of our magnetic scales are cheaper than HDH.”

Standard optical encoders use gratings on various carriers (such as glass, glass ceramic, solid steel or steel tape) as their measuring standard. The accuracy and thermal behaviour of the encoder can be optimised by the choice of carrier.

When light falls onto the graduation of a measuring standard it is diffracted. If the grating period is much larger than the wavelength of light, the diffracted beam components become insignificant. When collimate light is directed through such graduations, it projects an image of the grating pattern. If, however, the grating period approaches the wavelength of light, this generates an interference pattern.

Large gratings are scanned according the projected light principle. Moving the measuring standard relative to an index grating produces a periodic fluctuation of light intensity, which is detected by photovoltaic cells.

For finer grating periods, graduations of appropriate structure are used to evaluate phase shifts in interference patterns. Interferentially scanned encoders are employed when very fine measuring steps (down to a few nanometres) necessitate small grating periods of typically 8um, 4um or less. Initially, these encoders were limited to research and scientific applications, but they are now coming to be used in grinding machines, which of all the machine tools impose the highest demands on resolution and accuracy. Some manufacturers of grinders, however, such as Studer in Switzerland, use Sony magnetic encoders. Quasi-single-field scanning is a hybrid system and is used on the Heidenhain LIDA range of optical encoders. These employ a steel scale tape with a grating period of 40um. The transparent index grating consists of two interlaced phase gratings with differing diffraction characteristics. The method has two advantages: firstly, the scanning method is relatively insensitive to a slight waviness of the scale tape. Also the gap and gap tolerances between the scale and scanning reticle are much greater than with the conventional scanning method. The index grating with one scanning field of two interlaced phase gratings generates four images on the measuring standard, with each image phase-shifted by one quarter of the grating period. Since only one scanning field is used to generate all four scanning signals, fluctuations in the intensity of the light, such as are caused by local contamination of the scale, will have an equal effect on the four signals, so they retain a high quality even in the presence of a certain degree of contamination. This factor could help it compete with Renishaw’s open system.

With measurement systems as accurate as these, thermal errors remain a significant area of concern. Moving from a lamp to a single-source LED was partly stimulated by the need to avoid thermal errors, though higher reliability is another benefit. Even so, thermal errors are significant enough to pay close attention to the glass used on optical scales. Heidenhain buys glass a year ahead and stabilises it. There are three types of glass used, of which the most expensive is called Zerodur (Schott) and has (approximately) zero thermal expansion. There is a glass with matches ferrous alloys and a standard glass which is used in most applications.

The three forms of output signal used with optical encoders are sinusoidal current signals of approximately 11uA, 1V peak-to-peak and TTL square-wave. In the first case, current signals allow cable lengths up to 30m between the linear encoder and the interpolating electronics, using double-screened cables. Voltage signals can be transmitted to the electronics unit over cables up to 150m, without the need for double-screened cable. This requires maintenance of the 5V +/-5% supply voltage at the encoder. Linear encoders that produce voltage signals have sensor line connections for detection of the supply voltage at the encoder. Corresponding control systems in the interpolating electronics then maintain the voltage tolerance.

Thirdly, linear encoders with TTL square-wave signals have the electronics integrated in the mounting block, allowing in situ interpolation by 5-fold or 10-fold. The signals are then digitised and exported over cable up to 50m in length. Silcox notes that there are tens of thousands of optical scales out in the field sold by HDH, Acu-rite, Anilam, Givi and others using 5V TTL square wave. These, he says, are the more traditional scales and they are vulnerable to RF interference. 20 times resolution in the head is sufficient to achieve 1um, which is ideal for the vast majority of non-feedback applications. Relatively cheap, they do not require additional pulse shaping from a sine wave so they can feed directly into a digital counter or PC interface card. Sine waves scales are better for RF immunity, says Silcox, with phase modulated systems being the most immune to RF interference.