The need to support a widening range of languages and font sizes is pushing the bitmap approach up against the tight resource constraints in low-end devices.
In today's increasingly device-centric world, design engineers are facing a proliferation of low-end, small-screen product requirements that require big system performance.
Whether it's a portable medical instrument, wearable device, automobile cluster display, or any number of Internet of Thing (IoT) devices, the functionality specifications are becoming more sophisticated and the text rendering performance needs to keep pace.
User expectations already are very unforgiving because of their experience with high-end devices, such as iPhones, so engineers are being asked to do more with less to make the screens look good, even on low-end devices.
Another important trend impacting display functionality is the increase in connectivity between devices of all types.
For example, even low-end devices streaming audio are now expected to display a variety of associated information (title, artist, genre, etc.) with clarity and responsiveness that meets high user expectations. In some cases, such as connected automobile clusters, navigation systems, and medical devices, the need for text clarity and instant readability can be a crucial factor as well.
Until recently, designers generally have been forced to settle for a bitmap approach to displaying text and images on lower end devices because of the overall constraints on system resources.
However, bitmap displays are no longer able to meet the evolving challenges. For example, connected devices must deal with a wider range of unpredictable text and more font sizes, which makes it much harder to pre-configure a sufficient range of bitmaps to deal with all situations.
Figure 1: Low-end automotive dashboards are required to display of a more diverse set of information, everything from turn-by-turn directions to song titles, artists, and genre.
In addition, today's global market environment requires built-in support for a wide range of languages including complex and bidirectional scripts that must be accurately and legibly rendered in different sizes as well. Bitmaps are particularly ill-suited for supporting complex scripts such as Arabic, Thai, and Indic which require very accurate shaping to convey the proper meaning.
Figure 2: Lower-end platforms are required to meet market demands for supporting complex and bidirectional scripts to address any region of the globe.
The need to support a widening range of languages and font sizes is also pushing the bitmap approach up against the tight resource constraints in low-end devices, because every combination of language, font size and style must be separately stored in local device memory.
Other key challenges with bitmaps include the R&D costs and hassles of maintaining different design approaches for low-to-mid level devices vs. the high-end devices that are almost always implemented with scalable font technology.
This not only adds significant development cost on the front-end; it also makes it much more difficult to evolve these bifurcated product families forward because new functions must be implemented in both static bitmaps and scalable font rendering versions on different devices. Going forward, OEMs need to be able to build out full suites of devices from low to high end by leveraging shared, reusable and extensible font technologies across their entire product families.
Over the next year, as high quality text display becomes a de facto requirement on all devices, engineers will need to find ways to implement scalable font rendering while still living within the tight resource constraints that will always be a factor in lower end products.
The rest of this article drills down to look at some of the specific challenges with low-end devices and requirements that a scalable font system will need to meet in order to support this industry transition.
CPU bandwidth and run-time memory
The first major constraint is the small run-time memory and lower CPU bandwidth in these devices. For example, lower-end automotive clusters and wearable devices typically are built around a very lean processor like an ARM Cortex-M3 running at 50 MHz to 120 MHz with limited run-time memory. Multiple languages and font size requirements are forcing bitmap storage beyond these constraints and current generation scalable font technologies typically exceed the limits as well. To successfully serve this low-end market, the entire font engine must have very efficient code and consume less than 20 Kbytes of run-time memory.
Font size and quality
The fonts, while delivering high quality, must take up minimal memory as well. For example a hinted Simplified Chinese font with more than 28,000 glyphs can be as large as 20 megabytes for high end devices. In this new environment with limited ROM, that same font needs to have equal quality in only 20% of that size
Variety of displays and resolution
When supporting a family of products from low to high end, designers must take into account the unique characteristics of every display. For example, a product family could span display sizes from 1.5 inches to 13 inches, with resolutions ranging from 70 dpi to 300 dpi. Different media (LCD, OLED, etc.), anti-glare overlays, ambient light variations and other factors make it difficult or impossible to design bitmap fonts for these multiple situations. Next-gen small-footprint scalable font rendering engines must be able to address individual pixels on each screen in order to properly scale and shape fonts for legibility, while also living within the tight processing constraints.
Optimizing for quick-glance readability
Unlike e-reader screens that are tuned for comfortable immersive reading and reduced eye fatigue, automotive, wearables, medical, and other consumer devices need interface text that are optimized for "quick-glance recognition."
For example, MIT AgeLab and Monotype conducted research earlier this year utilized a new methodology called Stimulus Onset Asynchrony (SOA), for testing the legibility of typefaces under glance-like behavior. The typeface legibility results of the methodology study found that a humanist (Frutiger®) typeface could be read accurately in shorter (8.8%) exposure times than a square grotesque (Eurostile®) typeface, proving that certain typefaces are better suited for quick glance readability than others (Figure 3).
Figure 3: Using the adapted SOA methodology, on average, the humanist typeface (across gender) in white-on-black text could be read accurately in exposure times that are 9.1% shorter than with the square grotesque typeface.
Efficient multi-font rendering flexibility
Even though low-end devices have constrained resource limits, the display requirements are now pretty sophisticated. If the product is to ship globally, it has to include multi-lingual capabilities.
Supporting only Latin is no longer sufficient. Support for both black-and-white and grayscale is also becoming important. Today's product families, including low-end devices, typically need to include four to six sizes of every font in at least two styles. Since the low-end environments tend to use less expensive screens, it often requires each font to be accurately rendered in as few as 12 pixels (as compared to 28 or higher for higher end devices).
Figure 4: While the bitmap text has a jagged appearance around the curves and diagonals, the scalable technology enables the text to be resized without introducing distortion.
Bitmap quality is low (on the the first picture above) due to jagged appearance of curves and diagonals.
Moving forward, the answer to these challenges has to be a combination of fonts and technology to optimize for both legibility and performance. An approach of making either the perfect bitmap font or the perfect rasterizing engine will always fall short, especially when it comes to keeping up with new font requirements and spanning a range from high- to low-end devices.
For device designers the bottom line will require combining very small-footprint font rendering engines and efficient font files with the lean-processor, low-memory constraints of low-end devices. This will enable the core font files to be flexibly rendered across the growing range of requirements, while avoiding the proliferation of bitmap fonts that ultimately would swamp the available resources.
— Geoff Greve is VP of Type Operations at Monotype.