Fast feasibility study

Challenge

The following example describes a challenge from an internal customer to strengthen safety-aspects of a joystick that was used in an existing mature design (Figure 1).

Figure 1: The “legacy” joystick device

Regulations stipulate that UI-devices in a safety-affecting function must be implemented using redundant sensing, based on physically different sensing-principles. The legacy-device was based on a singular principle – a moving magnet with a magnetometer to detect the joystick’s motion – and did not fulfill this stipulation. The design was in an advanced stage and naturally, a part-change to a different jostick component was not prefered; instead a “patch” to introduce redundancy within the existing joystick was desirable.

The customer requested us to work together with the supplier of this joystick to rapidly evaluate the possibility to add such secondary (redundant) sensing features to the existing mechanical/electrical product-design. The secondary principle is needed only to provide a “qualifier” signal, to catch false-positives of user-interaction with the stick.

Approach

In the diverge-then-converge approach, the different stakeholders are also actively involved – in this example, both our client, as well as their supplier (i.e. the manufacturer of the joystick-device). In addition to creating a potent engineering solution, this approach often generates interesting new IP, and also IP to strengthen the existing proposition.

Our case-study: The “Diverge” phase

Figure 2: graphical depiction of the ‘diverge’ phase

Given the mechanical envelope afforded to the joystick, some locations are candidates for adding additional sensing modalities (Figure 2). The motion of the joystick is most “visible” at the top (where the stick exits the enclosure) and at the bottom of the stick (where the conventional joystick sensor is also located to sense the user-input). In the first instance, we therefore split the candidate options into two main groups – top-side sensing and bottom-side sensing.

Capacitive sensing of the moving parts is yet another potential principle, but was quick-failed, as collocating the capacitive electrodes near the existing magnetic circuitry would prove difficult, given the parasitic capacitance-paths to existing structures and circuits.

Quick prototypes and functional evaluation

Optical cage

The movement of the target object conditionally interrupts one or more light-beams. When using line-of-sight, multiple light emitting and sensing components can be employed. Alternatively, reflective paths may be constructed, needing fewer emitters and detectors. A construction using the former approach is illustrated below (Figure 3).

Figure 3: Optical cage concept implementation

Galvanic contact-detector

A low-profile set of mechanical “feelers” is placed around the joystick, to detect motion beyond a preset limit. When the limit is exceeded, subcomponents touch each other, and with a simple electrical circuit the excursion of the stick is confirmed.

Bottom-side inductive go/no-go sensing

Here inductors are placed in proximity of the existing moving parts. Motion would be detected as a change in inductance, and this can trigger a decision.

The available volume allowed for only a singular inductor to be constructed on the PCB near the joystick. Off-the-shelf induction sensing ICs were proposed to sense the (change in) inductance. However, from simulations, it is quickly evident that the inductance is at a minimum at the “centered” position of the joystick, and as such, the sensitivity of the self-inductance to a small stick-movement is low. Recent advances in off-the-shelf ICs allow measurement of multiple inductances. This would enable placing a cluster of inductors rather than a single one, and measuring differentially. However, these ICs were not available at that time, and space too being constrained, this concept was “quick-failed”.

Judiciously letting go of some concepts (when better options are clearly available) frees up resources to fine-tune those more promising alternatives.

Bottom-side optical reflective sensing

There was just barely enough space between the joystick and the original PCB to allow placement of a cluster of low-profile LEDs and photosensors. The expectation was that the tilt of the stick should be observable as modulation of the reflected light.

Is this indeed a correct expectation? Let’s evaluate this with a thought experiment.

The small form-factor and low-cost construction of the LED and photosensor prohibits the addition of optical features, like lenses or mirrors. This means that the emission-rate of the LED and acceptance-rate of the photosensor follow Lambert’s cosine-law, as illustrated above. Consider the left half light-path. The light-cone exiting the LED at a a ~45° angle gets projected on the photo-sensor on the left by the non-tilted surface (i.e. bottom of the joystick) directly in front of it.

Later, when the surface (joystick) gets tilted, the emission-cone moves to an angle that is closer to the normal direction – let’s say 42°. Obviously, a higher emission-rate is directed at the photosensor on the right via the tilted mirror. However, a second effect happens simultaneously. The incoming light on the photosensor now arrives from a more oblique angle (let’s say 48°). The Lambert’s cosine law now counteracts the previous increase, with a net effect that is almost zero. Does this mean that the optical principle is unusable?

Quite the contrary: Let’s instead flip the photosensors on their sides: the Lambertian patterns of the LED and phototransistors (in combination with a tilted reflector) now add-up instead of cancelling out (42° and 42° in the illustration below).

Fortunately, phototransistors are also available as sideways-looking parts. Using a simple SMT-process, and off-the-shelf parts, a low-profile optical sensing system with excellent angle-measuring performance can be realized. This too is proven by a rapid prototype that can be retrofitted to the existing joystick-parts. With the prototypes, we discover that the LED has strong sideways emission that would directly (i.e. not via the reflector) reach the photosensor (unknown unknowns). Light-screens may be added to prevent a direct illumination path. This can be seen as protective black bars in the illustration above as well as in the rapid-prototype below.

This last concept has indirectly led to IP-generation too for a different Philips-product.

Conclusion

The approach discussed in this article has a number of strengths: It allows for rapid scoping and evaluation of numerous concepts, without overly committing to any one of them. Thought-experiments, calculations & simulations are an integral part of this evaluation, and allow us to shortlist promising candidate-concepts and to quick-fail problematic ones.

Subsequent rapid-prototypes enable us to discover (and solve) unexpected issues and to prove the operation of the available options in tangible real-life circumstances. It also facilitates the stakeholders to objectively face the engineer’s duality dilemma:

It creates transparency in the decision-process so that all concerned parties can jointly make informed choices, while also creating opportunities to strengthen the client’s business proposition with unique IP-generation possibilities.