June 2011 - minifaros.eu · usual technique has been to ... barcode scanners, laser print-ers, endoscopes, laser scan- ... angle of high inertia mirrors.

issue 3 June 2011

Editorial

Welcome to the MiniFaros EC funded project third

newsletter. MiniFaros is con-tinuing successfully its ac-tivities. The work performed

so far has been dissemi-nated for the first time to

the wider public through two major events in Europe: the ITS European Congress

that took place in Lyon on June 8-9, 2011 and the

AMAA Conference specializ-ing on Microsystems for Automotive applications.

Minifaros featured 3 papers in the Conference achieving

thus a very strong repre-sentation to that particular conference that took place

in Berlin on June 29-30, 2011.

In this Newsletter various articles containing among

others information on the project advancements that

were presented in the past Conferences as well as up-dates on the core research

items are included. More in-formation can be found on

the project website (www.minifaros.eu), while Minifaros has also a page on

Facebook as a supplemen-tary communication chan-

nel. Enjoy reading.

Editorial 1

TDC11 (Time-to-

Digital Converter) (J.

Kostamovaara) 2

Omnidirectional

lenses for low cost la-

ser scanners

(M. Aikio ) 3

MEMS mirror for low

cost laser scanners (U.

Hofmann) 4

News and Events 6

MiniFaros Consortium

7

Inside this issue:

TDC11 (Time-to-Digital Converter)

functionality and performance now

verified

One of the project goals is to develop a multi-channel

time-to-digital converter integrated circuit, which measures the time intervals

between the emitted laser pulse and several succes-

sive echoes related to the transmitted pulse (TSP1-TSP3 in Fig.1). Moreover, the de-

vice is to measure the widths of the received ech-

oes, which can then be used for the walk error compensation (TW1-TW3 in

Fig. 1). The timing walk (dependence of the timing

moment on the echo ampli-tude, see Fig. 1.) is the

main source of systematic error in pulsed time-of-

flight laser radars. In fact, the accurate multi-channel TDC techniques to be de-

veloped enable in principle the realization of new

“multiple-threshold time-domain” RF/high-speed op-tical pulse detection princi-

ples and circuits. The latter make it possible to detect

with picosecond accuracy the time position of the re-ceived pulse over a wide

dynamic amplitude range exceeding that of the re-

ceiver. It is also believed that the use of these tech-

June 2011

niques may result in reduc-tion in power consumption

and complexity relative to the levels available with tra-ditional high-speed synchro-

nous receiver sampling and AD conversion techniques.

This CMOS time-to-digital converter (TDC11) developed

by the University of Oulu team has now been realized

and tested with respect to the main performance pa-rameters. The main perform-

ance parameters of the TDC11 are its measurement

precision (sigma value of the distribution of the single shot measurement results for a

constant time interval), measurement accuracy and

drift. The measurement pre-cision of the developed de-vice is shown in Figure 2 and

demonstrates a single shot precision of better than 10ps.

The TDC11 is capable of measuring also “negative time intervals” (time inter-

vals where stop signal pre-ceeds the start signal, which

may well be the case in a practical laser radar at short

measurement distances due

to the electronic delay in the

start pulse gen-eration). The

accuracy is at the level of a few pico sec-

onds and partly limited by the

performance of the measure-ment arrangement.

The temperature drifts of the TDC11 with respect to the

start-stop time interval and stop pulse width are shown in Fig. 3 indicating a drift of

~0,3ps/°C and ~0,4ps/°C, respectively.

The measurement results ver-

ify the operation of the TDC in different circumstances with

the state-of-the art perform-ance. Time interval measure-

ment is stable with respect to

variations in temperature and operating voltage, and the low internal jitter in the

delay lines makes it possible to use a low frequency exter-

nal crystal as a reference. A measurement precision bet-ter than about 10ps is

achieved over the whole temperature range (C).

Figure 1: TDC11 measurement ap-

proach.

Figure 2: Single shot measurement precision

Figure 3: Start-stop and Stop pulse width drifts

of the TDC11 in relation to Temperature (C)

June 2011

There is a need for small sensors that provide 360-

degree field of view in intelli-gent vehicle applications. The usual technique has been to

use a catadioptric system where a conical shaped mir-

ror is placed in front of a camera, providing 360-degree horizontal field of

view and a few decades of degrees of vertical view. The

downside of these kinds of systems has been their size, usually ranging around 20

centimetres. A so-called om-nidirectional lens can fold the

optical path inside the lens decreasing the volume re-quirements considerably,

while still providing compara-tive optical performance. In

this work, two different om-nidirectional lens systems are presented, more common

type of this lens images a whole surrounding scenery to

an image sensor, providing instant 360-degree field of view. The other lens can se-

lect a known position from the 360-degree scenery, and

provide an undistorted image of it. The other application

for this type of lens is a laser scanner that necessitates di-rection selectivity.

The general objective of the

current Minifaros-project is to replace a large rotating mirror from laser scanners

with a MEMS mirror. Instead of imaging a whole scenery

reflected at the lens, a rotat-ing mirror is used to select a portion of the scenery to be

imaged on the sensor – or to be measured with a laser

scanner. This kind of lens is new and no prior art work has been published. The

working principle of the

lens is shown in Figure 4, and one manufactured

lens is shown in Figure 5. A biaxial laser scanner

consisting of two lenses as shown in Figure 5 was

constructed, and the per-formance was evaluated. The divergence of the

sensor was 30 milliradians with a detector of diame-

ter 200 µm. The signal to noise ratio allowed the us-age of the sensor up to 10

metres, with a black dif-fuse target. Expanding

the measurement distance from this is one of the objec-

tives in Minifaros project. Omnidirectional vision and

sensor systems are important in autonomous vehicle opera-

tions if the amount of sensors needs to be reduced. By using a large field of view sensor,

there is no need to have mul-tiple sensors in a vehicle.

However, one constraint on using them has been the size, manufacturing tolerances and

the price of the resulting sys-tem.

The type of omnidirectional lens presented just above al-

lows also imaging of the sur-rounding scenery without

distortion, if multiple expo-sures are taken and the ava-lanche photo diode is re-

placed with a small image sensor.

The second important factor to be considered is the price

of the sensor and related op-tics. The omnidirectional

lenses are roughly 40 to 50 millimetres in diameter and

are made of plastic to allow for easier serial production of this type of optics. In serial

production when the produc-tion volume approaches hun-

dreds of thousands of pieces per year, the price for a sin-gle omnidirectional lens is

around several cents. In Minifaros project, the omnidi-

rectional lens is used in a la-ser scanner application (LIDAR) to prevent and miti-

gate the consequences of ve-hicle accidents.

Omnidirectional lenses for low cost laser scanners

Figure 4: A sketch of an omnidirectional lens that

has a beam direction capa-bility .

Figure 5: A manufactured om-nidirectional lens which is

used in conjunction with a beam steering mirror.

June 2011

LIDAR sensors are becoming

increasingly interesting for

the realization and improve-

ment of driver assistance

systems like pre-crash safety

systems, intersection assis-

tant, lane change assistant,

blind spot assistant, parking

assistant or traffic jam assis-

tant. A wide angular range

and high angular resolution

are key-features that scan-

ning LIDAR systems offer.

Existing scanning LIDAR sys-

tems use bulky servo motors

for rotation of a large aper-

ture scanning mirror making

it difficult to demonstrate the

required sensor dimensions

and sensor costs for a series

automotive product. But cost

reduction and a higher level

of miniaturization seem to be

possible by introduction of

MEMS technology. The con-

cept and the design of a low

cost two-axis MEMS scanning

mirror that aims at replacing

the bulky and expensive con-

ventional laser scanner in an

automotive LIDAR sensor ap-

plication is presented.

The key feature of the low-

cost LIDAR sensor is an om-

nidirectional lens that inte-

grates several reflective and

refractive functions within

one single component like

the lens presented in the

previous article. Omnidirec-

tional scanning is achieved

by first collimating the diver-

gent laser beam by passing

the refractive centre area of

the omnidirectional lens. The

collimated beam then im-

pinges on a 2-axis MEMS

scanning mirror.

The tilted mirror reflects the

beam back to propagate

trough the lens again. After

passing two internal reflec-

tions at two reflective lens

facets the beam exits the om-

nidirectional lens almost per-

pendicular to the optical axis

of the incoming divergent la-

ser beam. According to the

cylindrical symmetry of the

overall configuration the laser

beam can be scanned within

the whole range of 360 de-

grees. The optical concept re-

quires a two-axis MEMS scan-

ning mirror which performs a

circular scan at a constant tilt

angle of 15 degrees resulting

in a cylinder symmetric optical

deflection of 30 degrees. In

order to enable a long meas-

urement range of up to 80

metres the optical configura-

tion requires a mirror diame-

ter of 7mm.

MEMS mirror design

MEMS scanning mirrors have

been used in many different

applications as for instance

barcode scanners, laser print-

ers, endoscopes, laser scan-

ning microscopes or laser pro-

jection displays. Typically

MEMS mirrors have a mirror

aperture size within the range

of 0.5 to 2 millimetres. There

are two major reasons for the

limitation of MEMS mirrors to

such small dimensions:

Firstly, static and dynamic

mirror deformations rapidly

increase with increasing mir-

ror diameter and secondly,

the very low driving forces of

MEMS actuators usually do

not allow a reasonable tilt

angle of high inertia mirrors.

Hence, to design and fabri-

cate a 2D-MEMS scanning

mirror with an outstanding

mirror size of 7 mm and a

large mechanical tilt angle of

+/-15 degrees is a challenge.

Static and dynamic mirror

deformation

The optical conception of the

LIDAR sensor requires that

deformation of the MEMS

mirror plate does not exceed

+/-500 nanometres. Defor-

mations can be caused by

stress gradients within the

layers which the mirror is be-

ing made of. Typically the

uppermost reflective layer

introduces mechanical stress

that deforms the mirror to

some extent. But more often

deformation is predominantly

caused by the MEMS mirror

dynamics. The dynamic mir-

ror deformation is known to

scale proportional to the fifth

power of mirror diameter.

This scaling law indicates

that to keep the deformation

of a mirror of 7 millimetres

and tilt angle of 15 degrees

sufficiently low the thickness

of the mirror needs to be

correctly adjusted. For a

more detailed investigation

on how different mirror ge-

ometries may effect the dy-

MEMS mirror for low cost laser scanners

June 2011

namic mirror deformation fi-

nite element analysis (FEA)

was carried out. Three differ-

ent types of mirrors were

simulated: 1) a mirror plate

having a standard thickness

of 80 microns (typical MEMS

device layer thickness), 2) a

mirror plate identical to first

type but additionally rein-

forced by a 500 micron thick

and 200 microns wide stiff-

ening ring underneath the

mirror plate, 3) a solid mirror

plate with a thickness of 580

microns. For each type of

mirror the simulation of mir-

ror deformation was per-

formed for four different di-

ameters (figure 6).

The FEA showed that a 7mm-

mirror with a standard thick-

ness of 80 microns would ex-

perience unacceptably large

deformations exceeding +/-6

microns. Considerable reduc-

tion of mirror deformation to

only +/-1.2 microns can be

achieved by a narrow but

500 microns thick reinforce-

ment ring underneath the

mirror. Finally a solid mirror

plate with a thickness of 580

microns achieved the best re-

sult and showed a minimized

mirror deformation of only +/-

0.2 microns. Thus, further de-

sign assessments and simula-

tions only considered the two

reinforced mirror types.

Driving concept and fabri-

cation process

In principle electromagnetic

actuation would enable to

achieve the highest driving

forces and hence would be the

first choice for actuation of

such a high inertia MEMS mir-

ror. But the attractiveness is

lowered by the fact that it

requires mounting of large

permanent magnets on chip

level resulting in a too large

and too expensive scanning

device. A compact and cost

effective solution is an elec-

trostatically driven MEMS

mirror since the whole device

can be produced completely

on wafer level including her-

metic packaging. Figure 7

shows a two-axes MEMS scan-

ning mirror electrostatically

actuated by stacked vertical

comb drives.

To drive such a large MEMS

mirror with an aperture size of

7millimetres to the required

large tilt angles of +/-15 de-

grees it is necessary to apply

resonant actuation because it

allows to achieve higher oscil-

lation amplitudes. However, if

the MEMS mirror works in

standard atmosphere damp-

ing by gas molecules is so

high that even resonant ac-

tuation is not sufficient to

achieve the required scan

angles. To meet the require-

ments of large mirror size

and large tilt angle it is nec-

essary to minimize damping.

This can be achieved by

packaging the 2D-MEMS

scanning on wafer level in a

miniature vacuum environ-

ment. This allows the MEMS

mirror to accumulate driving

energy over many thousand

oscillation cycles.

The low-cost LIDAR MEMS

scanning mirror will be fabri-

cated in a dual layer thick

polysilicon process. Wafer

bonding techniques will be

applied to permanently pro-

tect each MEMS mirror

against contamination by

particles, fluids or gases. A

titanium getter will be inte-

grated into each MEMS scan-

Figure 6: Calculated mir-ror deformation versus

mirror diameter for three different mirror geome-tries.

Figure 7: Typical gimbal-mounted two-axes MEMS

scanning mirror electro-statically driven by stacked vertical comb

drives

June 2011

ner cavity in order to achieve

a permanent miniature vac-

uum environment.

Suspension concept

The standard design to allow

a MEMS mirror to scan a la-

ser beam in two dimensions

is a gimbal mounted device.

But the optical concept of the

targeted low-cost LIDAR sen-

sor requires a circular scan

trajectory and the MEMS mir-

ror has to provide two per-

pendicular scan axes that

have identical scan fre-

quency. Practically, this is

difficult to be achieved using

a gimbal mounted mirror de-

sign. For that reason a com-

pletely different design was

chosen which eliminates the

need for an outer gimbal

frame. Instead of suspending

the mirror by two torsional

beams the mirror plate is

movably kept by three long

and circular bending beams.

This allows achieving an ad-

vantageous ratio of mirror di-

ameter and chip size which is

an important factor for a low

cost scanner. Because of a

considerably lower total mass

with respect to a gimbal mir-

ror design such a tripod de-

sign shows higher robustness.

Finite element analysis has

shown that mechanical stress

in the bending beams can be

kept sufficiently low to enable

the required tilt angle of 15

degrees. Regardless of the

three beams which are spa-

tially separated by angles of

120 degree the mirror builds

two perpendicular tilt axes

(two eigenmodes) that have

almost identical resonant fre-

quencies. In comparison with

a gimbal mounted mirror de-

sign the tripod approach

shows a considerably lower

number of parasitic eigen-

modes. Different variants of

such a tripod MEMS mirror

design will be fabricated cov-

ering a range of scan fre-

quency of 600Hz to 1.6kHz.

This scan frequency depends

on the stiffness of the three

suspensions and by the mo-

ment of inertia which is dif-

ferent for a solid reinforced

mirror and for the ring rein-

forced mirror. The whole 360

degree scenery is thus

scanned at a rate of 600Hz

or higher.

News and Events

Minifaros managed to participate in two major conferences this period, initiating thus success-

fully the dissemination of its mid-term results.

ITS in Europe, Lyon France, June 8-9, 2011

Minifaros was represented by Florian Ahlers (SICK) to the special session “SS 42 / Avoiding ac-

cidents by enhanced perception and active interventions: a look into the future of intelligent

vehicles ” organized jointly by the IP interactIVe and

Minifaros. A presentation about the novel laser scan-

ners and its applications

AMAA 2011—15th International Forum on

Advanced Microsystems for Automotive

Applications

Minifaros had a strong presence featuring 4 papers ac-

companied by the respective presentations. Presenta-

tions were very attractive to the audience consisting of

key stakeholders from the automotive companies and

suppliers.

issue 3 June 2011