36th

International Electronic Manufacturing Technology Conference, 2014

Ultra Low Loop Wire Bonding of 20 µµµµm Palladium Coated Copper Wire for Very Thin Packages

Loh Kian Hwa, Loh Lee Jeng, Liew Siew Har, Eric Erfe

Carsem Technology Center (CTC)

Carsem (M) Sdn Bhd (S-site)

Lot 52986, Taman Meru Industrial Estate, Jelapang

P.O. Box 380, 30720 Ipoh, Perak, Malaysia

khloh@carsem.com

Abstract

In 1998, Carsem introduced the MLP (Micro Leadless

Package). Today, the MLP has become the semiconductor

industry’s package of choice for many devices with low-to-

medium I/O count. Coincidentally, the MLP came out at the

same time that smartphones were just starting to take off.

Following the tremendous increase in market demand for

smartphones, tablets and other handheld devices, the MLP

soon outpaced the growth (in billions of units shipped) of

every other leadframe based package in the industry. This

phenomenal growth is greatly attributed to the MLP’s low

cost, small size, and excellent thermal & electrical

performance characteristics.

Consumer handheld devices are constantly getting

thinner, smaller and cheaper. In recent years, there has been

a significant trend towards thinner and thinner packages.

Nowadays, the MLP package is available in different

package thicknesses ranging from 0.9 mm to as thin as 0.3

mm. The trend for thinner MLP presents several challenges

in the choice of packaging materials, most especially in the

choice of interconnect wire material. Gold wire has been

widely used by the industry to interconnect the silicon chip

to the metal leadframe. Thin MLPs will require wire bonds

with lower loop profiles. Numerous innovations have been

implemented in the gold wire bonding process. For instance,

forward bond loops with 50 µm loop height are already in

mass production. Other innovations such as die to die

bonding, cascade bonding, and bonding overhang die are

also being implemented.

But as the price of gold continues to sky rocket, copper

wire will gain widespread use, especially for price-sensitive

consumer electronic devices. Therefore, it is only inevitable

that the technological developments made on low loop wire

bonding of gold wires will have to be extended to copper

wires as well.

This paper presents the development work done on Ultra

Low Loop wire bonding using 20 µm (0.8 mil) palladium

coated copper wire. Actual results will be shared to

demonstrate capability to achieve a maximum loop height of

63.5 µm (2.5 mils). This paper will also discuss details on

the loop type selection, the appropriate test vehicles used,

and the relevant output responses after wirebond. And lastly,

results from the reliability stress tests will also be discussed.

1. Introduction

In order to produce wires with very low loops (for

example, 2.5 mils) using very fine (ex: 0.8 mil) palladium-

coated copper wire, there are a several looping profiles

available such as the flex loop, the escargot loop, and the

folded loop (just to name a few). For this study, the team

selected to use the folded loop method. The selection was

based on Carsem’s long experience using the folded loop

method to produce ultra-low loops using fine gold wire. And

between the 3 methods mentioned, the folded loop induces

the least amount of stress at the neck of the wire.

Fig 1: Flex loop

Fig 2 : Escargot loop

Fig 3 : Folded loop

2. First test – Dummy bonding on leadframe

To demonstrate proof of concept, we performed dummy

bonding on bare leadframe without any die attached. Wires

of varying lengths were bonded using the folded loop

method. The length of the wires ranged from 1mm to 4mm

since these are the typical wire lengths we would expect in a

9x9 QFN package. And the target output responses are as

follows:

1) Loop height below 2.5mils (63.5µm)

2) No neck crack for the first kink

3) Looping consistency

4) Min bond pull strength of 2.5gf

5) No wire sway

Fig 4: First test device bonding layout

3. First test result

We were able to achieve less than 2.5mils (63.5um) loop

height. This is based on a sample size of

wire length.

Fig 5: First test result table

Fig 6: JMP comparison of loop height results (1

on leadframe

performed dummy

bonding on bare leadframe without any die attached. Wires

using the folded loop

wires ranged from 1mm to 4mm

since these are the typical wire lengths we would expect in a

And the target output responses are as

Loop height below 2.5mils (63.5µm)

able to achieve less than 2.5mils (63.5um) loop

of 30 readings per

JMP comparison of loop height results (1

st test)

4. First test result discussion

The Heat Affected Zone (HAZ) has a direct

the minimum loop height you can produce. The HAZ of

gold wires are well documented, but there is n

defined HAZ for palladium coated copper wire. Despite this,

the results from our first test have demonstrated that

(20um) palladium coated copper wire can be bent as low as

gold wire in order to meet the loop height

50um at the neck.

Fig 7: Gold versus Palladium Coated Copper properties

comparison table

5. Second test study – Bonding on daisy chain die

For the second test, we bonded on a 1x1mm daisy chain

die on a 5x5mm MLP package with 3x3mm

paddle. The die was purposely attached at

order to assess different wire lengths

~4mm). The die thickness is 50um die

bondline thickness is 25um. This was chosen in order to

assess the most difficult bond condition

target responses are the same as the first test.

Fig 7: Second test device bonding layout.

6. Second test loop height result

We are able to achieve less than 2.5mils (63.5um) loop

height with folded loop for wire length fr

~4mm. This is based on a sample size of 613 wires.

The Heat Affected Zone (HAZ) has a direct impact on

the minimum loop height you can produce. The HAZ of

gold wires are well documented, but there is no clearly

defined HAZ for palladium coated copper wire. Despite this,

have demonstrated that 0.8mils

alladium coated copper wire can be bent as low as

gold wire in order to meet the loop height target of less than

Fig 7: Gold versus Palladium Coated Copper properties

Bonding on daisy chain die

, we bonded on a 1x1mm daisy chain

with 3x3mm die attach

attached at top left corner in

s (ranging from ~1mm to

50um die and the epoxy

. This was chosen in order to

condition. And the output

target responses are the same as the first test.

Fig 7: Second test device bonding layout.

We are able to achieve less than 2.5mils (63.5um) loop

for wire length from ~1mm to

This is based on a sample size of 613 wires.

36th

International Electronic Manufacturing Technology Conference, 2014

Fig 8: Loop height result table (2

nd test)

Fig 9: JMP comparison chart of loop height results (2

nd test)

7. Second test wire pull strength result

Aside from the samples that were prepared using folded

loop, we also assembled a few units with the traditional

square loop. And we compared the wire pull strength for

additional information.

Fig 10: Wire pull strength result table

Fig 11: JMP comparison chart of wire pull strength (2

nd test)

The wire pull strengths of both folded loop and square

loop can meet the minimum spec limit with Cpk > 1.67.

The average pull strength of folded loop is about 27% lower

compare to standard square loop (~4mils loop height). And

the failure mode for the folded loop 44% break at the neck

while the square had no failures at the neck. This difference

is possibly due to the reduced cross sectional area of the

neck of a folded loop.

Fig 12: Wire pull failure mode histogram

Fig 13: SEM image of a folded loop after wire pull test

36th

International Electronic Manufacturing Technology Conference, 2014

SEM inspection revealed no visible crack at wire neck

area, looping is consistent and no wire sway. Optical

inspection revealed that the Palladium coating is scratched

off by the wirebond capillary. There is partial exposure of

the base copper at the folded region (neck area). To assess

the reliability risk, samples were submitted for Reliability

testing.

Fig 14: SEM image on the ultra low loop

Fig 15: Optical inspection

Fig 16: Folded loop versus Square loop

8. Reliability test result

We have selected 2 test vehicles (32-lead MLPQ 5x5mm

with package thickness of 0.4mm and 16-lead MLPQ

2.3x2.3mm with package thickness of 0.32mm)

manufactured from 2 sites (Carsem Ipoh and Carsem

Suzhou) for reliability test. All samples passed MSL1,

260°C, uHAST 96hours, -65 +150°C condition C

Temperature Cycle 1000 cycles and 150°C High

Temperature Storage 1008hours.

Fig 17: Reliability Test summary table

To confirm the robustness of this 2.5mils (63.5um) ultra

low loop, we also subjected unmolded strips to Temperature

Cycling up to 1000 cycles under -65°C +150°C to see if

there will be any significant degradation of wire mechanical

properties. The results obtained also show wire pull Cpk >

1.67 with acceptable break mode.

Fig18: Unmolded strip wire pull test result after 1000

temperature cycles

Fig 19: JMP comparison chart of unmolded strip wire

pull test after 1000 Temp Cycles

36th

International Electronic Manufacturing Technology Conference, 2014

Fig 20: Wire pull test failure mode distribution after

1000 Temperature Cycles

9. Discussion

Based on our high volume manufacturing experience

with gold wire, we have characterized the Palladium Coated

copper wire thoroughly using appropriate test vehicles. The

results obtained from the first study on bare leadframe

bonding, the second study using 50um thickness test die

bonding and lastly the reliability test and aging test helped

us to develop the suitable bonding parameters for a robust

process for Palladium Coated copper wire.

10. Conclusion

Ultra low loop (folded loop) is feasible for Palladium

Coated Copper wire based on these assessment results:

a) Loop height below 2.5mils (63.5µm) with Ppk >

1.67

b) No neck crack for the 1st kink

c) Looping is consistent

d) Passed minimum bond pull strength of 2.5gf with

Ppk > 1.67

e) No wire sway

The average wire pull strength for folded loop is about

27% lower as compared to standard square loop (~4mils

loop height). This is possibly due to the reduced cross

sectional area of the neck of a folded loop.

Reliability tests and aging tests have confirmed there is

no significant degradation of the wire’s mechanical

properties after environmental stress test.

In closing, this new process capability of ultra low loop

wire bonding will enable the introduction of next-gen ultra-

thin QFN such as the X3.2 package: Carsem’s latest MLP

package introduced in 2014.

Fig 21: 0.9mm package with normal loop versus 0.32mm

package with ultra low loop

Acknowledgments

The authors would like to thank to Carsem Technology

Officer, Mr. LW Yong and Senior R & D Manager, Mr. KH

Lee to initiate this project..

References

1. SH Liew, WL Law “Reliable Ultra-low-loop Bonding

Approach on X2/X3 thin QFN”, 35th International

Electronic Manufacturing Technology Conference,

2012.

2. G. Harman, “Wire bonding in Microelectronics

material, process, reliability and yield”, 2nd

adition,

McGraw-Hill, New York (1997), pp. 203-207.

Normal loop height

4 to 5 mils

(100 to 125 um)

V package

thickness

0.9 mmUltra-low

loop height<2.5 mils

(<63.5 um)

X3.2

package thickness

0.32 mm

Ultra Thin QFN