High frequency DC:DC power conversion:

The influence of package parasitics

Mark Pavier*, Andrew Sawle*, Arthur Woodworth*, Ralph Monteiro**, Jason Chiu**, Carl Blake**

* International Rectifier, Holland Road, Hurst Green, Surrey, RH8 9BB, UK

** International Rectifier, 233 Kansas St., El Segundo, CA, 90245 USA

As presented at APEC 03

Abstract

Operating power MOSFET devices at frequencies over

1MHz will pose significant challenges to established

power electronic packages such as the D2-Pak and

wirebonded SO-8 devices. In this paper the high

frequency parasitic impedances of a range of power

electronic packages are presented. Results show that a

source mounted power package technology based upon

a copper clip type assembly has considerably lower

parasitic impedance compared to conventional power

packaging at frequencies in the range of 500KHz to over

1MHz. The resistance of conventional packages

recorded over this range of frequencies increases

significantly as the frequency approaches 1Mhz. This is

expected to be a result of skin effect related phenomena

occurring in wire bonds and package leads. Package

impedance data up to frequencies of 5MHz will be

presented for a range of packages along with efficiency

data recorded from devices operating in multi-phase

buck converter circuits.

I. INTRODUCTION

The level of integration in desktop and portable PC

microprocessors has increased over the past decade in

line with Moore’s law [1]. During the same time period

clock frequencies have also increased from a few tens of

MHz to over 2 GHz [1]. As a consequence of these two

factors the current requirements of microprocessors have

also significantly increased. Power is usually supplied to

a microprocessor from a fixed 12V supply rail using a

point of load converter or voltage regulator module

(VRM). In order to supply the fast transient current

response required by the processor, converter

frequencies have been increased to over 1MHz and in

many cases multiphase designs have been adopted.

Operating Power MOSFET devices at over 1MHz will

pose significant challenges to established power

electronic packages such as the D2-PAK and the

wirebonded SO-8 devices. In this paper the high

frequency parasitic impedances of D2-Pak, D-Pak, MLP,

SO-8 and a proprietary source mounted package,

DirectFET

, are presented. These packages are shown

in figure 1. Parasitic inductance and resistance are

studied over a frequency range of 500KHz to 5MHz.

Finally, in circuit efficiency data and switching waveforms

of near identical silicon operating in SO-8 and

DirectFET

packages in high frequency 2 phase VRM

circuits are presented in order to highlight the effects of

power package parasitics upon VRM performance.

Figure 1. Packages investigated. Going from left to right:

Top row: DirectFET, D2-Pak and D-Pak

Bottom row: SO-8 and MLP packages

II. PACKAGE ASSEMBLY

Packages were assembled with the silicon die removed

(e.g. die free). In the case of D-Pak, D2-Pak, MLP and

SO-8 devices this involved placing wirebonds on the

silicon bond pad of the device leadframe. Care was

taken to ensure that the wirebonds occupied the same x-

y coordinates on the die bond pads as they would

normally occupy in the packages containing silicon. Die

free samples of the source-mounted package

(DirectFET

) were assembled by replacing the silicon

die with a stamped copper die of equivalent dimensions

to the original silicon. This was required in order that the

underside of the DirectFET

clip could be brought into

contact with the source electrode printed circuit board

pads.

Devices were assembled onto test cards fabricated from

double-sided FR4 with 2 Ounce copper tracking.

Sn62Pb36Ag2 near eutectic no clean solder (Multicore

SN62MP100AGS90) was screen-printed onto test cards

prior to component pick and place. After component

placement test cards were re-flowed using a standard

JEDEC profile and visually inspected for defects.

III. TEST PROCEDURE

Resistance and inductance measurements were

performed using an Agilent/HP4285A high precision LCR

meter. A custom designed test fixture was used to

interface the device test cards to the LCR output current

source/voltage sense (Kelvin) terminals. The test

procedure was as follows:

1) Insert ‘open circuit test card’ into test fixture &

perform OPEN circuit correction

2) Insert ‘short circuit test card’ into test fixture &

measure SHORT circuit R and L values

3) Insert device test card into test fixture and

measure LOAD R and L

Note that the above procedure was carried out for each

device at each frequency tested. The HP4285A was

operated in the correction mode for improved low

resistance repeatability. Figures 2(a), (b) and (c) show

the OPEN circuit test card, short circuit test card, and

device test card respectively, for the DirectFET

package. It should be noted that in order to obtain low

resistance short circuit measurements the Kelvin drain

and source via holes were located as close together as

possible. The resultant short circuit obtained between

Kelvin vias was calculated to be in the region of 40

µOhms. Note that the track impedances from the Kevin

vias to the device under test are included in all of the

experimental results shown below. The results

presented below are therefore for devices mounted on a

board, not discrete devices. Figure 3 shows the test card

designs for each of the packages investigated. The

circle superimposed onto each of the images indicates

the location of each of the sense vias. In each case

these vias were located as close as possible to edge of

the package, and where possible underneath the

package leads.

Figures 2 from left to right

(a) OPEN, (b) SHORT and (c) DEVICE UNDER TEST Cards.

Figures 3 from left to right

D2-Pak, D-Pak, SO-8, MLP and DirectFET

test cards.

Notes circles highlight sense via location.

IV. PACKAGE RESISTANCE CHARACTERISTICS

Figure 4 shows the board mounted package resistance

versus frequency for each of the packages characterised.

The DirectFET

package shows the lowest board

mounted package resistance from DC up to 5MHz. At

frequencies of 1MHz and above it should be noted that

surface mount devices based upon conventional leaded

packages, such as the D2-Pak and D-Pak, show very

significant increases in package resistance in relation to

values measured at DC.

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 1 2 3 4 5

Frequency [MHz]

Board mounted package resistance [mOhm]

DFET

MLP

SO-8

D-Pak

D2-Pak

Figure 4. Measured package resistance versus frequency

With reference to figure 4 each mounted package

exhibits a characteristic increase in resistance with

Short circuit

G D S

Kelvin vias

circuit