Do You Know How Much Capacitance Do You Really Get? I Presenter: Istvan Novak

• A Real-life Scenario

• The Problem

• The Reasons

• Does It Matter?

• Possible remedies

• Conclusions

INTRODUCTION

A Real-Life Scenario

• You purchase a 1uF +-20% ceramic capacitor.

• When it comes in, you measure it: it reads 0.6 uF.

• You send it back to the vendor for retesting.

• It comes back from the vendor: they say the part is in spec.

• You measure it again, you still get 0.6 uF.

• You check the measurement setups at both ends with precise reference pieces and conclude that measurements at both ends were correct.

How is this possible?

So What Is Going On?

To get a clue, we need to look at the way how the measurements are done.

• Vendors use standardized measuring conditions: impedance bridge, source voltage of 1Vrms or 0.5Vrms with 100 or 120Hz frequency. First hint: a typical capacitor at the measurement frequency is high impedance: most of the source voltage appears across the capacitor.

• You more likely use a small pcb as test fixture and a VNA with 0dBm source power at frequencies above 1 kHz. Second hint: in VNA measurements and in typical bypass applications the voltage across the capacitor is single or at most double digit mV.

MLCC Material and Construction

Layer count

N = H/th

th

VE

th

GLWNC r

=

−=

)2(0

Class II and higher ceramic materials are ferroelectric

Ferroelectric materials have saturating hysteretic D-E curves

and C are voltage dependent!

D

E

Effect of Temperature on MLCC Capacitance

EIA Class II and Class III ceramics

First character:Z: + 10CY: - 30CX: - 55C

Second character:2: + 45C4: + 65C5: + 85C6: + 105C7: + 125C8: + 150C9: + 200C

Third character:F: +- 7.5%P: +- 10%R: +- 15%S: +- 22%T: + 22 / - 33%U: + 22 / - 56%V: + 22 / - 82%

DC Bias Dependence (Old School Info)

Percentage loss of capacitance [%]

-80

-60

-40

-20

0

20

1.E-01 1.E+00 1.E+01 1.E+02DC bias percentage of rated voltage [%]

Y5V

X5R

X7R

For some time, it was a

common assumption that X7R

MLCCs had less DC bias

sensitivity than X5R parts.

But lately…

X5R vs X7R at 10mV AC

Capacitance [F]

0.E+0

2.E-7

4.E-7

6.E-7

8.E-7

1.E-6

-20 -10 0 10 20

DC bias [V]

A5

B5

B7

C5

D7

F(1)5

F(1)7

F(2)5

Percentage capacitance [%]

-100

-80

-60

-40

-20

0

-20 -10 0 10 20DC bias [V]

F(2)5 D7

C5

A5

F(1)7

F(1)5B5

B7

1uF 0603 16V X5R and

X7R

Samples from Vendors A,B,C,D,F

Last digit refers to X5R or X7R

For more info see http://www.electrical-integrity.com/Paper_download_files/DCE11_200.pdf

AC Bias Dependence: The Problem

Testing and typical

usage apply very

different AC levels.

User application

Vendor testing

AC Bias Dependence

4.7uF 0805-size 16V X5R parts from Vendor F

Capacitance [F]

0.E+0

1.E-6

2.E-6

3.E-6

4.E-6

5.E-6

-20 -15 -10 -5 0 5 10 15 20DC bias [V]

1V

0.5V

0.2V

0.1V

50mV

20mV

10mV

5mV

2mV

1mV

AC bias [Vrms]

Percentage capacitance loss [%]

-100

-80

-60

-40

-20

0

-20 -15 -10 -5 0 5 10 15 20DC bias [V]

1m

2m

5m

10m

20m

30m

0.1

0.2

0.5

1

AC bias [Vrms]

Capacitance [F]

0.E+0

1.E-6

2.E-6

3.E-6

4.E-6

5.E-6

1 10 100 1000

AC bias voltage [mVrms]

0

2

4

6

8

10

12

14

16

18

20

DC bias [V]

High AC bias increases

capacitance at low DC bias.

At high DC bias the AC bias

sensitivity is lower.

ESR and ESL vs Bias

• ESR does not change above SRF• ESR increases below SRF as C drops• Piezo effect shows up with increasing bias• ESL shows no measurable difference

ESR [Ohm]

0.01

0.1

1

1.E+5 1.E+6 1.E+7

Frequency [Hz]

0V

20V

Beware of Details

• Sensitivity vs. body height

• Data from vendor

• Lower body height comes

with higher sensitivity

Percentage capacitance, Vendor-C [%]

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

10

0 2 4 6 8 10 12 14 16DC bias [V]

X5R 0.5mmX5R 0.8mm

X7R 0.8mm

OK, Less Capacitance, So What?

In SI (less of an issue)

If OK for the design, just oversize the capacitance (worst case: by 6x)

Design could become space limited

In differential DC block the two legs may become unbalanced at low frequencies

In PI (much more serious)

Results in more droop (in case of single capacitor bank)

Results in peaking (more noise)

May result in current amplification, increased dissipation

OK, Less Capacitance, So What?

1uF 0603-size 16V X7R part from Vendor-D and 47uF 1206-size 6.3V X5R part from Vendor-E

All three resonances

shift.

Impedance magnitude [dBOhm]

-55

-50

-45

-40

-35

-30

-25

1.E+5 1.E+6 1.E+7

Frequency [Hz]

10V8V

6V4V

2V

0V

For more info: http://www.electrical-integrity.com/Paper_download_files/DC16_Paper_ElectricalAndThermalConsequencesOf_Choi.pdf

Filter Response vs. DC Bias Voltage

No DC current bias through L1

No change below 10 kHz and above 1 MHz

No change in peaking

Peak frequency and cut-off frequency increases with increasing bias

Voltage transfer function [dB]

-100

-80

-60

-40

-20

0

20

1.E+2 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7

Frequency [Hz]

Voltage transfer function [dB]

-40

-30

-20

-10

0

10

1.E+4 1.E+5Frequency [Hz]

0 10VDC bias

Zoom

24 kHz 55 kHz

Change Due to DC Bias Voltage and Current

Voltage transfer function, 0-4V, 0-1A [dB]

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

10

1.E+3 1.E+4 1.E+5 1.E+6 1.E+7

Frequency [Hz]

50 dB

3.2 : 1

Voltage transfer function, 0-16V, 0-1.5A [dB]

-90

-80

-70

-60

-50

-40

-30

-20

-10

0

10

1.E+3 1.E+4 1.E+5 1.E+6 1.E+7

Frequency [Hz]

68 dB

7.2 : 1

For more info see http://www.electrical-integrity.com/Paper_download_files/DC19_Tutorial_SLIDES_HowToDesignGoodPDNFilter.pdf

What Else May Change the Capacitance

Percentage range [%] Relative multiplier

Initial tolerance +-10 0.9 … 1.1

Temperature effect +-15 0.85 … 1.15

DC bias effect +0 -70 0.3 … 1

AC bias effect +0 -30 0.7 … 1

Aging (over 3 years) +0 -7.5 0.925 … 1

When we multiply the worst-case contributors, we get 0.9*0.85*0.3*0.7*0.925 = 0.15, which means in worst case instead

of 1uF we have only 0.15uF capacitance.

Does This Happen In All Capacitors?

Luckily, NO.

This happens in high Dk MLCCs

It does not happen in

Low Dk MLCCs (COG/NPO), but we don’t get high density either

Electrolytics

Tantalum

Polymer

Film

How To Simulate

S-parameter models don’t easily scale

SPICE equivalent circuits scale easily, but in complex models, scaling is tricky due to multiple Cs describing a single device

Best option: dynamic models (available for most Murata MLCCs)

For more info on dynamic models: http://www.electrical-integrity.com/Quietpower_files/QuietPower-36.pdf

How To Mitigate

If the DC-AC bias effects have to be reduced, consider to

Create circuits, which are less sensitive to capacitance variations

Check carefully bias curves (if available) and select ‘better’ parts

Avoid using MLCC parts with the highest capacitance density

Avoid using MLCCs

Conclusions

DC and AC bias effect can reduce MLCC capacitance substantially

Total capacitance loss can be up to 80-90%

Capacitance loss can increase droop, resonance peaks and degrade filter

performance

For information about Samtec’s gEEk® spEEk presentations,contact: gEEkspEEk@samtec.com

For Signal Integrity questions, contact: SIG@samtec.com

To view previous gEEk® spEEk webinar recordings, go to www.samtec.com/geekspeek