• No results found

Design-In for EMC on digital circuit

N/A
N/A
Protected

Academic year: 2021

Share "Design-In for EMC on digital circuit"

Copied!
19
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

MICAS Department of Electrical Engineering (ESAT)

Design-In for EMC on digital circuit

December 5th, 2005

Low Emission Digital Circuit Design

Junfeng Zhou Wim Dehaene

KULeuven ESAT-MICAS

(2)

MICAS Department of Electrical Engineering (ESAT)

Outline

1. Introduction

2. Logic family selection 3. Clock strategy selection

SSCG - Delay cell array approach 4. Low noise power supply

(3)

MICAS Department of Electrical Engineering (ESAT)

Part I: Introduction

Electro-Magnetic Interference (EMI) and radiated emission have become a major problem for high speed digital circuit,

Most of them are due to power and ground fluctuation.

Although the detailed calculation of EMI noise is rather difficult , we can use the di/dt as the index, since the current loop

contributes the EMI.

(4)

MICAS Department of Electrical Engineering (ESAT)

Part 2: Logic Family Selection

SCMOS PNMOS RSBCMOS

CSL MCML FSCL

(5)

MICAS Department of Electrical Engineering (ESAT)

Comparison of di/dt ,power and area

Target : Mixed-Mode Automotive Electronics Design Key aspects : di/dt + Power + Area + Speed

Inv erter Power in different logic techniques

0.00E+00 5.00E+01 1.00E+02 1.50E+02 2.00E+02 2.50E+02

SCMOS PNMOS RSBMOS CSL MCML FSCL

Power [μW]

Inv erter area in different logic techniques

0 0.5 1 1.5 2 2.5 3 3.5 4

SCMOS PNMOS RSBMOS CSL MCML FSCL

Area m2]

Ring Oscillator of 21-stages

(Static + Dynamic)

Current Steering Logic

But there is static power !!

(6)

MICAS Department of Electrical Engineering (ESAT)

Detailed comparison of CSL and SCMOS

0 20 40 60 80 100 120 140 160 180

10-6 1x10-5 1x10-4 10-3

Frequency [ MHz ]

Power consumption [ W ]

CSL dynamic power

CSL static power CSL dynamic power CSL power

(activity=0.5) SCMOS power (activity=0.5)

SCMOS power activity=0.5 CSL power activity=0.5 CSL static power

VDD=1.5v

Note:

The curve of CSL 16-bit RCA was obtained by calculating the real speed F of the circuit, given the different supply current I.

CSL

One-bit Adder

IT is a static power problem, Switching off when standby ?

(7)

MICAS Department of Electrical Engineering (ESAT)

Detailed comparison of CSL and SCMOS

SCMOS

CSL

(8)

MICAS Department of Electrical Engineering (ESAT)

Problem with CSL

Mismatch sensitive, annoying for standard cells

rather slow/power hungry

Not full swing

Matching required!

M1 > M3

(9)

MICAS Department of Electrical Engineering (ESAT)

Can we do it better ?



C-CBL:

sizing for optimal current balance is really difficult ,process dependent

CBL

[Albuquerque, E.F.M.; Silva, M.M., Current-balanced logic for mixed-signal IC's]

(10)

MICAS Department of Electrical Engineering (ESAT)

Solution- Enhanced current steering logic

Still current source basing

Increase in logic level, hence increase the robustness

Reduced output capacitance, hence the speed is increased

Fig.3 E-CSL inverter

Minimum size

(11)

MICAS Department of Electrical Engineering (ESAT)

Comparison of CSL, C-CBL,

ECSL and SCMOS

10M 100M 1G 10G

1 10 100 1k 10k 100k 1M 10M 100M

CSL C-CBL E-CSL SCMOS VDD=3 v CLOAD=5 fF

F [Hz]

di/dt p-p [A/s]

Fig.5 di/dt vs. frequency Fig.4 power vs. frequency

10M 100M 1G 10G

1E-7 1E-6 1E-5 1E-4

1E-3 Area@500 MHz

CSL 7.2 um2 C-CBL 2.1 um2 E-CSL 1.53 um2 SCMOS 6.5 um2

VDD=3 v CLOAD=5 fF

Power [w]

F [Hz]

CSL C-CBL E-CSL SCMOS

Ring Oscillator of 21-stages

(12)

MICAS Department of Electrical Engineering (ESAT)

di/dt performance vs.

process variation

10M 100M 1G 10G

10 100 1k 10k 100k 1M 10M

100M VDD=3 v CLOAD=5 fF

TT FF FS SF SS F [Hz]

di/dt p-p [A/s]

CSL

10M 100M 1G 10G

10 100 1k 10k 100k 1M 10M

100M VDD=3 v CLOAD=5 fF

TT FF FS SF SS F [Hz]

di/dt p-p [A/s]

E-CSL

10M 100M 1G 10G

10 100 1k 10k 100k 1M 10M

100M VDD=3 v CLOAD=5 fF

TT FF FS SF SS

F [Hz]

di/dt p-p [A/s]

C-CBL

10M 100M 1G 10G

10 100 1k 10k 100k 1M 10M

100M VDD=3 v CLOAD=5 fF

TT FF FS SF SS di/dt p-p [A/s]

F [Hz]

SCMOS

Fig.6 di/dt vs. process corner

MAX di/dt change

MIN di/dt change

Ring Oscillator of 21-stages

(13)

MICAS Department of Electrical Engineering (ESAT)

Conclusion of Low noise Logic Families

Winner is E-CSL

CSL,E-CSL show a smaller area per logic function for complex digital gates and systems compared to SCMOS logic technique.

Current source ensures the major di/dt reduction,

Process variation sensitivity also becomes better due to the dominance of current source,

E-CSL gives comparable di/dt performance with CSL,

E-CSL is Faster and Less power consuming than CSL due to the lower area and lower capacitance.

Static power consumption remains the challenge for wide application of the CSL,E-CSL technique in very large

digital systems. Can be solved by using power down strategies, which is highly application dependent

(14)

MICAS Department of Electrical Engineering (ESAT)

Part II: Clock strategy (SSCG)

[Keith B. Hardin, Spread Spectrum Clock Generation for the Reduction of Radiated Emissions]





MOD o SSC

dB f

f

A n

log10

10

1. Deviating the period of the clock signal from its fundamental by a small

percentage(usually +/- 1% ) and in a predictable fashion(usually Triangular modulation profile )

2. The total power of the clock signal remains the same.

Implementation:

1. PLL-SSCG: VCO has its input voltage controlled by a modulation waveform.

2. DCA-SSCG: By controlling the temporal spacing of the edges, the clock’s frequency is indirectly

controlled

(15)

MICAS Department of Electrical Engineering (ESAT)

SSCG - PLL vs. DCA ?

Disadvantage of PLL-SSCG:

Basically analog circuit(VCO, charge pump, loop filter), more susceptible to noise

PLL-SSCG suffers from the drawback of reduction in maximum achievable EMI reduction due to the inherent random jitter of the circuitry(due to thermal noise, flicker noise).

Leads to large jitter in clock which is unacceptable

Advantage of DCA-SSCG

Digital circuits, good immunity to noise.

Leads to smaller random jitter, simpler implementation and reduction in area

The reduction in variance of unintentional jitter is key to the delay cell array technique being able to achieve greater reduction in EMI.

(16)

MICAS Department of Electrical Engineering (ESAT)

SSCG-DCA: How does it work ?

Each delay cell comprised of delay element and a positive latch:

EN D Q

EN D Q

EN D

Q

Delay Cell #1 Delay Cell #2 Delay Cell #N

Result: Edge-to-edge jitter varied in deterministic fashion.

f0

SSC: f0 + ∆f

N/2 Counter Q T

T Flip-Flop

Delay Cell Control

(17)

MICAS Department of Electrical Engineering (ESAT)

Clock Attenuation

Our results show:

Dominant power in odd harmonics

DCA-SSCG:

Delay cell based SSCG

implemented shows low power and simple circuit implementation

8dB of clock attenuation on fundamental

Improved design can be achieved by using differential delay cell

element

(18)

MICAS Department of Electrical Engineering (ESAT)

Part III: Low Noise Power supply design

However 2 problems still remain:

• Static power consumption

• New logic family standard cell must be designed and characterised

?? Is there any global approach ??

(19)

MICAS Department of Electrical Engineering (ESAT)

Questions

Thank you for your attention

Referenties

GERELATEERDE DOCUMENTEN

Figure 4.3: 4x4 Torus Topology Throughput VS Injection rate (flits/cycle) In Figure 4.4 a comprehensive comparison of average packet latency vs injection rate has been obtained

The methods developed in this thesis address three global problems: (1) the efficient and accurate reduction of multi-terminal circuits, (2) the appropriate synthesis of the

In the next section, we describe three simulation methods to price European call options: importance sampling, naive Monte Carlo, and antithetic variable.. The estimators obtained

In this paper we propose a Lanczos–like algorithm to trans- form a symmetric matrix into a similar semiseparable one that, similarly to the Lanczos tridiagonalization, relies on

behoud van stabiliteit dan de laatstgenoemde gewogen modelreductie, moet er een bijkomende reden zijn om relatieve fout modelreductie te verkiezen voor e en welbekende

Joos Vandewalle, Bart De Moor, Yves Moreau, Marc Moonen, Johan Suykens, Moritz Diehl, Lieven De Lathauwer, Sabine Van Huffel, Jan Engelen.. SISTA - COSIC

Joos Vandewalle, Bart De Moor, Yves Moreau, Marc Moonen, Johan Suykens, Moritz Diehl, Lieven De Lathauwer, Sabine Van Huffel, Jan Engelen.. SISTA - COSIC

Joos Vandewalle, Bart De Moor, Yves Moreau, Marc Moonen, Johan Suykens, Moritz Diehl, Lieven De Lathauwer, Sabine Van Huffel, Jan Engelen.. SISTA - COSIC