31
The new E-port interface circuits Filip Tavernier CERN

The new E-port interface circuits Filip Tavernier CERN

Embed Size (px)

Citation preview

Page 1: The new E-port interface circuits Filip Tavernier CERN

The new E-port interface circuits

Filip TavernierCERN

Page 2: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 2

Required modifications to the previous version

• modify the TX so that it has a high-impedant output when disabled

• add a programmable (ON/OFF) termination network to the RX

• modify the RX so that it can handle both SLVS and LVDS signals

• boost the RX operation speed so that it can handle a 320 MHz clock signal(‘bandwidth switch’ to trade speed for power?)

• make a bi-directional cell

Page 3: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 3

Architecture of the SLVS TX

• signals:– input: digital single-ended voltage– output: analog differential current

• building blocks:– predriver: converts the input signal into a differential

digital voltage– driver: converts the differential voltage into a

differential current

predriver driver

Page 4: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 4

The previous TX predriver

outn

outp

latp1

latp3

latn2

latn1

latn3

latp2

latp1

latp3

latn2

latn1

latn3

latp2

GND

GND

GND

GND

GND

GND

offb1

offb3

offb2in

off

offb1

offb3

offb2

cset<0:3>

cset<0:3>

cset<0:3>

off = HIGH → outp = LOW & outn = HIGH

Page 5: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 5

The TX driver

inn

inpinn

inp

at the RX

bias1

bias2zero-Vt

outn outp

• off = LOW– inp = HIGH & inn = LOW

→ current flows ‘from right to left’ in the termination resistor

– inp = LOW & inn = HIGH→ ‘from left to right’

• off = HIGH– bias1 = bias2= 0 V– a low-impedant path still exists

between outp and GND because inn = HIGH and the zero-Vt device is not completely off

Page 6: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 6

The new TX predriver

outn

outp

latn1

latn3

latn2latn2

latn1

latn3

GND

GND

GND

GND

GND

GND

offb1

offb3

offb2

off

offb1

offb3

offb2

cset<0:3>

cset<0:3>

cset<0:3> latn1

latn3

latn2latp2

latp1

latp3

offb1

offb3

offb2

in

off = HIGH → outp = LOW & outn = LOW

Page 7: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 7

Simulation of the new SLVS TX• Simulation of the TX in the bi-directional cell• Simulation of RC-extracted netlists• Simulation in 3 corners:

typical: VDD = 1.2 V (1.5 V) / T = 27° C / TT modelsslow: VDD = 1.1 V (1.4 V) / T = 100° C / SS modelsfast: VDD = 1.35 V (1.65 V) / T = -30° C / FF models

• Simulation with the ‘nominal’ current setting (2 mA)• Input signal: 27-1 PRBS @ 640 Mbit/s• Off-chip termination resistor of 100 Ω• Package modeled by 1 nH bondwire inductors and 1 pF

off-chip parasitic capacitance

TX

RX

Page 8: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 8

Simulation of the SLVS TX @ VDD = 1.2 V

current [mA] / power [mW]:• typical: 3.31 / 3.97• slow: 2.73 / 3.00• fast: 4.13 / 5.58

Page 9: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 9

Simulation of the SLVS TX @ VDD = 1.5 V

current [mA] / power [mW]:• typical: 4.01 / 6.02• slow: 3.41 / 4.77• fast: 4.93 / 8.13

Page 10: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 10

SLVS TX layout

Page 11: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 11

SLVS and LVDS signals for the RX

single-ended input signal levels when terminated with a differential 100 Ω resistor:

→ RX termination network needs nMOS (SLVS) and pMOS switch (LVDS)

time

volt

age

[V

]

1.2

0.2

LVDS

SLVS

0.4 V

0.2 V

term

termb

50 Ω 50 Ω

→ 4 mA in 100 Ω

→ 2 mA in 100 Ω

Page 12: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 12

Amplifier of the SLVS/LVDS RX

inverter buffer

outinn

inn

inp

inp

biasn

biasp

input pair for SLVS(off for LVDS)

input pair for LVDS(off for SLVS)

reduce sensitivityof the current wrt

variation of VDD

Page 13: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 13

Biasing network of the SLVS/LVDS RX

biasn biasp

M1 M2

M3M4

M5

(W/L)pk(W/L)p

(W/L)n (W/L)n

R

• current in the 2 branches is equal and completely determined by: (W/L)p, k and R:

• M5 is a start-up device to force that biasn ≠ 0 V and biasp ≠ VDD: → Vth,1 + |Vth,2| + Vth,5 < VDD,min = 1.1 V

→ Vgs,1 + |Vgs,2| + Vth,5 > VDD,max = 1.65 V

2

2

11

12

kRLWC

Ipoxp

Page 14: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 14

• Simulation of the RX in the bi-directional cell• Simulation of RC-extracted netlists• Simulation in 3 corners:

typical: VDD = 1.2 V (1.5 V) / T = 27° C / TT modelsslow: VDD = 1.1 V (1.4 V) / T = 100° C / SS modelsfast: VDD = 1.35 V (1.65 V) / T = -30° C / FF models

• Simulation with the on-chip termination network activated

• Input signal: 27-1 PRBS @ 640 Mbit/s• Package modeled by 1 nH bondwire inductors and 1 pF

off-chip parasitic capacitance

Simulation of the new SLVS/LVDS RX

TX

RX

Page 15: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 15

Simulation of the RX @ VDD = 1.2 V with SLVS input signals

current [µA] / power [µW]:• typical: 320 / 384• slow: 310 / 341• fast: 298 / 402

Page 16: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 16

Simulation of the RX @ VDD = 1.5 V with SLVS input signals

current [µA] / power [µW]:• typical: 391 / 587• slow: 439 / 615• fast: 390 / 644

Page 17: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 17

Simulation of the RX @ VDD = 1.5 V with LVDS input signals

current [µA] / power [µW]:• typical: 453 / 680• slow: 474 / 664• fast: 481 / 794

Page 18: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 18

Input CM range of the RX amplifier

• In theory, the RX amplifier has a rail-to-rail CM input range thanks to the input stage with nMOS and pMOS transistors.

• In reality, the voltage gain changes significantly with varying CM input voltage because the input transistors go out of saturation at ‘extreme’ CM voltages, as such reducing the gain of the amplifier.

• Note that the correct operation of the RX amplifier relies on the clipping of the signal at the input of the inverter (see slide 12) so that it can make a reliable decision. This means that the amplifier should not work in its linear region.

• Therefore, the effective CM input range varies with the input signal amplitude (a larger signal can work over a larger CM input range), or alternatively, the minimal signal amplitude depends on the CM input voltage (a smaller signal can be used at the ‘optimal’ CM input voltage than at the extremes).

Page 19: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 19

Input range of the RX @ VDD = 1.2 V for large signals in the typical corner

→ typical corner CM input range @ VDD = 1.2 V for ‘large’ signals : 0 V – 1 V

Page 20: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 20

Input range of the RX @ VDD = 1.2 V for small signals in the typical corner

→ typical corner CM input range @ VDD = 1.2 V for ‘small’ signals : 0.1 V – 0.9 V

Page 21: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 21

Input range of the RX @ VDD = 1.5 V for large signals in the typical corner

→ typical corner CM input range @ VDD = 1.5 V for ‘large’ signals : 0 V – 1.3 V

Page 22: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 22

Input range of the RX @ VDD = 1.5 V for small signals in the typical corner

→ typical corner CM input range @ VDD = 1.5 V for ‘small’ signals : 0.1 V – 1.2 V

Page 23: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 23

Input range of the RX @ VDD = 1.2 V for large signals in the slow corner

→ slow corner CM input range @ VDD = 1.2 V for ‘large’ signals : 0.1 V – 0.9 V

Page 24: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 24

Input range of the RX @ VDD = 1.2 V for small signals in the slow corner

offset frominverter threshold

→ slow corner CM input range @ VDD = 1.2 V for ‘small’ signals : 0.2 V – 0.8 V

Page 25: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 25

Input range of the RX @ VDD = 1.5 V for large signals in the slow corner

→ slow corner CM input range @ VDD = 1.5 V for ‘large’ signals : 0 V – 1.2 V

Page 26: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 26

Input range of the RX @ VDD = 1.5 V for small signals in the slow corner

→ slow corner CM input range @ VDD = 1.5 V for ‘small’ signals : 0.2 V – 1 V

Page 27: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 27

Input range of the RX @ VDD = 1.2 V for large signals in the fast corner

→ fast corner CM input range @ VDD = 1.2 V for ‘large’ signals : 0 V – 1.2 V

Page 28: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 28

Input range of the RX @ VDD = 1.2 V for small signals in the fast corner

→ fast corner CM input range @ VDD = 1.2 V for ‘small’ signals : 0 V – 1.1 V

Page 29: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 29

Input range of the RX @ VDD = 1.5 V for large signals in the fast corner

→ fast corner CM input range @ VDD = 1.5 V for ‘large’ signals : 0 V – 1.5 V

Page 30: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 30

Input range of the RX @ VDD = 1.5 V for small signals in the fast corner

→ fast corner CM input range @ VDD = 1.5 V for ‘small’ signals : 0.1 V – 1.5 V

Page 31: The new E-port interface circuits Filip Tavernier CERN

Filip Tavernier - CERN 31

SLVS/LVDS RX layout