Do We Have RF Performance Degradation Due To Substrate Noise?
Domine Leenaerts Philips Research Labs.
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Contents
Introduction to substrate noise-coupling. What can we do; guard rings. Is this enough? RF examples. Conclusions.
• • • •
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction
• Substrate is connecting layer between all circuits on a single die. • If a circuit produces spiky signals (‘noise’) on the substrate, other circuits will ‘feel’ this. • Amount of generated noise on substrate is therefore an issue. You can minimise the influence of substrate coupling but not eliminate it. Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction; injection paths • PN-junction of MOS (mainly in digital circuits) – capacitive coupling to substrate – can be several hundreds pico Farad for digital
• N-Well
$
+
)*
"
&
# (
&
(
'
%
#
!"
– forms PN-junction with substrate – capacitive coupling between Vdd and substrate
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction; injection paths
• Bond wires
/
6
45
-
1
. 3
1
3
2
0
.
,-
– MOS logic produces spikes on supply lines – Vss tightly connected to substrate – vdrop = Lbondwire ⋅ dI spike / dt – off-chip, digital ground connected to substrate Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction; reception paths • Bulk contacts – substrate is connected to ground via bulk contacts – resistive coupling
• PN-junction MOS / N-Well – capacitive coupling into supply lines – capacitive coupling into signal lines
• Backgate MOS :
A
?@
8
<
9 >
<
>
=
;
9
78
– injection into signal path Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction • Used technology is important
E
L
JK
C
G
D I
G
I
H
F
D
BC
– high resistive substrate – low resistive substrate on high resistive epi
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction; problem?
P
W
UV
N
R
O T
R
T
S
Q
O
MN
• 2nd order low pass Sigma-Delta Mod • 0.25 µm CMOS, 10 mΩ.cm substrate on top of 11 Ω.cm epi of 3 µm • 1.8 V supply voltage • separate digital/analogue power lines • design rules to minimise substrate bounces applied Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction; problem? 8 bit 1 MHz
13.5 MHz
Filter
Quantizer
[
1 bit 216 MHz
b
`a
Y
]
Z _
]
_
^
\
Z
XY
DAC
Decimation
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Introduction; problem?
20dB
Clock off
Clock on
f
m
kl
d
h
e j
h
j
i
g
e
cd
Measured spectrum of signal at substrate Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
What can we do? Some people claim that you must use p-guard rings and n-guard rings around your circuits. We have tested this opinion for a 0.18 µm CMOS technology using high and low resistive substrate
q
x
vw
o
s
p u
s
u
t
r
p
no
(J. Pluymakers et.al. PS-LTG AMOS)
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Guard rings • 10 mΩ.cm substrate of 200 µm on top of 11 Ω.cm epi of 3 • 10 Ω.cm substrate of 200 µm
|
z
~
{
~
}
{
yz
Vdd
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Guard rings
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Guard rings
13 dB improvement of suppression relative to the situation without P-ring
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Guard rings
Only 4 dB improvement!
¤
¢£
¡
¡
N-Ring deteriorates the effect of the P-ring
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Guard rings; low ohmic substrate • N-ring around RF circuit does not help you • P-ring around RF circuit gives -13 dB improvement compared to no ring – Nail the substrate as good as possible to the ground
• Use N-ring only around your source of noise
¨
¯
®
¦
ª
§ ¬
ª
¬
«
©
§
¥¦
– collect ‘wandering’ electrons
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Guard rings; high ohmic substrate • N-ring around RF circuit does not help you • P-ring around RF circuit gives -40 dB improvement compared to no ring • Place source as far away as possible – 50 µm: -34 dB, 100 µm: -40 dB (reference: -24 dB)
³
º
¸¹
±
µ
² ·
µ
·
¶
´
²
°±
• Use N-ring only around your source of noise
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Is this enough? • Consider the low ohmic substrate • Can the substrate noise influence the RF performance of a certain circuit?
¾
Å
ÃÄ
¼
À
½ Â
À
Â
Á
¿
½
»¼
• Test circuit in 0.25 µm CMOS, 10 mΩ.cm substrate on top of 11 Ω.cm epi of 3 µm
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: circuit 1
2
11
buffer
substrate
0.2
-15
0 -0.2 -0.4
-0.8
0
2
4
4
6 x 10
9
-0.8 s 22 (dB)
s 21 (dB) Ë
Ç ÎÏ
Ð
ÆÇ
É
È
È
Ê Ì
2
-0.6
-6
-1 -1.2
-8
Í
0
9
-4
-10
Ë
-25 -30
6 x 10
0
Í
-20
-0.6
-2
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
clock out C
s 12 (dB)
11-stage ring oscillator center freq. 772 MHz Unmatched LNA: 2-5 GHz Wafer probe measurements: substrate noise due to pnjunction mechanism
s 11 (dB)
• • • •
Vtune
0
2 4 fre que nc y
6 x 10
9
-1.4
0
2 4 freque nc y
6 x 10
9
Ô
Û
ÙÚ
Ò
Ö
Ó Ø
Ö
Ø
×
Õ
Ó
ÑÒ
RF example: circuit
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: circuit • Spectrum substrate • inside guard ring • LNA off
fundamental 2nd harmonic
ß
æ
äå
Ý
á
Þ ã
á
ã
â
à
Þ
ÜÝ
3rd harmonic
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: circuit 2.3 GHz 772 MHz
ê
LNA and clock operational; spectrum at output LNA
ñ
ïð
è
ì
é î
ì
î
í
ë
é
çè
Only LNA is operational @ 2.3 GHz; spectrum at output LNA
2.3 GHz & 3rd harmonic of 772 MHz
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: circuit; conclusion
õ
ü
úû
ó
÷
ô ù
÷
ù
ø
ö
ô
òó
• Not only amplified RF input signal at output LNA, but also (weakened) injected substrate noise at the drain. DISTORTION! • If LNA has gain, then substrate noise injected at RF input will be amplified. • S-parameters did not change, performance LNA is not changed due to substrate noise. Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: system • What if you have large digital circuitry and RF circuitry on the same substrate?
þ
ÿ
ÿ
ýþ
• Test circuit in 0.18 µm CMOS, 10 Ω.cm substrate
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: system
• RF front-end @ 2.45 GHz • 500 kgates digital • 1.8 V
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: system
• VCO runs at 2.45 GHz • Clock for digital is tuneable • Measure output spectrum VCO
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
RF example: system
(
#
!
&'
%
#
%
$
"
Clock digital off Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Clock digital: 13 MHz
RF example: system
,
3
12
*
.
+ 0
.
0
/
-
+
)*
Clock digital: 40 MHz Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Clock digital: 64 MHz
RF example: system; conclusion
7
>
<=
5
9
6 ;
9
;
:
8
6
45
• FM modulation between digital clock switching and VCO • Can not be prevented by guard ring techniques. • Only smart system design can minimise the effects of substrate noise on system performance
Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs
Concluding remarks
B
I
GH
@
D
A F
D
F
E
C
A
?@
• Yes, substrate noise coupling is an important issue in RF design • Yes, guard rings together with high ohmic substrate reduce the coupling • Yes, smart system and circuit design can reduce it too • No, you can not eliminate it Domine Leenaerts, Workshop on Substrate Noise-coupling in Mixed-Signal ICs