Professional Documents
Culture Documents
Integration, The VLSI Journal: Didem Erol As, Mustafa Berke Yelten
Integration, The VLSI Journal: Didem Erol As, Mustafa Berke Yelten
Keywords: This paper focuses on designing low-power, low-noise amplifiers (LNA) performances. Different LNA topologies
Ultra-low-power operating with sub-mW power consumption at 2.4 GHz have been implemented in a commercial 40 nm CMOS
Linearity improvement process. The LNA1 (cascode common source LNA) has a voltage gain of 12.22 dB, a noise figure (NF) of
LNA
4.35 dB, and a third-order input intercept point (IIP3) of −12.68 dBm at 995.6 μW while the proposed LNA2
Complementary metal-oxide–semiconductor
with improved linearity has a 5.68 dB gain, 5.13 dB NF, and a −0.12 dBm IIP3. The difference between both
(CMOS)
Intermodulation distortion (IMD)
final designs, which consist of improved linearity and gain, stems from the location of the gate inductance
Linearization (𝐿𝑔 ) in the chip. The proposed LNA3 with an on-chip 𝐿𝑔 has a voltage gain of 11.1 dB, an NF of 4.27 dB, and
RF circuit an IIP3 of -0.82 dBm. Moreover, the proposed LNA4 with an off-chip 𝐿𝑔 has 10.31 dB voltage gain, 3.68 dB
NF, and 0.89 dBm IIP3 at 989.6 μW in post-layout simulations. Comparing the LNAs, the proposed LNA4 with
an off-chip 𝐿𝑔 has the best figure-of-merit (FOM). This work aims to achieve improved linearity figures at
sub-mW power.
✩ This work was sponsored by the Scientific and Technological Research Council of Turkey under the project TÜBİTAK 1001 118E253.
∗ Corresponding author.
E-mail addresses: erold@itu.edu.tr (D. Erol As), yeltenm@itu.edu.tr (M.B. Yelten).
1
The article was written jointly by Didem Erol and Mustafa Berke Yelten. Both authors have contributed equally to this research.
https://doi.org/10.1016/j.vlsi.2022.09.010
Received 30 April 2022; Received in revised form 17 September 2022; Accepted 26 September 2022
Available online 20 October 2022
0167-9260/© 2022 Elsevier B.V. All rights reserved.
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
Fig. 1. The schematic of (a) LNA1 (Cascode LNA) (b) LNA2 with improved linearity and degraded voltage gain (c) LNA3 having both better linearity and voltage gain performance
(d) LNA4 that matches LNA3 except the off-chip 𝐿𝑔 .
279
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
Fig. 4. 𝐾 and |𝛥| values depicting the stability of the designed LNA.
Fig. 2. (a) The LNA with an NMOS IMD sinker (adopted in LNA2). (b) The third-order
and 𝑔𝑚,𝑀𝑏 , thereby bringing the total to 13.31 mS. Sizing of 𝑀𝑏 should
transconductances of 𝑀1 , 𝑀2 , and 𝑀𝑎 .
be done not to disturb the IMD current balance so that the linearity
performance can be preserved.
The output resistance of LNA2, 𝑅𝑂𝑈 𝑇 drops by the addition of 𝑀𝑎 .
𝑖𝑎 = 𝑔𝑚1,𝑀𝑎 𝑉𝑔𝑠𝑎 + 𝑔𝑚2,𝑀𝑎 (𝑉𝑔𝑠𝑎 )2 + 𝑔𝑚3,𝑀𝑎 (𝑉𝑔𝑠𝑎 )3 , (4) In Fig. 3, 𝑅𝑂𝑈 𝑇 ,𝑎 which equals to (1∕𝑔𝑚𝑎 ) ∥ 𝑟𝑜𝑎 appears in parallel
where 𝑉𝑔𝑠1 and 𝑉𝑔𝑠𝑎 stand for the gate-source voltage of 𝑀1 and 𝑀𝑎 , with 𝑟𝑜1 . While 𝑅𝑂𝑈 𝑇 roughly equals to 𝑔𝑚2 𝑟𝑜2 𝑟𝑜1 for LNA1, it becomes
respectively. 𝑉𝑔𝑠𝑎 is a function of 𝑉𝑔𝑠1 , so 𝑉𝑔𝑠𝑎 can be given using the 𝑔𝑚2 𝑟𝑜2 (𝑟𝑜1 ‖1∕𝑔𝑚𝑎 ‖𝑟𝑜𝑎 ) in LNA2. 𝑅𝑂𝑈 𝑇 of LNA2 is smaller than that of
power-series expansion, LNA1 since 1∕𝑔𝑚𝑎 ≪ 𝑟𝑜1 , 𝑟𝑜𝑎 . When 𝑀𝑏 is introduced, the operating
current of the cascode stage slightly decreases due to the maximum
𝑉𝑔𝑠𝑎 = 𝑐1 𝑉𝑔𝑠1 + 𝑐2 (𝑉𝑔𝑠1 )2 + 𝑐3 (𝑉𝑔𝑠1 )3 , (5) power requirement of 1 mW. Reduction of the cascode stage current
where 𝑐𝑖 is the frequency-dependent coefficient. A first-order approxi- provides an increase in 𝑅𝑂𝑈 𝑇 even though 𝑟𝑜𝑏 comes in shunt with
mation is enough to prove the cancellation, so the second- and third- the output resistance of the cascode. Also, 𝑔𝑚 of 𝑀𝑏 boosts 𝐺𝑚 of the
order components are neglected. Thus, 𝑖𝑎 can be found in terms of 𝑉𝑔𝑠1 : LNA since it provides a voltage gain path in shunt with the cascode
amplifier. An increase in both 𝑅𝑂𝑈 𝑇 and 𝐺𝑚 improves the voltage gain
of LNA3.
𝑖𝑎 = 𝑔𝑚1,𝑀𝑎 𝑐1 𝑉𝑔𝑠1 + 𝑔𝑚2,𝑀𝑎 𝑐2 (𝑉𝑔𝑠1 )2 + 𝑔𝑚3,𝑀𝑎 𝑐3 (𝑉𝑔𝑠1 )3 . (6) A spiral inductor has not been employed at the source since bond-
Taking (2)–(6) into (1), 𝑖2 will be equal to: wires already cause an inductive effect there. Typically, bondwires
have a higher quality factor than spiral inductors, which helps reduce
𝑖2 = (𝑔𝑚1,𝑀1 − 𝑐1 𝑔1,𝑀𝑎 )𝑉𝑔𝑠1 + (𝑔𝑚2,𝑀1 − 𝑐12 𝑔𝑚2,𝑀𝑎 )(𝑉𝑔𝑠1 )2 the parasitic resistance and contribution to noise. By eliminating the
(7)
+ (𝑔𝑚3,𝑀1 − 𝑐13 𝑔3,𝑀𝑎 )(𝑉𝑔𝑠1 )3 . spiral inductors, the total chip area is reduced as well. The bondwire
inductance value here is chosen to be 500 pH which may be obtained
The coefficients 𝑐1 , 𝑐2 , and 𝑐3 depend on frequency, and 𝑐1 has a
by using two parallel bond wires, each having a typical inductance of
positive sign since 𝑀1 pulls the current from 𝑀𝑎 commensurately [6].
1 nH.
Besides, it is aimed to cancel (𝑔𝑚3,𝑀1 − 𝑐13 𝑔𝑚3,𝑀𝑎 ) by adjusting the gate
𝑀𝑏 is directly connected to the input; hence, the stability must
biasing. However, IMD3 is not fully canceled due to the limited amount
be checked. The conditions for unconditional stability in terms of
of allocated power, so the circuit is designed to minimize the IMD3
S-parameters are shown in (8) and (9) [7].
component.
In Fig. 2(b), the graphs of the third-order transconductances (𝑔𝑚3 s) |𝛥| = |𝑆11 𝑆22 − 𝑆12 𝑆21 | < 1. (8)
for 𝑀1 , 𝑀2 , and 𝑀𝑎 were obtained by taking the third order derivative
1 − |𝑆11 |2 − |𝑆22 |2 + |𝑆11 𝑆22 − 𝑆12 𝑆21 |2
of the related transistor drain currents with respect to 𝑉𝑔𝑠1 . The value 𝐾= >1 (9)
2|𝑆12 𝑆21 |
of 𝑔𝑚3,𝑀2 decreases from 400 mA/V3 to 160 mA/V3 at 𝑉𝑔𝑠1 = 531 mV
and 𝑃 = 986.6 μW. However, the voltage gain is also reduced due to In Fig. 4, 𝐾𝑓 and 𝐵1𝑓 , which refer to 𝐾 and |𝛥|, equal to 3.24 and
lowering in the (𝑔𝑚1,𝑀1 − 𝑐1 𝑔𝑚1,𝑀𝑎 ) expression. Therefore, 𝑀𝑏 is added 0.162 at 2.4 GHz, respectively. Thus, 𝑀𝑏 does not degrade the stability
to increase the voltage gain, as shown in Fig. 1(d). of LNA.
LNA4 with high linearity is designed in Fig. 1(d), including the
versions of 𝐿𝑔 being on and off-chip. The aim of the off-chip 𝐿𝑔 is to 3. Simulation results and discussion
obtain a high-quality factor, thereby reducing the occupied chip estate.
𝑔𝑚 in the LNA without the voltage gain improvement technique is The proposed LNA with high linearity has been designed in a
equal to the transconductance of 𝑀1 , 𝑔𝑚,𝑀1 = 11.24 mS, while 𝑔𝑚 in the commercial CMOS 40 nm process technology with a supply voltage of
LNA with the voltage gain improvement technique is the sum of 𝑔𝑚,𝑀1 𝑉𝐷𝐷 = 1 V. The current consumption is 989.6 μA. All LNAs have been
280
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
Table 2
Device sizes 𝑊 ∕𝐿 and circuit component values.
𝑀1 32 μm | 40 nm 𝐶1 200 pF
𝑀2 16 μm | 150 nm 𝐿𝑏 750 pH
𝑀𝑎 8 μm | 40 nm 𝐶𝑒𝑥 712 fF
𝑀𝑏 44 μm | 40 nm 𝐿𝑑 8.5 nH
𝐿𝑔 2.2 nH 𝐶2 483 fF
281
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
Fig. 8. (a) Voltage gain & NF (b) IIP3 of LNA3 with improved linearity and voltage
gain.
Fig. 10. (a) 𝑆11 and (b) 𝑆22 of LNA4 with improved linearity and voltage gain (off-chip
𝐿𝑔 ).
282
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
Table 3
The performance of the LNA4 at different process corners and temperatures.
Temperature SS FF FS SF
0 ◦C 25 ◦ C 75 ◦ C 0 ◦C 25 ◦ C 75 ◦ C 0 ◦C 25 ◦ C 75 ◦ C 0 ◦C 25 ◦ C 75 ◦ C
𝑆11 (dB) −9 −9 −9 −8.75 −9.33 −10.6 −12 −13.0 −14.6 −12.7 −13.6 −15.4
𝑆22 (dB) −22.6 −21.1 −19 −13.4 −12.9 −12.3 −15.6 −15.0 −14 −16.9 −16.5 −15.8
𝑆21 (dB) 7 6.57 5.65 13.5 13.22 12.7 11 10.64 9.88 10.4 9.99 9.2
NF (dB) 4.5 4.99 5.97 2.36 2.64 3.2 3.22 3.6 4.35 3.33 3.72 4.5
IIP3 (dBm) 7.73 8.18 7 −2.56 −2.69 −2.43 −0.20 −0.32 −0.07 4.38 3.69 3.74
Fig. 12. IIP3 (a) (pre-layout) and (b) (post-layout) of LNA4 with improved linearity
and voltage gain (off-chip 𝐿𝑔 ).
Table 4
The MC analysis for LNA4.
Best value Worst value Mean
𝑆11 (dB) −27.2 −8.53 −14.01
𝑆22 (dB) −24.52 −12.09 −16.91
𝑆21 (dB) 12.19 7.99 10.24
NF (dB) 2.95 4.66 3.73
IIP3 (dBm) 9.44 −5.27 2.2
Table 5
Performances of the proposed LNAs.
LNA1 LNA2 LNA3 LNA4 (pre-layout) LNA4 (post-layout)
𝑆21 (dB) 12.22 5.68 11.1 11.49 10.31
NF (dB) 4.35 5.13 4.27 3.38 3.68
𝑆11 (dB) −27 −17.97 −21 −19.36 −13.35
𝑆22 (dB) −16 −23 −14.78 −21.21 −15.42
IIP3 (dBm) −12.68 −0.12 −0.82 −0.7 0.89
P1dB (dBm) −23.74 −12.13 −13.57 −11.54 −11.16
𝑃𝐷𝐶 (μW) 995.6 986.6 997.2 1079 989.6
FOM 0.198 1.36 2.82 6.10 7.05
decreases. 𝑆11 degrades for the SS and FF process corners. Except for Fig. 13. Histograms of different performance parameters in Monte Carlo simulations:
these issues, other performance parameters are acceptable. (a) 𝑆11 (b) 𝑆22 (c) 𝑆21 (d) NF (e) IIP3.
283
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
Table 6
Performance of the proposed LNA regarding its counterparts.
Gain (dB) NF (dB) IIP3 (dBm) 𝑃𝑑𝑐 (μW) 𝐹𝑐 (GHz) Tech. (nm) FOM
This Work (post-layout) 10.31 3.68 0.89 989.6 2.4 40 7.05
[12] (measurement) 11 6.8 −2.2 174 2.4 65 6.56
[13] (post-layout) 18.2 3.38 −4.32 967 2.4 180 2.93
[14] (post-layout) 14 3.45 −8 980 2.4 180 0.92
[8] (pre-layout) 14 5.2 −8.6 30 2.4 40 15.2
[15] (measurement) 12.6 5.5 ∼6.5 −9 750 0.1∼7 90 0.47
[16] (measurement) 12.3 4.9∼16 −9.5 400 0.1∼2.2 130 0.87
[17] (measurement) 14 4∼6 −10 250 0.6∼4.2 130 1.87
[18] (measurement) 17.4 2.8 −10.7 480 2.4 65 1.71
[9] (measurement) 13.9 8.9 −13 69 2.4 65 1.27
[19] (measurement) 12.2 1.9∼2.2 −16 350 3∼5 130 0.97
[10] (measurement) 21.5 6.3 −16 900 2.4 28 0.11
[11] (measurement) 6.9 3 −18 44 2.4 16 1.25
[20] (post-layout) 13.64∼15.64 4.5∼6 −5∼ -6 600 0.03∼3 180 1.30
[21] (measurement) 16.8 6.6 −16.4 350 4 28 0.20
best, and mean values of the performance parameters achieved in MC Data availability
simulations.
Table 5 summarizes the performances of four different LNAs. While Data will be made available on request.
LNA1 is not suitable to satisfy the linearity specification, the IIP3 per-
formance has been dramatically improved by approximately 12.5 dBm References
in LNA2, thereby meeting the requirements. However, the voltage gain
[1] C.-T. Hsiao, Design of a 2.4 GHz CMOS LNA For Bluetooth Low Energy
and NF of LNA2 are well below the specifications. Although the voltage Application Using 45 nm Technology (Master’s thesis), San Jose State Uni-
gain and IIP3 performances are enhanced in LNA3, this topology may versity, 2017, Available at https://scholarworks.sjsu.edu/cgi/viewcontent.cgi?
occupy a huge chip area due to the on-chip input matching circuit. By article=8349&context=etd_theses.
[2] H.-J. Liu, Z.-F. Zhang, An ultra-low power CMOS LNA for WPAN applications,
taking 𝐿𝑔 off-chip, all performance metrics of LNA4 are made compliant IEEE Microw. Wirel. Compon. Lett. 27 (2) (2017) 174–176.
with the initial design specifications. Since 𝐿𝑔 is off-ship, a high-quality [3] A. Nadia, H. Belgacem, F. Aymen, A low power low noise CMOS amplifier for
factor can be assumed, further improving overall performance. When bluetooth applications, in: 2013 International Conference on Applied Electronics,
the figure of merits (FOMs), calculated per the expression provided in IEEE, 2013, pp. 1–4.
[4] M.B. Yelten, K.G. Gard, A novel design procedure for tunable low noise
(10), of four LNAs are compared topologies, LNA4 is the most successful amplifiers, in: 2009 IEEE 10th Annual Wireless and Microwave Technology
of all in terms of achieved performance metrics. Conference, IEEE, 2009, pp. 1–5, http://dx.doi.org/10.1109/WAMICON.2009.
5207313.
Gain (dB) × IIP3 (μW) [5] A. Çağlar, M.B. Yelten, Design of cryogenic LNAs for high linearity in space
𝐹 𝑂𝑀 = . (10)
(NF (dB) − 1) × Power (μW) applications, IEEE Trans. Circuits Syst. I. Regul. Pap. 66 (12) (2019) 4619–4627,
http://dx.doi.org/10.1109/TCSI.2019.2936506.
In Table 6, performance metrics of LNA4 are compared with the [6] T.-S. Kim, B.-S. Kim, Post-linearization of cascode CMOS low noise amplifier
results of various low-power LNAs reported in the literature with an using folded PMOS IMD sinker, IEEE Microw. Wirel. Compon. Lett. 16 (4) (2006)
operating frequency in the range of 0.1 GHz to 7 GHz. Although the 182–184.
[7] G. Gonzalez, Analysis and design, Microw. Transistor Amplif. (1997).
FOM value of [8] is higher than that of LNA4, its IIP3 is almost equal
[8] E. Kargaran, D. Manstretta, R. Castello, Design and analysis of 2.4 GHz 30 𝜇w
to that of LNA1. Except for [8], LNA4 is the best of all compared studies CMOS LNAs for wearable wsn applications, IEEE Trans. Circuits Syst. I. Regul.
in terms of the FOM value. Moreover, the power consumption of [9] is Pap. 65 (3) (2017) 891–903.
remarkably low; however, its NF is extremely high, and the reported [9] A. Dissanayake, H.-G. Seok, O.-Y. Jung, S.-K. Han, S.-G. Lee, A 64 𝜇w, 23 dB
gain, 8 dB NF, 2.4 GHz RF front-end for ultra-low power internet-of-things
IIP3 value is excessively lower than the LNA4 design described in this
transceivers, in: 2017 IEEE Radio Frequency Integrated Circuits Symposium
work. Also, [10,11] are remarkable in terms of using deep nanoscale (RFIC), IEEE, 2017, pp. 184–187.
CMOS technologies. However, the NF of [10] is very high, while the [10] J. Zaini, F. Hameau, T. Taris, D. Morche, P. Audebert, E. Mercier, A tunable
voltage gain of [11] is significantly low. ultra low power inductorless low noise amplifier exploiting body biasing of 28
nm FDSOI technology, in: 2017 IEEE/ACM International Symposium on Low
Power Electronics and Design (ISLPED), IEEE, 2017, pp. 1–6.
4. Conclusion [11] Y.-T. Lu, J.-D. Jin, 100-MV 44-𝜇w 2.4-GHz LNA in 16 nm FinFET technology,
in: 2016 IEEE MTT-S International Microwave Symposium (IMS), IEEE, 2016,
pp. 1–3.
In this paper, four LNAs using two different techniques to improve [12] K. Xu, J. Yin, P.-I. Mak, R.B. Staszewski, R.P. Martins, A single-pin antenna
linearity and voltage gain are designed at sub-mW power in a com- interface RF front end using a single-MOS DCO-PA and a push–pull LNA, IEEE
J. Solid-State Circuits 55 (8) (2020) 2055–2068.
mercial 40 nm CMOS process at 2.4 GHz. The proposed LNA, LNA4,
[13] A. Karimlou, R. Jafarnejad, J. Sobhi, An inductor-less sub-mw low noise amplifier
powered by a 1 V supply, achieves an IIP3 figure of 0.89 dBm and for wireless sensor network applications, Integration 52 (2016) 316–322.
dissipates 989.6 μW at 2.4 GHz. It has a voltage gain of 10.31 dB [14] F. Rastegari, M. Dousti, B. Ghalamkari, A 0.75 V sub-mW CMOS LNA employing
and a noise figure of 3.68 dB. Based on its design metrics, LNA4 transmitted signal suppression technique in a full-duplex wireless brain-machine
interface transceiver, AEU-Int. J. Electron. Commun. 132 (2021) 153632.
can be employed in ultra-low-power, high-performance receivers, as [15] M. Parvizi, K. Allidina, M.N. El-Gamal, A sub-mw, ultra-low-voltage, wideband
frequently encountered in biomedical applications. low-noise amplifier design technique, IEEE Trans. Very Large Scale Integr. (VLSI)
Syst. 23 (6) (2014) 1111–1122.
[16] M. Parvizi, K. Allidina, M.N. El-Gamal, An ultra-low-power wideband inductorless
Declaration of competing interest CMOS LNA with tunable active shunt-feedback, IEEE Trans. Microw. Theory
Tech. 64 (6) (2016) 1843–1853.
[17] M. Parvizi, K. Allidina, M.N. El-Gamal, Short channel output conductance
The authors declare that they have no known competing finan-
enhancement through forward body biasing to realize a 0.5 V 250 𝜇w 0.6-
cial interests or personal relationships that could have appeared to 4.2 GHz current-reuse CMOS LNA, IEEE J. Solid-State Circuits 51 (3) (2015)
influence the work reported in this paper. 574–586.
284
D. Erol As and M.B. Yelten Integration 88 (2023) 278–285
[18] M. Rahman, R. Harjani, A 2.4-GHz, sub-1-V, 2.8-dB NF, 475-𝜇w dual-path noise [20] A.P. Tarighat, Ultra-low power inductorless differential LNA for WSN application,
and nonlinearity cancelling LNA for ultra-low-power radios, IEEE J. Solid-State Analog Integr. Circuits Signal Process. 108 (2) (2021) 409–419.
Circuits 53 (5) (2018) 1423–1430. [21] T. Taris, J. Desevedavy, F. Hameau, P. Audebert, D. Morche, Inductorless multi-
[19] A.M. Saied, M.M. Abutaleb, M.I. Eladawy, H. Ragai, Ultra-low power LNA mode RF-CMOS low noise amplifier dedicated to ultra low power applications,
design technique for UWB applications, AEU-Int. J. Electron. Commun. 97 (2018) IEEE Access 9 (2021) 83431–83440.
149–153.
285