IEEE JSSC 202009: Solid-State Circuits and Integrated Circuits Special Section

需积分: 14 187 浏览量更新于2024-07-15 收藏 126.11MB PDF 举报

"JSSC_202009.pdf" 这篇文档是2020年9月刊的《IEEE固体状态电路期刊》(IEEE Journal of Solid-State Circuits)，主要关注的是集成电路的晶体管级设计，特别是涉及固体状态电路的广泛领域。该期刊每月发布论文，涵盖的内容包括但不限于在2019年IEEE广播与通信固态电路技术会议（BCICTS Conference）的特辑。特辑部分介绍了2019年IEEE BCICTS会议的嘉宾编辑评论，由H. Wang撰写，讨论了会议的重要性和相关研究。接着是几篇专题论文： 1. 提出了一种基于250纳米InP DHBT技术的110 GHz带宽2:1模拟多路复用器的超过1 Tb/s的相干光发射前端。这项工作展示了在高速通信领域的技术创新，尤其是在光学传输的前端设计上。 2. 基于InP HEMT晶体管和混频器的300 GHz带宽120 Gb/s无线前端设计，由Hamada等人提出。这个无线前端展示了极高的频率操作能力和数据传输速率，对于无线通信系统的未来具有重要意义。 3. Kobayashi和Y.Z. McCleary介绍了一款使用SiGe HBT和InP DHBT的宽带三重叠分布式放大器，带有主动偏置终端，工作范围从基带到140 GHz，速度达到100 GHz。这种放大器设计提高了频率响应和性能。 4. Veni等人分析并设计了一个17 GHz全npn推挽类C压控振荡器，展示了在高频率下的高效能和线性特性。常规论文部分包含了一篇关于高度线性、高功率802.11ac/ax WLAN SiGe HBT功率放大器的文章，采用紧凑型第二谐波短路四路变压器。这个设计优化了功率效率和线性度，适用于无线局域网应用。这期JSSC涵盖了从基础组件到完整系统设计的各种先进固体状态电路技术，特别关注高速、高频和高数据传输能力的集成电路解决方案，对从事微电子、通信和集成电路设计的工程师和技术人员具有极高的参考价值。

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

HAMADA et al.: 300-GHz-BAND 120-Gb/s WIRELESS FRONT-END BASED ON InP-HEMT PAs AND MIXERs 11

Fig. 23. CG improvement by the RF–IF isolation obtained by the analysis.

Fig. 24. Simulated CG of the designed mixer with and without the RF–IF

isolation.

294 GHz. The measured S-parameters of the LO PA are shown

in Fig. 28. The small signal gain is 15 dB at 270 GHz. Because

the output power of the extender is around −7dBm,theLO

PA is cascaded to provide sufﬁcient LO power to the mixer.

The dependence between the CG and LO power is shown

in Fig. 29. The CG is saturated when the LO power is larger

than 5 dBm. In the following measurement, the LO power

was set to 7.5 dBm. The band-pass-ﬁlter (BPF) is used to

cut out the spurious emission and to apply the desired LO

signal of 270 GHz to the mixer. The measured CG for both

up-conversion (U-C) and D-C is shown in Fig. 30(a). The CG

is −15 dB for both U-C and D-C in the IF of 0–10 GHz. The

3 dB BW for the USB is very broad, 32 GHz. The measured

CG is approximately 5-dB lower than the simulation results

shown in Figs. 24 and 25. We consider that this is caused by

the IF packaging loss (1–2 dB, IC-to-V-connector junction),

RF packaging (1 dB, WG propagation loss inside the mixer

module), and the inaccuracy of the transistor model made from

the low-frequency (10 MHz–110 GHz) measurement data. The

linearity for the U-C at the IF of 21 GHz was also measured,

as shown in Fig. 30(b). The measured OP1dB is −16.5 dBm.

The LO-to-RF leakage of the mixer was also measured using

the power meter, VDI PM5. The input LO power of the mixer

(after the 270-GHz BPF shown in Fig. 27) and LO leakage

measured at the RF port of the mixer are 7.5 and −16.4 dBm,

respectively, as shown in Fig. 31. Therefore, LO leak sup-

pression is 23.9 dBc. This is not that bad for the single-

ended mixers compared to the value for balanced ones in

300-GHz-band (19.6 dBc in [35]).

V. TRX

AND DATA TRANSMISSION EXPERIMENTS

The proposed TRX is constructed using the RF PAs,

mixers, and LO PAs discussed in Section IV. To locate the

RF in our target frequency band, which is in the un-allocated

frequency of above 275 GHz, the LO frequency, IF, and RF are

Fig. 25. Simulated CG of the mixer module with ridge couplers for with and

without the RF–IF isolation. (a) Simulation model. (b) Simulation results.

set to 270, 20, and 290 GHz, respectively, i.e., the USB is used

as the RF signal. The characteristics of each transmitter (TX)

and receiver (RX) are discussed in this section.

A. TX

The schematic of the TX is shown in Fig. 32(a). It consists

of the individual WG modules of RF PAs, mixers and LO

PAs described in Section IV. As described in Section IV-B,

the mixers have substantial LO-to-RF leakage of −16.4 dBm

where the LO input power is 7.5 dBm at 270 GHz. This

LO leakage is almost the same as the mixers OP1dB of

−16.5 dBm shown in Fig. 30(b). Therefore, the HPF module

is used to cut out the LO leakage between the U-C mixer

and RF PA to avoid the undesired effect of the LO leakage

such as the saturation of the RF PA and the intermodulation

distortion. This HPF also cuts the image signal of the LSB

from the mixer output. The CG and I–O characteristics were

measured using the same measurement setup in Fig. 27. Note

that in the TX measurement, the WR3.4 WG-type attenuator

(ATT) with the attenuation of 14 dB was inserted between the

RF PA and the WR3.4 extender to protect the extender from

the large output power of the RF PA. The insertion loss of the

ATT is subtracted after the measurement. Fig. 32(b) shows

the CG of the TX in the USB. The frequency response of the

TX CG reﬂects those of the CG and the gains of the mixer

and PA. It gradually increases as the frequency increases and

takes its peak valu e of around 3 dB around 290–295 GHz and

degrades after 296 GHz. The 3-dB BW is 17 GHz. The I–O

characteristics at the RF and IF of 291 and 21 GHz are shown

in Fig. 32(c). The measured OP1dB of the TX is 1.5 dBm,

which matches the theoretically calculated OP1dB of 1.6 dBm

using the formula [36] in the cascaded form of the mixer

(OP1dB =−16.5 dBm) and RF PA (OP1dB = 6.8dBm)

described in Figs. 19(a) and 30(b).

Authorized licensed use limited to: University of Exeter. Downloaded on July 14,2020 at 01:57:55 UTC from IEEE Xplore. Restrictions apply.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

12 IEEE JOURNAL OF SOLID-STATE CIRCUITS

Fig. 26. Photographs of (a) mixer IC and (b) module.

Fig. 27. Schematic of the measurement setup of the mixer module.

Fig. 28. Measured S-parameters of the LO PA module.

Although this TX has relatively large BW, the output power

is not as high as the OP1dB of the RF PA which is around

6 dBm. This is because the output power of the mixer and/or

the gain o f the RF PA are not sufﬁcient to leverage the

superior output power characteristics of the RF PA illustrated

in Fig. 19(b). Accordingly, another TX conﬁguration, shown

in Fig. 33(a), is considered. In this conﬁguration, simply

an additional RF PA is cascaded to the RF PA (primary

RF PA) used in the former TX to enhance the output RF

power from the TX. The same PA module was used for

both the primary and the additional RF PA shown in Fig. 9.

We call the TXs in Figs. 32(a) and 33(a) as single-PA-TX

(SPTX) and dual-PA-TX (DPTX), respectively. The CG and

I–O characteristics of the DPTX were also measured, as shown

in Fig. 33(b) and (c). The CG and OP1dB increase to 21 dB

and 6 dBm. However, the 3-dB BW become narrower by

the PA cascading, 10 GHz. In the following wireless data

transmission experiments, we use both the SPTX and DPTX

with the aid of their broad BW and higher output power.

B. RX

The conﬁguration o f the RX is shown in Fig. 34(a). Gener-

ally, an LNA is used for the ﬁrst stage of the RX. We used the

Fig. 29. Measured CG versus LO input power of the mixer module.

Fig. 30. (a) Measured CG and (b) I–O (IF 21 GHz) of the mixer module.

Fig. 31. Measured LO leakage to the RF port of the mixer module.

same ampliﬁer as the RF PA in Fig. 9 for the LNA. Also, the

same mixer as the TX side was used as the D-C mixer with

the aid of the commutativity of the resistive mixer architecture.

Authorized licensed use limited to: University of Exeter. Downloaded on July 14,2020 at 01:57:55 UTC from IEEE Xplore. Restrictions apply.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

HAMADA et al.: 300-GHz-BAND 120-Gb/s WIRELESS FRONT-END BASED ON InP-HEMT PAs AND MIXERs 13

Fig. 32. (a) Schematic, (b) measured CG, and (c) I–O of the SPTX.

Unlike the TX, the HPF is not used because the LO leakage

and LSB signal are not applied to the input of the LNA under

normal operation of the RX. The measured frequency response

of the CG and I-O using the measurement setup described

in Fig. 27 are shown in Fig. 34(b) and (c), respectively. Similar

frequency response is obtained as the TX except for the CG

ripple occurring below 290 GHz in the RX. The cause of this

ripple is considered to be the impedance mismatch between

the PA output port and mixer RF port. The peak CG and 3-dB

BW are around 3 dB and 17 GHz which are similar to the TX

results. I–O measurement results show the lin ear response of

the RX output power with the RF input power up to −7dBm.

Due to the lack of the RF power of the WR3.4 extender in

the measurement setup (Fig. 27), the saturated output power

and the OP1dB of the RX were not measureable. The OP1dB

of the RX can be estimated by the OP1dB values of the RF

PA (6.8 dBm) and mixer (−16.5 dBm) as −17 dBm.

The NF of the LNA (RF PA in this case) was measured

by using the NF measurement system (NFMS) as shown

in Fig. 35(a). We measured the NF of RF PA by Y -factor

method using the WR3.4 noise source made by ELVA-1. Noise

power was measured by the spectrum analyzer (SPA, Agilent

PXA). To down-convert the 300-GHz-band noise to lower

frequency which can be measured in the SPA, we used the

Fig. 33. (a) Schematic, (b) measured CG, and (c) I–O of the DPTX.

resistive mixer module described in Section IV. To enhance the

noise power from the DUT to supply large enough power to

the SPA, total gain of th e NFMS was set to a very large value

(60–70 dB) by using two-cascaded RF PA which are the same

PA as the DUT (gain: 20 dB) and IF ampliﬁer (gain: 23 dB).

LO frequency of the D-C mixer in NFMS is 270 GHz.

To reject the image response, the HPF was inserted between

the two RF PAs. First, we measured the NF of the cascaded

system of RF PA (DUT) and NFMS. Then, the NF of NFMS

was measured, and the NF of the DUT was estimated by using

the Friis formula for noise. Fig. 35(b) shows the measured NF

of the RF PA and NFMS. The NF of the RF PA is around

15 dB in the vicinity of 300 GHz. It is allowable value to be

used as the LNA in the RX considering that the typical NF of

the LNA is 7–10 dB in 300-GHz band [41], [42]. The NF of

the RX can be estimated by considering the gain and NF values

of the RF PA (gain: 20 d B and NF: 15 dB) and mixer (gain:

−15 dB and NF: 15 dB) as 15.04 dB. Note that the NF of our

mixer is considered to be the inverse value of the CG because

of its passive topology.

C. Data Transmission Experiments

By using the TX and RX described in the previous sub-

sections, we conducted three data transmission experiments:

Authorized licensed use limited to: University of Exeter. Downloaded on July 14,2020 at 01:57:55 UTC from IEEE Xplore. Restrictions apply.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

14 IEEE JOURNAL OF SOLID-STATE CIRCUITS

Fig. 34. (a) Schematic, (b) measured CG, and (c) I–O of the RX.

(i) back-to -back data transmission with the SPTX and RX,

(ii) wireless data transmission over the link distance of 9.8 m

with the SPTX and RX, and (iii) wireless data transmission

over the link distance of 9.8 m with the DPTX and RX

depicted in Fig. 36. Wideband IF modulation signal g eneration

and detection for all experiments was carried out as follows.

To generate a wideband QAM IF signal, an AWG, Keysight

M8196A, was used. A DSO, Keysight Z634A was used as the

receiver of the IF signal. The IF ampliﬁer (SHF, S807B) was

used to amplify the output IF signal of the RX and supply the

sufﬁcient voltage amplitude to the DSO. A raised-cosine roll-

off ﬁlter (roll-off factor of 0.05) was applied to the IF signal

generated by the AWG. A pseudo-random-bit sequence of the

modulation data was set to 2

− 1. The signal-equalization

was applied to the received IF signal by using Keysight VSA

software (equalization ﬁlter length: 99). The constellation,

and SNR of the received IF signal for a certain baud rate

were evaluated in this setup. The required SNR where the

BER is equal to 10

−3

was used to evaluate maximum data

transmission.

(i) Back-to-Back Data Transmission With the SPTX and RX:

First, the back-to-back data transmission was carried out to

Fig. 35. (a) Schematic of NFMS and (b) measured NF o RF PA.

determine the maximum data rate obtained by using the SPTX

and RX. The ATT with the attenuation of 9 dB was inserted

between the SPTX and RX to avoid the saturation of the RX.

In this experiment, the 16QAM was used for the IF signal.

The dependence of SNR and TX output power was measured

to ﬁnd out the optimum IF input power of the TX for the

baudrates of 15, 20, 25, and 30 Gbaud. The measurement and

calculation results are shown in Fig. 37. For the SNDR calcu-

lation, the same parameters in Table I(A) and (C), which are

based on the stand-alon e measurement results, are utilized. The

maximum measured SNR matched well with the calculation

for 15 and 20 Gbaud. The mismatch becomes larger as the

baudrate becomes higher. This is because the frequency d epen-

dence of the TX and RX building blocks are not considered

in the SNDR calculation. The 3-dB BW of the TX and RX is

17 GHz as described in Figs. 32 and 34. Therefore, the SNR

degrades for b audrates exceeding above 20 Gbaud as shown in

Fig. 37. The measured optimum TX output power is 2–3 dB

larger than the calculation results in Fig. 37. This is caused by

the variation of the individual modules of PAs and mixers used

in our TX and RX. In the following experiments (ii), (iii), and

this back-to-back experiment, we used the TX output power of

−10 dBm (this corresponds to the TX input power of −9dBm)

as it is the optimum value to achieve maximum SNR shown

in Fig. 37. The SNR was measured for each baudrate of the

IF signal received in the DSO. The experimental results (SNR

and signal constellation) are shown in Fig. 38. At 5 Gbaud

(20 Gb/s), the SNR was quite high (>25 dB). The SNR

is inversely proportional to the baudrate, which agrees with

the theory, up to 25 Gbaud (100 Gb/s). Above 25 Gbaud,

the SNR decreases more rapidly than the 1/Baudrate.

This decreasing reﬂects the BW of the TX and RX, as shown

in Figs. 32(b) and 34(b). The measured maximum baudrate

where the SNR is equal to 16.5 dB, the required SNR value

Authorized licensed use limited to: University of Exeter. Downloaded on July 14,2020 at 01:57:55 UTC from IEEE Xplore. Restrictions apply.

剩余327页未读，继续阅读

netshell

粉丝: 11
资源: 185

IEEE JSSC 202009: Solid-State Circuits and Integrated Circuits S...

最新资源

IEEE JSSC 202009: Solid-State Circuits and Integrated Circuits S...

IEEE固态电路期刊特别版：欧洲固态电路会议论文集

10纳米工艺1.1V 6.4Gb/s DDR5 SDRAM设计：新技术与电路实现

2020 IEEE 固态电路期刊：国际固态电路会议特刊

JSSC_202007.pdf

JSSC_202006.pdf

JSSC_202012.pdf

JSSC_202004.pdf

JSSC_202001.pdf

JSSC_202005.pdf

15_sar.pdf

最新资源