Wireless digital data transmission from a 300 GHz CMOS transmitter

ELECT Wireless performance of a 300 GHz CMOS transmitter reported recently is presented. Wireless digital data transmission at 28 Gbit/s over 5 cm with 16-QAM and 1 Gbit/s over 1 m with quaternary phase-shift keying was achieved. A figure of merit (FoM) for transmit–receive systems that allows comparison of diverse configurations (with different modulation formats, antenna gains, distances, etc.) is introduced. The CMOS transmitter turns out to be comparable in performance to other 0.3 THz or higher systems involving compound-semiconductor or photo-mixing technologies. Analysis based on the FoM reveals how best use could be made of frequencies around 300 GHz with a CMOS technology.

Introduction: The frequency band above 275 GHz currently remains unallocated, and its spectrum allocation is due to be discussed.Some wireless transceivers operating above 275 GHz have been realised with compound semiconductors [1,2] or photonic devices [3][4][5][6].We recently reported a 300 GHz CMOS transmitter that operates above the transistor unity-power-gain frequency f max but nevertheless supports high-order digital modulation such as the quadrature amplitude modulation (QAM) [7,8].Fig. 1 shows its simplified block diagram and a snapshot of a wireless digital data transmission experiment.The experimental data presented in [7,8] were obtained mostly by directly connecting a measurement system to the transmitter through a waveguide.In this Letter, we present the transmitter's wireless capability and discuss how the relatively unexplored frequency band should be covered.Wireless performance of 300 GHz CMOS transmitter: In the measurement set-up, a WR3.4 waveguide probe leads to the transmitting horn antenna as shown in Fig. 1.The receiving horn antenna is mounted on a block downconverter (VDI WR3.4 MixAMC).The IF 1 , centred at 18 GHz, was generated by an arbitrary waveform generator (Keysight M8195A).The local oscillator (LO) and RF frequencies are 106 and 300 GHz, respectively.The frequencies of IF 1 and LO were chosen so that unwanted image in IF 2 is reduced.We measured the error-vector magnitude (EVM) of the received signal while changing the modulation format, the symbol rate, and the antenna-to-antenna distance.A real-time oscilloscope (Keysight DSA-Z 334A) and an associated vector signal analyser software were used for the measurement.A channel equaliser built into the software was applied.The modulation formats used in the experiment were quaternary phase-shift keying (QPSK), 16-QAM, and 64-QAM.The measurement results are shown in Fig. 2. The maximum symbol rate of 13 Gbaud was dictated by the bandwidth of an intermediate frequency (IF) amplifier in the measurement system.Table 1 summarises the longest distance and fastest transmission data for each modulation.Comparison of transmit-receive systems: To put the experimental wireless performance of the 300 GHz CMOS transmitter into perspective, we introduce a figure of merit (FoM) that allows comparison of transmit-receive systems with diverse configurations.Different systems have different transmitters, receivers, antennas, distances, modulation formats, and symbol rates.We propose that all such differences but the symbol rate, r s , be absorbed into an effective distance.For example, a wireless communication experiment performed using high-gain antennas can be interpreted to correspond to an experiment performed over a shorter effective distance with 0 dBi antennas.Similarly, a 64-QAM experiment, which requires a high-channel signal-to-noise ratio (SNR), can be regarded as equivalent to a QPSK experiment, for which a lower SNR suffices, performed over a longer distance.Assuming that the antenna-to-antenna distance d is long enough (far-field assumption), we use the Friis transmission formula where P t is the power fed into the transmitting antenna, G t its antenna gain, P r the power available from the receiving antenna, G r its antenna gain, and λ the wavelength.From (1), the following equation can be obtained: The left-hand side can be regarded as an effective distance squared.To see how a difference in modulation should be incorporated, suppose that the modulation format is changed from a low-order one to a higher-order one, which raises the requirement for channel SNR by a factor R (>1).Then, the measured achievable distance d becomes shorter (d/ R √ ).To make the effective distance invariant under the said simultaneous changes in modulation and distance (d d/ R √ ), both sides of (2) should be multiplied by R. To be able to compare completely different systems, R should be defined using a fixed reference value, SNR 0 , as ELECTRONICS LETTERS 21st July 2016 Vol.52 No. 15 pp.1353-1355 R W SNR/SNR 0 .We set SNR 0 = 1 and get d eff is our effective distance.It is the distance at which the channel SNR would become unity if a pair of 0 dBi antennas were used.Evaluation of SNR involves relating the measured bit-error rate (BER) to SNR.To do so, we make a simplifying assumption that the channel bandwidth B equals the r s (B = r s ).Then, the information-bit-energy-to-noise-density ratio [9], , where M is the signal order (number of points in a signal constellation).Put together with known theoretical relationships between BER and where erfc(x) is the complementary error function.
The signal bandwidth is proportional to the symbol rate r s .The power spectral density, therefore, is inversely proportional to r s .If the symbol rate is increased from a certain reference value r s0 to r s , the SNR degrades and SNR becomes (r s0 /r s ) times the original value.Dividing (3) by (r s0 /r s ) • (λ 2 /r s0 ), we arrive at The FoM represents the achievable symbol rate when d eff = λ.The Friis formula (or the far-field assumption) breaks down at d eff = λ if d eff is comparable to the actual distance d, but d eff ≫ d is often satisfied. [10]

Fig. 3 Comparison of digital data transmission experiments
a FoMs on r s against d eff /λ plane b List of symbols; OOK: on-off keying Fig. 3 plots FoMs of various experimental data on a symbol rate-against-effective distance plane.Our transmitter, when combined with the measurement system, shows comparable FoM values to other 0.3 THz or higher systems [1][2][3][4][5][6].In Fig. 3, (6) gives a straight line of slope −2 for a given value of r s .Our QPSK (°) and 16-QAM (Δ) results with (d eff /l) .4 appear to lie on such a line.Other shorter-distance results of ours are not on the line because d was too small and the far-field assumption was violated.A 240 GHz system employing high-performance GaAs metamorphic high-electron-mobility transistors (mHEMTs) [10] shows very high FoMs.To increase the FoM, the output power of the transmitter and/or r s should be increased.To increase the output power at 300 GHz (> f max ), where power amplification is not possible, our transmitter performs 32-way power combining [7].
Given the absence of any wireless communication standards at these frequencies, the upper-bound value of r s (vertical axis of Fig. 3) is dictated by the hardware.In our case, r s is limited by the bandwidth of the receiver IF amplifier.If receiver, too, is to be implemented with a CMOS technology with merely adequate terahertz performance, the maximum r s is likely to become lower.In such a system, highest data rates (r s log 2 M ) will be achieved by combining a moderate r s with QAM (Table 1).This backs up the proposal in [7] that a 30 GHz-wide band be covered with six 5 GHz-wide QAM channels.If, on the other hand, high-performance devices are available, it seems appropriate to aim for extremely high data rates (>50 Gbit/s) with very high r s and a simpler modulation, as in [10].

Conclusion:
We presented the wireless performance of a 300 GHz CMOS transmitter [7,8].The peak data rate reached 28 Gbit/s with 16-QAM (Fig. 2 and Table 1).We introduced an FoM (6) that allows comparison of transmit-receive systems and considered how the vast frequency band around 300 GHz could be best utilised.If a CMOS technology is to be adopted, it is best to cover a very wide bandwidth (several tens of GHz) with multiple QAM channels with a reasonable per-channel data rate [7].This underpins the importance of QAM capability even at 300 GHz, where very wide bandwidths are available [7].Superior integration capability of CMOS technology could facilitate possible eventual introduction of channel bonding.

Fig. 1 300
Fig. 1 300 GHz CMOS transmitter a Simplified block diagram b Snapshot of wireless digital data transmission experiment

Fig. 2
Fig. 2 Measured EVM as a function of distance a QPSK b 16-QAM c 64-QAM photo-mixing

Table 1 :
Performance summary of the 300 GHz CMOS transmitter QPSK 16-QAM 64-QAM Distance 1 m 5 cm 40 cm 5 cm 15 cm 5 cm Data rate 1 Gbit/s 26 Gbit/s 2 Gbit/s 28 Gbit/s 3 Gbit/s 12 Gbit/s Bold values indicate longest distance and highest data rate achieved for the given modulation format