Chapter 6 65nm Tuner Implementation and Verification
6.5 Considerations on PCB Layout
As mentioned in the section 5.6.1, noise coupling via the path either on the chip or the board may significantly degrade the signal quality. For this chip, the tuner needs to receive a -96.6dBm signal at its RF input port in the presence of 1.2V digital switching noise. Since the magnitude difference between the digital and RF signal can be as large as 0.1 million (100 dB), the sensitive RF signal may be corrupted by the large digital noise without proper separations and shielding. Special attentions on the reduction of on-chip noise coupling [71] is applied by use of the additional deep N-well shielding [72] and block separation as distant as possible. In addition to the on-chip issue, the printed circuit board (PCB) design is also important to an RF integrated system.
RF Circuit (VCO)
Analog Circuit (PLL) RF part
(LNA)
RF Circuit (Mixer)
Analog Circuit (ABB)
Digital Circuit (DSM, crystal) Center of star
Power Supply
1uF 1nF 10pF
Fig. 6.13 A star configuration of VDD pins.
The PCB layout significantly affects the performance, stability and reliability of the wireless system. One critical issue to be concerned with is the power supply routing on PCBs. Noise propagation via the low-impedance power and ground traces needs more attentions to alleviate. To minimize coupling between different domains of the IC, a star configuration is widely adopted for the power-supply layout [73], [74].
Chapter 6 65nm Tuner Implementation and Verification
134
As shown in Fig. 6.13, decoupling capacitors are placed at the central VDD node, and the power traces branch out from this node, with each trace going to separate supply pins. Typically, each supply pin must have a bypass capacitor placed as close as possible to the pin with low impedance to ground at the frequency of interest. To reduce the BOM, some experiments can be applied to determine the essentially required bypass capacitors. To our experiences, the bypass capacitor at the supply pin of the digital switching circuit should not be eliminated, which shows significant benefits on suppressing the noise source.
To evaluate the issues of PCB layout, two types of PCB are implemented. One directly shares common power/ground planes among the RF and digital circuits. The other one has separate power/ground planes among the RF and digital circuits as shown in Fig. 6.8. It follows the star routing rule in the supply connections of the RF and digital domains to sink the supply source. Fig. 6.14 and Fig. 6.15 depict the measured spectrum at the RF input port of these two boards, respectively. In this test, the tuner is configured in a single-ended mode and all circuit blocks are activated in the normal operation mode. High-order harmonics leakages from the clock switching to the RF input port are then measured. From the measurement results, it can be found that an extra isolation of 16dB can be achieved using separate power planes in contract to the common planes for the dominant components below 600MHz. It should be noted that the coupling mechanism is dominated by the board level, rather than the chip level from our experiments. Almost the same spectrum is observed as we remove the bonding wire of the RF input pad or disable the LNA bias current.
A star configuration has been demonstrated an efficient way of isolating power noise propagation from the digital part to the RF part. If the board contains an extremely noisy part, inductors or resistors can be placed in series with the power supply trace of these noisy parts to provide a higher level of isolation.
Chapter 6 65nm Tuner Implementation and Verification
135
Fig. 6.14 Clock leakage to the RF input with shared power/ground planes.
Fig. 6.15 Clock leakage to the RF input with separate power/ground planes.
Chapter 6 65nm Tuner Implementation and Verification
136
6.6 Measurement Results
The measurement setup is shown in Fig. 6.16. An external LDO chip is applied to provide a single 1.2V supply to the tuner core, and a 2.5V to the I/O supply. The measured performance is referred to the SMA connector input for either the single-ended or differential receiving mode. In other words, the balun effect is included in the measured performance of the differential mode. The external balun components used for the measurement are TDK TCM12B51-900-2P-T [75] and TDK HHM1525 [76] for the UHF and L-band, respectively. Fig. 6.17 depicts the photograph of the actual PCB for the differential mode test.
Fig. 6.18 shows the measured S11 of the receiver in the single-ended and differential mode, respectively. The broadband characteristic can be observed in the single-ended mode, while a slightly band-selecting characteristic is shown in the differential mode due to the external balun. At the band of interest, however, the S11 is below -10dB for both two modes. The measured S11 is more reliable across the band of interest compared to that of the previous work which use inductively source degeneration LNAs and feature S11 worse than -6.5dB [64]. Fig. 6.19 depicts the measured gain flatness of the receiver across the band of interest. As can be seen, the receiver has a gain flatness of ±1.5dB, much better than the previous work which use inductively source degeneration LNAs and feature gain flatness larger than ±4dB [16], [31], [18]. As shown in Fig. 6.20, the UHF receiver features an NF of 3.3－
3.8dB in the single-ended mode and 3.0－3.8dB (including the balun loss) in the differential mode. As the RF front-end has a gain back-off of 16dB, i.e., in the low-gain mode, the measured NF is shown in Fig. 6.21. The receiver features an NF of 8.8－9.6dB in the single-ended mode and 8.9－10.1dB in the differential mode.
Fig. 6.22 shows the measured IIP3 and IIP2 of the receiver in the single-ended mode. The receiver is set at the gain mode of RFE=max and ABB=50dB. The IIP3 values are measured by applying two-tone frequencies at 13.25MHz and 29.25MHz away from the desired frequency, which models from the test scenario of the L3 pattern. For the IIP2, whereas, a two-tone test with blockers at 13.25MHz and 16MHz offset is performed. This condition models from the test scenario of the S2 pattern. At the high-gain mode (RFE=max and ABB=50dB), the receiver achieves an
Chapter 6 65nm Tuner Implementation and Verification
137
IIP3 of -11dBm and an IIP2 of +29dBm.
To observe the clock leakage within the band of each desired channel, some experiments are made. First, the tuner is activated in the normal operation mode, but the input is terminated to a 50-ohm termination load. Fig. 6.23 shows the measured spectrum at the baseband output under a tuner setting of 666MHz channel and a 93dB gain. As can be seen, one -15dBm tone is measured at 2MHz. This corresponds to a leakage component of -108dBm at 664MHz or 668MHz as referred to the RF port. Next, we apply a QPSK signal of -96.6dBm to the RF input, and measure the EVM per subcarrier of the signals as shown in Fig. 6.24. As can be seen, the MER at subcarriers around 2MHz is significantly deteriorated due to this leakage noise. It should be noted that the measured MER is validated using Rohde & Schwarz SFU broadcast test system (modulation signal source) together with ETL TV analyzer (VSA). Fig. 6.25 depicts the measured leakage components across the desired 49 channels as the RF input is configured as a single-ended mode and a differential mode, respectively. The measurement result shows that the leakage component to the RF port is less than -98 and -112dBm in the single-ended and differential mode, respectively. Such components show minor impacts on the signal quality.
Fig. 6.26 and Fig. 6.27 respectively show the phase noise spectrum measured at the synthesizer and the receiver baseband output at the 474MHz channel. The synthesizer output centered at 2844MHz achieves an integrated phase noise less than 1.06° (or -34.6dBc) from 400Hz to 4MHz. When a sinusoidal signal with a -40dBm power level at 473MHz frequency was applied to the RF input, the IF signal at 1MHz can be measured at the receiver baseband output. The receiver now has a gain setting:
the maximum gain in the front-end and 0dB in the baseband. The integrated phase noise measured at 1MHz at the baseband output is 0.316° (or -45.2dBc) from 400Hz to 4MHz, which is about 10.6dB lower than the synthesizer output. In theory, the LO signal (after the /6 divider) should have a noise profile which is 15dB lower than that of the VCO signal. This can be found by comparing Fig. 6.26 with Fig. 6.27 at the in-band frequencies. At 400Hz, for example, the synthesizer output has a level of -76.9dBc/Hz, while the baseband output has a level of -91.9dBc/Hz. Indeed, the difference is 15dB. However, at the frequencies far away from the central frequency, which have noise level much lower than the central tone, the noise contribution from the receiver chain will dominate the overall noise level and raise the noise floor. This
Chapter 6 65nm Tuner Implementation and Verification
138
can explain the significant difference from the theoretical value of 15dB at frequencies beyond the loop bandwidth.
System performance has been measured on the tuner IC along with a measurement demodulator (VSA). To verify the impact of RF impairments on the signal quality, the MER or EVM is measured as the figures of merit (FOM). EVM is the root mean square of the sum of error vector magnitudes (across all data carriers).
In the case of additive white Gaussian (AWG) noise, MER and SNR are equal [48].
Because the BER test (system performance) depends not only on the radio chip but the baseband demodulator as well, the estimate is given according to the measured C/N of the radio chip not exceeding one specific value based on the modulation schemes in the MBRAI specification. In general, the validation with a companion demodulator chip achieves the reference BER equal to 2E-4 with a lower C/N requirement in the Gaussian channel. Thus, the listed performance related to sensitivity, linearity, and selectivity test would be better when the chip is validated with a companion demodulator.
Fig. 6.28 illustrates the measured constellation diagram as a modulated signal (64QAM, CR=3/4, GI=1/8, 8k-mode, BW=8MHz) is applied. The MER (EVM) per subcarrier is measured as shown in Fig. 6.29. The slight roll-off around the edges may result from the impairment of the analog baseband, such as frequency-dependent I/Q imbalance. The measured performance overview after the digital demodulation is shown in Fig. 6.30. The standard defines a minimum requirement on the BER before RS not exceeding 2E-4. The receiver MER (EVM) as a function of the input power level is shown in Fig. 6.31, where a 64-QAM 3/4 OFDM signal was applied. The measurement shows that –81.4dBm/-81.5dBm sensitivity for the single-ended and differential mode is achieved. Similarly, the measurement result with a QPSK 1/2 modulation signal is shown in Fig. 6.32. The measurement shows that -97.5dBm/-97.7dBm sensitivity in the single-ended and differential mode is achieved.
As can be seen, the MER better than 30dB ranges from -70dBm to -8dBm, allowing for robust operation in a mobile environment. Moreover, the differential mode shows a wider dynamic range due to its better linearity as the RF front-end is set at the minimum gain mode.
The measured performance is summarized in Table 6.3, including the single-end and differential modes. Table 6.4 presents the measurement results related to the
Chapter 6 65nm Tuner Implementation and Verification
139
selectivity and linearity patterns based on the MBRAI 2.0 specification. The result shows that this tuner chip complies with the requirement.
Fig. 6.16 Measurement setup with an external LDO.
Fig. 6.17 Photograph of actual PCB in the differential receiving mode.
Chapter 6 65nm Tuner Implementation and Verification
140
200 400 600 800 1000
-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0
Single-ended
Differential (incl. balun)
S 11 (dB)
Frequency (MHz)
Fig. 6.18 Measured input return loss in both two receiving modes.
474 514 554 594 634 674 714 754 794 834 84.5
85.0 85.5 86.0 86.5 87.0 87.5 88.0 88.5
Single-ended
Differential (incl. balun)
Voltage Gain (dB)
Channel (MHz)
Fig. 6.19 Measured gain flatness across the band (RFE=max, ABB=50dB).
Chapter 6 65nm Tuner Implementation and Verification
141
474 514 554 594 634 674 714 754 794 834 3.0
3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0
Single-ended
Differential (incl. balun)
NF (dB)
Channel (MHz)
Fig. 6.20 Measured NF in the high-gain mode (RFE=max, ABB=50dB).
474 514 554 594 634 674 714 754 794 834 8.0
8.5 9.0 9.5 10.0 10.5 11.0
Single-ended
Differential (incl. balun)
NF (dB)
Channel (MHz)
Fig. 6.21 Measured NF in the low-gain mode (RFE=max, ABB=50dB).
Chapter 6 65nm Tuner Implementation and Verification
142
-50 -40 -30 -20 -10 0 10 20 30
-140 -120 -100 -80 -60 -40 -20 0 20 40
P IMn (dBm)
Pin (dBm)
fund. (input referred) IM3 (input referred) IM2 (input referred)
Fig. 6.22 Measured IIP3 and IIP2 in the high-gain mode (RFE=max,
ABB=50dB).
Fig. 6.23 Measured spectrum at the baseband output with the RF port terminated to a 50ohm terminator; clock leakage component at 2MHz is measured at the gain setting of 93dB.
Chapter 6 65nm Tuner Implementation and Verification
143
Fig. 6.24 Signal quality deterioration due to the clock leakage, measured with the RF port having an input of -96.6dBm QPSK signal.
450 500 550 600 650 700 750 800 850 900 -125
-120 -115 -110 -105 -100
-95 Single-ended
Differential (incl. balun)
Leakage to RF (dBm)
Channel (MHz)
Fig. 6.25 Measured clock leakage level referred to the RF input across the
band of interest.
Chapter 6 65nm Tuner Implementation and Verification
144
Fig. 6.26 Measured phase noise spectrum at the VCO output (2.844GHz).
Fig. 6.27 Measured phase noise spectrum at the baseband output at the 474MHz channel (LO=2.844GHz/6).
Chapter 6 65nm Tuner Implementation and Verification
145
Fig. 6.28 Measured constellation diagram of DVB-H signal (64-QAM 3/4).
Fig. 6.29 Measured MER (EVM) per subcarrier for DVB-H signal (8k mode).
Chapter 6 65nm Tuner Implementation and Verification
146
Fig. 6.30 Measured performance overview for DVB-H signal (8k mode).
-90 -80 -70 -60 -50 -40 -30 -20 -10 0 15
20 25 30 35 40
Single-ended
Differential (incl. balun) estimates by hand cal.
MER (dB)
Pin (dBm)
64-QAM, CR=3/4, 8k, GI=1/8
Fig. 6.31 Measured MER versus the input power level (64-QAM 3/4).
Chapter 6 65nm Tuner Implementation and Verification
147
-110 -100 -90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 0
5 10 15 20 25 30 35 40
Single-ended
Differential (incl. balun) estimates by hand cal.
QPSK, CR=1/2, 8k, GI=1/8
MER (dB)
Pin (dBm)
Fig. 6.32 Measured MER versus the input power level (QPSK 1/2).
Chapter 6 65nm Tuner Implementation and Verification
148
TABLE 6.3
PERFORMANCE SUMMARY OF RFTUNER
Mode/Band Parameter
Single-ended Differential
UHF L-band UHF L-band
Input Return Loss (dB) < -12 < -12 < -11 < -11 Gain Max/Min (dB) 107 / -8 105 / 8 106.5 / -8.5 105 / 8
Gain Step (dB) 0.5 0.5 0.5 0.5
RF range/BB range (dB) 37 / 78 19 / 78 37 / 78 19 / 78
NF @(a)Max Gain (dB) 3.3－3.8 4.2 3.0－3.8 3.9
NF @ (b)Low Gain (dB) 8.7－9.7 11 8.8－10.2 12.2
IIP2 (N+2) @ (a)Max Gain/
(b)Low Gain (dBm) 29 / 50 34 / 55 35 / 54 40 / 58
IIP3 (N+2,N+4) @ (a)Max
Gain/ (b)Low Gain (dBm) -11 / 6 -10 / 15 -9.5 / 8 -8 / 16 Phase noise @
10k/100k/1M (dBc/Hz) 105/104/127 99/98/121 105/104/127 99/98/121 Integrated phase noise
(400-4MHz) < 0.4 < 1 < 0.4 < 1
Filter rejection (c) @
5.25/13.25MHz 32 / 46 32 / 46 32 / 46 32 / 46
DC offset < 6 < 6 < 6 < 6
I/Q matching < -37 < -37 < -37 < -37 Leakage to RF (dBm) <-98 < -115 -112 < -118
Supply Voltage (V) 1.2 1.2 1.2 1.2
Current Consumption (mA) 88 90 90 90
Die size (mm2) 7.8 7.8 7.8 7.8
(a) Low Gain = Max RF -16dB, (b) Low Gain = Max RF -16dB, (c) at 4MHz BW setting.
Chapter 6 65nm Tuner Implementation and Verification
149
TABLE 6.4
SELECTIVITY/LINEARITY AND SENSITIVITY MEASUREMENT RESULTS
Mode Single-ended Differential
Pattern Modulation Interferer location
U/D (dB) Spec.
U/D (dB) Measured
U/D (dB) Measured S1
(PAL-G) 8k 64-QAM 3/4
N+1 35 38.4 38.8
N-1 35 38.2 39
N+2 43 45.6 46.2
N-2 43 45.6 46.2
S2
(PAL-G) 8k 64-QAM 3/4
N+1 27 34.7 35.4
N-1 27 34.7 35.4
N+2 40 43.8 44
N-2 40 43.8 44.4
L1 8k 16-QAM 2/3 N+2, N+4 40/45 41.8/46.8 41.4/46.4
L2 8k 16-QAM 2/3 N+2, N+4 45 46.9 46.4
L3 8k 16-QAM 2/3 N+2, N+4 40 44.8 43
Modulation C/N
(dB) Spec. (dBm) Measured (dBm)
Measured (dBm) Sensitivity
8k 64-QAM 3/4 19.9 -81.3 -81.4 -81.5
8k 16-QAM 2/3 12.7 -88.5 -88.8 -89
8k QPSK 1/2 4.6 -96.6 -97.3 -97.5
Chapter 6 65nm Tuner Implementation and Verification
150
6.7 Conclusion
A 1.2V highly integrated RF tuner for DVB-T/H applications in 65nm CMOS technology is demonstrated. Based on the same architecture in the 0.13μm implementation, the tuner further integrates a wideband LNA compatible for differential and single-ended inputs to meet the requirements either on RF-alone or system-on-a-chip (SoC) developments. The impacts of technology scaling on RF/mixed-signal design have been explored by the study of the proposed LNA design.
In addition, the critical issues of the board-level design are discussed. The tuner consumes only 88mA from a single 1.2V supply in the continuous mode and occupies a silicon area of 7.8 mm2. The measurement results show that this tuner complies with the MBRAI 2.0 requirement.
Chapter 6 65nm Tuner Implementation and Verification
151