SlideShare a Scribd company logo

Analysis of GSM ORFS Issue

criterion123, profile picture
criterion123

This document discusses various topics related to optimizing radio frequency (RF) performance in wireless transmitters, including: 1. Adjacent channel power ratio (ACPR) testing parameters and their importance in characterizing distortion and interference. 2. Output RF spectrum (ORFS) measurements for GSM systems, including modulation spectrum and effects of rapid power ramping. 3. Techniques for timing optimization of the RF power amplifier and antenna in GSM transmitters. 4. Issues that can arise from improper transmitter timing settings, such as lower than expected output power or calibration/registration failures. 5. Factors that impact the linearity and output RF spectrum performance of power amplifiers,

Engineering
1 of 144
Downloaded 270 times
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Most read
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
Most read
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
Most read
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
Analysis of GSM ORFS Issue
Introduction Calibration PA Power Management IC TX Timing Optimization RF Power Amplifier RF Connector
VCO Supply Voltage I/Q Signal Power Management IC Common GND
By Criterion 4 Introduction
By Criterion 5 ACPR(Adjacent Channel Power Ratio) nn For many of the current and future transmission standards(GSM, CDMA, TD-SCDMA, WCDMA, and LTE), ACPR (sometimes also termed adjacent channel leakage ratio-ACLR) is an important test parameter for characterizing the distortion of subsystems and the likelihood that a given system may cause interference with a neighboring radio[32].
By Criterion 6 ORFS(Output Radio Frequency Spectrum) nn In GSM, As part of the out-of-channel measurements, the ACPR is defined by two measurements: spectrum due to modulation and wideband noise, and spectrum due to switching. These two measurements are usually grouped together and referred to as output RF spectrum (ORFS)[34].
By Criterion 7 ORFS due to Modulation nn The modulation process in a transmitter causes continuous wave carriers to spread spectrally. The spectrum due to modulation and wideband noise measurement is used to ensure that the modulation process does not cause excessive spread. This would cause interference to adjacent channel users[34]. With max PCL, why is the carrier power merely 22 dBm? With max PCL, why is the carrier power merely 22 dBm?
By Criterion 8 ORFS due to Modulation nn The analyzer is tuned to a spot frequency and then time-gated across 50% ~ 90% part of the modulated burst to avoid training sequence[34]. nn Besides, spectral components that result from the effect of bursting do NOT appear because the ramps are gated out[34].
By Criterion 9 ORFS due to Switching nn GSM/EDGE transmitters ramp RF power rapidly due to burst mode. If RF power is ramped too quickly, undesirable spectral components exist in the transmission. Users operating on different frequencies, especially those close to the channel of interest, will experience significant interference[34].
By Criterion 10 ORFS due to Switching nn In this case, no time gating is used, so power from both the ramping and modulation processes affect the measurement[34]. nn Nevertheless, because ORFS due to switching level is larger than modulation level, the effect of ramping dominates the spectrum due to switching measurements[34].
By Criterion 11 Calibration
By Criterion 12 DPD nn DPD(Digital Predistortion) is applied to linearize the PA to improve the linearity in the linear region and increase the linear region by compensating the compression with an inverse function[39]. nn As shown below, DPD can be used to lower spectral regrowth to improve both ACLR and ORFS as well[1,41,66].
By Criterion 13 GMSK and 8PSK nn GMSK is a constant-envelope modulation, but 8PSK is NOT[59]. nn Thus, 8PSK has larger PAPR than GMSK. In other words, 8PSK has more strict linearity than GMSK[61]. Average Power Peak Power Time Power Average Power Peak Power Time Power
By Criterion 14 GMSK and 8PSK nn With the identical output power, 8PSK has worse ORFS- SW than GMSK[62]. nn Thus, Pre-distortion calibration helps EDGE 850/900 ORFS-SW margin, not ORFS-MOD[12].
By Criterion 15 Path Delay for EDGE nn A source of distortion is a delay that might occur between the amplitude and phase path[42,74].
By Criterion 16 Path Delay for EDGE nn The existence of two separate waveform paths could lead to different signal delays on the phase path and the envelope path, affecting the EDGE waveform quality[1,42]. nn As shown below, the larger the delay is, the worse the ORFS due to modulation and EVM will be[42].
By Criterion 17 Path Delay for EDGE nn Hence, while calibration, the chipset solution allows for delays to be added to either path to make sure that the phase and envelope paths remain in sync to improve ORFS due to modulation[1,9].
By Criterion 18 Path Delay for EDGE nn Path Delay Cal Tree is as below[9]:
By Criterion 19 Path Delay for EDGE nn The best-case delay is determined by finding the delay value resulting in the lowest ORFS due to the modulation measurement at +400 kHz and -400 kHz[1,9].
By Criterion 20 Path Delay for EDGE nn As shown below, the figure shows a 8PSK modulation spectrum for various values of delay mismatch. With a delay mismatch more than 30ns, the spectrum fails. And the spectrum passes with a delay mismatch less than 30ns[54].
By Criterion 21 PA Power Management IC
By Criterion 22 QFE2101 nn The main QFE2101 functional blocks are[4]: uuAPT buck converter – essentially a programmable buck converter, or switched-mode power supply (SMPS). uuBypass switch – routes battery or primary phone power (VPH_PWR) directly to the PA(s). uuGSM switches – bypass switches (or optional extra GSM capacitor switch). ll Bypass – a separate bypass switch for GSM with a very low resistance that connects the GSM PA directly to battery. ll GSM capacitor switch – C_GSM pin should be connected to primary phone power (VPH_PWR); utilize this switch as the third bypass switch to further reduce the IR drop.
By Criterion 23 QFE2101
By Criterion 24 QFE2101 nn To switch-in extra load capacitance at the VPA node (C_GSM to capacitor only) during GSM Tx. This option can be used to improve the margin for GSM ORFS/RxBN due to less IR drop, and APT mode instead of bypass mode.[2-4].
By Criterion 25 TX Timing Optimization
By Criterion 26 Ramping Profile nn Effect of different Ramp timing on Switching Transients[44].
By Criterion 27 Ramping Profile nn Nevertheless, there is a trade-off between GSM PvT and GSM ORFS due to switching @400kHz with high PCL[6]. nn Hence, tune ramping profile to pass PVT and ORFS specification simultaneously[6,8].
By Criterion 28 GSM RF Timing Adjustment nn In general, there are six types of GSM RF timing adjustment, as shown below[10]:
By Criterion 29 GSM RF Timing Adjustment
By Criterion 30 GSM RF Timing Adjustment nn pa_en_start/stop_offset_adj:The PA enable signal start/stop time adjustment value can be altered to move the rising edge of the PA enable signal forward or backward to get the optimal configuration[9]. nn pa_start/stop_offset_adj :The PA turn-on/off time adjustment value can be altered to adjust the rising edge of the TDMA Tx power burst[9]. nn ant_timing_start/stop_offset_adj : The antenna select control signal start/stop time adjustment value can be altered to move the falling edge of antenna select control signal forward or backward relative to the PA enable signal falling edge to get the optimal configuration[9].
By Criterion 31 PA ON Timing[9]
By Criterion 32 PA OFF Timing[9]
By Criterion 33 PA Ramp Up Timing[9]
By Criterion 34 PA Ramp Down Timing[9]
By Criterion 35 Antenna ON Timing[9]
By Criterion 36 Antenna OFF Timing[9]
By Criterion 37 Potential Issues of Improper Tx Timing nn In addition to PvT and ORFS, Improper Tx timing value may cause improper power in certain hardware boards[9]. nn For MIPI PA device, PA ON signal is expected earlier than the PA range signal, since PA ON script sets PA to the lowest gain state and overrides the settings in the PA range script[9].
By Criterion 38 Potential Issues of Improper Tx Timing nn If PA ON is later than PA range timing, the power will be lower than expected; e.g. Max Tx power is merely ~20 dBm[9] nn The error message can be observed in QXDM log[9].
By Criterion 39 Potential Issues of Improper Tx Timing nn And the improper power may cause EDGE calibration or registration failure in certain hardware boards[9] nn Failed to calibrate GSM850 due to improper PA timing applied[9]. nn Failed to calibrate GSM900 per improper PA timing applied[9].
By Criterion 40 Potential Issues of Improper Tx Timing nn In the case, PCS1900, there is ORFS issue at low temperature, and which is independent of PCL. nn Nevertheless, the issue was NOT observed in DCS1800.
By Criterion 41 Potential Issues of Improper Tx Timing nn First, we did NOT observe this issue in FTM(Factory Test Mode). In other words, the issue disappears without RX operation. nn Second, we did NOT observe this issue at transceiver output, as shown below :
By Criterion 42 Potential Issues of Improper Tx Timing nn Because there are merely two blocks(transceiver and PAMiD) in TX chain, so we conclude that this issue is related to PAMiD[74], especially ASM timing.
By Criterion 43 Potential Issues of Improper Tx Timing nn There are two statuses for TX mode while RX is operating : uuDisable Status(Default) : uuIsolation Status: RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time ASM RX TX PA ASM RX TX PA RX OFF RX ON ASM_TX ON Level Time RX OFF RX ON ASM_TX ON Level Time
By Criterion 44 Potential Issues of Improper Tx Timing nn For disable status, when switching to Tx state under low temperature, it may take up to 100 µS for the switch to be fully operational and RF compliant. That’s to say, there will be delay in TX timing while switching from TX OFF to TX ON[28]. nn As mentioned earlier, the antenna select control signal start/stop time affects ORFS performance. That’s why ORFS fails. RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time
By Criterion 45 Potential Issues of Improper Tx Timing nn For isolation status, TX is still ON while RX is ON, but TX port will shunt to ground while RX is ON. That’s why it’s called “isolation” status. nn Because TX is always ON, there will be NO delay in TX timing while switching from TX OFF to TX ON. Hence the issue passes after introducing Isolation Status. RX OFF RX ON ASM_TX ON Level Time RX OFF RX ON ASM_TX ON Level Time ASM RX TX PA ASM RX TX PA
By Criterion 46 Potential Issues of Improper Tx Timing nn Besides, DCS1800 RX path is different from PCS1900 RX path. Chances are that PCS1900 takes longer time switching from RX port to TX port than DCS1800, so we did NOT observe this issue in DCS1800. RX OFF RX ON Level Time ASM_TX OFF ASM_TX ON ASM_TX OFF ASM_TX ON PCS DCS RX OFF RX ON Level Time ASM_TX OFF ASM_TX ON ASM_TX OFF ASM_TX ON PCS DCS
By Criterion 47 RF Power Amplifier
By Criterion 48 Linearity nn As mentioned above, ORFS is just the concept of ACLR. And the formula of ACLR is : ACLR = 2(Pin - OIP3 + G)+C, where C is correction coefficient nn Hence, the better the linearity is, the better the ORFS due to switching will be[45,46].
By Criterion 49 Temperature nn In general, the higher the temperature is, the worse the OIP3 will be[47]. nn Thus, chances are that high temperature leads to worse ORFS due to switching[15].
By Criterion 50 Layout nn Thus, the GND vias should be as numerous as possible to spread the heat[48,63].
By Criterion 51 Temperature nn As shown below, the lower the temperature is, the higher the gain will be[47]. nn Thus, with constant Pin, lower temperature leads to higher Pout[49].
By Criterion 52 Temperature nn As shown below, the larger Pout is, the worse the ACLR will be. nn That’s to say, chances are that low temperature leads to low ORFS due to switching margin[43].
By Criterion 53 Temperature Compensation nn Hence, we have to do temperature compensation to make Pout identical under all temperatures. nn By doing this, the Pout under low temperature will not be too high, thereby improving ORFS due to switching under low temperature.
By Criterion 54 Temperature Compensation nn Nevertheless, with temperature compensation, the ORFS due to switching under high temperature may aggravate more.
By Criterion 55 Temperature Compensation nn As mentioned earlier, high temperature aggravates linearity, thereby degrading ORFS due to switching. nn With TC(Temperature Compensation), the Pout under high temperature becomes larger. With worse linearity and larger Pout, that’s why ORFS due to switching under high temperature may aggravate more with TC.
By Criterion 56 Ramping Profile Adjustment nn Hence, with TC, adjust the ramping profile under high temperature to improve ORFS due to switching because Power versus time (PVT) variation is one of the most important features for describing electrothermal performance of RF power amplifiers (PAs)[51,52].
By Criterion 57 PA Output Matching Network Adjustment nn Besides, adjust PA output matching network to improve linearity and lower noise floor, thereby improving ORFS due to switching and RxBN under high temperature with TC[6].
By Criterion 58 PA Output Matching Network Adjustment nn As mentioned earlier, 8PSK has more strict linearity than GMSK. In other words, if ORFS issue is related to PA linearity, the symptom should be:
By Criterion 59 PA Input Matching Adjustment nn As shown below, the PA input matching networks is DA’s(Driver Amplifier) load-pull as well. nn In other words, good ORFS@PA input leads to good ORFS@PA output[50,53]. nn Hence, to optimize PA input matching network can be used to improve ORFS as well.
By Criterion 60 PA Post loss nn According to the following formula : with constant target power, larger PA post loss leads to larger PA output power, thereby aggravating ORFS-SW @ +- 400 kHz, even though PA’s output impedance is close to 50Ω[11]. Target Power = PA Output Power – PA Post loss nn Thus, minimize the PA post loss to have more headroom in Tx power. Because LB has larger max power than HB, headroom equal to 0.7 dB from 32.5 dBm(i.e. 33.2 dBm) is preferred[11].
By Criterion 61 Stability nn In the case, LB EDGE ORFS test fails in both modulation/switching spectrum when PCL is set < 10(High Power Mode), and passes when PCL is set ≥ 11(Low Power Mode)[11].
By Criterion 62 Stability nn Firstly, the best ORFS due to modulation result was found to be only approximately -43 dB, even after tuning the path delay, but it can reach -60 dB[11]. ORFS-MO Before After -400KHz -43.1158 dB < -60 dB +400KHz -43.0755 dB < -60 dB ORFS-MO Before After -400KHz -43.1158 dB < -60 dB +400KHz -43.0755 dB < -60 dB nn Besides, many spurs are found(shown as yellow traces in the graph), according to the analysis, the PA is oscillating[11].
By Criterion 63 Stability nn According to the K-factor definition, the stability depends on S-parameter[56]. Chances are that the components marked as red circle affect the S-parameter and stability of QFE2320, thereby oscillating[11] nn Because the issue occurs in LB, the components marked as red circle must be changed to remove oscillation[11].
By Criterion 64 Stability nn Therefore, after tuning the path delay and changing the BOM, LB EDGE ORFS margin is sufficient for all PCL[11].
By Criterion 65 Switch Point nn For this case, EDGE LB ORFS fails at PCL8, PCL12, and PCL16. nn As mentioned above, If failure is at maximum power and improves as power is reduced, the issue is likely due to linearity. Thus, it fails at PCL8 but passes at PCL9 ~ PCL11[10,16]. PCL 8 9 10 11 Result Fail Pass Pass Pass PCL 12 13 14 15 Result Fail Pass Pass Pass PCL 16 17 18 19 Result Fail Pass Pass Pass
By Criterion 66 Switch Point nn In general, there is hysteresis from the switching point where the PA rises from a lower gain to a higher gain, and vice versa to avoid the PA oscillating at the switching point[64]. nn In other words, as illustrated below, for EDGE single- slot, 19 dBm(PCL12) can be achieved by high gain mode or mid gain mode; 11 dBm(PCL16) can be achieved by mid gain mode or low gain mode.
By Criterion 67 Switch Point nn Hence, as illustrated below, if 11 dBm is achieved by low gain mode, the PA will be in saturation region, thereby degrading linearity and aggravating ORFS. So is 19 dBm(PCL 12). nn In order to solve this issue, make these output power @ switch points be in proper gain mode to be in linear region[16].
By Criterion 68 RF Connector
By Criterion 69 Material issue nn If the connector is abnormal, the antenna matching network may alter PA’s output impedance, thereby aggravating ORFS. nn Besides, abnormal connector leads to leakage. And as shown below, with constant target power(e.g. LB 32.5 dBm or HB 29.5 dBm), the larger the leakage is, the larger the PA output power will be, thereby aggravating ORFS.
By Criterion Layout issue nn In the case, it is very difficult to pull load impedance to 50 Ω for the common path[11]. nn Hence, ORFS due to switching performance is not good enough.
By Criterion Layout issue nn Because the signal pad size of the RF connector is very large(0.6 mm × 0.7 mm). According to simulation, if using L9 as RF GND, it introduces serious mismatching. nn Thus, areas under the pads of the RF connector in L9, L8, and L7 should be cut out, and set L6 as the reference GND, to get better impedance[11]. Signal Pad GND Pad L9 L8 L7 L6 Impedance 14 Ω 32 Ω 44 Ω 53 Ω
By Criterion 72 VCO
By Criterion 73 VCO Pulling nn In the case, GSM ORFS (modulation and switching) is observed to be failing by about 5−6 dB on quad-band GSM some channels at high power level (LB: PCL5; HB: PCL0)[30]. nn The max power at connector can be 34 dBm for LB, and 31 dBm for HB. Besides, the PA’s load-pull is close to 50Ω for both LB and HB. In other words, the issue is NOT related to PA linearity.
By Criterion 74 VCO Pulling nn As shown below, the issue can be solved with SPDT.
By Criterion 75 VCO Pulling nn Without SPDT, the issue is root caused too high Tx power during GSM operation leaking into FBRx path through front-end coupler and later coupling into TX quadrature up-converter of transceiver[16,17,30,72].
By Criterion 76 VCO Pulling nn An SPDT (with bypass option) is recommended at front end, as shown here to isolate GSM Tx power from leaking into FBRx path[17,30]. nn Since FBRx is not operational in GSM mode. FBRX only works in 3G/4G mode. So this switch (SPDT) does not impact GSM operation[16,30]. nn Two possible implementations during GSM operation are shown here[30]:
By Criterion 77 VCO Pulling nn In the case, as shown below, ORFS-MO passes if we press the shielding can, and it fails without press.
By Criterion 78 VCO Pulling nn As shown below, PA will couple TX signal onto the shielding can. In other words, there will be residual TX signal on the shielding can. nn With good grounding, the residual TX signal will flow to GND completely. Without good grounding, the shielding can will reflect the residual TX signal.
By Criterion 79 VCO Pulling nn Thus, with press, the residual TX signal flows to GND completely due to good grounding. nn If possible, don’t put PA and transceiver in the same shielding area to avoid VCO pulling.
By Criterion 80 VCO Pulling nn As mentioned earlier, the PA input matching networks is DA’s(Driver Amplifier) load-pull as well. nn Besides, for Direct-Conversion transmitter architecture, the RF frequency is the same as LO. So if PA input matching is not close to 50Ω, the reflection due to poor S11 may interfere VCO, thereby aggravating ORFS-MO.
By Criterion 81 VCO Pulling nn In the case, the relationship between ORFS-MO and ASM is as below : Antenna Switch Module Vendor A (With Original RX Matching) Vendor B (With Original RX Matching) Vendor B (With New RX Matching) Pass Fail Pass
By Criterion 82 VCO Pulling nn If we set TX port as Port 1, and RX port as Port 2, S21 is just the isolation between TX port and RX port, as shown below : nn Hence, the isolation depends on TX matching, RX matching, and ASM.
By Criterion 83 VCO Pulling nn Therefore, vendor B ASM degrades isolation, and ORFS-MO fails. nn With vendor B ASM, we introduce new RX matching, and isolation improves, thereby improving ORFS-MO. nn As shown below, the isolation affects VCO pulling through RX path.
By Criterion 84 Keep Out Area nn Besides, keep-out areas on PCB layer 1 are required to avoid sensitive on-chip components such as VCO[73].
By Criterion 85 Keep Out Area nn Otherwise, the parasitic effect aggravates phase noise, thereby degrading ORFS-MO[10]. nn In general, check the LO phase noise at 400 kHz offset, it should be better than -118 dBc/Hz[10].
By Criterion 86 Supply Voltage
By Criterion 87 IR Drop nn IR drop refers to a voltage drop that appears at the resistive component of any impedance. This voltage drop across any resistance is the product of current (I) passing through resistance and resistance value (R). nn If there is IR drop issue in the supply voltage of any TX chain blocks, such as transceiver, PA, and ASM, the TX performance including ORFS will aggravate[24].
By Criterion 88 IR Drop nn In the case, LB GSM ORFS-MO fails for high power mode In low temperature. nn We found the ORFS-MO issue appears not only in connector, but also in transceiver output. As mentioned above, poor ORFS performance in PA input leads to worse ORFS performance in PA output.
By Criterion 89 IR Drop nn Thus, there are three possible factors related transceiver: Layout, supply voltage, and material issue. nn After changing the BOM for transceiver supply voltage, the issue is solved.
By Criterion 90 IR Drop nn As mentioned above, low temperature makes gain increase. Higher gain leads to higher Icc, thereby increasing IR drop. nn In addition, with max PCL, low band has larger output power than high band. So the issue occurred in LB instead of HB[57]. LB 14 dBm HB 11 dBm
By Criterion 91 IR Drop nn As shown below, power is getting lower as supply voltage drops[11]. In other words, the lower the supply voltage is, the lower power will be.
By Criterion 92 IR Drop nn In other words, with constant target power(e.g. LB 32.5 dBm or HB 29.5 dBm), the larger the IR drop is, the larger PA output power and current consumption(i.e. Icc) will be, thereby aggravating ORFS-SW performance. IR drop PA Output Power & Icc IR drop PA Output Power & Icc
By Criterion 93 IR Drop nn Besides, as shown below, the larger the temperature is, the larger the resistance including chip resistor or trace will be, thereby increasing IR drop.
By Criterion 94 IR Drop nn As mentioned above, with TC(Temperature Compensation), the Pout under high temperature becomes larger. With worse linearity and larger Pout, that’s why ORFS due to switching under high temperature may aggravate more with TC. nn Thus, with TC and IR drop, ORFS-SW aggravates more in high temperature.
By Criterion 95 Decoupling Capacitor nn Due to burst mode characteristic, GSM PA is being ON and OFF all the time. As shown below, while PA is ON, there will be unwanted 20KHz signal riding on DC, which is just ripple. Any supply disturbances caused by on/off activities could directly convert to ORFS issue[10].
By Criterion 96 Decoupling Capacitor nn As shown below, larger decoupling capacitor can help improve spectral regrowth, thereby improving ORFS[31].
By Criterion 97 Decoupling Capacitor nn Besides, the decoupling capacitor should be as close to PA as possible. Otherwise, PA supply transient current may leak into other ICs, and transient current from other ICs may leak into GSM PA, thereby aggravating ORFS performance[10].
By Criterion 98 Decoupling Capacitor nn Besides, the supply voltage of LO is important as well. nn As shown below, change C3115 from 220nF to 4.7uF, ORFS-MO at 1.6 MHz offset improves.
By Criterion 99 Decoupling Capacitor nn The lower ESL is, the wider capacitive range will be. nn The lower ESR is, the lower ripple will be
By Criterion 100 Decoupling Capacitor nn The path the noise flows inside 3-terminal capacitor is shorter than which inside 2-terminal capacitor. That’s to say, the 3-terminal capacitor has lower ESL and ESR.
By Criterion 101 Decoupling Capacitor nn Compared to multiple capacitors, the 3-terminal capacitor has better noise suppression, and it doesn’t have anti-resonance issue. Besides, the 3-terminal capacitor can save PCB area due to less components[75].
By Criterion 102 Star Routing nn Use star routing from battery to PMIC and PA. Besides, the branch point should be as close to battery as possible[10]. nn As shown below, the length of coupling path 2 is shorter than coupling path 1, so the noise from GSM PA may leak to PMIC. That’s why branch point should be close to battery.
By Criterion 103 Star Routing nn In the case, at high power level of GSM low bands, ORFS modulation performance is marginal, especially at the 400 kHz offset. nn The issue is dependent on PCL. It disappears at low power level such as PCL19. Chances are that the issue is related to PA linearity, so we checked transceiver output performance to eliminate the influence of PA.
By Criterion 104 Star Routing nn The issue already appears at transceiver output, so there are three possibilities : material, layout, and voltage supply. nn With an external voltage supply, the issue improved.
By Criterion 105 Star Routing nn Due to burst mode characteristic, GSM PA has strong transient current. That’s why the issue disappears at low power level because low power level leads to low transient current. And that’s why the issue appears at high power level of GSM low bands because PCL5 of LB (33 dBm) is larger than PCL0 of HB(30 dBm). nn As shown below, the coupling path : GSM PA => PMIC => transceiver
By Criterion 106 Star Routing nn In the case, as shown below, ORFS modulation performance fails when monitor is ON, but performance is good when monitor is OFF. nn Besides, the issue appears at GSM quad-band, especially at the 400 kHz offset. And the issue is independent on power level.
By Criterion 107 Star Routing nn This is because backlight driver IC has strong transient current as well. That’s why the issue disappears when monitor is OFF. nn As shown below, Coupling path 1 : Backlight driver IC => PA Coupling path 2 : Backlight driver IC => PMIC => Transceiver
By Criterion 108 Star Routing nn Thus, increase the decoupling capacitor and power inductor value for Backlight driver IC, and the performance improved.
By Criterion 109 Star Routing nn Similarly, use star routing from PMIC to VBATT pin and Vcc_GSM pin of GSM PA separately, and the branch point should be as close to PMIC as possible. nn At least, the branch point should NOT be close to GSM PA, as shown below :
By Criterion 110 Star Routing nn If the branch point is close to GSM PA, make the VPH_PWR common part(marked as pink) and separate part(marked as green) on different layer. As shown below, the coupling path(marked as yellow) has high impedance to avoid noise coupling between the 2 pins.
By Criterion 111 High Frequency Noise nn In the case, as shown below, ORFS-MO passes if we press the PA shielding can, and it fails without press. Besides, after removing shielding can, it passes as well. nn PA and transceiver are in separate shielding area, so the issue is NOT related to VCO pulling.
By Criterion 112 High Frequency Noise nn As mentioned earlier, without good grounding, the shielding can will reflect the residual TX signal. With press, the residual TX signal flows to GND completely due to good grounding. nn In addition, no shielding can, no residual TX signal. That’s why the issue disappears after removing shielding can.
By Criterion 113 High Frequency Noise nn Thus, the issue is due to that residual RF TX signal reflects to PA Vcc by shielding can. In terms of DC voltage, RF signal is regarded as noise. nn There’re usually large capacitor for decoupling and small capacitor for filtering high frequency noise. For the issue, we can modify small capacitor value to filter residual RF TX signal.
By Criterion 114 I/Q Signal
By Criterion 115 XO Harmonics nn In the case, on certain channels of GSM high bands, ORFS switching and modulation performance is marginal, especially at the 400 kHz offset. nn The most affected channels are DCS channel 599, 600, 602, 791, 792, 794, and 795; and PCS channel 668.
By Criterion 116 XO Harmonics nn As shown in the table below, these channels are all 90 times 19.2 MHz approximately in terms of frequency. In other words, these channel are related to XO harmonics. nn Baseband chip has an internal issue related to XO harmonics coupling into TX I/Q lines that cause degradation of ORFS performance only in GSM high bands[13,14]. Band Channel Frequency 19.2 MHz * N 599 1727.6 89.97916667 600 1727.8 89.98958333 602 1728.2 90.01041667 791 1766 91.97916667 792 1766.2 91.98958333 794 1766.6 92.01041667 795 1766.8 92.02083333 PCS 668 1741.4 90.69791667 DCS Band Channel Frequency 19.2 MHz * N 599 1727.6 89.97916667 600 1727.8 89.98958333 602 1728.2 90.01041667 791 1766 91.97916667 792 1766.2 91.98958333 794 1766.6 92.01041667 795 1766.8 92.02083333 PCS 668 1741.4 90.69791667 DCS
By Criterion 117 XO Harmonics nn As shown above, the solution is to add series L and shunt C onto each of the TX_I/Q lines[63]. nn As the spur level is reduced, and therefore the ORFS modulation is improved[63]. nn This LC filter can also help reduce the number of exceptions for the GSM Rx band noise[63]. w/o solution w/ solution
By Criterion 118 XO Harmonics nn The trace between transceiver and PMIC is rich in XO harmonics because it is XO reference clock. nn Thus, we need to put a R-C filter to reject XO harmonics and a DC block to reject DC offset. nn Besides, we need to keep the trace away from I/Q signal to avoid ORFS issue.
By Criterion 119 XO Harmonics nn The digital XO signal and the analog XO signal can corrupt each other; the layout should provide isolation between them. Isolation is highly recommended [77]. nn As shown below, they are too close.
By Criterion 120 I/Q Imbalance nn I/Q imbalance can be characterized by: a phase mismatch that can happen between the I and the Q components of the local oscillator signals, which becomes not exactly 90 degrees, and an amplitude mismatch due to the gain difference of the mixers of the I and Q branches[70].
By Criterion 121 I/Q Imbalance nn As shown below, I/Q imbalance aggravates image, thereby aggravating sideband suppression and ORFS- MO. nn Thus, we need to adjust phase and amplitude offsets between I and Q channel by means of calibration to lower image to improve ORFS-MO[69].
By Criterion 122 I/Q Imbalance nn Besides, in terms of modulator, high temperature aggravates sideband suppression as well, thereby aggravating ORFS-MO. So good thermal design of transceiver is important[69]. nn This is especially important for direct up-conversion transmitter because I/Q imbalance are the inherent shortcomings of it[69].
By Criterion 123 I/Q Imbalance nn Ideally, the four traces on the IQ signal path from the DAC output to the modulator input should be symmetrical between the I channel and Q channel and between the positive side and negative side within a channel[69]. nn In reality, due to PCB layout, trace lengths are not perfectly matched. The mismatches cause the signal in one channel to be skewed from the other, and, therefore, result in IQ imbalance(both gain and phase)[69].
By Criterion 124 LO Leakage(Carrier Leakage) nn Excessive DC offsets in I/Q channels cause high levels of LO leakage, thereby aggravating ORFS-MO[69]. nn According to the formula shown below, the more the DC offsets in I/Q channels, the larger the LO leakage will be[69]. = + +
By Criterion 125 LO Leakage(Carrier Leakage) nn Thus, we need to remove DC Offsets by means of calibration to lower LO leakage to improve ORFS- MO[69]. nn In terms of modulator, high temperature aggravates carrier suppression as well, thereby aggravating ORFS- MO. So good thermal design of transceiver is important[69].
By Criterion 126 Power Management IC
By Criterion 127 Charging nn In the case, when charging is enabled, GSM850, ch190, ORFS-MO failure (0.5 dB) is observed at 600 kHz offset measurement[18].
By Criterion 128 Charging nn For a 5 V DCP(Dedicated Charging Port) plug-in, the charging frequency is set to 600 kHz. nn Spur coupling path: PMIC buck switching (600 kHz) spurs leaking into the Vbatt and appears at the Tx-RF output through QET4101 and PA[18]. nn Hence, there will be +- 600 KHz offset ORFS-MO issue due to 2nd order intermodulation mechanism.
By Criterion 129 Charging nn Adding a Pi (π) filter on the VPH_PWR rail attenuates the charger spur by 20 dB and passes the GSM ORFS specification[18]. nn As a workaround, the placeholder for the π filter should be provided at the QET4101 input to attenuate spur coupling from the PMIC charger and also a 0 Ω (0201 package) placeholder at power supply line of PA (VCC1_PA) for noise filtering[18].
By Criterion 130 Charging nn As shown below[18]:
By Criterion 131 SMPS nn In the case, GSM ORFS modulation spectrum periodicity jump in all bands and all power levels, especially the mid-power level[28]. nn IC NCP6335 is external SMPS(Switching Mode Power Supply) to the MSM8X25Q core. The NCP6335 is the main interference source to ORFS[28,75-76].
By Criterion 132 SMPS nn NCP6335 has PFM and PWM operation for optimum increased efficiency, The transition between PWM/PFM modes can occur Automatically, that’s why ORFS-MO spectrum periodicity jump[28]. nn Set NCP6335 to forced PWM, and ORFS-MO passed.
By Criterion 133 Common GND
By Criterion 134 Common GND nn In the case, with all decoupling capacitor and bypass capacitors, ORFS fails; without all decoupling capacitor and bypass capacitor, ORFS passes.
By Criterion 135 Common GND nn This is because transient current and high frequency noise may leak into PA Vcc through these capacitors by means of common GND on top layer.
By Criterion 136 Common GND nn Thus, make these shunt capacitors GND pad separate from common GND on top layer(i.e. GND Island), and add GND via as many as possible. nn If GND island is bound by PCB area, at least, the GND via should be as many as possible.
By Criterion Noise 137 Common GND nn As mentioned above, we’re able to use Pi (π) filter to attenuate the noise. Nevertheless, the GND of capacitors should be separate. Otherwise, the noise may bypass the capacitors through common ground, thereby aggravating filtering effect.
Analysis of GSM ORFS Issue
Analysis of GSM ORFS Issue
Analysis of GSM ORFS Issue
Analysis of GSM ORFS Issue
Analysis of GSM ORFS Issue
Analysis of GSM ORFS Issue
Analysis of GSM ORFS Issue

More Related Content

PDF
Carrier Aggregation Discussion
criterion123
 
PDF
The ABCs of ADCs Understanding How ADC Errors Affect System Performance
criterion123
 
PDF
GNSS De-sense By IMT and PCS DA Output
criterion123
 
PDF
ABCs of Carrier Aggregation
criterion123
 
PDF
Challenges In Designing 5 GHz 802.11 ac WIFI Power Amplifiers
criterion123
 
PDF
802.11ac WIFI Fundamentals
criterion123
 
PDF
EVM Degradation in LTE systems by RF Filtering
criterion123
 
PDF
Some issue due to incorrect PA and transceiver configuration
criterion123
 
Carrier Aggregation Discussion
criterion123
 
The ABCs of ADCs Understanding How ADC Errors Affect System Performance
criterion123
 
GNSS De-sense By IMT and PCS DA Output
criterion123
 
ABCs of Carrier Aggregation
criterion123
 
Challenges In Designing 5 GHz 802.11 ac WIFI Power Amplifiers
criterion123
 
802.11ac WIFI Fundamentals
criterion123
 
EVM Degradation in LTE systems by RF Filtering
criterion123
 
Some issue due to incorrect PA and transceiver configuration
criterion123
 

What's hot (20)

PDF
How to solve ACLR issue
criterion123
 
PDF
Performance requirement and lessons learnt of LTE terminal---transmitter part
criterion123
 
PDF
RF Issue Due To PA Timing
criterion123
 
PDF
Performance Requirement and Lessons Learnt of LTE Terminal_Transmitter Part
criterion123
 
PDF
System(board level) noise figure analysis and optimization
criterion123
 
PDF
IIP2 requirements in 4G LTE Handset Receivers
criterion123
 
PDF
Sensitivity or selectivity - How does eLNA impact the receriver performance
criterion123
 
PDF
OXX B66 Rx sensitivity and desense analysis issue debug
Pei-Che Chang
 
PDF
One Case Study For GSM Unstable Output Power Issue
criterion123
 
PDF
Why Ferrite Beads Aggravates ACLR
criterion123
 
PDF
Receiver Desense Common Issue
criterion123
 
PPTX
B2 desence
Pei-Che Chang
 
PDF
DDR Desense Issue
criterion123
 
PDF
SAW-less Direct Conversion Receiver Consideration
criterion123
 
PDF
One LTE B7 Desense Case Study
criterion123
 
PDF
PA Output Notch Filter Consideration
criterion123
 
PDF
A Study On TX Leakage In 4G LTE Handset Terminals
criterion123
 
PDF
Introduction to PAMiD
criterion123
 
PDF
Introduction to modern receiver
criterion123
 
PDF
1 RB sensitivity at middle RBs poor than other RBs
Pei-Che Chang
 
How to solve ACLR issue
criterion123
 
Performance requirement and lessons learnt of LTE terminal---transmitter part
criterion123
 
RF Issue Due To PA Timing
criterion123
 
Performance Requirement and Lessons Learnt of LTE Terminal_Transmitter Part
criterion123
 
System(board level) noise figure analysis and optimization
criterion123
 
IIP2 requirements in 4G LTE Handset Receivers
criterion123
 
Sensitivity or selectivity - How does eLNA impact the receriver performance
criterion123
 
OXX B66 Rx sensitivity and desense analysis issue debug
Pei-Che Chang
 
One Case Study For GSM Unstable Output Power Issue
criterion123
 
Why Ferrite Beads Aggravates ACLR
criterion123
 
Receiver Desense Common Issue
criterion123
 
B2 desence
Pei-Che Chang
 
DDR Desense Issue
criterion123
 
SAW-less Direct Conversion Receiver Consideration
criterion123
 
One LTE B7 Desense Case Study
criterion123
 
PA Output Notch Filter Consideration
criterion123
 
A Study On TX Leakage In 4G LTE Handset Terminals
criterion123
 
Introduction to PAMiD
criterion123
 
Introduction to modern receiver
criterion123
 
1 RB sensitivity at middle RBs poor than other RBs
Pei-Che Chang
 
Ad

Similar to Analysis of GSM ORFS Issue (20)

PDF
Computer aided design of communication systems / Simulation Communication Sys...
Makan Mohammadi
 
PDF
Ltu ex-05238-se
Aircel Shimla
 
PPTX
Lte power control
Pranay Akul
 
PPT
Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)
BorsYanchev
 
PPT
Network Optimization PPTour score increases as you pick a category, fill out ...
chinkupandey1992
 
PDF
Novel Cyclic Prefix Selection to Improve Spectral Efficiency and Signal Stren...
idescitation
 
PPTX
Lecture Notes: EEEC6440315 Communication Systems - Inter Symbol Interference...
AIMST University
 
DOCX
Philip V. Mariano_2016
Philip Mariano
 
PDF
Master_Thesis_Report
Tejas subramanya
 
DOCX
Gsm 2 g rollout dt
INAYAT ULHAQ
 
PDF
Initial tuning umts (2)
Emerson Eduardo Rodrigues Von Staffen
 
PDF
ANALYSIS AND REDUCTION OF TIMING JITTER USING HYBRID OFDM - DFMA PONS
IRJET Journal
 
PDF
Drive Test and Optimization Tutorial - I.pdf
hamdi_saif
 
PDF
cdma2000_Fundamentals.pdf
CheikhAhmetTidianeDi1
 
PDF
MIPI DevCon 2016: MIPI D-PHY - Physical Layer Test & Measurement Challenges
MIPI Alliance
 
DOCX
High Spectral Efficiency Field Trial Using Time-Packed Terabit/s DP-DQPSK Sup...
simonhackett1
 
PPTX
3G Network Drive Testing detailed Learning
VermaJagdeep
 
PPTX
Lte air-interface
Arshad Alam
 
PDF
Optimizing VoLTE Accessibility including Basic Parameters
hmalikzte
 
PDF
IRJET- Reduction in Peak to Average Power Ration using Proposed Approach for ...
IRJET Journal
 
Computer aided design of communication systems / Simulation Communication Sys...
Makan Mohammadi
 
Ltu ex-05238-se
Aircel Shimla
 
Lte power control
Pranay Akul
 
Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)
BorsYanchev
 
Network Optimization PPTour score increases as you pick a category, fill out ...
chinkupandey1992
 
Novel Cyclic Prefix Selection to Improve Spectral Efficiency and Signal Stren...
idescitation
 
Lecture Notes: EEEC6440315 Communication Systems - Inter Symbol Interference...
AIMST University
 
Philip V. Mariano_2016
Philip Mariano
 
Master_Thesis_Report
Tejas subramanya
 
Gsm 2 g rollout dt
INAYAT ULHAQ
 
Initial tuning umts (2)
Emerson Eduardo Rodrigues Von Staffen
 
ANALYSIS AND REDUCTION OF TIMING JITTER USING HYBRID OFDM - DFMA PONS
IRJET Journal
 
Drive Test and Optimization Tutorial - I.pdf
hamdi_saif
 
cdma2000_Fundamentals.pdf
CheikhAhmetTidianeDi1
 
MIPI DevCon 2016: MIPI D-PHY - Physical Layer Test & Measurement Challenges
MIPI Alliance
 
High Spectral Efficiency Field Trial Using Time-Packed Terabit/s DP-DQPSK Sup...
simonhackett1
 
3G Network Drive Testing detailed Learning
VermaJagdeep
 
Lte air-interface
Arshad Alam
 
Optimizing VoLTE Accessibility including Basic Parameters
hmalikzte
 
IRJET- Reduction in Peak to Average Power Ration using Proposed Approach for ...
IRJET Journal
 
Ad

More from criterion123 (8)

PDF
WIFI Spectrum Emission Mask Issue
criterion123
 
PDF
LTE carrier aggregation technology development and deployment worldwide
criterion123
 
PDF
RF Matching Guidelines for WIFI
criterion123
 
PDF
Introduction To Antenna Impedance Tuner And Aperture Switch
criterion123
 
PDF
Introduction to 3 terminal capacitor
criterion123
 
PDF
GPS RF Front End Considerations
criterion123
 
PDF
CDMA Zero-IF Receiver Consideration
criterion123
 
PDF
Reverse IMD
criterion123
 
WIFI Spectrum Emission Mask Issue
criterion123
 
LTE carrier aggregation technology development and deployment worldwide
criterion123
 
RF Matching Guidelines for WIFI
criterion123
 
Introduction To Antenna Impedance Tuner And Aperture Switch
criterion123
 
Introduction to 3 terminal capacitor
criterion123
 
GPS RF Front End Considerations
criterion123
 
CDMA Zero-IF Receiver Consideration
criterion123
 
Reverse IMD
criterion123
 

Recently uploaded (20)

PPTX
FINAL REVIEW FOR COPD DIANOSIS FOR PULMONARY DISEASE.pptx
rubeshbme
 
PPTX
OOP with Java - Java Introduction (Basics)
Rashmi T V
 
DOCX
573137875-Attendance-Management-System-original
AYUSHMANAV
 
PPTX
Construction Project Organization Group 2.pptx
vj5agdales
 
PDF
Model Code of Practice - Construction Work - 21102022 .pdf
AinieButt1
 
PPTX
Foundation to blockchain - A guide to Blockchain Tech
Davies Chacko
 
PPT
Project quality management in manufacturing
BarMusaTetik
 
PDF
TFEC-4-2020-Design-Guide-for-Timber-Roof-Trusses.pdf
AinieButt1
 
PDF
Embodied AI: Ushering in the Next Era of Intelligent Systems
Yu Huang
 
PDF
Operating System & Kernel Study Guide-1 - converted.pdf
mranoninvisib16
 
PPTX
Internet of Things (IOT) - A guide to understanding
Davies Chacko
 
PPTX
CARTOGRAPHY AND GEOINFORMATION VISUALIZATION chapter1 NPTE (2).pptx
abhi1361yadav
 
PDF
The CXO Playbook 2025 – Future-Ready Strategies for C-Suite Leaders Cerebrai...
Cerebraix Technologies
 
PPTX
Sustainable Sites - Green Building Construction
mhdumerali
 
PPTX
MET 305 2019 SCHEME MODULE 2 COMPLETE.pptx
VinayB68
 
PDF
keyrequirementskkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk
mojeeburahmanwardak
 
PDF
SM_6th-Sem__Cse_Internet-of-Things.pdf IOT
smceramu
 
PPTX
IOT PPTs Week 10 Lecture Material.pptx of NPTEL Smart Cities contd
ssusera400e8
 
PPTX
M Tech Sem 1 Civil Engineering Environmental Sciences.pptx
ShriNRPrasad
 
PPT
CRASH COURSE IN ALTERNATIVE PLUMBING CLASS
rene militante
 
FINAL REVIEW FOR COPD DIANOSIS FOR PULMONARY DISEASE.pptx
rubeshbme
 
OOP with Java - Java Introduction (Basics)
Rashmi T V
 
573137875-Attendance-Management-System-original
AYUSHMANAV
 
Construction Project Organization Group 2.pptx
vj5agdales
 
Model Code of Practice - Construction Work - 21102022 .pdf
AinieButt1
 
Foundation to blockchain - A guide to Blockchain Tech
Davies Chacko
 
Project quality management in manufacturing
BarMusaTetik
 
TFEC-4-2020-Design-Guide-for-Timber-Roof-Trusses.pdf
AinieButt1
 
Embodied AI: Ushering in the Next Era of Intelligent Systems
Yu Huang
 
Operating System & Kernel Study Guide-1 - converted.pdf
mranoninvisib16
 
Internet of Things (IOT) - A guide to understanding
Davies Chacko
 
CARTOGRAPHY AND GEOINFORMATION VISUALIZATION chapter1 NPTE (2).pptx
abhi1361yadav
 
The CXO Playbook 2025 – Future-Ready Strategies for C-Suite Leaders Cerebrai...
Cerebraix Technologies
 
Sustainable Sites - Green Building Construction
mhdumerali
 
MET 305 2019 SCHEME MODULE 2 COMPLETE.pptx
VinayB68
 
keyrequirementskkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkk
mojeeburahmanwardak
 
SM_6th-Sem__Cse_Internet-of-Things.pdf IOT
smceramu
 
IOT PPTs Week 10 Lecture Material.pptx of NPTEL Smart Cities contd
ssusera400e8
 
M Tech Sem 1 Civil Engineering Environmental Sciences.pptx
ShriNRPrasad
 
CRASH COURSE IN ALTERNATIVE PLUMBING CLASS
rene militante
 

Analysis of GSM ORFS Issue

  • 2. Introduction Calibration PA Power Management IC TX Timing Optimization RF Power Amplifier RF Connector
  • 3. VCO Supply Voltage I/Q Signal Power Management IC Common GND
  • 5. By Criterion 5 ACPR(Adjacent Channel Power Ratio) nn For many of the current and future transmission standards(GSM, CDMA, TD-SCDMA, WCDMA, and LTE), ACPR (sometimes also termed adjacent channel leakage ratio-ACLR) is an important test parameter for characterizing the distortion of subsystems and the likelihood that a given system may cause interference with a neighboring radio[32].
  • 6. By Criterion 6 ORFS(Output Radio Frequency Spectrum) nn In GSM, As part of the out-of-channel measurements, the ACPR is defined by two measurements: spectrum due to modulation and wideband noise, and spectrum due to switching. These two measurements are usually grouped together and referred to as output RF spectrum (ORFS)[34].
  • 7. By Criterion 7 ORFS due to Modulation nn The modulation process in a transmitter causes continuous wave carriers to spread spectrally. The spectrum due to modulation and wideband noise measurement is used to ensure that the modulation process does not cause excessive spread. This would cause interference to adjacent channel users[34]. With max PCL, why is the carrier power merely 22 dBm? With max PCL, why is the carrier power merely 22 dBm?
  • 8. By Criterion 8 ORFS due to Modulation nn The analyzer is tuned to a spot frequency and then time-gated across 50% ~ 90% part of the modulated burst to avoid training sequence[34]. nn Besides, spectral components that result from the effect of bursting do NOT appear because the ramps are gated out[34].
  • 9. By Criterion 9 ORFS due to Switching nn GSM/EDGE transmitters ramp RF power rapidly due to burst mode. If RF power is ramped too quickly, undesirable spectral components exist in the transmission. Users operating on different frequencies, especially those close to the channel of interest, will experience significant interference[34].
  • 10. By Criterion 10 ORFS due to Switching nn In this case, no time gating is used, so power from both the ramping and modulation processes affect the measurement[34]. nn Nevertheless, because ORFS due to switching level is larger than modulation level, the effect of ramping dominates the spectrum due to switching measurements[34].
  • 12. By Criterion 12 DPD nn DPD(Digital Predistortion) is applied to linearize the PA to improve the linearity in the linear region and increase the linear region by compensating the compression with an inverse function[39]. nn As shown below, DPD can be used to lower spectral regrowth to improve both ACLR and ORFS as well[1,41,66].
  • 13. By Criterion 13 GMSK and 8PSK nn GMSK is a constant-envelope modulation, but 8PSK is NOT[59]. nn Thus, 8PSK has larger PAPR than GMSK. In other words, 8PSK has more strict linearity than GMSK[61]. Average Power Peak Power Time Power Average Power Peak Power Time Power
  • 14. By Criterion 14 GMSK and 8PSK nn With the identical output power, 8PSK has worse ORFS- SW than GMSK[62]. nn Thus, Pre-distortion calibration helps EDGE 850/900 ORFS-SW margin, not ORFS-MOD[12].
  • 15. By Criterion 15 Path Delay for EDGE nn A source of distortion is a delay that might occur between the amplitude and phase path[42,74].
  • 16. By Criterion 16 Path Delay for EDGE nn The existence of two separate waveform paths could lead to different signal delays on the phase path and the envelope path, affecting the EDGE waveform quality[1,42]. nn As shown below, the larger the delay is, the worse the ORFS due to modulation and EVM will be[42].
  • 17. By Criterion 17 Path Delay for EDGE nn Hence, while calibration, the chipset solution allows for delays to be added to either path to make sure that the phase and envelope paths remain in sync to improve ORFS due to modulation[1,9].
  • 18. By Criterion 18 Path Delay for EDGE nn Path Delay Cal Tree is as below[9]:
  • 19. By Criterion 19 Path Delay for EDGE nn The best-case delay is determined by finding the delay value resulting in the lowest ORFS due to the modulation measurement at +400 kHz and -400 kHz[1,9].
  • 20. By Criterion 20 Path Delay for EDGE nn As shown below, the figure shows a 8PSK modulation spectrum for various values of delay mismatch. With a delay mismatch more than 30ns, the spectrum fails. And the spectrum passes with a delay mismatch less than 30ns[54].
  • 21. By Criterion 21 PA Power Management IC
  • 22. By Criterion 22 QFE2101 nn The main QFE2101 functional blocks are[4]: uuAPT buck converter – essentially a programmable buck converter, or switched-mode power supply (SMPS). uuBypass switch – routes battery or primary phone power (VPH_PWR) directly to the PA(s). uuGSM switches – bypass switches (or optional extra GSM capacitor switch). ll Bypass – a separate bypass switch for GSM with a very low resistance that connects the GSM PA directly to battery. ll GSM capacitor switch – C_GSM pin should be connected to primary phone power (VPH_PWR); utilize this switch as the third bypass switch to further reduce the IR drop.
  • 24. By Criterion 24 QFE2101 nn To switch-in extra load capacitance at the VPA node (C_GSM to capacitor only) during GSM Tx. This option can be used to improve the margin for GSM ORFS/RxBN due to less IR drop, and APT mode instead of bypass mode.[2-4].
  • 25. By Criterion 25 TX Timing Optimization
  • 26. By Criterion 26 Ramping Profile nn Effect of different Ramp timing on Switching Transients[44].
  • 27. By Criterion 27 Ramping Profile nn Nevertheless, there is a trade-off between GSM PvT and GSM ORFS due to switching @400kHz with high PCL[6]. nn Hence, tune ramping profile to pass PVT and ORFS specification simultaneously[6,8].
  • 28. By Criterion 28 GSM RF Timing Adjustment nn In general, there are six types of GSM RF timing adjustment, as shown below[10]:
  • 29. By Criterion 29 GSM RF Timing Adjustment
  • 30. By Criterion 30 GSM RF Timing Adjustment nn pa_en_start/stop_offset_adj:The PA enable signal start/stop time adjustment value can be altered to move the rising edge of the PA enable signal forward or backward to get the optimal configuration[9]. nn pa_start/stop_offset_adj :The PA turn-on/off time adjustment value can be altered to adjust the rising edge of the TDMA Tx power burst[9]. nn ant_timing_start/stop_offset_adj : The antenna select control signal start/stop time adjustment value can be altered to move the falling edge of antenna select control signal forward or backward relative to the PA enable signal falling edge to get the optimal configuration[9].
  • 31. By Criterion 31 PA ON Timing[9]
  • 32. By Criterion 32 PA OFF Timing[9]
  • 33. By Criterion 33 PA Ramp Up Timing[9]
  • 34. By Criterion 34 PA Ramp Down Timing[9]
  • 35. By Criterion 35 Antenna ON Timing[9]
  • 36. By Criterion 36 Antenna OFF Timing[9]
  • 37. By Criterion 37 Potential Issues of Improper Tx Timing nn In addition to PvT and ORFS, Improper Tx timing value may cause improper power in certain hardware boards[9]. nn For MIPI PA device, PA ON signal is expected earlier than the PA range signal, since PA ON script sets PA to the lowest gain state and overrides the settings in the PA range script[9].
  • 38. By Criterion 38 Potential Issues of Improper Tx Timing nn If PA ON is later than PA range timing, the power will be lower than expected; e.g. Max Tx power is merely ~20 dBm[9] nn The error message can be observed in QXDM log[9].
  • 39. By Criterion 39 Potential Issues of Improper Tx Timing nn And the improper power may cause EDGE calibration or registration failure in certain hardware boards[9] nn Failed to calibrate GSM850 due to improper PA timing applied[9]. nn Failed to calibrate GSM900 per improper PA timing applied[9].
  • 40. By Criterion 40 Potential Issues of Improper Tx Timing nn In the case, PCS1900, there is ORFS issue at low temperature, and which is independent of PCL. nn Nevertheless, the issue was NOT observed in DCS1800.
  • 41. By Criterion 41 Potential Issues of Improper Tx Timing nn First, we did NOT observe this issue in FTM(Factory Test Mode). In other words, the issue disappears without RX operation. nn Second, we did NOT observe this issue at transceiver output, as shown below :
  • 42. By Criterion 42 Potential Issues of Improper Tx Timing nn Because there are merely two blocks(transceiver and PAMiD) in TX chain, so we conclude that this issue is related to PAMiD[74], especially ASM timing.
  • 43. By Criterion 43 Potential Issues of Improper Tx Timing nn There are two statuses for TX mode while RX is operating : uuDisable Status(Default) : uuIsolation Status: RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time ASM RX TX PA ASM RX TX PA RX OFF RX ON ASM_TX ON Level Time RX OFF RX ON ASM_TX ON Level Time
  • 44. By Criterion 44 Potential Issues of Improper Tx Timing nn For disable status, when switching to Tx state under low temperature, it may take up to 100 µS for the switch to be fully operational and RF compliant. That’s to say, there will be delay in TX timing while switching from TX OFF to TX ON[28]. nn As mentioned earlier, the antenna select control signal start/stop time affects ORFS performance. That’s why ORFS fails. RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time RX OFF RX ON ASM_TX OFF ASM_TX ON Level Time
  • 45. By Criterion 45 Potential Issues of Improper Tx Timing nn For isolation status, TX is still ON while RX is ON, but TX port will shunt to ground while RX is ON. That’s why it’s called “isolation” status. nn Because TX is always ON, there will be NO delay in TX timing while switching from TX OFF to TX ON. Hence the issue passes after introducing Isolation Status. RX OFF RX ON ASM_TX ON Level Time RX OFF RX ON ASM_TX ON Level Time ASM RX TX PA ASM RX TX PA
  • 46. By Criterion 46 Potential Issues of Improper Tx Timing nn Besides, DCS1800 RX path is different from PCS1900 RX path. Chances are that PCS1900 takes longer time switching from RX port to TX port than DCS1800, so we did NOT observe this issue in DCS1800. RX OFF RX ON Level Time ASM_TX OFF ASM_TX ON ASM_TX OFF ASM_TX ON PCS DCS RX OFF RX ON Level Time ASM_TX OFF ASM_TX ON ASM_TX OFF ASM_TX ON PCS DCS
  • 47. By Criterion 47 RF Power Amplifier
  • 48. By Criterion 48 Linearity nn As mentioned above, ORFS is just the concept of ACLR. And the formula of ACLR is : ACLR = 2(Pin - OIP3 + G)+C, where C is correction coefficient nn Hence, the better the linearity is, the better the ORFS due to switching will be[45,46].
  • 49. By Criterion 49 Temperature nn In general, the higher the temperature is, the worse the OIP3 will be[47]. nn Thus, chances are that high temperature leads to worse ORFS due to switching[15].
  • 50. By Criterion 50 Layout nn Thus, the GND vias should be as numerous as possible to spread the heat[48,63].
  • 51. By Criterion 51 Temperature nn As shown below, the lower the temperature is, the higher the gain will be[47]. nn Thus, with constant Pin, lower temperature leads to higher Pout[49].
  • 52. By Criterion 52 Temperature nn As shown below, the larger Pout is, the worse the ACLR will be. nn That’s to say, chances are that low temperature leads to low ORFS due to switching margin[43].
  • 53. By Criterion 53 Temperature Compensation nn Hence, we have to do temperature compensation to make Pout identical under all temperatures. nn By doing this, the Pout under low temperature will not be too high, thereby improving ORFS due to switching under low temperature.
  • 54. By Criterion 54 Temperature Compensation nn Nevertheless, with temperature compensation, the ORFS due to switching under high temperature may aggravate more.
  • 55. By Criterion 55 Temperature Compensation nn As mentioned earlier, high temperature aggravates linearity, thereby degrading ORFS due to switching. nn With TC(Temperature Compensation), the Pout under high temperature becomes larger. With worse linearity and larger Pout, that’s why ORFS due to switching under high temperature may aggravate more with TC.
  • 56. By Criterion 56 Ramping Profile Adjustment nn Hence, with TC, adjust the ramping profile under high temperature to improve ORFS due to switching because Power versus time (PVT) variation is one of the most important features for describing electrothermal performance of RF power amplifiers (PAs)[51,52].
  • 57. By Criterion 57 PA Output Matching Network Adjustment nn Besides, adjust PA output matching network to improve linearity and lower noise floor, thereby improving ORFS due to switching and RxBN under high temperature with TC[6].
  • 58. By Criterion 58 PA Output Matching Network Adjustment nn As mentioned earlier, 8PSK has more strict linearity than GMSK. In other words, if ORFS issue is related to PA linearity, the symptom should be:
  • 59. By Criterion 59 PA Input Matching Adjustment nn As shown below, the PA input matching networks is DA’s(Driver Amplifier) load-pull as well. nn In other words, good ORFS@PA input leads to good ORFS@PA output[50,53]. nn Hence, to optimize PA input matching network can be used to improve ORFS as well.
  • 60. By Criterion 60 PA Post loss nn According to the following formula : with constant target power, larger PA post loss leads to larger PA output power, thereby aggravating ORFS-SW @ +- 400 kHz, even though PA’s output impedance is close to 50Ω[11]. Target Power = PA Output Power – PA Post loss nn Thus, minimize the PA post loss to have more headroom in Tx power. Because LB has larger max power than HB, headroom equal to 0.7 dB from 32.5 dBm(i.e. 33.2 dBm) is preferred[11].
  • 61. By Criterion 61 Stability nn In the case, LB EDGE ORFS test fails in both modulation/switching spectrum when PCL is set < 10(High Power Mode), and passes when PCL is set ≥ 11(Low Power Mode)[11].
  • 62. By Criterion 62 Stability nn Firstly, the best ORFS due to modulation result was found to be only approximately -43 dB, even after tuning the path delay, but it can reach -60 dB[11]. ORFS-MO Before After -400KHz -43.1158 dB < -60 dB +400KHz -43.0755 dB < -60 dB ORFS-MO Before After -400KHz -43.1158 dB < -60 dB +400KHz -43.0755 dB < -60 dB nn Besides, many spurs are found(shown as yellow traces in the graph), according to the analysis, the PA is oscillating[11].
  • 63. By Criterion 63 Stability nn According to the K-factor definition, the stability depends on S-parameter[56]. Chances are that the components marked as red circle affect the S-parameter and stability of QFE2320, thereby oscillating[11] nn Because the issue occurs in LB, the components marked as red circle must be changed to remove oscillation[11].
  • 64. By Criterion 64 Stability nn Therefore, after tuning the path delay and changing the BOM, LB EDGE ORFS margin is sufficient for all PCL[11].
  • 65. By Criterion 65 Switch Point nn For this case, EDGE LB ORFS fails at PCL8, PCL12, and PCL16. nn As mentioned above, If failure is at maximum power and improves as power is reduced, the issue is likely due to linearity. Thus, it fails at PCL8 but passes at PCL9 ~ PCL11[10,16]. PCL 8 9 10 11 Result Fail Pass Pass Pass PCL 12 13 14 15 Result Fail Pass Pass Pass PCL 16 17 18 19 Result Fail Pass Pass Pass
  • 66. By Criterion 66 Switch Point nn In general, there is hysteresis from the switching point where the PA rises from a lower gain to a higher gain, and vice versa to avoid the PA oscillating at the switching point[64]. nn In other words, as illustrated below, for EDGE single- slot, 19 dBm(PCL12) can be achieved by high gain mode or mid gain mode; 11 dBm(PCL16) can be achieved by mid gain mode or low gain mode.
  • 67. By Criterion 67 Switch Point nn Hence, as illustrated below, if 11 dBm is achieved by low gain mode, the PA will be in saturation region, thereby degrading linearity and aggravating ORFS. So is 19 dBm(PCL 12). nn In order to solve this issue, make these output power @ switch points be in proper gain mode to be in linear region[16].
  • 68. By Criterion 68 RF Connector
  • 69. By Criterion 69 Material issue nn If the connector is abnormal, the antenna matching network may alter PA’s output impedance, thereby aggravating ORFS. nn Besides, abnormal connector leads to leakage. And as shown below, with constant target power(e.g. LB 32.5 dBm or HB 29.5 dBm), the larger the leakage is, the larger the PA output power will be, thereby aggravating ORFS.
  • 70. By Criterion Layout issue nn In the case, it is very difficult to pull load impedance to 50 Ω for the common path[11]. nn Hence, ORFS due to switching performance is not good enough.
  • 71. By Criterion Layout issue nn Because the signal pad size of the RF connector is very large(0.6 mm × 0.7 mm). According to simulation, if using L9 as RF GND, it introduces serious mismatching. nn Thus, areas under the pads of the RF connector in L9, L8, and L7 should be cut out, and set L6 as the reference GND, to get better impedance[11]. Signal Pad GND Pad L9 L8 L7 L6 Impedance 14 Ω 32 Ω 44 Ω 53 Ω
  • 73. By Criterion 73 VCO Pulling nn In the case, GSM ORFS (modulation and switching) is observed to be failing by about 5−6 dB on quad-band GSM some channels at high power level (LB: PCL5; HB: PCL0)[30]. nn The max power at connector can be 34 dBm for LB, and 31 dBm for HB. Besides, the PA’s load-pull is close to 50Ω for both LB and HB. In other words, the issue is NOT related to PA linearity.
  • 74. By Criterion 74 VCO Pulling nn As shown below, the issue can be solved with SPDT.
  • 75. By Criterion 75 VCO Pulling nn Without SPDT, the issue is root caused too high Tx power during GSM operation leaking into FBRx path through front-end coupler and later coupling into TX quadrature up-converter of transceiver[16,17,30,72].
  • 76. By Criterion 76 VCO Pulling nn An SPDT (with bypass option) is recommended at front end, as shown here to isolate GSM Tx power from leaking into FBRx path[17,30]. nn Since FBRx is not operational in GSM mode. FBRX only works in 3G/4G mode. So this switch (SPDT) does not impact GSM operation[16,30]. nn Two possible implementations during GSM operation are shown here[30]:
  • 77. By Criterion 77 VCO Pulling nn In the case, as shown below, ORFS-MO passes if we press the shielding can, and it fails without press.
  • 78. By Criterion 78 VCO Pulling nn As shown below, PA will couple TX signal onto the shielding can. In other words, there will be residual TX signal on the shielding can. nn With good grounding, the residual TX signal will flow to GND completely. Without good grounding, the shielding can will reflect the residual TX signal.
  • 79. By Criterion 79 VCO Pulling nn Thus, with press, the residual TX signal flows to GND completely due to good grounding. nn If possible, don’t put PA and transceiver in the same shielding area to avoid VCO pulling.
  • 80. By Criterion 80 VCO Pulling nn As mentioned earlier, the PA input matching networks is DA’s(Driver Amplifier) load-pull as well. nn Besides, for Direct-Conversion transmitter architecture, the RF frequency is the same as LO. So if PA input matching is not close to 50Ω, the reflection due to poor S11 may interfere VCO, thereby aggravating ORFS-MO.
  • 81. By Criterion 81 VCO Pulling nn In the case, the relationship between ORFS-MO and ASM is as below : Antenna Switch Module Vendor A (With Original RX Matching) Vendor B (With Original RX Matching) Vendor B (With New RX Matching) Pass Fail Pass
  • 82. By Criterion 82 VCO Pulling nn If we set TX port as Port 1, and RX port as Port 2, S21 is just the isolation between TX port and RX port, as shown below : nn Hence, the isolation depends on TX matching, RX matching, and ASM.
  • 83. By Criterion 83 VCO Pulling nn Therefore, vendor B ASM degrades isolation, and ORFS-MO fails. nn With vendor B ASM, we introduce new RX matching, and isolation improves, thereby improving ORFS-MO. nn As shown below, the isolation affects VCO pulling through RX path.
  • 84. By Criterion 84 Keep Out Area nn Besides, keep-out areas on PCB layer 1 are required to avoid sensitive on-chip components such as VCO[73].
  • 85. By Criterion 85 Keep Out Area nn Otherwise, the parasitic effect aggravates phase noise, thereby degrading ORFS-MO[10]. nn In general, check the LO phase noise at 400 kHz offset, it should be better than -118 dBc/Hz[10].
  • 87. By Criterion 87 IR Drop nn IR drop refers to a voltage drop that appears at the resistive component of any impedance. This voltage drop across any resistance is the product of current (I) passing through resistance and resistance value (R). nn If there is IR drop issue in the supply voltage of any TX chain blocks, such as transceiver, PA, and ASM, the TX performance including ORFS will aggravate[24].
  • 88. By Criterion 88 IR Drop nn In the case, LB GSM ORFS-MO fails for high power mode In low temperature. nn We found the ORFS-MO issue appears not only in connector, but also in transceiver output. As mentioned above, poor ORFS performance in PA input leads to worse ORFS performance in PA output.
  • 89. By Criterion 89 IR Drop nn Thus, there are three possible factors related transceiver: Layout, supply voltage, and material issue. nn After changing the BOM for transceiver supply voltage, the issue is solved.
  • 90. By Criterion 90 IR Drop nn As mentioned above, low temperature makes gain increase. Higher gain leads to higher Icc, thereby increasing IR drop. nn In addition, with max PCL, low band has larger output power than high band. So the issue occurred in LB instead of HB[57]. LB 14 dBm HB 11 dBm
  • 91. By Criterion 91 IR Drop nn As shown below, power is getting lower as supply voltage drops[11]. In other words, the lower the supply voltage is, the lower power will be.
  • 92. By Criterion 92 IR Drop nn In other words, with constant target power(e.g. LB 32.5 dBm or HB 29.5 dBm), the larger the IR drop is, the larger PA output power and current consumption(i.e. Icc) will be, thereby aggravating ORFS-SW performance. IR drop PA Output Power & Icc IR drop PA Output Power & Icc
  • 93. By Criterion 93 IR Drop nn Besides, as shown below, the larger the temperature is, the larger the resistance including chip resistor or trace will be, thereby increasing IR drop.
  • 94. By Criterion 94 IR Drop nn As mentioned above, with TC(Temperature Compensation), the Pout under high temperature becomes larger. With worse linearity and larger Pout, that’s why ORFS due to switching under high temperature may aggravate more with TC. nn Thus, with TC and IR drop, ORFS-SW aggravates more in high temperature.
  • 95. By Criterion 95 Decoupling Capacitor nn Due to burst mode characteristic, GSM PA is being ON and OFF all the time. As shown below, while PA is ON, there will be unwanted 20KHz signal riding on DC, which is just ripple. Any supply disturbances caused by on/off activities could directly convert to ORFS issue[10].
  • 96. By Criterion 96 Decoupling Capacitor nn As shown below, larger decoupling capacitor can help improve spectral regrowth, thereby improving ORFS[31].
  • 97. By Criterion 97 Decoupling Capacitor nn Besides, the decoupling capacitor should be as close to PA as possible. Otherwise, PA supply transient current may leak into other ICs, and transient current from other ICs may leak into GSM PA, thereby aggravating ORFS performance[10].
  • 98. By Criterion 98 Decoupling Capacitor nn Besides, the supply voltage of LO is important as well. nn As shown below, change C3115 from 220nF to 4.7uF, ORFS-MO at 1.6 MHz offset improves.
  • 99. By Criterion 99 Decoupling Capacitor nn The lower ESL is, the wider capacitive range will be. nn The lower ESR is, the lower ripple will be
  • 100. By Criterion 100 Decoupling Capacitor nn The path the noise flows inside 3-terminal capacitor is shorter than which inside 2-terminal capacitor. That’s to say, the 3-terminal capacitor has lower ESL and ESR.
  • 101. By Criterion 101 Decoupling Capacitor nn Compared to multiple capacitors, the 3-terminal capacitor has better noise suppression, and it doesn’t have anti-resonance issue. Besides, the 3-terminal capacitor can save PCB area due to less components[75].
  • 102. By Criterion 102 Star Routing nn Use star routing from battery to PMIC and PA. Besides, the branch point should be as close to battery as possible[10]. nn As shown below, the length of coupling path 2 is shorter than coupling path 1, so the noise from GSM PA may leak to PMIC. That’s why branch point should be close to battery.
  • 103. By Criterion 103 Star Routing nn In the case, at high power level of GSM low bands, ORFS modulation performance is marginal, especially at the 400 kHz offset. nn The issue is dependent on PCL. It disappears at low power level such as PCL19. Chances are that the issue is related to PA linearity, so we checked transceiver output performance to eliminate the influence of PA.
  • 104. By Criterion 104 Star Routing nn The issue already appears at transceiver output, so there are three possibilities : material, layout, and voltage supply. nn With an external voltage supply, the issue improved.
  • 105. By Criterion 105 Star Routing nn Due to burst mode characteristic, GSM PA has strong transient current. That’s why the issue disappears at low power level because low power level leads to low transient current. And that’s why the issue appears at high power level of GSM low bands because PCL5 of LB (33 dBm) is larger than PCL0 of HB(30 dBm). nn As shown below, the coupling path : GSM PA => PMIC => transceiver
  • 106. By Criterion 106 Star Routing nn In the case, as shown below, ORFS modulation performance fails when monitor is ON, but performance is good when monitor is OFF. nn Besides, the issue appears at GSM quad-band, especially at the 400 kHz offset. And the issue is independent on power level.
  • 107. By Criterion 107 Star Routing nn This is because backlight driver IC has strong transient current as well. That’s why the issue disappears when monitor is OFF. nn As shown below, Coupling path 1 : Backlight driver IC => PA Coupling path 2 : Backlight driver IC => PMIC => Transceiver
  • 108. By Criterion 108 Star Routing nn Thus, increase the decoupling capacitor and power inductor value for Backlight driver IC, and the performance improved.
  • 109. By Criterion 109 Star Routing nn Similarly, use star routing from PMIC to VBATT pin and Vcc_GSM pin of GSM PA separately, and the branch point should be as close to PMIC as possible. nn At least, the branch point should NOT be close to GSM PA, as shown below :
  • 110. By Criterion 110 Star Routing nn If the branch point is close to GSM PA, make the VPH_PWR common part(marked as pink) and separate part(marked as green) on different layer. As shown below, the coupling path(marked as yellow) has high impedance to avoid noise coupling between the 2 pins.
  • 111. By Criterion 111 High Frequency Noise nn In the case, as shown below, ORFS-MO passes if we press the PA shielding can, and it fails without press. Besides, after removing shielding can, it passes as well. nn PA and transceiver are in separate shielding area, so the issue is NOT related to VCO pulling.
  • 112. By Criterion 112 High Frequency Noise nn As mentioned earlier, without good grounding, the shielding can will reflect the residual TX signal. With press, the residual TX signal flows to GND completely due to good grounding. nn In addition, no shielding can, no residual TX signal. That’s why the issue disappears after removing shielding can.
  • 113. By Criterion 113 High Frequency Noise nn Thus, the issue is due to that residual RF TX signal reflects to PA Vcc by shielding can. In terms of DC voltage, RF signal is regarded as noise. nn There’re usually large capacitor for decoupling and small capacitor for filtering high frequency noise. For the issue, we can modify small capacitor value to filter residual RF TX signal.
  • 115. By Criterion 115 XO Harmonics nn In the case, on certain channels of GSM high bands, ORFS switching and modulation performance is marginal, especially at the 400 kHz offset. nn The most affected channels are DCS channel 599, 600, 602, 791, 792, 794, and 795; and PCS channel 668.
  • 116. By Criterion 116 XO Harmonics nn As shown in the table below, these channels are all 90 times 19.2 MHz approximately in terms of frequency. In other words, these channel are related to XO harmonics. nn Baseband chip has an internal issue related to XO harmonics coupling into TX I/Q lines that cause degradation of ORFS performance only in GSM high bands[13,14]. Band Channel Frequency 19.2 MHz * N 599 1727.6 89.97916667 600 1727.8 89.98958333 602 1728.2 90.01041667 791 1766 91.97916667 792 1766.2 91.98958333 794 1766.6 92.01041667 795 1766.8 92.02083333 PCS 668 1741.4 90.69791667 DCS Band Channel Frequency 19.2 MHz * N 599 1727.6 89.97916667 600 1727.8 89.98958333 602 1728.2 90.01041667 791 1766 91.97916667 792 1766.2 91.98958333 794 1766.6 92.01041667 795 1766.8 92.02083333 PCS 668 1741.4 90.69791667 DCS
  • 117. By Criterion 117 XO Harmonics nn As shown above, the solution is to add series L and shunt C onto each of the TX_I/Q lines[63]. nn As the spur level is reduced, and therefore the ORFS modulation is improved[63]. nn This LC filter can also help reduce the number of exceptions for the GSM Rx band noise[63]. w/o solution w/ solution
  • 118. By Criterion 118 XO Harmonics nn The trace between transceiver and PMIC is rich in XO harmonics because it is XO reference clock. nn Thus, we need to put a R-C filter to reject XO harmonics and a DC block to reject DC offset. nn Besides, we need to keep the trace away from I/Q signal to avoid ORFS issue.
  • 119. By Criterion 119 XO Harmonics nn The digital XO signal and the analog XO signal can corrupt each other; the layout should provide isolation between them. Isolation is highly recommended [77]. nn As shown below, they are too close.
  • 120. By Criterion 120 I/Q Imbalance nn I/Q imbalance can be characterized by: a phase mismatch that can happen between the I and the Q components of the local oscillator signals, which becomes not exactly 90 degrees, and an amplitude mismatch due to the gain difference of the mixers of the I and Q branches[70].
  • 121. By Criterion 121 I/Q Imbalance nn As shown below, I/Q imbalance aggravates image, thereby aggravating sideband suppression and ORFS- MO. nn Thus, we need to adjust phase and amplitude offsets between I and Q channel by means of calibration to lower image to improve ORFS-MO[69].
  • 122. By Criterion 122 I/Q Imbalance nn Besides, in terms of modulator, high temperature aggravates sideband suppression as well, thereby aggravating ORFS-MO. So good thermal design of transceiver is important[69]. nn This is especially important for direct up-conversion transmitter because I/Q imbalance are the inherent shortcomings of it[69].
  • 123. By Criterion 123 I/Q Imbalance nn Ideally, the four traces on the IQ signal path from the DAC output to the modulator input should be symmetrical between the I channel and Q channel and between the positive side and negative side within a channel[69]. nn In reality, due to PCB layout, trace lengths are not perfectly matched. The mismatches cause the signal in one channel to be skewed from the other, and, therefore, result in IQ imbalance(both gain and phase)[69].
  • 124. By Criterion 124 LO Leakage(Carrier Leakage) nn Excessive DC offsets in I/Q channels cause high levels of LO leakage, thereby aggravating ORFS-MO[69]. nn According to the formula shown below, the more the DC offsets in I/Q channels, the larger the LO leakage will be[69]. = + +
  • 125. By Criterion 125 LO Leakage(Carrier Leakage) nn Thus, we need to remove DC Offsets by means of calibration to lower LO leakage to improve ORFS- MO[69]. nn In terms of modulator, high temperature aggravates carrier suppression as well, thereby aggravating ORFS- MO. So good thermal design of transceiver is important[69].
  • 126. By Criterion 126 Power Management IC
  • 127. By Criterion 127 Charging nn In the case, when charging is enabled, GSM850, ch190, ORFS-MO failure (0.5 dB) is observed at 600 kHz offset measurement[18].
  • 128. By Criterion 128 Charging nn For a 5 V DCP(Dedicated Charging Port) plug-in, the charging frequency is set to 600 kHz. nn Spur coupling path: PMIC buck switching (600 kHz) spurs leaking into the Vbatt and appears at the Tx-RF output through QET4101 and PA[18]. nn Hence, there will be +- 600 KHz offset ORFS-MO issue due to 2nd order intermodulation mechanism.
  • 129. By Criterion 129 Charging nn Adding a Pi (π) filter on the VPH_PWR rail attenuates the charger spur by 20 dB and passes the GSM ORFS specification[18]. nn As a workaround, the placeholder for the π filter should be provided at the QET4101 input to attenuate spur coupling from the PMIC charger and also a 0 Ω (0201 package) placeholder at power supply line of PA (VCC1_PA) for noise filtering[18].
  • 130. By Criterion 130 Charging nn As shown below[18]:
  • 131. By Criterion 131 SMPS nn In the case, GSM ORFS modulation spectrum periodicity jump in all bands and all power levels, especially the mid-power level[28]. nn IC NCP6335 is external SMPS(Switching Mode Power Supply) to the MSM8X25Q core. The NCP6335 is the main interference source to ORFS[28,75-76].
  • 132. By Criterion 132 SMPS nn NCP6335 has PFM and PWM operation for optimum increased efficiency, The transition between PWM/PFM modes can occur Automatically, that’s why ORFS-MO spectrum periodicity jump[28]. nn Set NCP6335 to forced PWM, and ORFS-MO passed.
  • 134. By Criterion 134 Common GND nn In the case, with all decoupling capacitor and bypass capacitors, ORFS fails; without all decoupling capacitor and bypass capacitor, ORFS passes.
  • 135. By Criterion 135 Common GND nn This is because transient current and high frequency noise may leak into PA Vcc through these capacitors by means of common GND on top layer.
  • 136. By Criterion 136 Common GND nn Thus, make these shunt capacitors GND pad separate from common GND on top layer(i.e. GND Island), and add GND via as many as possible. nn If GND island is bound by PCB area, at least, the GND via should be as many as possible.
  • 137. By Criterion Noise 137 Common GND nn As mentioned above, we’re able to use Pi (π) filter to attenuate the noise. Nevertheless, the GND of capacitors should be separate. Otherwise, the noise may bypass the capacitors through common ground, thereby aggravating filtering effect.