- Email: [email protected]
Q mismatch calibration of a transmitter using local quadrature oscillator
Microelectronics Journal 55 (2016) 82–91
Contents lists available at ScienceDirect
Microelectronics Journal journal homepage: www.elsevier.com/locate/mejo
I/Q mismatch calibration of a transmitter using local quadrature oscillator Gholamreza Zare Fatin a,n, Mostafa Savadi Osgooei b, Ali Fotowat-Ahmady c a
Department of Electrical Engineering, University of Mohaghegh Ardabili, Ardabil, Iran Infineon Technologies AG, Linz, Austria c Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran b
art ic l e i nf o
a b s t r a c t
Article history: Received 27 January 2015 Received in revised form 21 February 2016 Accepted 5 July 2016
In this paper a calibration technique for I/Q mismatch of a transmitter is introduced. The calibration technique is based on the fact that all mismatches in I/Q paths can be modeled as the mismatch of the local oscillator quadrature outputs. Based on this fact, a simple tuning scheme for quadrature output of the oscillator is used to calibrate the mismatch of the I/Q transmitter. In addition to quadrature oscillator, gain mismatch of the I/Q paths and LO feedthorugh and leakage is calibrated by using a tunable linear Gm cell in the base band part. In order to demonstrate the proposed technique, a transmitter with 1.8 GHz carrier frequency is designed with auxiliary blocks to extract and cancel out the I/Q mismatches. The simulation results in 0.18 mm RFCMOS process show that the proposed technique can reduce amplitude of the image signal resulting from mismatches in I/Q paths about 18.5 dB in the transmitter output. & 2016 Elsevier Ltd. All rights reserved.
Keywords: Transmitter I/Q calibration Quadrature oscillator Mismatch detector Synchronous detector
1. Introduction THE I/Q modulation scheme is ubiquitous in most of today wireless transmitters due to spectral efficiency of this modulation. The I/Q paths in a transmitter are subject to phase and gain mismatches which are the result of imbalance between different blocks and interconnections in I/Q paths. This imbalance can be from any block in transmitter chain and usually is rooted to layout and components mismatch in the circuit [1–3]. This mismatch in I/ Q paths will generate an image signal in the spectrum of the output signal which will degrade the error vector magnitude (EVM) of output constellation. A few methods have been introduced in literatures for mismatch calibration of I/Q transmitters. The loop back detector with a recursive algorithms [1] and 2D search algorithms [4,5] are two popular methods of I/Q calibration. The first method can take long calibration time and will consume large power and area concerning the IC implementation. This paper is utilizing the second method of calibration by using iterative search algorithm to calibrate the transmitter. In this paper, with a linear transconductance in the n
Corresponding author. E-mail addresses: [email protected] (G. Zare Fatin), mostafa.savadiosgooei@infineon.com (M. Savadi Osgooei), [email protected] (A. Fotowat-Ahmady). http://dx.doi.org/10.1016/j.mejo.2016.07.004 0026-2692/& 2016 Elsevier Ltd. All rights reserved.
transmit mixer similar to the one used in [5] and also using a phase tunable oscillator with the quadrature outputs, a low power and small area calibration circuit has been devised. This paper contributions can be categorized in three distinct sections: 1) This paper demonstrates the feasibility of transmitter I/Q calibration by using quadrature oscillator. The simulation results show that considerable rejection in the output image signal is achieved by phase tuning of the quadrature oscillator. 2) Another important contribution of the paper is introduction of a technique for phase tuning of the quadrature outputs of the oscillator with minimum impact on the phase noise. 3) Third contribution of the paper includes the design and validation of different auxiliary blocks for I/Q mismatch detection and cancellation including the transmitter itself and elaborating the calibration process. Meanwhile some circuit design techniques in different blocks of the mismatch detection path was introduced. This paper is organized as follows, Section 2 of the paper explains the overall block diagram of the transmitter including the calibration blocks and procedure. Section 3 presents the quadrature oscillator used as the key calibration element. Section 4 explains the detector block used in the mismatch detection and calibration process. Section 5 of the paper presents the simulation results of the quadrature oscillator and calibration process and demonstrates the operation of the proposed method; finally some conclusions are drawn in Section 6.
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
83
Fig. 1. Block diagram of the transmitter with calibration paths.
2. Calibration process
Fig. 2. Transmitter output spectrum with images signal resulted from mismatches in I/Q paths and LO feedthrough.
Fig. 1 shows the transmitter block diagram along with calibration loops. The direct-conversion architecture using only one I/ Q transmit mixer is opted for the transmitter. Transmitter circuit includes two calibration loops which take the signal back from the output of the up conversion mixer through a block entitled as detector. The detector block is used to down convert the output signal for further filtering and extraction of the necessary tone; this block will be explained further in continue of the paper. The signal extracted from output of the detector is used as the residue of the image signal and criteria for effect of the calibration process on the I/Q mismatches. In addition to I/Q calibration, the output of this detector block can be used for LO feedthrough (LOFT) reduction which is shown in Fig. 1 as a red calibration signal and
Fig. 3. The circuit schematic of the linear transconductance used in the transmitter.
84
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 4. The current sources are added to linear Gm cell for the calibration purpose.
Table 1 The adjustment of oscillator output phases and the resulted amplitude difference. ΔVctr(mV)
Δϕ(°)
ΔA(mV)
0 10 50 100 200
0.7 m 0.3 1.47 2.75 4
0.040 0.393 2.4 5.8 8.2
Fig. 5. The quadrature oscillator used in the transmitter along with output I/Q phase tuning technique.
Fig. 7. The block diagram of the detector circuit.
Fig. 6. The simulated quadrature output signals of the oscillator.
performs the calibration through the linear transconductor (Gm) of the transmitter. The I/Q calibration loop is closed either by Gm tuning or quadrature oscillator output phase tuning, which are utilized depending on the calibration purpose, whether it is a gain imbalance tuning or phase mismatch tuning. Regardless of the nature of the mismatch (gain or phase imbalance) between I/Q paths, they will generate an image signal in the output spectrum which indeed will degrade the EVM of the transmitter. This effect is shown in Fig. 2 which shows the image signal resulted from I/Q mismatches along with LO feedthrough and main output signal for a transmitter with a carrier frequency at 1.8 GHz. The calibration algorithm as mentioned before is based on 2D recursive search method which is similar to the method explained in [4]. The Gm cell as shown in block diagram of Fig. 1, is very important block in this calibration process. The circuit schematic of this Gm cell is shown in Fig. 3 which was introduced in [5] as a linear transmit transconductor. The linearity feature of this transconductance has been thoroughly explained in a literature [6]
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
85
Fig. 8. The circuit schematic of the synchronous detector.
Fig. 9. The synchronous detector simulation result (a) input signal which is taken from transmitter output, (b) down converted output signal of the synchronous detector.
published before. The transconductance of this block in ideal case 1 R9 is equal to Gm = R1 R7 which is determined based on resistors proportion. The current sources RF_LOFT and BB_LOFT were added to the circuit in order to facilitate the calibration process.
The function of the different elements in this linear Gm cell during the calibration process is shown in Fig. 4. The transistor M9 is working as a linear resistor and shorts two differential outputs of the transconductor and adjusts the gain of the Gm cell which is
86
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 10. The bandpass filter used in the detector circuit which is based on a biquad cell.
Fig. 11. The circuit schematic of the rectifier circuit used in the logarithmic amplifier.
Fig. 12. The simulated phase noise of the quadrature oscillator used in the transmit mixer.
used for correction of the gain imbalance in the I/Q paths. Transistors M10-M17 are used for gain adjustment of the baseband signal which can be controlled by 3-bits digital word. This digital word controls the baseband gain of the transmitter. The DC offset in the baseband signal can create the LO-feedthrough which in this circuit is cancelled by using the BB_LOFT current sources. On the other hand the LO-feedthrough due to LO to RF leakage is canceled
Fig. 13. Phase noise of the quadrature oscillator with and without calibration signal at the tail current.
by using RF_LOFT currents sources. These current sources are connected to drain of the M7 and M8 transistors which provide the tail current of the mixer. The mixer cell is a tuned active double balanced Gilbert cell.
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 14. Phase noise of the oscillator in different corners of the process.
87
base band part while the gain is set to minimum value. Then, the detector circuit as shown in Fig. 1, measures the LO-feedthrough due to leakage. After measurement of LO-feedthrough, the RF_LOFT current sources are used to cancel out the unwanted tone in the output. In the next step of the calibration process, baseband gain is set to maximum value. In this state, the detector will measure the LOfeedthrough due to baseband offset voltage. Then BB_LOFT current sources are utilized to cancel this effect. In continue of the calibration process and for I/Q imbalance cancellation, a 2 MHz baseband signal is injected into the I/Q paths; then detector measures the image signal power which is the result of the I/Q mismatches. This extracted image signal power is reduced by adjustment of the quadrature oscillator output phases which will be explained in the next section of the paper. The tuning of the quadrature oscillator outputs phase can adjust the phase mismatch of the I/Q paths. Although, phase tuning changes the amplitude of the quadrature output signals in addition to their phase, but this effect is negligible and can positively contribute to gain imbalance correction of the transmitter. In this design the amplitude limiting stages were designed at the output of the oscillator to limit and match the amplitude of the quadrature outputs after phase tuning. After phase mismatch tuning and in order to further reduce the amplitude of the image signal, the gain of the I/Q paths is tuned; this is carried out by using linear Gm cell gain adjustment capability which was explained before.
3. The quadrature oscillator
Fig. 15. Phase noise variations due to mismatches in devices size and threshold voltage.
Fig. 16. The simulation result of the logarithmic amplifier.
As mentioned before the calibration process is very similar to the algorithm used in [4]; this process is explained here for further details and also to clarify the algorithm. The calibration process starts with LO-feedthrough cancellation. A signal (a 4 MHz signal in this design) is injected into the
As explained in the previous section and in the I/Q mismatch calibration, the phase imbalance between I/Q paths is corrected through a quadrature oscillator. This quadrature oscillator drives the transmit mixer. The proposed oscillator which is used in the calibration process is shown in Fig. 5. It is a complementary LC oscillator which was designed in quadrature form to drive the I/Q mixers of the transmitter. The complementary LC oscillator is preferred over the NMOS only realization because of better phase noise performance [7]. The complementary oscillator has higher transconductance for the same bias current which results in faster switching and shorter crossing time which is very important in reducing the timing jitter or phase noise of the oscillator. As mentioned before and explained in Section 2, in order to calibrate the phase mismatch of the I/Q paths and remove the image signal, it is necessary to adjust the output phases of the quadrature oscillator. As shown in Fig. 5, the differential pair made by M1 and M2 transistors is added to oscillator to tune the phase of the output signals. By adjusting the output phases of the oscillator, the phase mismatches resulting from other parts of the transmitter can be canceled. This deliberate quadrature phase mismatch in oscillator outputs can negate the I/Q phase imbalances which are originated from other parts of the circuit. The gate-source voltage (Vgs) of the transistors M1 and M2 are changed in contrast to each other by a control voltage Vctr which is applied in differential form. This control voltage in addition to phase change will result in amplitude variation of the output signals as well. But, the impact of oscillator quadrature outputs amplitude difference in mixer performance is less severe than phase difference. Also, in order to remove these amplitude mismatches resulted from phase adjustment of the oscillator, we have designed a limiter at the quadrature outputs of the oscillator which can balance the amplitudes of the output signals. The quadrature output signals of the oscillator are shown in Fig. 6. Table 1 represents the differences in phases and amplitudes
88
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 17. The output spectrum of the transmitter with some intentional mismatches in I/Q paths.
of the outputs of the quadrature oscillator created by adjustment of the control voltage (Vctr).
4. The detector As shown in Fig. 1 and in the calibration process of the transmitter, a detector block is used to extract the power of the targeted component from output spectrum. This is a key block in detecting and cancelling the image signal or oscillator feedthrough. The block diagram of the detector [4] is shown in Fig. 7. It has a synchronous detector at the input which down samples the transmitter output signal. The down converted signal will be then filtered by a bandpass filter to extract the targeted tone from the output spectrum. The output of the bandpass filter is given to a logarithmic amplifier to produce a dc voltage proportional to extracted tone for calibration process.
The circuit schematic of the synchronous detector is shown in Fig. 8. It is based on the design reported in [4]. In this paper we have redesigned the circuit in CMOS process and optimized the circuit for this application. The synchronous detector circuit in Fig. 8 is based on Gilbert cell structure. The input transistors M1 and M2 operate as the input transconductance. In this design, the common-gate structure has been selected for input transconductance with two series resistors at the input, R1 and R2. Using these input resistors in addition to common-gate structure boosts the linearity of the detector which then increases the dynamic range of the detector as well. The switching transistors M5-M8 do the mixing function; as shown in Fig. 8, the input signal in addition to input transconductance also drives the switching transistors. This is like a selfmixing of an RF signal. The voltage division between the input resistors (R1, R2) and input resistance(1/gm) of the M1 and M2
Fig. 18. Transmitter output spectrum (a) before and (b) after the calibration for LO to RF leakage.
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
transistors attenuates the RF signal before reaching the input transconductance. The simulation results of the synchronous detector for an input spectrum which has an RF signal at 1.8 GHz along with an LO feedthorugh and image signal is shown in Fig. 9. The down converted signal has the main signal at dc with LO feedthrough at 4 MHz and image signal at 8 MHz which are shown in Fig. 9(b). This arrangement of the signals has been adopted for LO feedthrough cancellation. The down converted signal at 4 MHz needs further filtering and amplification to be used as the control signal in the calibration process. In order to select the desired tone from down-converted spectrum at the output of the synchronous detector, a bandpass filter is added to detector circuit which was shown in the block diagram of Fig. 7. The integrated filter was designed based on gmC method and is a biquad cell which is shown in Fig. 10. The bandpass output of the biquad cell is used here for filtering purpose which has the transfer function as in (1),
VBPF −Gm1 = Vinput C1 S 2 +
S+
(
1 R1C1
+
1 R 2 C2
1 R 2 C2
)S +
1 +Gm2 Gm3 R1R 2 R1R 2 C1C2
(1)
The bandpass filter is designed based on the previous work reported in [8]. The circuit schematic of the gm cell utilized in this biquad is also shown in Fig. 10 which is a differential pair with resistive loads. The logarithmic amplifier is the final stage in the detector circuit and is also shown in Fig. 7. The amplifier includes the limiting stages L1-L8 and rectifier circuits which add their output currents in an RC circuit [9,10]. The R resistor is used to convert the output currents of the rectifier blocks to voltage signal; the capacitor (C) is also added to perform the low pass filtering function. The limiting stages (L1-L8) utilized here are differential pairs with 12 dB dynamic range each which add up to total dynamic range of 96 dB for the logarithmic amplifier. The RC time constant is chosen equal to 1 ms.
89
The rectifier circuit is shown in Fig. 11 which is a modified form of the Gilbert cell. The operation of the circuit is based on a selfmixing of the input signal which resembles the rectification task.
5. Simulation results The different blocks of the proposed calibration system which were explained in previous sections are put together to validate their functionality in the calibration of the transmitter. These circuits are designed and simulated in 0.18 mm RFCMOS process. The individual blocks are also simulated to validate their operation and also to be optimized before using in the calibration system. The simulation result for phase noise of the quadrature oscillator is shown in Fig. 12. The phase noise of the oscillator is about 110 dBc at 1 MHz offset frequency. In order to evaluate the effect of the calibration signal in the phase noise of the oscillator, the phase noise was simulated in presence of the calibration signal. The amplitude of the calibration signal was set based on the overall calibration system requirement and signal level that is necessary to suppress the image signal below a certain level in compare to the desired signal. The simulation results of this test is shown in Fig. 13. As shown in this figure increase in the phase noise value in compare to the state without the calibration signal is negligible. The phase noise of the oscillator has been simulated in different corners of the process and the results are shown in Fig. 14. The largest shift in phase noise value comes from FS corner and is about 4 dB higher in compare to TT state. The phase imbalance between quadrature outputs of the oscillator has also been simulated in different corners. The maximum change in quadrature phase is happening in FS corner, but the value is less than 0.5 degree. The phase noise and phase imbalance have been simulated with a 720% mismatch in devices size and threshold voltage; this accounts for total 25 runs with different mismatch specification. The output
Fig. 19. Transmitter output spectrum (a) before and (b) after the calibration for LO leakage due to offset.
90
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 20. Transmitter output spectrum (a) before and (b)–(c) after calibration for the I/Q paths mismatches.
Table 2 The comparison of the proposed calibration system with two other designs. CMOS Process (nm) [11] 65 [12] 40 This Work 180
TX Frequency (GHz)
TX IRR (dB)
TX þ Calibration Power (mW)
1.4 1.6 1.8
10 24 18.5
40a – 37.7
a includes the power consumption of the receiver which is used for the calibration.
quadrature phase imbalance in subject to these variations is negligible. The phase noise simulation results for this 25 runs is shown in Fig. 15. The phase noise variations resulted from these mismatches in devices size and threshold voltage are insignificant. The logarithmic amplifier which also plays the vital role in the calibration system is designed and tested as a separate block to ensure its functionality. The simulation result for the logarithmic amplifier is shown in Fig. 16 which confirms its ability in responding to input signals with small amplitudes. The calibration process as explained before is performed in several steps depending on which component is being calibrated and what is the major source of the unwanted tone. In order to
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
validate the calibration system operation, we have put some intentional mismatches in the transmitter whichincludes 10 mV offset voltage in the mixer, 1.5 degree phase and 1% gain imbalances between I/Q paths of the transmitter. The transmitter output spectrum is shown in Fig. 17 which has image signal and LO feedthrough alongside with main output signal. The function of the calibration system is to cancel and reduce the power of these unwanted signals in the output spectrum. In the calibration process, first the transmitter is set to low gain mode and the effect of LO to RF leakage is calibrated by using the RF current sources (RF_LOFT) in the circuit of the linear transconductance (Fig. 3). The result of this stage of calibration is shown in Fig. 18 which indicates a reduction of about 6 dB in the magnitude of the LO feedthrough. Then, transmitter is set to high gain mode which is aimed for calibration of the LO leakage due to the offset voltage. In this case, the current sources BB_LOFT in linear transconductance are utilized to perform this calibration task. Fig. 19 shows the output of the transmitter before and after this calibration which indicates an improvement of about 8 dB in magnitude of the LO leakage. In the next stage of the calibration and in order to cancel the I/Q mismatches, a baseband signal at frequency of 2 MHz is applied to the transmitter which has some deliberately mademismatches. The output of the transmitter with this baseband signal and above mentioned mismatches is shown in Fig. 20 (a) which indicates an image signal about 37.5 dB below the main output signal. In order to reduce the power of this image signal, first the quadrature oscillator with phase tuning capability is utilized. The result of this tuning is shown in Fig. 20(b) which indicates an improvement of about 8 dB in magnitude of the image signal. Then and in continue of the I/Q mismatch correction, the gain of the I/Q paths are corrected by small gain adjustments in the linear Gm cell. The result of these adjustments is shown in Fig. 20(c) which indicates an improvement of about 10.5 dB in compare to the previous step. The total improvement can be measured by comparing the Fig. 20(a) and (c) which shows a decrease of about 18.5 dB in the amplitude of the image signal. The results of the proposed calibration system is compared with two other designs in Table 2. The proposed design and the design in [12] are based on 2D search algorithm; but the method proposed in [11] is utilizing a non-recursive algorithm without using a test tone and exploit the receiver hardware for detection and calibration process. Although the results in [11] and [12] are from measurements, but the simulations of the proposed circuit in different corners and also for device and threshold voltage mismatches, confirm the reliability of the proposed method.
6. Conclusions In this paper a mismatch calibration method for an I/Q Transmitter based on phase tuning of a quadrature oscillator and gain adjustment of a linear transconductance was introduced. The
91
linear transconductance is located in the baseband part of the transmitter. The calibration process reduces the power of the unwanted tones in the transmitter output spectrum. These are the LO feedthrough and image signal resulted from mismatches in I/Q paths. The LO feedthrough is calibrated by using current sources that are devised into the linear transconductance. The calibration process is performed in several steps depending on the targeted tone in the output spectrum and also the source of the undesired tone. The power of the image signal was reduced by utilizing the phase tuning of the quadrature oscillator and gain tuning of the linear Gm cell. The calibration process was applied to a direct conversion transmitter with some deliberate mismatches introduced to I/Q paths. The results of the simulations indicate a reduction of about 18.5 dB in the magnitude of the transmitter image signal after applying the proposed calibration procedure.
References [1] Wei-Han Yu, Chak-Fong Cheang, Pui-In Mak, Weng-Fai Cheng, Ka-Fai Un, U-Wai Lok, and Rui P. Martins, “A Nonrecursive Digital Calibration Technique for Joint Elimination of Transmitter and Receiver I/Q Imbalances With Minimized Add-On Hardware” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 60, no. 8, pp. 462-466, Aug. 2013. [2] M. Valkama, M. Renfors, “A novel image rejection architecture for quadrature radio receivers,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 51, no. 2, pp. 61– 68, Feb. 2004. [3] J. Yim, T. Kim, S. Ko, W. Ko, H. Shin, I. Ryu, S.-G. Yang, J.-D. Bae, H. Park, “Design and analysis of an ultra-wideband automatic self calibrating up-converter in 65-nm CMOS,” IEEE Trans. Microw. Theory Tech., vol. 60, no. 7, pp. 2178–2191, Jul. 2012. [4] Ali Parsa, Ali Fotowat, Ali Faghfuri, Mahta Jenabi, Emmanauel Riou, Steffan Hahn, “A new method for Quadrature control and feed-through rejection based on Double-Synchronous detection and Logarithmic amplification” US Patent, 7,548,591, USA.. [5] C.P. Lee, A. Behzad,, D. Ojo, M. Kappes, S. Au, Meng-An Pan, K. Carter, S. Tian, “A Highly Linear Direct-Conversion Transmit Mixer Transconductance Stage with Local Oscillation Feedthrough and I/Q Imbalance Cancellation Scheme” IEEE International Solid-State Circuits Conference, 2006. ISSCC 2006. Digest of Technical Papers. [6] I. Mehr, D.R. Welland, “A CMOS continuous-time Gm-C filter for PRML read channel applications at 150 Mb/s and beyond” IEEE Journal of Solid-State Circuits, vol. 32, no.4, pp.499-513, Apr. 1997. [7] A. Hajimiri, T.H. Lee, “Design issues in CMOS differential LC oscillators,” IEEE Journal of Solid-State Circuits, vol. 34 , no.5, pp.717-724, May 1999. [8] Gh. Zareh Fatin, Z.D. Koozeh kanani “A very low power bandpass filter for lowIF applications” Journal of Circuits, Systems and Computers, vol. 17, no. 4, pp.685-702, Aug. 2008. [9] S. Khorram, A. Rofougaran, A.A. Abidi, “A CMOS limiting amplifier and signalstrength indicator,” Symposium on VLSI Circuits, 1995. Digest of Technical Papers. [10] S. Khorram, et al., “A fully integrated SoC for 802.11b in 0.18 μm CMOS” IEEE Journal of Solid-State Circuits, vol. 40, no. 12, pp. 2492 – 2501, Dec. 2005. [11] Wei-Han Yu, et al., “A Nonrecursive Digital Calibration Technique for Joint Elimination of Transmitter and Receiver I/Q Imbalances With Minimized AddOn Hardware” IEEE Transaction on Circuits and Systems—II: Express briefs, vol. 60, no. 8, Aug. 2013. [12] Emanuele Lopelli, Silvian Spiridon and Johan van der Tang “A 40nm Wideband Direct-Conversion Transmitter with Sub-Sampling-Based Output Power, LO Feedthrough and I/Q Imbalance Calibration” IEEE International Solid-State Circuits Conference, 2011. ISSCC 2011. Digest of Technical Papers.
Contents lists available at ScienceDirect
Microelectronics Journal journal homepage: www.elsevier.com/locate/mejo
I/Q mismatch calibration of a transmitter using local quadrature oscillator Gholamreza Zare Fatin a,n, Mostafa Savadi Osgooei b, Ali Fotowat-Ahmady c a
Department of Electrical Engineering, University of Mohaghegh Ardabili, Ardabil, Iran Infineon Technologies AG, Linz, Austria c Department of Electrical Engineering, Sharif University of Technology, Tehran, Iran b
art ic l e i nf o
a b s t r a c t
Article history: Received 27 January 2015 Received in revised form 21 February 2016 Accepted 5 July 2016
In this paper a calibration technique for I/Q mismatch of a transmitter is introduced. The calibration technique is based on the fact that all mismatches in I/Q paths can be modeled as the mismatch of the local oscillator quadrature outputs. Based on this fact, a simple tuning scheme for quadrature output of the oscillator is used to calibrate the mismatch of the I/Q transmitter. In addition to quadrature oscillator, gain mismatch of the I/Q paths and LO feedthorugh and leakage is calibrated by using a tunable linear Gm cell in the base band part. In order to demonstrate the proposed technique, a transmitter with 1.8 GHz carrier frequency is designed with auxiliary blocks to extract and cancel out the I/Q mismatches. The simulation results in 0.18 mm RFCMOS process show that the proposed technique can reduce amplitude of the image signal resulting from mismatches in I/Q paths about 18.5 dB in the transmitter output. & 2016 Elsevier Ltd. All rights reserved.
Keywords: Transmitter I/Q calibration Quadrature oscillator Mismatch detector Synchronous detector
1. Introduction THE I/Q modulation scheme is ubiquitous in most of today wireless transmitters due to spectral efficiency of this modulation. The I/Q paths in a transmitter are subject to phase and gain mismatches which are the result of imbalance between different blocks and interconnections in I/Q paths. This imbalance can be from any block in transmitter chain and usually is rooted to layout and components mismatch in the circuit [1–3]. This mismatch in I/ Q paths will generate an image signal in the spectrum of the output signal which will degrade the error vector magnitude (EVM) of output constellation. A few methods have been introduced in literatures for mismatch calibration of I/Q transmitters. The loop back detector with a recursive algorithms [1] and 2D search algorithms [4,5] are two popular methods of I/Q calibration. The first method can take long calibration time and will consume large power and area concerning the IC implementation. This paper is utilizing the second method of calibration by using iterative search algorithm to calibrate the transmitter. In this paper, with a linear transconductance in the n
Corresponding author. E-mail addresses: [email protected] (G. Zare Fatin), mostafa.savadiosgooei@infineon.com (M. Savadi Osgooei), [email protected] (A. Fotowat-Ahmady). http://dx.doi.org/10.1016/j.mejo.2016.07.004 0026-2692/& 2016 Elsevier Ltd. All rights reserved.
transmit mixer similar to the one used in [5] and also using a phase tunable oscillator with the quadrature outputs, a low power and small area calibration circuit has been devised. This paper contributions can be categorized in three distinct sections: 1) This paper demonstrates the feasibility of transmitter I/Q calibration by using quadrature oscillator. The simulation results show that considerable rejection in the output image signal is achieved by phase tuning of the quadrature oscillator. 2) Another important contribution of the paper is introduction of a technique for phase tuning of the quadrature outputs of the oscillator with minimum impact on the phase noise. 3) Third contribution of the paper includes the design and validation of different auxiliary blocks for I/Q mismatch detection and cancellation including the transmitter itself and elaborating the calibration process. Meanwhile some circuit design techniques in different blocks of the mismatch detection path was introduced. This paper is organized as follows, Section 2 of the paper explains the overall block diagram of the transmitter including the calibration blocks and procedure. Section 3 presents the quadrature oscillator used as the key calibration element. Section 4 explains the detector block used in the mismatch detection and calibration process. Section 5 of the paper presents the simulation results of the quadrature oscillator and calibration process and demonstrates the operation of the proposed method; finally some conclusions are drawn in Section 6.
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
83
Fig. 1. Block diagram of the transmitter with calibration paths.
2. Calibration process
Fig. 2. Transmitter output spectrum with images signal resulted from mismatches in I/Q paths and LO feedthrough.
Fig. 1 shows the transmitter block diagram along with calibration loops. The direct-conversion architecture using only one I/ Q transmit mixer is opted for the transmitter. Transmitter circuit includes two calibration loops which take the signal back from the output of the up conversion mixer through a block entitled as detector. The detector block is used to down convert the output signal for further filtering and extraction of the necessary tone; this block will be explained further in continue of the paper. The signal extracted from output of the detector is used as the residue of the image signal and criteria for effect of the calibration process on the I/Q mismatches. In addition to I/Q calibration, the output of this detector block can be used for LO feedthrough (LOFT) reduction which is shown in Fig. 1 as a red calibration signal and
Fig. 3. The circuit schematic of the linear transconductance used in the transmitter.
84
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 4. The current sources are added to linear Gm cell for the calibration purpose.
Table 1 The adjustment of oscillator output phases and the resulted amplitude difference. ΔVctr(mV)
Δϕ(°)
ΔA(mV)
0 10 50 100 200
0.7 m 0.3 1.47 2.75 4
0.040 0.393 2.4 5.8 8.2
Fig. 5. The quadrature oscillator used in the transmitter along with output I/Q phase tuning technique.
Fig. 7. The block diagram of the detector circuit.
Fig. 6. The simulated quadrature output signals of the oscillator.
performs the calibration through the linear transconductor (Gm) of the transmitter. The I/Q calibration loop is closed either by Gm tuning or quadrature oscillator output phase tuning, which are utilized depending on the calibration purpose, whether it is a gain imbalance tuning or phase mismatch tuning. Regardless of the nature of the mismatch (gain or phase imbalance) between I/Q paths, they will generate an image signal in the output spectrum which indeed will degrade the EVM of the transmitter. This effect is shown in Fig. 2 which shows the image signal resulted from I/Q mismatches along with LO feedthrough and main output signal for a transmitter with a carrier frequency at 1.8 GHz. The calibration algorithm as mentioned before is based on 2D recursive search method which is similar to the method explained in [4]. The Gm cell as shown in block diagram of Fig. 1, is very important block in this calibration process. The circuit schematic of this Gm cell is shown in Fig. 3 which was introduced in [5] as a linear transmit transconductor. The linearity feature of this transconductance has been thoroughly explained in a literature [6]
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
85
Fig. 8. The circuit schematic of the synchronous detector.
Fig. 9. The synchronous detector simulation result (a) input signal which is taken from transmitter output, (b) down converted output signal of the synchronous detector.
published before. The transconductance of this block in ideal case 1 R9 is equal to Gm = R1 R7 which is determined based on resistors proportion. The current sources RF_LOFT and BB_LOFT were added to the circuit in order to facilitate the calibration process.
The function of the different elements in this linear Gm cell during the calibration process is shown in Fig. 4. The transistor M9 is working as a linear resistor and shorts two differential outputs of the transconductor and adjusts the gain of the Gm cell which is
86
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 10. The bandpass filter used in the detector circuit which is based on a biquad cell.
Fig. 11. The circuit schematic of the rectifier circuit used in the logarithmic amplifier.
Fig. 12. The simulated phase noise of the quadrature oscillator used in the transmit mixer.
used for correction of the gain imbalance in the I/Q paths. Transistors M10-M17 are used for gain adjustment of the baseband signal which can be controlled by 3-bits digital word. This digital word controls the baseband gain of the transmitter. The DC offset in the baseband signal can create the LO-feedthrough which in this circuit is cancelled by using the BB_LOFT current sources. On the other hand the LO-feedthrough due to LO to RF leakage is canceled
Fig. 13. Phase noise of the quadrature oscillator with and without calibration signal at the tail current.
by using RF_LOFT currents sources. These current sources are connected to drain of the M7 and M8 transistors which provide the tail current of the mixer. The mixer cell is a tuned active double balanced Gilbert cell.
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 14. Phase noise of the oscillator in different corners of the process.
87
base band part while the gain is set to minimum value. Then, the detector circuit as shown in Fig. 1, measures the LO-feedthrough due to leakage. After measurement of LO-feedthrough, the RF_LOFT current sources are used to cancel out the unwanted tone in the output. In the next step of the calibration process, baseband gain is set to maximum value. In this state, the detector will measure the LOfeedthrough due to baseband offset voltage. Then BB_LOFT current sources are utilized to cancel this effect. In continue of the calibration process and for I/Q imbalance cancellation, a 2 MHz baseband signal is injected into the I/Q paths; then detector measures the image signal power which is the result of the I/Q mismatches. This extracted image signal power is reduced by adjustment of the quadrature oscillator output phases which will be explained in the next section of the paper. The tuning of the quadrature oscillator outputs phase can adjust the phase mismatch of the I/Q paths. Although, phase tuning changes the amplitude of the quadrature output signals in addition to their phase, but this effect is negligible and can positively contribute to gain imbalance correction of the transmitter. In this design the amplitude limiting stages were designed at the output of the oscillator to limit and match the amplitude of the quadrature outputs after phase tuning. After phase mismatch tuning and in order to further reduce the amplitude of the image signal, the gain of the I/Q paths is tuned; this is carried out by using linear Gm cell gain adjustment capability which was explained before.
3. The quadrature oscillator
Fig. 15. Phase noise variations due to mismatches in devices size and threshold voltage.
Fig. 16. The simulation result of the logarithmic amplifier.
As mentioned before the calibration process is very similar to the algorithm used in [4]; this process is explained here for further details and also to clarify the algorithm. The calibration process starts with LO-feedthrough cancellation. A signal (a 4 MHz signal in this design) is injected into the
As explained in the previous section and in the I/Q mismatch calibration, the phase imbalance between I/Q paths is corrected through a quadrature oscillator. This quadrature oscillator drives the transmit mixer. The proposed oscillator which is used in the calibration process is shown in Fig. 5. It is a complementary LC oscillator which was designed in quadrature form to drive the I/Q mixers of the transmitter. The complementary LC oscillator is preferred over the NMOS only realization because of better phase noise performance [7]. The complementary oscillator has higher transconductance for the same bias current which results in faster switching and shorter crossing time which is very important in reducing the timing jitter or phase noise of the oscillator. As mentioned before and explained in Section 2, in order to calibrate the phase mismatch of the I/Q paths and remove the image signal, it is necessary to adjust the output phases of the quadrature oscillator. As shown in Fig. 5, the differential pair made by M1 and M2 transistors is added to oscillator to tune the phase of the output signals. By adjusting the output phases of the oscillator, the phase mismatches resulting from other parts of the transmitter can be canceled. This deliberate quadrature phase mismatch in oscillator outputs can negate the I/Q phase imbalances which are originated from other parts of the circuit. The gate-source voltage (Vgs) of the transistors M1 and M2 are changed in contrast to each other by a control voltage Vctr which is applied in differential form. This control voltage in addition to phase change will result in amplitude variation of the output signals as well. But, the impact of oscillator quadrature outputs amplitude difference in mixer performance is less severe than phase difference. Also, in order to remove these amplitude mismatches resulted from phase adjustment of the oscillator, we have designed a limiter at the quadrature outputs of the oscillator which can balance the amplitudes of the output signals. The quadrature output signals of the oscillator are shown in Fig. 6. Table 1 represents the differences in phases and amplitudes
88
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 17. The output spectrum of the transmitter with some intentional mismatches in I/Q paths.
of the outputs of the quadrature oscillator created by adjustment of the control voltage (Vctr).
4. The detector As shown in Fig. 1 and in the calibration process of the transmitter, a detector block is used to extract the power of the targeted component from output spectrum. This is a key block in detecting and cancelling the image signal or oscillator feedthrough. The block diagram of the detector [4] is shown in Fig. 7. It has a synchronous detector at the input which down samples the transmitter output signal. The down converted signal will be then filtered by a bandpass filter to extract the targeted tone from the output spectrum. The output of the bandpass filter is given to a logarithmic amplifier to produce a dc voltage proportional to extracted tone for calibration process.
The circuit schematic of the synchronous detector is shown in Fig. 8. It is based on the design reported in [4]. In this paper we have redesigned the circuit in CMOS process and optimized the circuit for this application. The synchronous detector circuit in Fig. 8 is based on Gilbert cell structure. The input transistors M1 and M2 operate as the input transconductance. In this design, the common-gate structure has been selected for input transconductance with two series resistors at the input, R1 and R2. Using these input resistors in addition to common-gate structure boosts the linearity of the detector which then increases the dynamic range of the detector as well. The switching transistors M5-M8 do the mixing function; as shown in Fig. 8, the input signal in addition to input transconductance also drives the switching transistors. This is like a selfmixing of an RF signal. The voltage division between the input resistors (R1, R2) and input resistance(1/gm) of the M1 and M2
Fig. 18. Transmitter output spectrum (a) before and (b) after the calibration for LO to RF leakage.
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
transistors attenuates the RF signal before reaching the input transconductance. The simulation results of the synchronous detector for an input spectrum which has an RF signal at 1.8 GHz along with an LO feedthorugh and image signal is shown in Fig. 9. The down converted signal has the main signal at dc with LO feedthrough at 4 MHz and image signal at 8 MHz which are shown in Fig. 9(b). This arrangement of the signals has been adopted for LO feedthrough cancellation. The down converted signal at 4 MHz needs further filtering and amplification to be used as the control signal in the calibration process. In order to select the desired tone from down-converted spectrum at the output of the synchronous detector, a bandpass filter is added to detector circuit which was shown in the block diagram of Fig. 7. The integrated filter was designed based on gmC method and is a biquad cell which is shown in Fig. 10. The bandpass output of the biquad cell is used here for filtering purpose which has the transfer function as in (1),
VBPF −Gm1 = Vinput C1 S 2 +
S+
(
1 R1C1
+
1 R 2 C2
1 R 2 C2
)S +
1 +Gm2 Gm3 R1R 2 R1R 2 C1C2
(1)
The bandpass filter is designed based on the previous work reported in [8]. The circuit schematic of the gm cell utilized in this biquad is also shown in Fig. 10 which is a differential pair with resistive loads. The logarithmic amplifier is the final stage in the detector circuit and is also shown in Fig. 7. The amplifier includes the limiting stages L1-L8 and rectifier circuits which add their output currents in an RC circuit [9,10]. The R resistor is used to convert the output currents of the rectifier blocks to voltage signal; the capacitor (C) is also added to perform the low pass filtering function. The limiting stages (L1-L8) utilized here are differential pairs with 12 dB dynamic range each which add up to total dynamic range of 96 dB for the logarithmic amplifier. The RC time constant is chosen equal to 1 ms.
89
The rectifier circuit is shown in Fig. 11 which is a modified form of the Gilbert cell. The operation of the circuit is based on a selfmixing of the input signal which resembles the rectification task.
5. Simulation results The different blocks of the proposed calibration system which were explained in previous sections are put together to validate their functionality in the calibration of the transmitter. These circuits are designed and simulated in 0.18 mm RFCMOS process. The individual blocks are also simulated to validate their operation and also to be optimized before using in the calibration system. The simulation result for phase noise of the quadrature oscillator is shown in Fig. 12. The phase noise of the oscillator is about 110 dBc at 1 MHz offset frequency. In order to evaluate the effect of the calibration signal in the phase noise of the oscillator, the phase noise was simulated in presence of the calibration signal. The amplitude of the calibration signal was set based on the overall calibration system requirement and signal level that is necessary to suppress the image signal below a certain level in compare to the desired signal. The simulation results of this test is shown in Fig. 13. As shown in this figure increase in the phase noise value in compare to the state without the calibration signal is negligible. The phase noise of the oscillator has been simulated in different corners of the process and the results are shown in Fig. 14. The largest shift in phase noise value comes from FS corner and is about 4 dB higher in compare to TT state. The phase imbalance between quadrature outputs of the oscillator has also been simulated in different corners. The maximum change in quadrature phase is happening in FS corner, but the value is less than 0.5 degree. The phase noise and phase imbalance have been simulated with a 720% mismatch in devices size and threshold voltage; this accounts for total 25 runs with different mismatch specification. The output
Fig. 19. Transmitter output spectrum (a) before and (b) after the calibration for LO leakage due to offset.
90
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
Fig. 20. Transmitter output spectrum (a) before and (b)–(c) after calibration for the I/Q paths mismatches.
Table 2 The comparison of the proposed calibration system with two other designs. CMOS Process (nm) [11] 65 [12] 40 This Work 180
TX Frequency (GHz)
TX IRR (dB)
TX þ Calibration Power (mW)
1.4 1.6 1.8
10 24 18.5
40a – 37.7
a includes the power consumption of the receiver which is used for the calibration.
quadrature phase imbalance in subject to these variations is negligible. The phase noise simulation results for this 25 runs is shown in Fig. 15. The phase noise variations resulted from these mismatches in devices size and threshold voltage are insignificant. The logarithmic amplifier which also plays the vital role in the calibration system is designed and tested as a separate block to ensure its functionality. The simulation result for the logarithmic amplifier is shown in Fig. 16 which confirms its ability in responding to input signals with small amplitudes. The calibration process as explained before is performed in several steps depending on which component is being calibrated and what is the major source of the unwanted tone. In order to
G. Zare Fatin et al. / Microelectronics Journal 55 (2016) 82–91
validate the calibration system operation, we have put some intentional mismatches in the transmitter whichincludes 10 mV offset voltage in the mixer, 1.5 degree phase and 1% gain imbalances between I/Q paths of the transmitter. The transmitter output spectrum is shown in Fig. 17 which has image signal and LO feedthrough alongside with main output signal. The function of the calibration system is to cancel and reduce the power of these unwanted signals in the output spectrum. In the calibration process, first the transmitter is set to low gain mode and the effect of LO to RF leakage is calibrated by using the RF current sources (RF_LOFT) in the circuit of the linear transconductance (Fig. 3). The result of this stage of calibration is shown in Fig. 18 which indicates a reduction of about 6 dB in the magnitude of the LO feedthrough. Then, transmitter is set to high gain mode which is aimed for calibration of the LO leakage due to the offset voltage. In this case, the current sources BB_LOFT in linear transconductance are utilized to perform this calibration task. Fig. 19 shows the output of the transmitter before and after this calibration which indicates an improvement of about 8 dB in magnitude of the LO leakage. In the next stage of the calibration and in order to cancel the I/Q mismatches, a baseband signal at frequency of 2 MHz is applied to the transmitter which has some deliberately mademismatches. The output of the transmitter with this baseband signal and above mentioned mismatches is shown in Fig. 20 (a) which indicates an image signal about 37.5 dB below the main output signal. In order to reduce the power of this image signal, first the quadrature oscillator with phase tuning capability is utilized. The result of this tuning is shown in Fig. 20(b) which indicates an improvement of about 8 dB in magnitude of the image signal. Then and in continue of the I/Q mismatch correction, the gain of the I/Q paths are corrected by small gain adjustments in the linear Gm cell. The result of these adjustments is shown in Fig. 20(c) which indicates an improvement of about 10.5 dB in compare to the previous step. The total improvement can be measured by comparing the Fig. 20(a) and (c) which shows a decrease of about 18.5 dB in the amplitude of the image signal. The results of the proposed calibration system is compared with two other designs in Table 2. The proposed design and the design in [12] are based on 2D search algorithm; but the method proposed in [11] is utilizing a non-recursive algorithm without using a test tone and exploit the receiver hardware for detection and calibration process. Although the results in [11] and [12] are from measurements, but the simulations of the proposed circuit in different corners and also for device and threshold voltage mismatches, confirm the reliability of the proposed method.
6. Conclusions In this paper a mismatch calibration method for an I/Q Transmitter based on phase tuning of a quadrature oscillator and gain adjustment of a linear transconductance was introduced. The
91
linear transconductance is located in the baseband part of the transmitter. The calibration process reduces the power of the unwanted tones in the transmitter output spectrum. These are the LO feedthrough and image signal resulted from mismatches in I/Q paths. The LO feedthrough is calibrated by using current sources that are devised into the linear transconductance. The calibration process is performed in several steps depending on the targeted tone in the output spectrum and also the source of the undesired tone. The power of the image signal was reduced by utilizing the phase tuning of the quadrature oscillator and gain tuning of the linear Gm cell. The calibration process was applied to a direct conversion transmitter with some deliberate mismatches introduced to I/Q paths. The results of the simulations indicate a reduction of about 18.5 dB in the magnitude of the transmitter image signal after applying the proposed calibration procedure.
References [1] Wei-Han Yu, Chak-Fong Cheang, Pui-In Mak, Weng-Fai Cheng, Ka-Fai Un, U-Wai Lok, and Rui P. Martins, “A Nonrecursive Digital Calibration Technique for Joint Elimination of Transmitter and Receiver I/Q Imbalances With Minimized Add-On Hardware” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 60, no. 8, pp. 462-466, Aug. 2013. [2] M. Valkama, M. Renfors, “A novel image rejection architecture for quadrature radio receivers,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 51, no. 2, pp. 61– 68, Feb. 2004. [3] J. Yim, T. Kim, S. Ko, W. Ko, H. Shin, I. Ryu, S.-G. Yang, J.-D. Bae, H. Park, “Design and analysis of an ultra-wideband automatic self calibrating up-converter in 65-nm CMOS,” IEEE Trans. Microw. Theory Tech., vol. 60, no. 7, pp. 2178–2191, Jul. 2012. [4] Ali Parsa, Ali Fotowat, Ali Faghfuri, Mahta Jenabi, Emmanauel Riou, Steffan Hahn, “A new method for Quadrature control and feed-through rejection based on Double-Synchronous detection and Logarithmic amplification” US Patent, 7,548,591, USA.. [5] C.P. Lee, A. Behzad,, D. Ojo, M. Kappes, S. Au, Meng-An Pan, K. Carter, S. Tian, “A Highly Linear Direct-Conversion Transmit Mixer Transconductance Stage with Local Oscillation Feedthrough and I/Q Imbalance Cancellation Scheme” IEEE International Solid-State Circuits Conference, 2006. ISSCC 2006. Digest of Technical Papers. [6] I. Mehr, D.R. Welland, “A CMOS continuous-time Gm-C filter for PRML read channel applications at 150 Mb/s and beyond” IEEE Journal of Solid-State Circuits, vol. 32, no.4, pp.499-513, Apr. 1997. [7] A. Hajimiri, T.H. Lee, “Design issues in CMOS differential LC oscillators,” IEEE Journal of Solid-State Circuits, vol. 34 , no.5, pp.717-724, May 1999. [8] Gh. Zareh Fatin, Z.D. Koozeh kanani “A very low power bandpass filter for lowIF applications” Journal of Circuits, Systems and Computers, vol. 17, no. 4, pp.685-702, Aug. 2008. [9] S. Khorram, A. Rofougaran, A.A. Abidi, “A CMOS limiting amplifier and signalstrength indicator,” Symposium on VLSI Circuits, 1995. Digest of Technical Papers. [10] S. Khorram, et al., “A fully integrated SoC for 802.11b in 0.18 μm CMOS” IEEE Journal of Solid-State Circuits, vol. 40, no. 12, pp. 2492 – 2501, Dec. 2005. [11] Wei-Han Yu, et al., “A Nonrecursive Digital Calibration Technique for Joint Elimination of Transmitter and Receiver I/Q Imbalances With Minimized AddOn Hardware” IEEE Transaction on Circuits and Systems—II: Express briefs, vol. 60, no. 8, Aug. 2013. [12] Emanuele Lopelli, Silvian Spiridon and Johan van der Tang “A 40nm Wideband Direct-Conversion Transmitter with Sub-Sampling-Based Output Power, LO Feedthrough and I/Q Imbalance Calibration” IEEE International Solid-State Circuits Conference, 2011. ISSCC 2011. Digest of Technical Papers.