Disclosure of Invention
To overcome the above-mentioned shortcomings of the prior art, the present invention aims to provide a wide-speed high-linearity phase interpolator suitable for 1-28Gbps SerDes, which makes the output clock phase and the input control code have a linear relationship, thereby reducing the deterministic jitter of the clock data recovery current; and enables the phase interpolator to be adapted for use in a wide-rate operating environment.
In order to achieve the purpose, the invention adopts the technical scheme that:
a wide-speed high-linearity phase interpolator suitable for 1-28Gbps SerDes comprises a buffer arranged at an input end to enhance the driving capability of a clock signal, and is characterized in that a digital control capacitor array connected in parallel is arranged at an output end of the buffer to form a low-pass filter, and the frequency of a main pole point of the buffer is controlled at the frequency point of an input signal, so that the input frequency is single.
Preferably, the buffer comprises a MOS transistor M1、M2And M3,M1And M2G pole of the transformer is respectively connected with two paths of input signals Vin1And Vin2,M1D pole of the output signal Vout1,M2D pole of the output signal Vout2,Vout1Through a resistance RD1Is connected with VDD,Vout2Through a resistance RD2Is connected with VDD,M1S pole, M of2S pole and M3D of (A)Polar connection, M3The S-pole of the digitally controlled capacitor array is grounded, and each path of the digitally controlled capacitor array is composed of a capacitor and a control switch connected in series.
Preferably, the control switch adopts an NMOS tube.
Preferably, the digitally controlled capacitor array is at V of the bufferout1And Vout2Three paths are respectively arranged on the frequency conversion circuit, namely the three paths are respectively controlled by three control bits, so that the capacitance of the output end of the buffer is adjusted according to the frequency of the input clock, the frequency range of 3.5-7GHz is quantized into 8 frequency points, and 0.5GHz is covered between every two frequency points.
Correspondingly, the invention also provides a clock data recovery circuit based on the wide-speed high-linearity phase interpolator applicable to the 1-28Gbps SerDes.
Compared with the prior art, the invention accurately calculates the capacitance values corresponding to different frequency points by carrying out thevenin equivalence on the buffer, and realizes the capacitance values in the circuit, so that the linearity between the output clock phase and the input control code is approximately in a linear relation in the range of 1-28Gb/s of the input data signal, and the deterministic jitter of the clock data recovery circuit is effectively reduced.
Drawings
Fig. 1 is a system architecture of a phase interpolator based clock data recovery circuit.
Fig. 2 is a schematic diagram of a conventional buffer.
Fig. 3 is a schematic diagram of a buffer with a digitally controlled capacitor array designed according to the present invention.
Fig. 4 is a davinin equivalent circuit at the output of the buffer of fig. 2.
FIG. 5 shows the improved linearity contrast of the pre-and post-phase interpolators when the input data is 28 Gb/s. Wherein (a) is PI output linearity before improvement, and (b) is PI output linearity after improvement.
FIG. 6 is a comparison of the recovered clock eye before and after improvement when the input data is 28 Gb/s. Wherein (a) is the PI output clock eye pattern before improvement, and (b) is the PI output clock eye pattern after improvement.
FIG. 7 is a graph showing a comparison of jitter performance of recovered clocks before and after improvement when the input data is 28Gb/s and the frequency difference is +200 ppm. Wherein, the first is PI control code before improvement, (b) is PI control code after improvement, (c) is clock eye pattern recovered before improvement, and (d) is clock eye pattern recovered after improvement.
FIG. 8 shows the improved linearity contrast of the pre-and post-phase interpolators when the input data is 22 Gb/s. Wherein (a) is PI output linearity before improvement, and (b) is PI output linearity after improvement.
FIG. 9 is a comparison of the recovered clock eye before and after improvement when the input data is 22 Gb/s. Wherein (a) is the PI output clock eye pattern before improvement, and (b) is the PI output clock eye pattern after improvement.
FIG. 10 is a graph showing a comparison of jitter performance of recovered clocks before and after improvement when input data is 22Gb/s and the frequency difference is +200 ppm. Wherein, the first is PI control code before improvement, (b) is PI control code after improvement, (c) is clock eye pattern recovered before improvement, and (d) is clock eye pattern recovered after improvement.
FIG. 11 shows the improved linearity contrast of the pre-and post-phase interpolators when the input data is 18 Gb/s. Wherein (a) is PI output linearity before improvement, and (b) is PI output linearity after improvement.
FIG. 12 is a comparison of the recovered clock eye before and after improvement when the input data is 18 Gb/s. Wherein (a) is the PI output clock eye pattern before improvement, and (b) is the PI output clock eye pattern after improvement.
FIG. 13 is a graph showing a comparison of jitter performance of recovered clocks before and after improvement when the input data is 18Gb/s and the frequency difference is +200 ppm. Wherein, the first is PI control code before improvement, (b) is PI control code after improvement, (c) is clock eye pattern recovered before improvement, and (d) is clock eye pattern recovered after improvement.
FIG. 14 shows the improved linearity contrast of the pre-and post-phase interpolators when the input data is 14 Gb/s. Wherein (a) is PI output linearity before improvement, and (b) is PI output linearity after improvement.
FIG. 15 is a comparison of the recovered clock eye diagrams before and after improvement when the input data is 14 Gb/s. Wherein (a) is the PI output clock eye pattern before improvement, and (b) is the PI output clock eye pattern after improvement.
FIG. 16 is a graph showing a comparison of jitter performance of recovered clocks before and after improvement when the input data is 14Gb/s and the frequency difference is +200 ppm. Wherein, the first is PI control code before improvement, (b) is PI control code after improvement, (c) is clock eye pattern recovered before improvement, and (d) is clock eye pattern recovered after improvement.
Detailed Description
The preferred embodiments will be described in detail below with reference to the accompanying drawings. It should be emphasized that the following description is merely exemplary in nature and is not intended to limit the scope of the invention or its application.
FIG. 3 shows an embodiment of a wide-speed high-linearity phase interpolator according to the present invention, in which a buffer for enhancing the driving capability of a clock signal is disposed at an input terminal of the phase interpolator. The invention arranges a digital control capacitor array connected in parallel at the output end of the buffer to form a low-pass filter, and controls the dominant pole frequency of the buffer at the frequency point of the input signal, thereby ensuring the single input frequency.
In this embodiment, the buffer includes a MOS transistor M1、M2And M3,M1And M2G pole of the transformer is respectively connected with two paths of input signals Vin1And Vin2,M1D pole of the output signal Vout1,M2D pole of the output signal Vout2,Vout1Through a resistance RD1Is connected with VDD,Vout2Through a resistance RD2Is connected with VDD,M1S pole, M of2S pole and M3D pole connection of (1), M3The S-pole of (a) is grounded.
Each route of the digital control capacitor array is composed of a capacitor and a control switch which are connected in series, and the control switch can adopt an NMOS tube.
In this embodiment, the digitally controlled capacitor array is controlled by three control bits to adjust the capacitance at the output of the buffer stage according to the input clock frequency. I.e. the digitally controlled capacitor array is at V of the bufferout1And Vout2The frequency range of 3.5-7GHz is quantized into 8 frequency points in this way, and 0.5GHz is covered between every two frequency points, so that the clock data recovery circuit can adapt to wide-speed data reception and realize the function of supporting continuous data rate transmission. .
The detailed principle of the invention is as follows:
the circuit of fig. 2 can be simplified to fig. 4 for analysis by the davinin equivalent theorem.
When an excitation signal is input to the circuit, its frequency domain response is:
it is known that the output contains two poles, one determined by the stimulus signal and the other determined by the circuit itself. The corresponding time domain response signal is:
wherein E is a forced response, determined by the excitation signal,
the transient response is determined by the circuit itself. R
eqC is the time constant of the circuit, determines the main pole frequency of the circuit, and s is the laplace transform factor.
The following can be obtained:
therefore, according to different data rates, the required capacitance value can be calculated, and each capacitance value in the capacitor array is digitally controlled, and can be the same or different according to the requirement
Such as:
when the receiving data rate is 22Gb/s and the local clock rate is 5.5GHz
When the receiving data rate is 18Gb/s and the local clock rate is 4.5GHz
When the receiving data rate is 14Gb/s and the local clock rate is 3.5GHz
C5.5G、C4.5GAnd C3.5GThe equivalent values of the capacitor array are respectively 5.5GHz, 4.5GHz and 3.5GHz of the local clock, G is a frequency unit, and f is a capacitance unit.
It should be noted that the method is universal, is not limited to 3 digital capacitor array control bits, and can be applied to a phase interpolator for any digitally controlled capacitor control bit; and is not limited to wide rate data of only 1-28Gb/s, and is applicable to any wide rate data range.
From (a) and (b) of fig. 5, it can be seen that the linearity of the control code and the output phase is significantly improved by using the non-equivalent current source phase interpolator.
From (a) and (b) of fig. 6, it can be seen that, when the local clock is 7GHz, the ideal PI output eye interval is 2.232ps, the maximum time interval of the improved front eye is 3.053ps, and the maximum error is (3.053-2.232)/2.232 ═ 36.8%; the maximum time interval of the improved eye pattern is 2.346ps, the maximum error is (2.346-2.232)/2.232-5.1%, and the linearity is improved by 31.7%.
From (a), (b), (c), (d) of fig. 7, it can be seen that when the received data rate is 28.0056Gb/s (frequency difference +200ppm from 28Gb/s data), and the local clock rate is 7 GHz: the clock jitter before improvement is 5.9ps, 0.083 UI; after improvement, the clock jitter is 5.3ps, 0.074 UI; the jitter performance of the recovered clock is improved by 10.2%.
From fig. 8 (a) and (b), it can be seen that, when the local clock is 5.5GH, and the numerical control capacitor array value is 547.3f, the linearity of the phase interpolator is significantly improved compared to the phase interpolator without the capacitor array.
At 5.5GHz of local clock, the ideal PI output eye interval is 2.84ps, and according to (a) and (b) of fig. 9, it can be seen that the maximum time interval of the improved front eye is 4.709ps, and the maximum error is 65.8%; the maximum time interval of the improved eye pattern is 3.144ps, the maximum error is 10.7%, and the linearity is improved by 55.1%.
From (a), (b), (c), (d) of fig. 10, it can be seen that when the received data rate is 22.0044Gb/s (frequency difference +200ppm from 22Gb/s data), and the local clock rate is 5.5 GHz: the clock jitter before improvement is 8.9ps, 0.098 UI; the clock jitter after improvement is 6.9ps, 0.076 UI; the jitter performance of the recovered clock is improved by 22.5%.
From fig. 11 (a) and (b), it can be seen that, when the local clock is 4.5GH, and the numerical control capacitor array value is 669f, the linearity of the phase interpolator is significantly improved compared to the phase interpolator without the capacitor array.
At a local clock of 4.5GHz, the ideal PI output eye interval is 3.47ps, and according to (a) and (b) of fig. 12, it can be seen that the maximum time interval of the improved front eye is 6.793ps, and the maximum error is 95.8%; the maximum time interval of the improved eye pattern is 4.238ps, the maximum error is 22%, and the linearity is improved by 73.8%.
From (a), (b), (c), (d) of fig. 13, it can be seen that when the received data rate is 18.0036Gb/s (frequency difference from 18Gb/s data is +200ppm), the local clock rate is 4.5 GHz: the clock jitter before improvement is 13.87ps, 0.125 UI; the improved clock jitter is 8.59ps and 0.073 UI; the jitter performance of the recovered clock is improved by 38.1%.
From fig. 14 (a) and (b), it can be seen that, when the local clock is 3.5GH, and the numerical control capacitor array value is 860f, the linearity of the phase interpolator is significantly improved compared to the phase interpolator without the capacitor array.
From (a) and (b) of fig. 15, it can be seen that the ideal PI output eye interval is 4.464ps for a local clock of 3.5 GHz. The maximum time interval of the improved front eye diagram is 10.63ps, and the maximum error is 138.1%; the maximum time interval of the improved eye pattern is 5.576ps, the maximum error is 24.9%, and the linearity is improved by 113.2%.
From (a), (b), (c), (d) of fig. 16, it can be seen that when the received data rate is 14.0028Gb/s (frequency difference from 14Gb/s data is +200ppm), the local clock rate is 3.5 GHz: the clock jitter before improvement was 22.26ps, 0.156 UI; the improved clock jitter is 12.75ps, 0.089 UI; the jitter performance of the recovered clock is improved by 42.72%.
By adopting the phase interpolator of the invention, based on fig. 1, it is obvious that a corresponding clock data recovery circuit can be obtained, and the sampling precision is further improved.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.