SerDes equalization for short reflective channels
By configuring the high-cut filter behavior in the serializer-deserializer and adjusting the equalizer parameters, the problem of insufficient equalization performance on low-loss channels is solved, and effective suppression of reflection intensity and performance improvement is achieved.
Patent Information
- Application Number
- CN202110106349.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2020-11-18
- Filing Date
- 2021-01-26
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-01-26
AI Technical Summary
Existing high-rate serializer-deserializer blocks exhibit insufficient equalization performance on low-loss channels, resulting in intersymbol interference and noise problems, especially in reflective low-loss channels. performance.
A serializer-deserializer scheme including a preequalizer, a continuous time linear equalizer and a decision feedback equalizer is designed. By configuring the frequency response of the preequalizer and CTLE, the detection and manual attenuation of the channel reflection intensity are realized. , adjust the cursor coefficient and transmit driver bandwidth to reduce high frequency components, and convert to high cutoff filter behavior to suppress interference caused by reflection.
It effectively improves the equalization performance on low-loss channels, reduces the bit error rate, improves the signal-to-noise ratio, and reduces the interference caused by reflection through manual attenuation, improving the performance of the serializer-deserializer.
Smart Images

Figure CN113541695B_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application is related to U.S. Application No. 16 / 691,523, entitled “A multi-function signal measurement circuit for ADC-based SerDes”, filed by inventor Junqing Sun on November 21, 2019; and U.S. Application No. 16 / 552,927, entitled “SerDes pre-equalizer having adaptable preset coefficient registers”, filed by inventors Junqing Sun and Haoli Qian on August 27, 2019. Both applications are hereby referenced in their entirety. Both applications are hereby incorporated by reference in their entirety. Background Art
[0003] Digital communication occurs between a transmitting device and a receiving device through an intermediate communication medium or "channel" (e.g., optical fiber, insulated wire, printed circuit board ("PCB") traces). Each transmitting device typically transmits symbols at a fixed symbol rate, while each receiving device detects a (possibly corrupted) sequence of symbols from which the transmitted data is reconstructed. A "symbol" is a state or valid condition of a channel that lasts for a fixed period of time, which is called a "symbol interval." A symbol can be, for example, a voltage or current level, an optical power level, a phase value, or a specific frequency or wavelength. The change from one channel state to another is called a symbol transition. Each symbol can represent (i.e., encode) one or more binary bits of data. Alternatively, data can be represented by a symbol transition or by a sequence of two or more symbols.
[0004] Many digital communication links use only one bit per symbol; a binary "0" is represented by one symbol (e.g., a voltage or current signal within a first range), and a binary "1" is represented by another symbol (e.g., a voltage or current signal within a second range), but higher order signal constellations are known and frequently used. In 4-level pulse amplitude modulation ("PAM4"), each symbol interval can carry any of four symbols labeled -3, -1, +1, and +3. Two binary bits can thus be represented by each symbol.
[0005] Channel imperfections create dispersion, which can cause each symbol to interfere with its neighbors, a result known as inter-symbol interference ("ISI"). ISI can make it difficult for a receiving device to determine which symbols were sent in each interval, especially when such ISI is combined with additive noise. As symbol rates continue to increase, the effects of channel dispersion and ISI become more severe. To combat noise and ISI, transmitters and receivers can employ a variety of equalization techniques, including linear equalizers and decision feedback equalizers.
[0006] A serializer / deserializer ("SerDes") block is a functional module that converts between a (low-rate) parallel symbol stream and a (high-rate) serial symbol stream. Integrated circuits often contain SerDes blocks to reduce the number of physical input / output ("I / O") pins and / or external signal lines required for communication. SerDes blocks are typically designed to operate at the highest possible symbol rate to maximize their benefits, and therefore they are the modules that benefit most from the use of efficient equalization techniques.
[0007] FIG. 1A is a block diagram of a host device 100, such as, for example, a server, a storage device, a switch, or a router. One or more packaged integrated circuit chips 101 are interconnected by traces on a printed circuit board (“PCB”) to achieve the intended operation of the device. At least one of the chips 101 includes a serializer-deserializer block 102 for high-rate digital communication. A collection 103 of PCB traces couples the serializer-deserializer block 102 to an external port connector 104. The external port connector 104 can be designed to interface with a standard-compliant network interface module 106, such as a quad small form factor pluggable double density (“QSFP-DD”) module. In order to drive and equalize communication signals across cable conductors or optical fibers 108, the network interface module 106 may include a data recovery and remodulation (“DRR”) device 110.
[0008] 1B is a block diagram of an illustrative DRR device 110. The illustrative device 110 includes a serializer-deserializer module with pins 111 for exchanging high-rate serial bit streams with the host device 100 across four channels, pins 112 for sending and receiving high-rate serial bit streams via cable conductors 108, and core logic 113 for implementing channel communication protocols and first-in, first-out buffering of the bit streams in each direction. The core logic 113 may include a programmable processor and associated firmware that configures the processor to provide the desired functionality.
[0009] Also included are various support modules and contacts 114, 115, such as power regulation and distribution for control, clock generation, digital input / output lines, and a JTAG module for built-in self-test. A chip designer can design a device by placing predefined modular cells of serializers, deserializers, power supplies, clock generators, I / O cells, and JTAG; and using some support logic to route the interconnections between the modular cells.
[0010] The deserializer block implements the receive functionality of the host 100 and the DRR device 110, including any suitable equalization techniques, such as linear equalization and decision feedback equalization (DFE). The serializer block implements the transmit functionality of the host 100 and the DRR device 110, optionally including spectrum shaping or "pre-equalization" to at least partially compensate for channel dispersion. It is expected that the serializer-deserializer blocks for the cable-oriented channel 112 will cope with longer channels (e.g., one meter long or more) with significant degrees of attenuation and dispersion, and in practice, each of the serializer-deserializer blocks can be designed to cope with such channels as a "worst case" scenario.
[0011] FIG. 2A shows a serializer communicating with a deserializer via a signal transmission channel 211. The serializer includes a parallel-to-serial ("P2S") converter 201, a pre-equalizer 202, and a transmit driver 203. The P2S converter 201 converts a set of parallel data symbol streams into a serial data symbol stream, for example, by time-multiplexing the parallel set onto a single signal line. Channel 211 typically exhibits dielectric loss and a skin effect that preferentially attenuates the high-frequency components of the transmitted signal. To partially offset this effect, pre-equalizer 202 is a linear equalizer, sometimes referred to as a feed-forward equalizer ("FFE"), configured as a high-pass filter ("HPF") to enhance the high-frequency components of the transmitted signal. Transmit driver 203 supplies the filtered transmit signal to channel 211.
[0012] At the far end of channel 211, a continuous time linear equalizer ("CTLE") 221 is configured as another HPF or bandpass filter to further boost the high frequency components of the received signal. Receiver FFE 222 and decision feedback equalizer ("DFE") 223 operate to minimize ISI and perform symbol detection. (Receiver FFE 222 and DFE 223 can be implemented in the digital domain, preceded by an analog-to-digital converter). A serial-to-parallel ("S2P") converter 224 converts the serial data stream into a collection of parallel data symbol streams originally received by the serializer. Although shown as separate blocks, the S2P functionality can be implemented at least in part by parallelization of the equalizer blocks 222, 223. Additional implementation details of the equalizer can be found, for example, in U.S. application Ser. No. 16 / 552,927, filed on August 27, 2019 by inventors Junqing Sun and Haoli Qian, and entitled “SerDes pre-equalizer having adaptable preset coefficient registers.”
[0013] FIG2B shows an illustrative pre-equalizer 202 implemented as a digital filter having delay elements 231, 232 that sequentially delay the input signal by one and two symbol intervals, respectively. -1 , C0, C1 scale the input signal, the input signal delayed once, and the input signal delayed twice. Then the multiplied signals are summed to form the output signal V.
[0014] One of the coefficients (C0) at or near the center of the delay chain is the "cursor" coefficient with the maximum value. In FIG. 2B , the cursor coefficient C0 is marked as being greater than 0.5. Typically, the cursor coefficient is normalized (set equal to unity), but it may take other values for implementation convenience. The illustrative pre-equalizer includes a pre-cursor coefficient (C -1 ) and a post-vernier coefficient (C1), both of which are set to be less than or equal to zero to boost high frequency signal components as needed to counteract the high frequency attenuation effects of the channel response.
[0015] FIG2C shows the (approximate) frequency response of CTLE 221. CTLE 221 can be designed to implement a transfer function having a polynomial ratio form in the Laplace transform domain. The numerator polynomial is ZThe Laplace transform has at least one zero at each frequency, so that the amplitude of the frequency response increases at a rate of 20 dB per decade for higher frequencies. To prevent excessive noise amplification at frequencies above the range of interest, the denominator polynomial of the Laplace transform has two zeros (they are "poles" because they are in the denominator), shown here as the frequency f P1 and f P2 , to cancel and then suppress signals above the Nyquist frequency (i.e., the symbol rate f of the transmitted signal). S Using standard filter design techniques, the zero and pole frequencies can be adjusted to tailor the CTLE response to provide the desired amount and range of high frequency boost.
[0016] The CTLE frequency response and coefficients of the pre-equalizer, receiver FFE, and DFE are preset based on the expected channel response and / or adapted during use to optimize performance. When the SerDes blocks are mass produced, the preset coefficients and / or initial settings for adaptation are selected based in part on a "worst case" scenario to ensure performance over as wide a range as possible. The resolution and range of programmable / adaptable parameter values are typically limited, in part because suboptimal equalizer performance can often be tolerated on higher quality channels, and in part because a smaller "configuration space" allows for more efficient hardware implementations and operation.
[0017] For these reasons, certain standards (such as the IEEE Standard for Ethernet, IEEE Std 802.3-2015) may set channel specifications and may accordingly restrict the range of allowable coefficient values, such as post-vernier coefficients. The authors have found that high-rate SerDes blocks designed with such channel assumptions and parameter restrictions exhibit insufficient performance on certain low-loss channels. Summary of the invention
[0018] Therefore, a technique for correcting such unexpectedly poor equalization performance over a reflective low-loss channel is provided herein. An illustrative short high-rate communication link includes: a serializer that provides a signal having a symbol rate of at least 10 GHz; and a deserializer that receives the signal via a printed circuit board ("PCB") trace that is coupled to the serializer at a first impedance mismatch and to the deserializer at a second impedance mismatch. At least one of the serializer and the deserializer includes an equalizer that attenuates frequency components of the signal at half the symbol rate relative to frequency components of the signal at one-third the symbol rate.
[0019] An illustrative serializer embodiment includes: a pre-equalizer that produces a filtered data stream having a symbol rate of at least 10 GB, the filtered data stream having frequency components at half the symbol rate, the pre-equalizer attenuating frequency components at half the symbol rate relative to frequency components at one-third the symbol rate; and a driver that converts the filtered data stream into a transmit signal.
[0020] An illustrative deserializer embodiment includes: at least one of a continuous time linear equalizer and a digital domain linear equalizer for converting a received signal into an equalized signal; a detector that converts the equalized signal into a symbol stream; and a controller that combines the equalized signal with the symbol stream or with a predetermined training sequence to evaluate channel reflection strength. If the channel reflection strength exceeds a threshold, the controller implements artificial attenuation of a frequency component of the received signal at half the symbol rate relative to a frequency component of the received signal at one third the symbol rate.
[0021] An illustrative communication method embodiment includes: coupling a communication channel to pass a signal from a serializer to a deserializer, the signal having a symbol rate of at least 10 GHz; characterizing reflection intensity in the communication channel; and configuring the serializer to attenuate or boost a frequency component of the signal at half the symbol rate relative to a frequency component at one-third the symbol rate based on whether the reflection intensity exceeds a predetermined threshold.
[0022] Each of the foregoing embodiments may be implemented individually or in combination, and may be implemented in any suitable combination with any one or more of the following optional features: 1. The length of the PCB trace is between 2.5 cm and 25 cm. 2. The first impedance mismatch is the connection of the PCB trace to a packaged integrated circuit chip including a serializer, and the second impedance mismatch is the connection of the PCB trace to a connector including a deserializer that mates with a standard-compliant network interface module (such as a quad small form factor pluggable double density ("QSFP-DD") module). 3. The equalizer is a digital domain pre-equalizer having a pre-vernier coefficient (C -1 ) of the magnitude of the post-vernier coefficient (C1). 4. The equalizer is a digital domain pre-equalizer having an alternating polarity sum ∑ greater than the pre-equalizer's pre-vernier coefficient i<0 (-1) i C i5. The deserializer includes a decision feedback equalizer that at least partially compensates for trailing intersymbol interference caused by the trailing vernier coefficients. 6. The equalizer is a continuous time linear equalizer. 7. The controller implements artificial attenuation by adjusting the pole frequencies of the continuous time linear equalizer. 8. The controller implements artificial attenuation by adjusting the trailing vernier coefficients of the digital domain linear equalizer. 9. The controller implements artificial attenuation by providing reverse channel information to the source of the received signal, the reverse channel information adjusting the trailing vernier coefficients of the pre-equalizer at the source. 10. The controller implements artificial attenuation by reducing the transmit driver bandwidth. 11. The deserializer is configured to attenuate or boost frequency components at half the symbol rate also based on whether the reflection intensity exceeds a predetermined threshold. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] 1A is a block diagram of a network interface port in a host device.
[0024] FIG. 1B is a block diagram of a data recovery and remodulation apparatus.
[0025] FIG. 2A is a block diagram of a prior art serializer-deserializer (“SerDes”) communication link.
[0026] FIG. 2B is a block diagram of a prior art digital domain pre-equalizer.
[0027] FIG. 2C is an approximate frequency response of a prior art continuous time linear equalizer (“CTLE”).
[0028] Figure 3 is a block diagram of the reflective low-loss channel model.
[0029] Figure 4 is a block diagram of an illustrative serializer-deserializer communication link embodiment.
[0030] Figure 5A-Figure 5B is an illustrative pre-equalizer configured to attenuate signal frequency components at half the symbol rate.
[0031] Figure 6 The approximate frequency response of an illustrative CTLE configured to attenuate signal frequency components at half the symbol rate is shown.
[0032] Figure 7 is an illustrative transmit driver embodiment with adjustable bandwidth.
[0033] Figure 8 is a second illustrative transmit driver embodiment with adjustable bandwidth.
[0034] Figure 9A-9Bis a graph comparing the responses of transmit drivers configured for different bandwidths.
[0035] Fig.10 is a flow chart of an illustrative communication method that takes into account reflective low-loss channels. DETAILED DESCRIPTION
[0036] Although specific embodiments are presented in the drawings and the following description, it should be kept in mind that they do not limit the present disclosure. Instead, they provide a basis for one of ordinary skill in the art to identify alternatives, equivalents, and modifications that are included within the scope of the appended claims.
[0037] While manufacturing processes for electronic devices vary widely, they generally involve the use of printed circuit boards ("PCBs") to interconnect packaged integrated circuit chips to each other and to various mechanisms for interacting with the user and environment of the electronic device. The chips have contacts that are soldered to corresponding contacts on the PCB. These connections can cause impedance mismatches for signals transmitted to or from the chip through the PCB traces. These impedance mismatches are typically insignificant over the frequency range used by most electronic devices, but serializer-deserializer (SerDes) communication links are pushing symbol rates to 10 GHz or higher, where such impedance mismatches can result in significant reflections and attenuation.
[0038] Figure 3 A model of an illustrative channel representing a short, low-loss path coupling a transmit chip 301 to a receive chip 303 is shown, such as a PCB trace connecting a network interface port to a packaged chip(s) of a host device. Block 302 represents the frequency domain transfer function of the PCB trace. Where the PCB trace is connected to the receive chip 303, an impedance mismatch 304 causes some of the signal energy to be reflected. Block 305 represents the frequency domain long-range reflection function R R (f). As represented by block 306, the reflection travels through the PCB trace in the reverse direction until it encounters an impedance mismatch 308 at the connection with the transmit chip 301. Block 307 represents the frequency domain near reflection function R N (f). The near reflections are combined with the original signal as represented by summing node 309. When combined, the model relates the received signal Y(f) to the transmitted signal X(f) as:
[0039]
[0040] Impedance mismatch can be characterized as parasitic capacitance or inductance, which preferentially reflects higher frequency signal components. PCB traces may typically have a length in the range of between 2.5 cm and 25 cm, where in the case of a low loss channel transfer function H(f), the intensity of the reflected signal component cannot be fully attenuated despite passing through the trace multiple times. The travel time associated with the reflections causes its ISI effect to fall outside the range of any feasible equalization filter, resulting in unexpectedly poor equalization performance, and a high-pass filter-type equalizer exacerbates this situation. As an example, the author has observed that a 5 cm PCB trace connecting a host device chip to a pluggable module connector may have an insertion loss of less than 6 dB at the Nyquist frequency, compared to an insertion loss of more than 14 dB at the Nyquist frequency for a 23 cm trace. Although shorter channels have less loss, the bit error rate ("BER") performance of the serializer-deserializer block may be significantly degraded due to strong reflections.
[0041] Figure 4 An illustrative serializer-deserializer communication link with a reflective low-loss channel 411 is shown. The high-pass filter pre-equalizer 202 of the prior art is replaced by a high-cut filter pre-equalizer 402 to artificially suppress rather than enhance the high-frequency components of the transmitted (and therefore received) signal. Alternatively or additionally, the high-pass filter CTLE 221 of the prior art is replaced by a high-cut filter CTLE 421, which also acts to artificially suppress rather than enhance the high-frequency components of the received signal. As another alternative or additional solution, the receiver FFE 222 can be configured to operate as a high-cut filter. As another alternative or additional solution, the transmit driver 203 can be replaced by a transmit driver 403 with a configurable bandwidth, which is reduced as needed to fully attenuate the high-frequency components of the signal.
[0042] Although such suppression actually makes the signal-to-noise ratio ("SNR") of the received signal worse, it attenuates reflection-induced ISI that would fall outside the equalization capabilities of the DFE 223, thereby reproducing the attenuation effect of a longer PCB trace. As with the longer trace, the attenuated high-frequency signal components are restored by the DFE, which operates by subtracting the estimated distortion based on the most recently received data value without amplifying the high-frequency noise.
[0043] Additionally, DFE 223 may be replaced with another type of detector having the capability of detecting symbols from a signal having attenuated frequency components at half the symbol rate, such as a Maximum Likelihood Sequence Detector (MLSD) or a Viterbi detector.
[0044] The channel characterization block 425 combines the symbol stream (generated by the DFE 223) with the equalized signal from the CTLE 421 and / or from the receiver FFE 222 to measure the characteristics of the channel. One suitable implementation of the channel characterization block can be found in co-pending application Ser. No. 16 / 691,523, entitled “A multi-function signal measurement circuit for ADC-based SerDes,” which is incorporated herein by reference. Other implementations are also contemplated, including conventional training controllers and filter coefficient adaptation modules. The channel characterization block 425 can determine whether the received signal includes an “echo” indicating a strong reflection. When the measured signal amplitude is only relatively large compared to the possible range of signal amplitudes allowed by the specification (e.g., greater than a threshold, such as 50% of the maximum signal amplitude), or the measured raw BER is relatively high compared to the expected BER (e.g., greater than a threshold, such as 10 -4 , 10 -5 or 10 -6 ), the reflection may be considered “strong”. When such strong reflections are detected, the channel characterization block 425 may switch the pre-equalizer, CTLE, or receiver FFE from high-pass filter (high frequency boost) behavior to high-cut filter (high frequency attenuation) behavior.
[0045] Figure 5A is a block diagram of an illustrative 3-tap FFE (which can be used as a digital domain pre-equalizer or receiver FFE) configured as a high cutoff filter. Although the transfer function of the FFE can be manipulated in various ways, we believe that by setting the post-vernier tap coefficient (C1) to a value greater than (negative of) the pre-vernier tap coefficient (C -1 ) to achieve a high cutoff filter characteristic, since the DFE can easily compensate for post-vernier ISI introduced in this way. Figure 5B The 5-tap FFE shown can be achieved by setting the post-vernier tap coefficient (C1) to a value greater than the alternating polarity sum of the pre-vernier coefficients of the pre-equalizer: i<0 (-1) i C i , to provide a high cutoff filter characteristic. .Similarly, the DFE can easily compensate for post-vernier ISI. We note here that this can be achieved by making the post-vernier tap coefficients smaller than |C -1 Changing the value of | (or less than the sum of alternating polarities for longer filters) to a value greater than this amount causes the illustrated FFE to switch between high-pass filter (high frequency boost) and high-cut filter (high frequency attenuation) behavior.
[0046] Figure 6 is a graph of the (approximate) frequency response of the high cutoff filter CTLE 421. In order to attenuate frequencies close to and above the Nyquist frequency (i.e., the symbol rate f S The frequency response is at the Nyquist frequency f S / 2 At or below the Nyquist frequency f S / 2 Has one or more poles. The frequency response is below the first pole frequency f P is essentially flat when , and if only one pole is used, the response is P It has a negative -20dB / dec slope, which is used to suppress the Nyquist frequency f S / 2 The signal frequency components at and above the Nyquist frequency f S / 2 signal frequency components. CTLE implementations with tunable zero and pole frequencies can be found in the open literature. For example, if the high-pass filter CTLE221 uses a tunable zero frequency f Z To achieve the response of Figure 2B, f Z is tuned to be substantially aligned with one of the pole frequencies (e.g., f P2 ) overlap, causing their effects to cancel out and produce Figure 6 Response shown.
[0047] Figure 7 An illustrative transmit driver 403 with configurable bandwidth is shown. Driver 403 includes an input amplifier 702 (also referred to as a "pre-driver") coupled to an output amplifier 704 (also referred to as a "final driver"). To controllably reduce the bandwidth of driver 403, the electrical connection between the (current-limited) input amplifier 702 and the output amplifier 704 is loaded with an adjustable capacitor 706. Capacitor 706 acts as a low-pass filter, reducing the maximum slew rate of the signal. Increasing the capacitance further reduces the slew rate of the signal, thereby reducing the bandwidth of the low-pass filter by reducing the cutoff frequency.
[0048] Figure 8 An alternative embodiment of a transmit driver 403 with configurable bandwidth is shown. It comprises a plurality of input amplifiers 702A-702D arranged in parallel for driving an output amplifier 704. Figure 8 Rather than changing the input capacitance of the final amplifier 704, an embodiment of the present invention disables one or more of the input amplifiers to further limit the current supplied to the final amplifier 704, thereby reducing the maximum slew rate of the signal. The transmit driver bandwidth is determined by the number (and size) of enabled input amplifiers, where fewer enabled input amplifiers lower the cutoff frequency and thereby reduce the bandwidth of the low pass filter.
[0049] Fig.9Ais a graph comparing the step response of a transmit driver configured with a nominal bandwidth (dashed line) and a transmit driver configured with a reduced bandwidth (solid line) to attenuate frequency components at half the symbol rate. The corresponding rise times Tr (as measured by the delay between the 20% and 80% levels of the step response) are 0.35 and 0.70 symbol intervals, respectively. The nominal bandwidth configuration achieves a complete zero-to-one symbol transition in a single interval, while the reduced bandwidth design doubles the rise time and significantly increases the time required for a complete transition.
[0050] Fig. 9B is a graph comparing the corresponding frequency domain gain curves for the transmit driver with nominal and reduced bandwidth configurations. At the Nyquist frequency, the nominal configuration has less than 2dB of attenuation, while the reduced bandwidth configuration provides over 6dB of attenuation.
[0051] Fig.10 1001, the channel response is evaluated to measure the strength of the reflections. For example, the channel characterization block 425 can determine the presence of strong echoes in the received signal by measuring the post-vernier ISI outside the range of the DFE, possibly in conjunction with the adaptation of the filter parameters to optimize performance. The adaptation of the filter parameters can be constrained as needed to provide any high frequency suppression determined to be necessary.
[0052] In block 1002, the reflection intensity is evaluated, for example, by comparing the ISI to a predetermined threshold representing the desired BER. If the reflection intensity exceeds the threshold, then in block 1003, the method artificially increases high frequency attenuation, for example, by increasing the value of the post-vernier tap coefficient (C1), tuning the pole and zero positions of the CTLE, and / or reducing the bandwidth of the transmit driver. Block 1003 can be iteratively performed until the reflection intensity is sufficiently suppressed by gradually increasing the post-vernier tap coefficient, gradually adjusting the CTLE parameters, and / or gradually reducing the transmit driver bandwidth. Once the reflection intensity drops below the threshold, the method continues in block 1004 to store the filter parameters in firmware or otherwise finalize the configuration of the serializer-deserializer equalizer.
[0053] The above embodiments can solve the problem of serializer-deserializer performance on reflective, low-loss channels by adding artificial insertion loss, which attenuates the high-frequency components of the signal and the high-frequency components of the reflected noise. The attenuated high-frequency components of the signal can be restored by the DFE. This attenuation can be achieved by modifying the existing equalizer (e.g., pre-equalizer, transmit driver, CTLE, or receiver FFE) to achieve high-cut filter behavior.
[0054] Once the above disclosure is fully understood, numerous alternatives, equivalents, and modifications will become apparent to those skilled in the art. For example, a pre-equalizer or receiver FFE may have additional front and rear vernier taps, wherein the modification of the response is achieved using any known filter design techniques. It is intended that the claims be interpreted as encompassing all of these alternatives, equivalents, and modifications that are included within the scope of the appended claims.
Claims
1. A circuit for a short high-rate communication link, the communication link comprising: a serializer providing a signal at a symbol rate; as well as a deserializer that receives the signal via a printed circuit board (PCB) trace that is coupled to the serializer at a first impedance mismatch and to the deserializer at a second impedance mismatch, the deserializer being operative to convert the signal into a symbol stream, At least one of the serializer and the deserializer includes an equalizer that attenuates frequency components of the signal at half the symbol rate relative to frequency components of the signal at one third the symbol rate.
2. The circuit according to claim 1, characterized in that The length of the PCB trace is between 2.5 centimeters and 25 centimeters, and wherein the first impedance mismatch is a connection of the PCB trace to a packaged integrated circuit chip including the serializer, and the second impedance mismatch is a connection of the PCB trace to a connector that is mated to a network interface module and includes the deserializer.
3. The circuit according to claim 1, characterized in that The equalizer is a digital domain pre-equalizer having a pre-cursor coefficient C greater than -1 The magnitude of the rear cursor coefficient C1.
4. The circuit according to claim 1, characterized in that The equalizer is a digital domain pre-equalizer having an alternating polarity sum ∑ i<0 (-1) i C i The rear cursor coefficient C1.
5. The circuit according to claim 4, characterized in that The deserializer includes a decision feedback equalizer that at least partially compensates for post-intersymbol interference caused by the post-vernier coefficients.
6. The circuit according to claim 1, characterized in that The equalizer is a continuous-time linear equalizer.
7. A serializer, comprising: a pre-equalizer that produces a filtered data stream having a symbol rate, the filtered data stream having frequency components at half the symbol rate, the pre-equalizer attenuating frequency components at half the symbol rate relative to frequency components at one third the symbol rate; as well as A driver converts the filtered data stream into a transmit signal.
8. The serializer according to claim 7, wherein: The pre-equalizer is a digital domain filter having a pre-vernier coefficient C greater than -1 The magnitude of the rear cursor coefficient C1.
9. The serializer according to claim 7, wherein: The pre-equalizer is a digital domain filter having an alternating polarity sum ∑ of pre-vernier coefficients greater than the pre-equalizer. i<0 (-1) i C i The rear cursor coefficient C1.
10. A deserializer, the deserializer comprising: at least one of a continuous time linear equalizer and a digital domain linear equalizer for converting a received signal having a symbol rate into an equalized signal wherein frequency components at half the symbol rate are attenuated relative to frequency components at one third the symbol rate; a decision feedback equalizer that converts the equalized signal into a symbol stream; as well as A controller that combines the equalized signal with the symbol stream or with a predetermined training sequence to evaluate channel reflection strength and, if the channel reflection strength exceeds a threshold, achieves artificial attenuation of frequency components of the received signal at half the symbol rate relative to frequency components of the received signal at one third the symbol rate.
11. The deserializer according to claim 10, wherein: The controller implements the artificial attenuation by adjusting the pole frequency of the continuous-time linear equalizer.
12. The deserializer according to claim 10, wherein: The controller implements the artificial attenuation by adjusting the post cursor coefficients of the digital domain linear equalizer.
13. The deserializer according to claim 10, wherein: The controller implements the artificial attenuation by providing backchannel information to a source of the received signal, the backchannel information adjusting post-verbose coefficients of a pre-equalizer at the source.
14. The deserializer according to claim 10, wherein: The controller implements the artificial attenuation by reducing the bandwidth of the transmit driver.
15. A communication method, the communication method comprising: providing a communication channel to transmit a signal from the serializer to the deserializer, the signal having a symbol rate; characterizing a reflection intensity in the communication channel; as well as The serializer is configured to attenuate or boost a frequency component of the signal at half the symbol rate relative to a frequency component at one third the symbol rate based on whether the reflection intensity exceeds a predetermined threshold.
16. The communication method of claim 15, further comprising configuring the deserializer to attenuate or boost the frequency component at half the symbol rate also based on whether the reflection intensity exceeds the predetermined threshold.
17. The communication method according to claim 16, characterized in that: The configuring the deserializer includes reducing a pole frequency of a continuous-time linear equalizer if the reflection intensity exceeds the threshold.
18. The communication method according to claim 17, characterized in that: Configuring the deserializer includes making the post-vernier coefficients of the digital domain linear equalizer greater than the alternating polarity sum of the pre-vernier coefficients of the equalizer: ∑ i<0 (-1) i C i .
19. The communication method according to claim 15, characterized in that: Configuring the serializer includes making the post-vernier coefficients of the digital domain linear equalizer greater than the alternating polarity sum of the pre-vernier coefficients of the equalizer: ∑ i<0 (-1) i C i .
20. The communication method according to claim 15, characterized in that: The configuring the serializer includes adjusting a bandwidth of a transmit driver.
21. A communication method, the communication method comprising: receiving a signal having a symbol rate from a serializer via a communication channel; characterizing a reflection intensity in the communication channel; as well as An equalizer in the deserializer is configured to attenuate or boost a frequency component of the signal at half the symbol rate relative to a frequency component at one third the symbol rate based on whether the reflection intensity exceeds a predetermined threshold.
22. The communication method according to claim 21, characterized in that: Configuring the equalizer includes reducing a pole frequency of a continuous-time linear equalizer if the reflection intensity exceeds the threshold.
23. The communication method according to claim 21, characterized in that: The configuration of the equalizer includes making the post cursor coefficient C1 of the digital domain linear equalizer greater than the pre cursor coefficient C -1 The amplitude.
24. The communication method according to claim 21, characterized in that: The configuration of the equalizer includes making the post cursor coefficient C1 of the digital domain linear equalizer greater than the alternating polarity sum of the pre cursor coefficients of the equalizer: ∑ i<0 (-1) i C i .
