Connecting interface suitable for bare chip-to-bare chip and wire channel connecting method thereof
By introducing multiple use lines and spare lines into the die-to-die connection interface, and using the controller to monitor and switch the lines, the problem of attenuation and damage of the bridge substrate connection lines is solved, achieving more efficiency and reliability data transmission.
Patent Information
- Application Number
- CN202310898559.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-21
- Publication Date
- 2025-01-21
AI Technical Summary
In the connection technology of the die-to-die (D2D) technology, the connection channel of the bridge substrate may experience transmission attenuation and damage after the transmission system is running for a period of time, affecting the transmission quality and causing data transmission interruption.
A connection interface is designed, which includes multiple first use lines, first backup lines, receiving lines and a controller. The controller monitors the data transmission status of each first used line channel, and when detecting an unhealthy line channel, changes the receiving line channel to the first backup line channel to ensure the quality and continuity of data transmission.
By monitoring and switching to the backup line, the quality of data transmission is improved, data transmission interruptions caused by damage to the connection line are avoided, and high-speed transmission between dies is maintained.
Smart Images

Figure CN119342789A_ABST
Abstract
Description
Technical Field
[0001] The present invention provides a connection interface and a line connection method thereof, in particular, a high-speed connection interface and a line connection method thereof suitable for using a die-to-die (D2D) connection technology. Background Art
[0002] With the rapid evolution of high performance computing (HPC) and artificial intelligence (AI) technologies, die-to-die (D2D), which can package a large circuit into multiple small modules to form a multi-chip module, has become one of the important technologies for current semiconductor chip packaging. Generally speaking, high-speed transmission is performed between two bare chips through the connecting lines of the bridge substrate. Then, usually after the transmission system has been running for a period of time, the connecting lines of the bridge substrate may have problems such as transmission attenuation, which affects the transmission quality. In addition, when the connecting lines of the bridge substrate are damaged, high-speed transmission between the two bare chips cannot be performed and must be interrupted. Summary of the invention
[0003] In one embodiment, the present invention provides a connection interface suitable for die-to-die. The connection interface includes a plurality of first use channels, a plurality of first backup channels, a plurality of receiving channels, and a controller. The controller is used to connect each receiving channel to one of the plurality of first use channels and monitor each first use channel respectively. When the first use channel is detected to be an unhealthy channel, the controller changes the receiving channel connected to the unhealthy channel to one of the plurality of first backup channels.
[0004] In one embodiment, the present invention provides a die-to-die channel connection method. The channel connection method includes: connecting a plurality of receiving channels of a connection interface to a plurality of first use channels of the connection interface, wherein each receiving channel is connected to one of the first use channels; monitoring each first use channel respectively; and when the first use channel is detected to be an unhealthy channel, connecting the receiving channel connected to the unhealthy channel to one of the plurality of first backup channels of the connection interface.
[0005] In summary, any embodiment of the bare-die connection interface and its line connection method can monitor the data transmission status of the first use line connected to each receiving line, and when a first use line is detected to be an unhealthy line, the receiving line connected to the unhealthy line is changed to a first backup line. In this way, the data can be changed from the first use line with a poor data transmission status to the first backup line with a better data transmission status, so that the data transmission quality can be improved. In addition, due to the monitoring of the data transmission status of each first use line, the processing module of the connection interface can find out the connection line that is about to be damaged in the bridge substrate in advance, and use another connection line and the first backup line for transmission, so that the data transmission between the two bare chips can not be interrupted due to the damage of the connection line, and the high-speed transmission between the two bare chips can be maintained.
[0006] The detailed features and advantages of the present invention are described in detail in the implementation mode below, and the content is sufficient to enable anyone familiar with the relevant skills to understand the technical content of the present invention and implement it accordingly. Moreover, according to the content disclosed in this specification, the scope of the patent application and the drawings, any person familiar with the relevant skills can easily understand the relevant purposes and advantages of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0007] Figure 1 is a block diagram of an embodiment of a transmission system.
[0008] Figure 2 is a block diagram of an embodiment of a transmission system.
[0009] Figure 3 The present invention is a flow chart of an embodiment of a line connection method.
[0010] Figure 4 FIG. 1 is a schematic diagram of an embodiment of sampling clock and training data.
[0011] Figure 5 FIG. 1 is a schematic diagram of an embodiment of sampling clock and training data.
[0012] Figure 6 FIG. 1 is a schematic diagram of an embodiment of sampling clock and training data.
[0013] Figure 7 FIG. 1 is a schematic diagram of an embodiment of sampling clock and training data.
[0014] Figure 8 A schematic diagram of an embodiment of a transmission system.
[0015] Fig. 9 A schematic diagram of an embodiment of a transmission system.
[0016] Fig.10A schematic diagram of an embodiment of a transmission system.
[0017] Fig.11 A partial flow chart of an embodiment of a line connection method.
[0018] Wherein, the reference numerals are:
[0019] 1:Transmission system
[0020] 100,200: Bare chip
[0021] 110,210: Connection interface
[0022] 111: Controller
[0023] 1111: Connection control module
[0024] 1112: Processing module
[0025] 120,220: System circuit
[0026] 211: Controller
[0027] 2111: Link control module
[0028] 2112: Processing module
[0029] 300: Bridge substrate
[0030] 1A1~1AX: First line of use
[0031] 1B1~1BY: First backup line
[0032] 1R1~1RX: receiving line
[0033] 2A1~2AX: Second line for use
[0034] 2B1~2BY: Second backup line
[0035] 2T1~2TX: Output line
[0036] CLK: sampling clock
[0037] 1D1~1DZ: signal contact
[0038] 2D1~2DZ:Signal contact
[0039] E1: Training Data
[0040] G1: Left indicator
[0041] G11-G1M: Receiving channel group
[0042] G21-G2M: Output channel group
[0043] G2: Right indicator
[0044] H1: Output change request
[0045] H2: Enter the replacement command
[0046] L1~LZ: Connecting line
[0047] LT1: Communication Line
[0048] M1~MX: Multiplexer
[0049] M21~M24: 2 to 1 multiplexer
[0050] M31~M34: 3 to 1 multiplexer
[0051] TO: Output data
[0052] W1: Monitoring window
[0053] S10~S80: Steps DETAILED DESCRIPTION
[0054] In order to make the above-mentioned purposes, features and advantages of the embodiments of the present invention more obvious and understandable, the following is a detailed description with reference to the accompanying drawings.
[0055] It should be understood that the word "comprising" used in this specification is used to indicate the existence of specific technical features, values, method steps, operation processing, elements and / or components, but does not exclude the addition of more technical features, values, method steps, operation processing, elements, components, or any combination of the above.
[0056] Figure 1 FIG. 1 is a block diagram of an embodiment of the transmission system 1. Figure 1 The transmission system 1 includes at least two bare chips (hereinafter referred to as bare chip 100 and bare chip 200 respectively), and the bare chip 100 is coupled to the bare chip 200. The bare chip 100 includes at least one connection interface 110 (hereinafter one is taken as an example, but the number is not limited thereto) and a system circuit 120, and the system circuit 120 is coupled to the connection interface 110. The bare chip 200 includes at least one connection interface 210 (hereinafter one is taken as an example, but the number is not limited thereto) and a system circuit 220, and the system circuit 220 is coupled to the connection interface 210.
[0057] In some embodiments, the transmission system 1 further includes a bridge substrate 300. The die 100 and the die 200 are disposed on the bridge substrate 300, and the die 100 can be coupled to the die 200 through the bridge substrate 300. Specifically, the bridge substrate 300 includes a plurality of connection lines L1-LZ, where Z is a positive integer greater than 1. The connection interface 110 of the die 100 and the connection interface 210 of the die 200 both adopt a die-to-die (D2D) connection technology, and the connection lines L1-LZ of the bridge substrate 300 are connected between the connection interface 110 and the connection interface 210, so that the system circuit 120 of the die 100 and the system circuit 220 of the die 200 can communicate with each other through the connection interface 110, the connection interface 210 and the connection lines L1-LZ of the bridge substrate 300.
[0058] Figure 2 FIG. 1 is a block diagram of an embodiment of the transmission system 1. Figure 1 to Figure 2 In some embodiments, the connection interface 110 includes a plurality of first use channels 1A1-1AX, a plurality of first backup channels 1B1-1BY, a plurality of receiving channels 1R1-1RX, and a controller 111. The connection interface 210 includes a plurality of second use channels 2A1-2AX, a plurality of second backup channels 2B1-2BY, a plurality of output channels 2T1-2TX, and a controller 211. X and Y are positive integers greater than 1, respectively.
[0059] In some embodiments, the connection interface 110 may further include a plurality of signal contacts 1D1-1DZ, and the connection interface 210 may further include a plurality of signal contacts 2D1-2DZ, wherein Z is a positive integer greater than 1, and Z is the sum of X and Y. In addition, the controller 111 may include a connection control module 1111 and a processing module 1112, and the controller 211 may include a connection control module 2111 and a processing module 2112.
[0060] In some embodiments, in the connection interface 110, a plurality of receiving channels 1R1-1RX are coupled between the connection control module 1111 of the controller 111 and the system circuit 120. A plurality of first use channels 1A1-1AX and a plurality of first backup channels 1B1-1BY are coupled between the connection control module 1111 of the controller 111 and a plurality of signal contacts 1D1-1DZ. In addition, the processing module 1112 of the controller 111 is coupled to the connection control module 1111 and the receiving channels 1R1-1RX. In some embodiments, in the connection interface 210, a plurality of output channels 2T1-2TX are coupled between the connection control module 2111 of the controller 211 and the system circuit 220. A plurality of second use channels 2A1-2AX and a plurality of second backup channels 2B1-2BY are coupled between the connection control module 2111 of the controller 211 and a plurality of signal contacts 2D1-2DZ. In addition, the processing module 2112 of the controller 211 is coupled to the connection control module 2111 .
[0061] In some embodiments, the first active lanes 1A1-1AX and the first backup lanes 1B1-1BY of the connection interface 110 are coupled to the signal contacts 1D1-1DZ in a one-to-one manner. The second active lanes 2A1-2AX and the second backup lanes 2B1-2BY of the connection interface 210 are coupled to the signal contacts 2D1-2DZ in a one-to-one manner. In addition, the connection lanes L1-LZ of the bridge substrate 300 are also coupled to the signal contacts 1D1-1DZ of the connection interface 110 and the signal contacts 2D1-2DZ of the connection interface 210 in a one-to-one manner.
[0062] In some implementations, in an initial state, the controller 111 of the connection interface 110 may configure all the lanes coupled to the signal contacts 1D1 - 1DZ in a manner that one first used lane is spaced apart from one first spare lane. Figure 2 As shown, the line coupled to the signal contact 1D1 is configured as the first used line 1A1, the line coupled to the signal contact 1D2 is configured as the first spare line 1B1, the line coupled to the signal contact 1D3 is configured as the first used line 1A2, and so on. Accordingly, the controller 211 of the connection interface 210 also configures all the lines coupled to the signal contacts 2D1-2DZ in a manner that one second used line is spaced by one second spare line. For example, Figure 2As shown, the line coupled to the signal contact 2D1 is configured as the second used line 2A1, the line coupled to the signal contact 2D2 is configured as the second backup line 2B1, the line coupled to the signal contact 2D3 is configured as the second used line 2A2, and so on. Therefore, the first used line 1A1 of the connection interface 210 is coupled to the second used line 2A1 of the connection interface 210 through the connection line L1, and the first backup line 1B1 of the connection interface 210 is coupled to the second backup line 2B1 of the connection interface 210 through the connection line L2, and so on. However, the present invention is not limited to this.
[0063] In other embodiments, the controller 111 of the connection interface 110 may also configure all the channels coupled to the signal contacts 1D1-1DZ in a manner that one spare channel is spaced between every two channels in use. Correspondingly, the controller 211 of the connection interface 210 may also configure all the channels coupled to the signal contacts 2D1-2DZ in a manner that one spare channel is spaced between every two channels in use. For example, the controller 111 configures the channel coupled to the signal contact 1D1 as the first channel in use 1A1, configures the channel coupled to the signal contact 1D2 as the first channel in use 1A2, configures the channel coupled to the signal contact 1D3 as the first spare channel 1B1, and so on. Accordingly, the controller 211 configures the line coupled to the signal contact 2D1 as the second used line 2A1, configures the line coupled to the signal contact 2D2 as the second used line 2A2, configures the line coupled to the signal contact 2D3 as the second backup line 2B1, and so on. Therefore, the first used line 1A1 of the connection interface 210 is coupled to the second used line 2A1 of the connection interface 210 through the connection line L1, the first used line 1A2 of the connection interface 210 is coupled to the second used line 2A2 of the connection interface 210 through the connection line L2, and the first backup line 1B1 of the connection interface 210 is coupled to the second backup line 2B1 of the connection interface 210 through the connection line L3, and so on.
[0064] Figure 3 FIG. 1 is a flow chart of an embodiment of a line connection method. Figures 1 to 3. The connection interface 110 and / or the connection interface 210 can execute the line connection method of any embodiment. In some embodiments, before the formal transmission, the controller 111 can perform a correction training (step S10) on each of the first use line channels 1A1-1AX and the first backup line channels 1B1-1BY, so that a positive edge and / or a negative edge of a sampling clock CLK can hit (be located at) the center position of each data. Specifically, taking the positive edge of the sampling clock CLK as an example to perform correction training on the first use line channel 1A1, a training data E1 with a known data pattern, such as "010", can be transmitted on the first use line channel 1A1 first, and at this time, the positive edge of the sampling clock CLK hits the data "1" in the training data E1, such as Figure 4 First, the processing module 1112 of the controller 111 can respectively set a left index G1 and a right index G2 on the left and right sides of the positive edge of the sampling clock CLK, and respectively adjust the left index G1 and the right index G2 until the current data "1" boundary, such as Figure 5 As shown. The processing module 1112 can determine whether the data has been adjusted to the left boundary of the data "1" according to whether the data sampled at the left pointer G1 is "0", and determine whether the data has been adjusted to the right boundary of the data "1" according to whether the data sampled at the right pointer G2 is "0". Afterwards, the processing module 1112 can know how to adjust the positive edge of the sampling clock CLK according to the distance between the left pointer G1 and the right pointer G2, so that the positive edge of the sampling clock CLK can hit (be located at) the center position of the data "1", as shown. Figure 6 As shown. Since the calibration training of the remaining first use channels 1A2-1AX and the first backup channels 1B1-1BY can refer to the calibration training of the first use channel 1A1, it will not be described in detail. Similarly, before the formal transmission, the controller 211 can also perform a calibration training on each of the second use channels 2A1-2AX and the second backup channels 2B1-2BY, so that a positive edge and / or a negative edge of a sampling clock CLK can hit (be located at) the center position of each data. Here, since the calibration training of the second use channels 2A1-2AX and the second backup channels 2B1-2BY can refer to the above, it will not be described in detail.
[0065] In some embodiments, in an initial state, each of the receiving channels 1R1-1RX of the connection interface 110 corresponds to one of the first use channels 1A1-1AX. For example, the receiving channel 1R1 corresponds to the first use channel 1A1, the receiving channel 1R2 may correspond to the first use channel 1A2, the receiving channel 1R3 may correspond to the first use channel 1A3, and so on. In addition, the processing module 1112 of the controller 111 causes the connection control module 1111 to connect the receiving channels 1R1-1RX to the corresponding first use channels 1A1-1AX in a one-to-one manner (step S20). Similarly, each of the output channels 2T1-2TX of the connection interface 210 corresponds to one of the second use channels 2A1-2AX. For example, the output channel 2T1 corresponds to the second used channel 2A1, the output channel 2T2 may correspond to the second used channel 2A2, the output channel 2T3 may correspond to the second used channel 2A3, and so on. In addition, the processing module 2112 of the controller 211 causes the connection control module 2111 to connect the output channels 2T1-2TX to the corresponding second used channels 2A1-2AX in a one-to-one manner.
[0066] In some embodiments, when transmission starts, the system circuit 220 of the bare chip 200 may generate a plurality of output data TO and output them through the output lines 2T1-2TX and the second use lines 2A1-2AX. Here, the plurality of output data TO output by the bare chip 200 may be transmitted to the bare chip 100 through the connection lines coupled to the second use lines 2A1-2AX in the bridge substrate 300. In addition, the plurality of output data TO transmitted to the bare chip 100 through the connection lines coupled to the second use lines 2A1-2AX may be sequentially transmitted to the system circuit 120 through the first use lines 1A1-1AX and the receiving lines 1R1-1RX.
[0067] In some embodiments, after the transmission is performed, the processing module 1112 of the controller 111 in the bare chip 100 monitors the data transmission status of each of the first use channels 1A1-1AX (step S30) to determine whether each of the first use channels 1A1-1AX is an unhealthy channel (i.e., a channel with a poor data transmission status) according to the data transmission status of each of the first use channels 1A1-1AX. In some implementations, since the first use channels 1A1-1AX are connected to the receiving channels 1R1-1RX at this time, the processing module 1112 can monitor the data transmission status of each of the first use channels 1A1-1AX by monitoring the data transmission status of each of the receiving channels 1R1-1RX.
[0068] In some implementations, the processing module 1112 may monitor all the first use lanes 1A1 - 1AX simultaneously. In other implementations, the processing module 1112 may monitor all the first use lanes 1A1 - 1AX sequentially in a predetermined order.
[0069] In some embodiments, taking monitoring of the first use channel 1A1 as an example, the processing module 1112 can monitor the data transmission status of the first use channel 1A1 through a monitoring window W1. The width of the monitoring window W1 can be obtained in the previous calibration training. Specifically, in the previous calibration training, the processing module 1112 can adjust the positive edge of the sampling clock CLK to hit (be located at) the center position of the data "1", and then adjust the left pointer G1 and the right pointer G2 to the boundary of the data "1" with the positive edge of the sampling clock CLK as the center to obtain the data eye. Afterwards, the processing module 1112 adjusts the positions of the left pointer G1 and the right pointer G2 according to the data eye with the positive edge of the sampling clock CLK as the center to set the width of the monitoring window W1, such as Figure 7 For example, when the data eye obtained in the previous calibration training is 100 picoseconds (ps), the processing module 1112 can adjust the left index G1 and the right index G2 to positions 20 picoseconds (i.e., 1 / 4 of the data eye) away from the positive edge of the sampling clock CLK, so that the width of the monitoring window W1 is 40 picoseconds (i.e., the distance between the left index G1 and the right index G2).
[0070] During monitoring, the processing module 1112 can use the left indicator G1, the sampling clock CLK and the right indicator G2 to sample the data on the first use channel 1A1 respectively, and determine the data transmission status of the first use channel 1A1 based on whether the data sampled by the left indicator G1 is the same as the data sampled by the sampling clock CLK and whether the data sampled by the sampling clock CLK is the same as the data sampled by the right indicator G2.
[0071] Specifically, the processing module 1112 can use an exclusive OR gate to determine whether the data sampled by the left indicator G1 is the same as the data sampled by the sampling clock CLK, and use another exclusive OR gate to determine whether the data sampled by the sampling clock CLK is the same as the data sampled by the right indicator G2. When the outputs of the two exclusive OR gates are both logic "0", it indicates that the data transmission status of the first use line 1A1 is good. When the output of any exclusive OR gate is logic "1", it indicates that the data transmission status on the first use line 1A1 has data deviation or narrowing, and the processing module 1112 will determine that the first use line 1A1 is an unhealthy line. Among them, if the exclusive OR gate for determining whether the data sampled by the left pointer G1 is the same as the data sampled by the sampling clock CLK outputs "0" and the exclusive OR gate for determining whether the data sampled by the sampling clock CLK is the same as the data sampled by the right pointer G2 outputs "1", it indicates that data forward shift occurs; if the exclusive OR gate for determining whether the data sampled by the left pointer G1 is the same as the data sampled by the sampling clock CLK outputs "1" and the exclusive OR gate for determining whether the data sampled by the sampling clock CLK is the same as the data sampled by the right pointer G2 outputs "0", it indicates that data backward shift occurs; and if the exclusive OR gate for determining whether the data sampled by the left pointer G1 is the same as the data sampled by the sampling clock CLK outputs "1" and the exclusive OR gate for determining whether the data sampled by the sampling clock CLK is the same as the data sampled by the right pointer G2 outputs "1", it indicates that data narrowing occurs.
[0072] It should be noted that the data transmission status of each first use channel 1A1-1AX at this time also represents the data transmission status of the connection channel connected to each first use channel 1A1-1AX. Therefore, when the first use channel, such as the first use channel 1A1, is determined to be an unhealthy channel, it means that the connection channel L1 connected to the first use channel 1A1 is also an unhealthy channel (that is, the connection channel L1 may have problems such as transmission attenuation or even damage). Since the monitoring of the remaining first use channels 1A2-1AX can refer to the monitoring of the first use channel 1A1, it will not be repeated here.
[0073] In some embodiments, when the processing module 1112 of the connection interface 110 detects that the first use channel, for example, the first use channel 1A1 is an unhealthy channel (i.e., a channel with poor data transmission status), the first use channel 1A1 may be set as an unhealthy channel (for example, but not limited to, setting the status of the first use channel 1A1 to an unhealthy state in the channel status table), and causes the connection control module 1111 to change the receiving channel connected to the unhealthy channel (i.e., the receiving channel 1R1 connected to the first use channel 1A1) to one of the first backup channels 1B1-1BY (step S40). In this way, the data (e.g., output data TO or input data described later) can be transferred from the first use channel 1A1 with poor data transmission status to the first backup channel with better data transmission status (i.e., the output data TO of the bare chip 200 or the input data described later can be transferred from the connection channel L1 with poor data transmission status to the connection channel connected to the first backup channel with better transmission status to the bare chip 100), so that the data transmission quality can be improved. In addition, due to the monitoring of the data transmission status of the first use channel, the connection channel that is about to be damaged can be discovered in advance, and another connection channel and the first backup channel can be used for transmission, so that the data transmission between the bare chip 100 and the bare chip 200 can not be interrupted due to the damage of the connection channel, and the high-speed transmission of the transmission system 1 can be maintained. In some embodiments, the transmission speed can reach at least 12G bps.
[0074] In some embodiments, each of the receiving lanes 1R1-1RX of the connection interface 110 may correspond to at least one dedicated first backup lane, and the first backup lanes to which each of the receiving lanes 1R1-1RX corresponds are different from each other. For example, when each receiving lane corresponds to one first backup lane, for example, the receiving lane 1R1 corresponds to the first backup lane 1B1, the receiving lane 1R2 corresponds to the first backup lane 1B2, the receiving lane 1R3 corresponds to the first backup lane 1B3, and so on. In contrast, each of the output lanes 2T1-2TX of the connection interface 210 may correspond to at least one dedicated second backup lane, and the second backup lanes to which each of the output lanes 2T1-2TX corresponds are also different from each other. Taking each output channel corresponding to a second backup channel as an example, for example, output channel 2T1 corresponds to second backup channel 2B1, output channel 2T2 corresponds to second backup channel 2B2, output channel 2T3 corresponds to second backup channel 2B3, and so on.
[0075] In some embodiments, the connection control module 2111 of the connection interface 210 can directly connect each of the output channels 2T1-2TX to the corresponding second use channel and at least one dedicated second backup channel. For example, when each output channel corresponds to one dedicated second backup channel, Figure 8 As shown, the output channel 2T1 can be connected to the second use channel 2A1 and the second backup channel 2B1, the output channel 2T2 can be connected to the second use channel 2A2 and the second backup channel 2B2, the output channel 2T3 corresponds to the second backup channel 2B3, and so on. In other words, the output data TO on each output channel of the connection interface 210 has two transmission channels, one of which is transmitted to the first use channel connected to the connection interface 110 through the second use channel connected to the output channel and then through the connected connection channel in the bridge substrate 300; and the other is transmitted to the first backup channel connected to the connection interface 110 through the second backup channel connected to the output channel and then through the connected connection channel in the bridge substrate 300. In other words, each first use channel of the connection interface 110 and its exclusive first backup channel will receive the output data TO outputted via the same output channel in the connection interface 210.
[0076] In addition, the connection control module 1111 in the connection interface 110 may include a plurality of multiplexers M1-MX, and each of the multiplexers M1-MX corresponds to one of the receiving channels 1R1-1RX. Each multiplexer is coupled between the corresponding receiving channel and the first used channel and the first spare channel corresponding to the receiving channel. Figure 8 As shown, the two input ends of the multiplexer M1 are respectively coupled to the first used line 1A1 and the first spare line 1B1, and the output end of the multiplexer M1 is coupled to the receiving line 1R1; the two input ends of the multiplexer M2 are respectively coupled to the first used line 1A2 and the first spare line 1B2, and the output end of the multiplexer M2 is coupled to the receiving line 1R2; and so on. Here, in the initial state, the processing module 1112 causes the multiplexer M1-MX to connect the receiving line 1R1-1RX to the first used line 1A1-1AX. In addition, the processing module 1112 causes the multiplexer connected to the unhealthy line to change the receiving line to its corresponding first spare line in step S40. For example, when the unhealthy channel is the first use channel 1A3, the processing module 1112 causes the multiplexer M3 to switch the connection in step S40, so that the receiving channel 1R3 is changed from being connected to the first use channel 1A3 to being connected to the dedicated first backup channel 1B3. In this way, the receiving channel 1R3 of the connection interface 110 can receive the output data TO outputted by the output channel 2T3 of the connection interface 210 via the connection channel L6 from the first backup channel 1B3.
[0077] In some embodiments, Fig. 9 As shown, in the connection interface 110, each of the receiving channels 1R1-1RX corresponds to one of the first use channels 1A1-1AX, and the receiving channels 1R1-1RX can be divided into a plurality of receiving channel groups G11-G1M. Each of the receiving channel groups G11-G1M includes at least two receiving channels, and each of the receiving channel groups G11-G1M can correspond to at least one first backup channel. In other words, at least one first backup channel corresponding to the same receiving channel group is shared by the receiving channels in the receiving channel group. In contrast, in the connection interface 210, each of the output channels 2T1-2TX corresponds to one of the second use channels 2A1-2AX, and the output channels 2T1-2TX can be divided into a plurality of output channel groups G21-G2M. Each of the output channel groups G21-G2M includes at least two output channels, and each of the output channel groups G21-G2M may correspond to at least one second backup channel. In other words, the at least one second backup channel corresponding to the same output channel group is shared by the output channels in the output channel group. In addition, each of the receiving channel groups G11-G1M may correspond to one of the output channel groups G21-G2M. The number of receiving channels included in each receiving channel group is the same as the number of output channels included in the corresponding output channel group, and the number of first backup channels corresponding to each receiving channel group is the same as the number of second backup channels corresponding to the corresponding output channel group.
[0078] Taking the example that each of the receiving channel groups G11-G1M includes three receiving channels and corresponds to two first backup channels, and each of the output channel groups G21-G2M includes three output channels and corresponds to two second backup channels, in one example, the receiving channel group G11 may include receiving channels 1R1, 1R2, 1R3 and correspond to first backup channels 1B1, 1B2; the receiving channel group G12 may include receiving channels 1R4, 1R5, 1R6 and correspond to first backup channels 1B3, 1B4; and so on. In contrast, the output channel group G21 may include output channels 2T1, 2T2, 2T3 and correspond to second backup channels 2B1, 2B2; the output channel group G22 may include output channels 2T4, 2T5, 2T6 and correspond to second backup channels 2B3, 2B4; and so on.
[0079] Fig.10 is a schematic diagram of an embodiment of a transmission system. Figure 1 , Figure 3 and Fig.10In one embodiment, the connection interface 110 includes four receiving channels 1R1-1R4, four first use channels 1A1-1A4 and four first backup channels 1B1-1B4, and the connection interface 210 includes four output channels 2T1-2T4, four second use channels 2A1-2A4 and four second backup channels 2B1-2B4.
[0080] In the connection interface 110, the receiving channel group G11 may include receiving channels 1R1 and 1R2 corresponding to the first backup channels 1B1 and 1B2, and the receiving channel group G12 may include receiving channels 1R3 and 1R4 corresponding to the first backup channels 1B3 and 1B4. In addition, the connection control module 1111 may include four 3-to-1 multiplexers M31-M34. Among them, the input side of the 3-to-1 multiplexer M31 is coupled to the first used line 1A1 and the first spare lines 1B1 and 1B2, and the output side of the 3-to-1 multiplexer M31 is coupled to the receiving line 1R1; the input side of the 3-to-1 multiplexer M32 is coupled to the first used line 1A2 and the first spare lines 1B1 and 1B2, and the output side of the 3-to-1 multiplexer M32 is coupled to the receiving line 1R2; the input side of the 3-to-1 multiplexer M33 is coupled to the first used line 1A3 and the first spare lines 1B3 and 1B4, and the output side of the 3-to-1 multiplexer M31 is coupled to the receiving line 1R3; and the input side of the 3-to-1 multiplexer M34 is coupled to the first used line 1A4 and the first spare lines 1B3 and 1B4, and the output side of the 3-to-1 multiplexer M34 is coupled to the receiving line 1R4.
[0081] In the connection interface 210, the output channel group G21 includes output channels 2T1 and 2T2 and corresponds to the second backup channels 2B1 and 2B2, and the output channel group G22 includes output channels 2T3 and 2T4 and corresponds to the second backup channels 2B3 and 2B4. Here, the connection control module 2111 directly connects the output channel 2T1 to the second use channel 2A1, directly connects the output channel 2T2 to the second use channel 2A2, directly connects the output channel 2T3 to the second use channel 2A3, and directly connects the output channel 2T4 to the second use channel 2A4. In addition, the connection control module 2111 may include four 2-to-1 multiplexers M21-M24. Among them, the input sides of the 2-to-1 multiplexers M21 and M22 are both coupled to the output lines 2T1 and 2T2, the output side of the 2-to-1 multiplexer M21 is coupled to the second spare line 2B1, and the output side of the 2-to-1 multiplexer M22 is coupled to the second spare line 2B2; and the input sides of the 2-to-1 multiplexers M23 and M24 are both coupled to the output lines 2T3 and 2T4, the output side of the 2-to-1 multiplexer M23 is coupled to the second spare line 2B3, and the output side of the 2-to-1 multiplexer M24 is coupled to the second spare line 2B4.
[0082] In the initial state, in the connection interface 110, the processing module 1112 causes the 3-to-1 multiplexer M31 to connect the receiving channel 1R1 to the first use channel 1A1, causes the 3-to-1 multiplexer M32 to connect the receiving channel 1R2 to the first use channel 1A2, causes the 3-to-1 multiplexer M33 to connect the receiving channel 1R3 to the first use channel 1A3, and causes the 3-to-1 multiplexer M34 to connect the receiving channel 1R4 to the first use channel 1A4. In this way, the output data TO on the output channel 2T1 of the connection interface 210 will be transmitted to the receiving channel 1R1 via the second use channel 2A1, the connection channel L1 and the first use channel 1A1 in sequence; the output data TO on the output channel 2T2 of the connection interface 210 will be transmitted to the receiving channel 1R2 via the second use channel 2A2, the connection channel L3 and the first use channel 1A2 in sequence; the output data TO on the output channel 2T3 of the connection interface 210 will be transmitted to the receiving channel 1R3 via the second use channel 2A3, the connection channel L5 and the first use channel 1A3 in sequence; and the output data TO on the output channel 2T4 of the connection interface 210 will be transmitted to the receiving channel 1R4 via the second use channel 2A4, the connection channel L7 and the first use channel 1A4 in sequence.
[0083] Afterwards, if the processing module 1112 in the connection interface 110 detects that the first use line 1A1 is an unhealthy line, the processing module 1112 will generate an output change request H1 according to the first use line 1A1, and transmit the output change request H1 to the connection interface 210 through the communication line LT1 in the bridge substrate 300 connected between the connection interface 110 and the connection interface 210. After receiving the output change request H1, the processing module 2112 in the connection interface 210 can know that the output path of the output line 2T1 should be changed accordingly according to the output change request H1, so the processing module 2112 will choose to cause the 2-to-1 multiplexer M21 to connect the output line 2T1 to the second backup line 2B1, or choose to cause the 2-to-1 multiplexer M22 to connect the output line 2T1 to the second backup line 2B2. Here, the processing module 2112 can make a selection according to a preset arbitration setting (for example, but not limited to, implemented by an arbitrator). When the processing module 2112 selects to cause the 2-to-1 multiplexer M22 to connect the output line 2T1 to the second backup line 2B2, the processing module 2112 will generate an input change command H2 according to the second backup line 2B2, and transmit the input change command H2 to the connection interface 110 through the communication line LT1 in the bridge substrate 300. After receiving the input change command H2, the processing module 1112 of the connection interface 110 can know that the output data TO of the output line 2T1 will be transmitted via the second backup line 2B2 and the connection line L4, so the processing module 1112 will cause the 3-to-1 multiplexer M31 to change the connection of the first backup line 1B1 to the receiving line 1R1. Since the connection switching of the receiving lines 1R2-1R4 can refer to the connection switching of the previous receiving line 1R1, it will not be repeated here.
[0084] Fig.11 FIG. 1 is a partial flow chart of an embodiment of a line connection method. Figures 1 to 11. In some embodiments, in the connection interface 210, the processing module 2112 of the controller 211 may first receive a plurality of output data TO from the system circuit 220 via the output channels 2T1-2TX, and generate an input data based on at least one of the output channels, such as the output data TO on the output channel 2T1 and a test data. Afterwards, the processing module 2112 transmits the input data to the first use channel 1A1 or the first spare channel connected to the receiving channel 1R1 in the connection interface 110 via the second use channel 2A1 and / or the second spare channel connected to the output channel 2T1 and the connected connection channel. Here, the test data may be data for repairing an unhealthy channel. In some embodiments, the processing module 2112 may generate at least one required test data through a pattern generator, and intersperse each test data in the output data TO to form the input data. For example, in the output data TO, a test data is inserted every 1 nanosecond (ns). Thus, in the connection interface 110, after receiving the input data via the receiving channel 1R1, the processing module 1112 of the controller 111 can execute the repair procedure for all channels set as unhealthy channels according to the test data when reading the test data portion (step S50).
[0085] After executing the repair procedure, the processing module 1112 of the controller 111 will determine whether the health of the unhealthy channel after the repair procedure is greater than a preset value (step S60). When the processing module 1112 determines that the health is greater than the preset value, it means that the repair is successful. At this time, the processing module 1112 can change the unhealthy channel to the first backup channel (step S70). On the contrary, when it is determined that the health is less than or equal to the preset value, it means that the repair fails. At this time, the processing module 1112 can change the unhealthy channel to a damaged channel (step S80). Here, once the channel is set as a damaged channel, the processing module 1112 will no longer perform the repair procedure on this channel.
[0086] Specifically, when executing the repair procedure for an unhealthy channel, such as the first use channel 1A1, the processing module 1112 of the controller 111 can first readjust the left pointer G1 and the right pointer G2 to, for example, the boundary of data "1" according to the test data of the known data pattern, such as "010". Afterwards, the processing module 1112 makes adjustments according to the spacing between the left pointer G1 and the right pointer G2 so that the positive edge of the sampling clock CLK can reach (be located at) the center position of the data "1". Here, the repair procedure is roughly the same as the aforementioned calibration training. In other words, executing the repair procedure for an unhealthy channel is equivalent to re-executing the calibration training for the unhealthy channel.
[0087] Next, the processing module 1112 adjusts the left index G1 and the right index G2 to the boundary of data "1" with the positive edge of the sampling clock CLK as the center to obtain the current data eye. Afterwards, the processing module 1112 compares the current data eye with a preset value, such as 1 / 2 of the original data eye (i.e., the original data eye of the test data). If the current data eye is greater than 1 / 2 of the original data eye, the processing module 1112 can change the first used channel 1A1 from an unhealthy channel to a healthy channel, so that the first used channel 1A1 can be used as a backup channel. For example, the state value corresponding to the first used channel 1A1 in the channel state table is changed from the second value to the first value, wherein the first value indicates a healthy channel and the second value indicates an unhealthy channel. If the current data eye is less than or equal to 1 / 2 of the original data eye, the processing module 1112 can change the first used channel 1A1 to a damaged channel. For example, the first used channel 1A1 is registered in the damaged channel table.
[0088] In some embodiments, the connection interface 110 may be referred to as a receiving interface, and the connection interface 210 may be referred to as a transmitting interface.
[0089] In some embodiments, the die 100 may include a connection interface 210 in addition to the connection interface 110. In contrast, the die 200 may include a connection interface 110 in addition to the connection interface 210. In addition, the connection interface 210 of the die 100 may be connected to the connection interface 110 of the die 200 through another set of connection lines L1-LZ of the bridge substrate 300 for communication. In other words, each die may include a receiving interface and a transmitting interface, and the receiving interface of one die may be connected to the transmitting interface of another die through a set of connection lines L1-LZ.
[0090] In some embodiments, in the bare die 100, the processing module 1112 of the connection interface 110 and the processing module 2112 of the connection interface 210 may be integrated into a single processing module. In some implementations, the processing module 1112 and / or the processing module 2112 may be implemented by, for example, an embedded controller (EC), a system on a chip (SoC), a central processing unit (CPU), a microcontroller (MCU), an application specific integrated circuit (ASIC), an application processor (AP), a digital signal processor (DSP), a programmable logic device (PLD), or any other suitable electronic component.
[0091] In summary, any embodiment of the connection interface 110, 210 and the line connection method for bare die to bare die can monitor the data transmission status of each of the first use lines 1A1-1AX connected to the receiving lines 1R1-1RX, and when a first use line is detected as an unhealthy line, the receiving line connected to the unhealthy line is changed to one of the first backup lines 1B1-1BY. In this way, because the data can be changed from the first use line with a poor data transmission status to the first backup line with a better data transmission status, the data transmission quality can be improved. In addition, due to the monitoring of the data transmission status of each first use line, the processing module 1112 can find out the connection line that is about to be damaged in the bridge substrate 300 in advance, and use another connection line and the first backup line for transmission, so that the data transmission between the bare die 100 and the bare die 200 can not be interrupted due to the damage of the connection line, and the high-speed transmission of the transmission system 1 can be maintained.
[0092] Although the technical content of this case has been disclosed as above in the preferred embodiment, it is not used to limit this case. Any slight changes and modifications made by anyone familiar with this art without departing from the spirit of this case should be included in the scope of this case. Therefore, the scope of protection of this case shall be determined by the scope of the attached patent application.
Claims
1. A die-to-die connection interface, characterized in that: Include: A plurality of first use lanes; A plurality of first backup lines; a plurality of receive channels; and A controller is used to connect each of the receiving channels to one of the first use channels and monitor each of the first use channels respectively. When the first use channel is detected to be an unhealthy channel, the controller changes the receiving channel connected to the unhealthy channel to one of the first backup channels.
2. The die-to-die connection interface according to claim 1, wherein: When receiving input data via the first used channel or the first backup channel, the controller executes a repair program on the unhealthy channel according to a test data in the input data, and determines whether a health level of the unhealthy channel after executing the repair program is greater than a preset value. When it is determined that the health level is greater than the preset value, the controller changes the unhealthy channel to the first backup channel.
3. The die-to-die connection interface according to claim 1, wherein: Also includes: A plurality of second use lines; a plurality of second backup lines; and A plurality of output channels, each of which is connected to one of the second use channels and at least one of the second backup channels.
4. The die-to-die connection interface according to claim 1, wherein: Also includes: A plurality of second use lines; a plurality of second backup lines; and A plurality of output channels, each of which is connected to one of the second use channels, and each of which corresponds to at least one of the second backup channels; When the controller receives an output change request from another connection interface, it connects the output channel corresponding to the output change request to one of the corresponding at least one second backup channel, generates an input change command according to the second backup channel to which the output channel is connected, and outputs the input change command to the other connection interface.
5. The die-to-die connection interface according to claim 1, wherein: When it is detected that the first used channel is the unhealthy channel, the controller generates an output change request according to the unhealthy channel and outputs the output change request to another connection interface, and when an input change command is received from the other connection interface, the controller changes the connection of the receiving channel connected to the unhealthy channel to the first standby channel corresponding to the input change command.
6. A die-to-die line connection method, characterized in that: Include: Connecting a plurality of receiving channels of a connection interface to a plurality of first use channels of the connection interface, wherein each of the receiving channels is connected to one of the first use channels; respectively monitoring the first usage channels; and When it is detected that the first used channel is an unhealthy channel, the receiving channel connected to the unhealthy channel is connected to one of the plurality of first backup channels of the connection interface.
7. The die-to-die line connection method according to claim 6, wherein: Also includes: receiving an input data via the first used channel or the first spare channel, wherein the input data includes a test data; Execute a repair procedure on the unhealthy line according to the test data; Determining whether a health level of the unhealthy line after executing the repair procedure is greater than a preset value; and When it is determined that the health level is greater than the preset value, the unhealthy channel is changed to the first backup channel.
8. The die-to-die line connection method according to claim 6, wherein: Also includes: The plurality of output channels of the connection interface are connected to the plurality of second use channels and the plurality of second spare channels of the connection interface, wherein each of the output channels is connected to one of the second use channels and at least one of the second spare channels.
9. The die-to-die line connection method according to claim 6, wherein: Also includes: Connecting a plurality of output channels of the connection interface to a plurality of second use channels of the connection interface, wherein each of the output channels is connected to one of the second use channels, and each of the output channels corresponds to at least one of the second backup channels; When receiving an output change request from another connection interface, connecting the output channel corresponding to the output change request to one of the corresponding at least one second backup channel; generating an input replacement command according to the second backup channel to which the output channel is connected; and The input change command is output to the other connection interface.
10. The die-to-die line connection method according to claim 6, wherein: The step of reconnecting the receiving channel connected to the unhealthy channel to one of the first backup channels of the connection interface comprises: generating an output replacement request according to the unhealthy line; outputting the output change request to another connection interface; and When an input change command is received from the other connection interface, the receiving channel connected to the unhealthy channel is reconnected to the first backup channel corresponding to the input change command.
Citation Information
Cited By
Self-adaptive adjustment method for inhibiting PCB electron migration and PCB automatic control system
CN119828844A
An Adaptive Adjustment Method for Suppressing PCB Electromigration and a PCB Automatic Control System
CN119828844B