Receiver bit alignment for multi-lane asynchronous high-speed data interface

1. A system comprising:

multiple transmitters each having a pseudorandom binary sequence (PRBS) generator for providing a training pattern, and a selection logic for selecting either the training pattern during a training phase or regular data and passing the selection onto a plurality of serial multi-lanes;

multiple receivers each having a synchronous capturing module for capturing multiple ones of the plurality of serial multi-lanes with data simultaneously, the captured data being used to calculate a PRBS distance or bit skew for each of the plurality of serial multi-lanes, where the PRBS distance with the largest latency is used as a reference to calculate a relative latency of each of the serial multi-lanes to other serial multi-lanes in the plurality of serial multi-lanes, the relative latency is used to calculate a number of shifts for at least one of a barrel shifter and word shifter in the receivers; and

a bit shifter for each of the plurality of serial multi-lanes of at least one receiver, that shifts the number of data bits in a lane as a calculated relative latency to achieve alignment among each of the serial multi-lanes;

wherein the transmitters and the receivers cooperate to provide bit alignment for one or more serial multi-lane asynchronous high speed data interfaces between the transmitters and receivers;

wherein for the bit alignment between the receivers and the transmitters, two training patterns are provided, and represent two phases of alignment, one pattern being periodic pattern having a sequence length equal to parallel word width and another pattern being a PRBS pattern; and

wherein each of the synchronous capturing modules comprise a PRBS shift register having a shift register length for capturing one or multiple parallel words simultaneously under control of capturing and shifting control logic, the number of words to capture depends on a PRBS shift register length to make sure that the captured length is no less than the shift register length, and a unique location in a bits PRBS sequence.

2. The system of claim 1 , wherein each of the synchronous capturing modules comprise a shifting control is based on the calculation of relative bit latency among the lanes which is done by loading a captured value C i from lane i, to calculate a number of cycles to reach a captured value C j from lane j, if the number of cycles c is within a pre-defined threshold, such as a maximum skew, then lane j has a larger delay than lane i and is shifted forward.

3. The system of claim 2 , wherein each of the synchronous capturing modules employs synchronous capturing logic to load C j into the PRBS shift register to calculate the cycles to reach C i , or to use a PRBS sequence length 2 L −1 minus a number of cycles c as the latency.

4. The system of claim 1 , wherein each of the synchronous capturing modules, for bit error tolerance, performs multiple capturing/calculating to check consistency, when the latencies are obtained, looks for the earliest lane or latest lane, depending on the shift direction, to use as reference, and then calculates the relative latency for other lanes.

5. The system of claim 2 , wherein each of the synchronous capturing modules performs the capturing by assuming that the earliest lane is j, and the latency for lane i (i≠j) is d i , then the number of shifts for the barrel shifter is (r−d i %r)%r where r is the parallel data width and “%” is modulo operation.

6. The system of claim 2 , wherein each of the synchronous capturing modules performs the capturing by assuming that the earliest lane is j, and the latency for lane i (i≠j) is d i , the number of shifts for word shifter is given by ┌(d i +(r−1))/r┐ where ‘┌ . . . ┐’ means “rounds down”, and where r is the parallel data width.

7. A method comprising the steps of:

providing a training pattern by a pseudorandom binary sequence (PRBS) generator in each of a plurality of transmitters;

selecting, by a selection logic in each of the plurality of transmitters, either the training pattern during a training phase or regular data, and passing the selection onto a plurality of serial multi-lanes and

capturing multiple ones of the plurality of serial multi-lanes with data simultaneously, with multiple receivers each with a synchronous capturing module, the captured data being used to calculate a PRBS distance or bit skew for each of the plurality of serial multi-lanes, of the distances calculated the distance with the largest latency is used as a reference to calculate a relative latency of each of the serial multi-lanes to other serial multi lanes in the plurality of serial multi-lanes, the relative latency is used to calculate a number of shifts for at least one of a barrel shifter and word shifter in the receivers; and

shifting bits of each lane of at least one receiver by the number of bits as a calculated relative latency to achieve alignment among each of the plurality of serial multi-lanes;

wherein the transmitters and receivers cooperate to provide receiver bit alignment for one or more serial multi-lane asynchronous high speed data interfaces between the transmitters and receivers;

wherein for the bit alignment between the receivers and the transmitters, two training patterns are provided, and represent two phases of alignment, one pattern being a periodic pattern with sequence length equal to parallel word width and another pattern being a PRBS pattern; and

wherein each of the synchronous capturing modules comprise a PRBS shift register having a shift register length for capturing one or multiple parallel words simultaneously under control of capturing and shifting control logic, the number of words to capture depends on a PRBS shift register length to make sure that the captured length is no less than the shift register length, and a unique location in a bits PRBS sequence.

8. The method of claim 7 , wherein each of the synchronous capturing modules comprise a shifting control is based on the calculation of relative bit latency among the lanes which is done by loading a captured value Ci from lane i, to calculate a number of cycles to reach a captured value C j from lane j, if the number of cycles c is within a pre-defined threshold, such as a maximum skew, then lane j has a larger delay than lane i and is shifted forward.

9. The method of claim 8 , wherein each of the synchronous capturing modules employs synchronous capturing logic to load C j into the PRBS shift register to calculate the cycles to reach C i , or to use a PRBS sequence length 2 L −1 minus a number of cycles c as the latency.

10. The method of claim 7 , comprising performing, for bit error tolerance, multiple capturing/calculating to check consistency, when the latencies are obtained, looking for the earliest lane or latest lane, depending on the shift direction, to use as reference, and then calculates the relative latency for other lanes.

11. The method of claim 8 , wherein each of the synchronous capturing modules comprises assuming that the earliest lane is j, and the latency for lane i (i≠j) is d i , then the number of shifts for the barrel shifter is (r−d 1 %r)%r where r is the parallel data width and “%” is modulo operation.

12. The method of claim 8 , wherein each of the synchronous capturing modules performs the capturing by assuming that the earliest lane is j, and the latency for lane i (i≠j) is d i , the number of shifts for word shifter is given by ┌(d i +(r−1))/r┐ where ‘┌ . . . ┐’ means “rounds down”, and where r is the parallel data width.