Browse Prior Art Database

High-Speed Serial Auto Skew Correction Circuit

IP.com Disclosure Number: IPCOM000241256D
Publication Date: 2015-Apr-09
Document File: 4 page(s) / 98K

Publishing Venue

The IP.com Prior Art Database

Abstract

Described is a high-speed serial auto skew correction circuit.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 40% of the total text.

Page 01 of 4

High-

-Speed Serial Auto Skew Correction Circuit

Speed Serial Auto Skew Correction Circuit

High speed serial (HSS) data is pushing the threshold of our current technology. The speed at which the HSS protocols are communicating has introduced new problems with signal integrity. A major challenge to signal integrity is, as the speed increases, the attenuation increases; the distances between communicating parts has not changed. These two items combined mean the receivers are seeing less signal strength, which has lead to broadband amplification of the input signal (peaking amplifier) and decision feedback circuitry (DFE) inside the receiver. For these two items to work properly, it is imperative that the incoming differential signal is pristine, with minimal skew (arrival time differences) between the P (positive) and N (negative) legs.

    There are a number of HSS protocols that are used over cabled structures without converting the electrical signal to an optical format. One of the issues with using a purely electrical signal is that non-linearities from the cable material affect the signal integrity. A common specification for cable induced skew is 10 pS/ meter. A common HSS protocol, serial attached SCSI (SAS), uses industry-standard cables that are 8 meters in length. IBM*, however, uses cable lengths up to 15 meters in length. Skew, especially when amplified by a peaking circuit, can trick the DFE circuit into changing a perceived 1 into a 0 or 0 into a 1 erroneously. Too many of these instances may cause: transmitter and receiver to never negotiate or link, errors in the data, dropped links or outages, specific sorting or rejection of parts. Each of these effects has a variety of costs and implications to a given project. Cables will not exhibit the same skew, and because cables in systems can be interchangeable, a one time skew adjustment will not fix the issue.

    For example, say the user has three cables all the same length: A, B and C. Cable A's differential pair is skewed, with the P side of the differential pair 100 pS longer than the N side. Cable B has no skew P to N. Cable C's differential pair is skewed, with the N leg 75 pS longer than the P side. All of these cables fell within the specification for manufacturing, but each will affect the receiver circuit differently.

    What is needed are receivers that can dynamically fix the skew at the receiver input by sensing the incoming data. These receivers need to be able to handle skew caused by different edge rates or cables. Currently there is IP that does allow for skew correction in the receiver but does so in an inefficient manner. There is prior art [1] that addresses variations in the transmitter from lane to lane to accomplish a skew correction across lanes. This does not help correct for the variations caused by the channel. Another [2] has an overly complicated skew detection where the P and N are being detected, and their center voltages are being used to determine...