Uncertainty about custom implementation of GMSL2 Eye Mapper Test (MAX9296A and MAX9295B)

Question

This is a followup to my GMSL2 Signal Integrity Tests (MAX9296A and MAX9295B) forum post. I was able to reproduce the link margin and eye mapper tests in Linux. Link margin test looks great for both forward and reverse channels. The eye mapper test looks good, but maybe not as good as I expected. Right now I'm uncertain if this is a bug in my implementation or if it's just a discrepancy in the hardware. 
 This is an example of what the forward channel eye diagrams look like for the MAX96724F/MAX96717F EVKs (csv files were generated in Windows using Analog Device's GMSL SerDes Public GUI software): 
 
 This is an example of what the forward channel eye diagrams look like for the MAX9296A and the MAX9295B (csv files were generated in Linux using my own c++ implementation): 
 
 As you can see, although both eye diagrams look good, the MAX96724F/MAX96717F EVKs have less jitter and a better SNR than the MAX9296A/MAX9295B parts that we are using in our design. 
 Q1: 
 Do the MAX9296A/MAX9295B eye diagrams look reasonable? 
 Q2: 
 Is the MAX96724F/MAX96717F SerDes pair a newer generation than the MAX9296A/MAX9295B SerDes pair, and does it offer any signal integrity improvements? According to the Selection Table for Gigabit Multimedia Serial Link (GMSL) , the MAX96724F/MAX96717F SerDes pair is recommended for newer designs. I can't find any reference to the MAX9296A/MAX9295B SerDes pair, so I am wondering if it is obsolete. 
 Edit: I spoke with an EE on my team who had more knowledge about the lifecycle of these products. He said that the MAX9296A/MAX9295B SerDes pair is much older than the MAX96724F/MAX96717F SerDes pair. We'd still be interested in knowing whether there were any substantial signal integrity or eye opening monitor improvements. 
 Q3: 
 This is kind of a catchall question regarding Equalizer Coefficients. The GMSL2 SerDes Public Software displays the following Equalizer Coefficients: AGC, BST, OSN, DFE1, DFE2, DFE3, and DFE4. I have no idea what they mean or how the values are captured or if they have any impact on the eye diagram. The Eye Map Algorithm document makes no references to them, so I don't think they impact the generation of the Eye Map from the CSV file, but perhaps these coefficients impact signal integrity. Could someone explain these values and, if possible, give advice on how to adjust them to improve signal integrity?

LiMason · Accepted Answer

Q1: 
 Do the MAX9296A/MAX9295B eye diagrams look reasonable? 
 It looks great. 
 
 Q2: 
 I can't find any reference to the MAX9296A/MAX9295B SerDes pair, so I am wondering if it is obsolete. 
 MAX9296 is an older part than 96717. Yet, the reason why 9296 is hidden is that it requires NDA signed before you can see the details. 
 
 Q3: 
 AGC, BST, OSN, DFE1, DFE2, DFE3, and DFE4. I have no idea what they mean or how the values are captured or if they have any impact on the eye diagram. 
 These are coefficients related to the PHY layer of GMSL. To be more concrete, they're equalizer coefficients. The only way you understand how to master these COEs is to learn something about analog and digital Equalizers, however, which is unnecessary because the adaptive equalizer will tune the COEs to the optimal set without your manual adjustment.)