DIY TG for HP8568B

[tkamiya]

Hello folks.  I'm considering a DIY of tracking generator for my HP8568B.  There are many pages for DIY TG but all are dealing with HP8566, not 8568.

Looking at HP8568B documentations and also HP8444A with option 58/59, I am quite confused partly because both documentation covers multiple models with slightly different schemes.  Especially the doc for SpecAn contains left over verbiage from HP8566B.  So if someone can help me clear up my thoughts, I'd be appreciative.  Initial order of things to do is to get frequency relationship correct.
  If there are any errors in the following, please point that out.

1)  Local out from SA is 2.0 to 3.7GHz, at +4dbm.  (1.7GHz bandwidth??)

2)  TG expects 500MHz and it mixes with 1.55GHz +/- 2Mhz (adjustment range) internally to produce 2.05GHz.

3)  They both mix to create 0 to 1.5GHz range.  Actually, TG documentation shows mixer taking in 2 - 3.4GHz and 2.05GHz +/- 2Mhz and producing 0 - 1.35GHz.  (huh?  the math is 50MHz off)

Here, math already doesn't make sense.  Diagram is pretty explicit about 1.55GHz and 500Mhz.  That will generate 2.05GHz.  If I mix this with 2.0GHz, I get 50MHz.  HP8568B starts from 100Hz.  So one of them has to be off.  But which?  I'm quite surprised that HP8568B can actually tune up to 2.49GHz.  (same as 8566) with much reduced RX sensitivity.  THAT, combined above mentioned frequencies, range doesn't make sense. 

I've read Kerry Wong's excellent Blog few times but his is 8566B.  As I understand it and having brief time with few units, I see they are quite different. 

[tkamiya]

I just found out a recently acquired frequency counter actually can go up to 3GHz, so I measured the 1st LO output.

At 0Mhz and no scan, 2.053GHz
At 0.5GHz and no scan, 2.553GHz
At 1GHz and no scan 3.053GHz.

It appears the LO is actually from 2.053GHz, rather than 2.0GHz.