Sanyo Denki ABS encoder

I'm planning to retrofit an old OTC-Daihen 6-axis robot with LinuxCNC, this robot is equipped with Sanyo Denki servo motors.
I have acquired Sanyo Denki servoamps (RS1) for each axis and for Mesa hardware I have 5i20+7i48+7i44+7i73.

These servoamps can output encoder data in different formats, serial and pulse. As I need absolute position information I would like to use serial data. For serial data there are three different options; Binary code (9.2ms), ASCII (16.7ms) and encoder signal direct output (21.6us). (Page 210 from RS-1 instruction manual)
So for robot application I would choose the fastest option ->encoder signal direct output.

What would be the best way to read this encoder? i.e. how to send 18bit request and then listen for encoder response. 

Specification for these battery-backup multiturn absolute encoders:
Single turn resolution= 17bit
Multiturn count= 16bit
Half-duplex RS-485, NRZ code, Asynchronous.
Baud rate is 2.5MBd, LSB is transferred first.
One request frame and three response fields.
Frames have 16 data bits + 1 start + 1 stop bit = 18 bits. 

I found a detailed description of this encoder protocol from here:
www.mikrocontroller.net/attachment/58813...B_PA035-017BM001.pdf

Different request codes are described in page 13, interesting requests would be absolute full data request CDF0 or Absolute lower 24-bit data request CDF1.
In page 14 there is an example of CDF1 (absolute lower 24-bit data request):
Start SC FC EA CC CRC Stop
0 010 00 000 10000 001 1

Encoder response would include three fields: Information field, data field 0 and data field 1.
Information field:
Start SC EA CC 0 ES Stop
0 001 000 10000 0 ES0 ES1 ES2 ES3 1
Data field 0:
Start D0[0:15]=ABS[0:15] Stop
Data field 1:
Start D1[0:7]=ABS[16:23] D1[8:15]=CRC[0:7] Stop

Encoder data would be interpreted as:
ABS[0:16]=ST[0:16]
ABS[17:32]=MT[0:15]
ABS[33:39]="0000000"

I have tried to study fanucabs, uart, pktuart, ssi and biss but I would really need some advice on how to move forward with this.  
User ExcessiveO has modified fanucabs for RS-485, but I would assume that there are some hurdles to overcome for modifying code for these encoders.
I also found sigma5enc.vhd from hostmot2-firmware that could be modified for this purpose, but I couldn't find any documentation or drivers for that from LCNC-side.

Any help is highly appreciated.

I have had some problems creating bitfile for my setup. I have tried to make file with 6xPktUART 6PWM SS, but apparently this will not fit to 5i20. I also tried to make bitfile with only 6xPktUART, but it won't fit. When I reduce the number of PktUarts to 1 (just for testing) I get error message:- The specification "TSLowToMed" is invalid because its TO- The specification "TSMedToLow" is invalid because itsSo I have had no luck in testing PktUart approach. I'm using Xilinx 9.2 for creating bit files.Out of the box sigma5enc is not going to work, because it includes HDLC encoding and manchester coding that are not used in this Sanyo Denki protocol.Maybe completely new .vhd file would need  to be written? I have no background in coding, but here's what I have tried to do. I made a new vhd-file "sanmotionenc.vhd" based on sigma5enc. See attached file sanmotionenc.txt. There could be some peculiarities, but bare with me. The first goal for me would be just to send correct bitstream to amplifier. I made also adjustment to idromconst and idromtools files for replacing parts with sigma to sanmotionenc. Then I created bitfile with 6xPWM 6xsanmotionenc SS. For the next steps I would need some guidance, what kind of adjustments should be done for existing absolute encoder drivers?   

Not yet, I'll have to do that.
This 5i20 is already quite old. I got it for my small milling machine conversion, from steppers to servo, but my mill is still equipped with steppers...

I'm also open to hardware suggestions, would for example 7i80HD-25 make sense for this kind of project?

Response from Mesa was that I shouldn't use PktUART because:
"1. You dont need or want  a buffered interface (it also wastes space as
you noticed)
2. The position request should be sent automatically _before_ the nominal
read time (this is what the Fanuc,SSI and BISS interfaces do)"

For Sanyo Denki protocol a specific module should be created (similar as fanuc) accompanied by the driver code to support this.

So I'm going to check that if I can re-use fanuc code for sending the request (16bit data instead of pulse request) and after that try to figure out how to read data back from amplifier.

I have made modifications to fanucabs and after quite many iterations I was able to send the request string out. (Picture shows tx and tx_en signals, measured directly from 5i20.) This happened just yesterday so I haven't had time to test it with actual amplifier. Next I will send the request string to amplifier and hopefully get a response back. After that I'll have to find out how the amplifier response is read and converted to position data.