RealWorld Certifier PC protocol
The Byte Brothers (now Triplett) RealWorld Certifier is an old handheld Ethernet cable certifier. It has an FT232R inside, and the official PC software is an Excel workbook with VBA macros. I wanted the results without Excel.
First thing I learned: when the tester starts talking to a PC, it shows TALKING TO PC, UNSAVED READINGS ARE ERASED and does exactly that. I found this out by poking at it blind and losing a reading. Save on the tester before you plug it in.
Finding the software
The "33z CD image" on techeia.com is a renamed NSIS installer with nothing in it but the 2010 FTDI driver. The workbook is in RWC_TriplettSW.zip from Newtech Industries, a dealer, as RWC1000_Rev48z.xlsm.
I didn't open it in Excel. The macros are in xl/vbaProject.bin, the module source is MS-OVBA compressed, and olevba or ~30 lines of Python gets it back out. The protocol is all in the SerialAccess module, and the record layout is in RawData.
Protocol
9600 8N1, no flow control. Every command is one byte, no terminator. The prompt is 2F 0D 0A 00.
| Byte | Does |
|---|---|
s then b |
handshake, replies WOW |
l |
first reading |
n |
next reading |
h |
last reading |
Each reading comes back as 48 bytes followed by an 8-byte trailer (0D 0A 00 00 + prompt). Past the end you get 48 bytes of FF, which fails the checksum, and that's how the vendor code knows to stop.
Record
Big-endian:
| Off | Field |
|---|---|
| 0 | valid, FF |
| 1 | status bits: 0x02 split, 0x04 NEXT/FEXT, 0x10 map, 0x20 delay/skew |
| 2 | reading number, u16 |
| 4, 5 | headroom % for cert 1, cert 2 |
| 6 | pairs 12, 36, 45, 78: u16 delay + status byte each |
| 18 | cable type: 5 CAT5, 6 CAT5E, 7 CAT6 |
| 19 | wire type: 0 solid, 1 stranded |
| 27 | cert 1 speed: 0x40 1 Gig, 0x20 100M, 0x10 10M |
| 28 | cert 2 speed and pairs |
| 29 | average quality, u16 |
| 31 | checksum, sum of bytes 0-30 mod 256 |
| 32 | name, 16 bytes ASCII |
Pair status 5-8 means the pair landed on 12/36/45/78 at the far end, so a straight-through cable reads 5, 6, 7, 8. 1 is open, 2 is short, 12 is a split pair.
The "length" fields are delay in 1/16 ns. The vendor converts at 24.39 units per foot, which is a hard-coded NVP of ~0.667, and reports the shortest good pair as the cable length. Their overall skew number divides by 16 twice, so use per-pair skew instead.
I had grabbed one raw record before the erase, and it decodes cleanly: CAT6 solid, 67 ft, 1 Gig at 82% headroom, 5 ns worst skew. So the cable was fine too.
Reader
import serial, struct
FMT = ">BBHBBHBHBHBHBBBBBBBBBBBBHB" # 32 bytes, then 16 bytes of name
def readings(port):
with serial.Serial(port, 9600, timeout=2.5) as s:
s.reset_input_buffer()
s.write(b"sb")
assert s.read(13).startswith(b"WOW")
cmd = b"l"
while True:
s.reset_input_buffer()
s.write(cmd)
rec = s.read(56)[:48]
if len(rec) < 48 or sum(rec[:31]) % 256 != rec[31]:
return
yield struct.unpack(FMT, rec[:32]), rec[32:].decode("latin-1").strip()
cmd = b"n"
Anyone with the older serial-cable version: I'd like to know if it speaks the same protocol. Nothing in the VBA checks hardware revision, so I suspect it does.