reverse engineering TDIF

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

tmbg

Well-known member
Joined
Nov 7, 2004
Messages
438
Location
Atlanta, GA
anyone have any opinions on reverse engineering TDIF, to try to decipher its protocol? I'm probably going to throw a logic analyzer on it tonight, and see if I can get it to reveal its deep dark secrets..


*awaiting the lawsuit* :sad:
 
Can't you get specifications from Tascam?

http://emusician.com/mag/emusic_spare_interchange/

http://www.prosoundweb.com/studyhall/net/transport/audio.php
 
SSLtech: don't suppose you have a pinout?

A tdif cable actually has 8 separate pieces of coax in it? :p

An AES/EBU is stereo, correct? So it's 4 AES/EBU lines in each direction?

hmmm....
 
[quote author="tmbg"]SSLtech: don't suppose you have a pinout?

A tdif cable actually has 8 separate pieces of coax in it? :p

An AES/EBU is stereo, correct? So it's 4 AES/EBU lines in each direction?

hmmm....[/quote]
Yes 8 each way, 4 pairs. I never needed to look it up, as you look at the 25-pin, row 1-13 pin 1 is 1/2, (out or in, I forget, but a scope will tell you... the outs will have signal on them, the ins will not!) pin 2 is 3/4, pin 3 is 5/6, pin 4 is 7/8. Pin 10 is 7/8 going the other way, pin 11 is 5/6, pin 12 is 3/4 and pin 13 is 1/2.

The shield for pin 1 is on pin 14, the shield for pin 2 is on pin 15, the shield for pin 3 is on pin 16, the shield for pin 4 is on pin 17.

The shield for pin 10 is on pin 22, the shield for pin 11 is on pin 23, the shield for pin 12 is on pin 24, the shield for pin 13 is on pin 25. Draw this down and you'll see a pattern. A TDIF-to TDIF cable is left-right reversed so that ins connect to outs.

Clock out is on one of the middle pins, along with LRCK out and S.R. flag (S.R. = Sample Rate). There are also S.R. flag in, LRCK in, clock in and a ground or two...

The Audio format is AES/EBU, but unbalanced, 75Ω. I think the clock is stripped out so that it appears only on the clock out pin as a word clock and not in the bitstream, but I'm not certain... you'd have to test it, I've never been that interested that I would spend that much time.

I know that the AE-88 interface might be rather more than just a level shift and balancing interface, so I suspect there are some bitstream modifications.

Keith
 
Ah. Here's a pinout. (20 seconds on google; "TDIF pinout")

http://www.newbox.co.uk/Wiring%20Bible%20pdf/TDIF.pdf

The pins marked "LACK" in this document are incorrectly labeled, it should be "LRCK"... I suspect that the writer didn;t know what "LRCK" was...

whaddayaknow... looks like a lot of people have been copying from each other... All of the pinout results have 'LACK' instead of "LRCK"!

FYI, "LRCK" is the left-right interleave indentification.

Keith
 
now I just need to find some info on the data in AES/EBU...

It's 32 bits per sample always, correct? And I'm guessing it's sample interleaved, so it would be 32 bits of Left, then 32 bits of Right?

Does the clock vary with sampling rate, or does it pad?

48khz * 32bits per sample * 2 channels = 3mbit/sec. Thats a pretty hefty amount of data! :/ and lightpipe would have to be 4 times that much data, since it's serial.

I was thinking about trying to use a 40MHz pic to shuffle all that data around, but a 40MHz pic is 10 MIPS, so it might not even be fast enough to read the aes/ebu, and almost definitely not fast enough to speak lightpipe :/
 
More reading about it,

http://www.sydec.be/Support/Faq/TechNotes/0007/
http://www.sydec.be/Support/Faq/TechNotes/0003/

There should also be the layout of a TDIF cable upthere somewhere, but i didn't see it (yet) . It is very simple in fact, i made all of my TDIF-cables and it involves something like all the numbers on one side of the cable in normal order and then starting somewhere shifted 1-2-3... to 13-14-15 or so, butt i will post this later when i am completely sure about it.

Tony
 
[quote author="tmbg"]Does the clock vary with sampling rate, or does it pad?[/quote]
Since you always have 32 bits per sample (24 audio bits), the clock frequency will have to be variable.

I was thinking about trying to use a 40MHz pic to shuffle all that data around, but a 40MHz pic is 10 MIPS, so it might not even be fast enough to read the aes/ebu, and almost definitely not fast enough to speak lightpipe :/
Why shuffle the data in a PIC, when you can get cheap ICs from Crystal and Alesis that do the work for you?
If you want a fast microcontroller, have a look at the silabs 100MIPS 8051 clones...

Best regards,

Mikkel C. Simonsen
 
TDIF cables

The TDIF port has a 25-pin sub-D connector. Connecting cables for short distances (e.g., 1m - 3m) can be simple 25-way flat ribbon cables with one of the 25-way male IDC (Insulation Displacement) connectors reversed, so that pin 1 on one end connects to pin 13 on the other. For longer distances (up to 5m), shielded twisted pair cable is recommended and cables of this type are supplied by Tascam or other cable suppliers.

TDIF Cables can be split so that inputs and outputs can connect to different devices. To do this, the flat cable is separated around wire 13. The following tables show the connections provided by TDIF and can help in making cables that split the inputs and outputs:

TDIF pin Connections

DB25--------------------Ribbon------------------DB25

Pin---------Function----Wire-------Pin----------Function

1-----------Dout 1/2-------1--------13----------Din 1/2
2-----------Dout 3/4-------3--------12----------Din 3/4
3-----------Dout 5/6-------5--------11----------Din 5/6
4-----------Dout 7/8-------7--------10----------Din 7/8
5-----------LRCK out------9---------9-----------LRCK in
6-----------FS0 out--------11--------8-----------S0 in
7-----------GND-----------13--------7-----------GND
8-----------FS0 in---------15---------6-----------FS0 out
9-----------LRCK in-------17---------5-----------LRCK out
10----------Din 7/8-------19---------4-----------Dout 7/8
11----------Din 5/6-------21---------3-----------Dout 5/6
12----------Din 3/4-------23---------2-----------Dout 3/4
13----------Din 1/2-------25---------1-----------Dout 1/2
14----------GND----------2---------25------------GND
15----------GND----------4---------24------------GND
16----------GND----------6---------23------------GND
17----------GND----------8---------22------------GND
18----------EMPH out----10---------21-----------EMPH in
19----------FS1 out------12---------20-----------FS1 in
20----------FS1 in-------14---------19-----------FS1 out
21----------EMPH in-----16---------18-----------EMPH out
22----------GND---------18---------17-----------GND
23----------GND---------20---------16-----------GND
24----------GND---------22---------15-----------GND
25----------GND---------24---------14-----------GND

It's a pitty this forum doesn't allow multiple spaces :sad:
I had everything lined up but it didn't work?
 
[quote author="tony dB"]It's a pitty this forum doesn't allow multiple spaces :sad:
I had everything lined up but it didn't work?[/quote]
But it does! Use
Code:
[ code]
kdshf       100
sdfjlsd     200
[ /code]

Best regards,

Mikkel C. Simonsen
 
[quote author="SSLtech"]8X AES/EBU over 75Ω coaxes. Sync and S.R. flags stripped out to seperate conductors in the middle of the connnector.

Keith[/quote]

I'm ready to built one like these, but i'm a bit unsure where to put the sleeves (ground) and what to combine for minimum crosstalk as i want to make cables 6 meter and more. I'm using coax Sommercable 75Ω (isopod)

Cheers,

Tony
 
[quote author="tony dB"]
...i'm a bit unsure where to put the sleeves (ground) and what to combine for minimum crosstalk as i want to make cables 6 meter and more. I'm using coax Sommercable 75Ω (isopod)
[/quote]

So, to answer my own question, after some research, and a phonecall to my friends at Soundscape-development, here what seems to be the best way of getting these connected:

I've have 8 balanced 75Ω cables named A, B,.... H

Connect on one end to the other end

1 hotwire A to 13
2 coldwire A to 12
14 ground A to 25

3 hotwire B to 11
4 coldwire B to 10
16 ground B to 23

5 hotwire C to 9
6 coldwire C to 8
17 ground C to 22

19 hotwire D to 20
20 coldwire D to 19
7 ground D to 7

8 hotwire E to 6
9 coldwire E to 5
22 ground E to 17

10 hotwire F to 4
11 coldwire F to 3
23 ground F to 16

12 hotwire G to 2
13 coldwire G to 1
25 ground G to 14

18 hotwire H to 21
21 coldwire H to 18
15 ground H to 24

I hope this is clear.

PS 24 not used (to be connected to ground if you want this) to 15 not used
 
That looks like the Tascam analog pinout - also a DB25 but a different pinout. 75 ohm coax should have only a 'hot' and a 'shield', no 'Cold'. Note that LRCLK are basically word clock, so I would expect them to require a shield as well, especially for a longer run.

Pin 1 -> hot A
Pin 14 -> shield A

Pin 2 -> hot B
Pin 15 -> shield B

Pin 3 -> hot C
Pin 16 - > shield C

Pin 4 -> hot D
Pin 17 -> shield D

Pin 5 -> hot E (LRCLK)
Pin 7 -> shield E (LRCLK)

Pin 6 -> just a wire
Pin 8 -> just a wire

Pin 9 -> hot F (LRCLK)
Pin 7 -> shield F (LRCLK)

Pin 10 -> hot G
Pin 22 -> shield G

Pin 11 -> hot H
Pin 23 -> shield H

Pin 12 -> hot I
Pin 24 -> shield I

Pin 13 -> hot J
Pin 25 -> shield J

Pin 18 -> just a wire
Pin 19 -> just a wire
Pin 20 -> just a wire
Pin 21 -> just a wire

You'll need to swap the wires on the opposite end as shown in an earlier post.
 
Back
Top