ADAT over Twisted Pair

[Rochey]

Hia folks,

quick question for those of you who know more about driving signals over cat5 twisted pair.

What kind of driver logic would be best to drive an ADAT signal over a twisted pair?
My thoughts so far is that a differential circuit would be best, maybe something like LVDS could be good. However, I read that a lot of RS-485 is used over Cat5 as well.

Can anyone shed any light on this? Any links to websites where I can learn more would be appreciated.

cheers

R

 

[jdbakker]

[quote author="Rochey"]What kind of driver logic would be best to drive an ADAT signal over a twisted pair?[/quote]
Do you want to stay <whatever>-compatible ? Over not too long spans, I suppose you could use the physical/electrical layer of AES (110Ohm differential, if memory serves me well, not too different from the 100Ohm nominal offered by TP). The advantage is that the problem of large common mode voltages is solved for you by the existing implementations.

My choice would be LVDS -- if you can guarantee that all systems are roughly at the same potential. Have a look at http://focus.ti.com/lit/an/slyt163/slyt163.pdf and http://focus.ti.com/lit/an/slyt172/slyt172.pdf. The latter is about ribbon cable, but much of it applies to TP as well. LVDS is way faster than you need, though, which violates the principle of simplicity (since you get to worry much more about impedances and termination).

RS-485 may be a bit too slow for your needs. Check out driver/receiver datasheets; DS26LS31/Am26LS31/DS26C31 are the standard drivers, the '32s are the corresponding receivers.

I don't remember if ADAT is guaranteed to have a 50% duty cycle like S/PDIF and AES/EBU. If so, I'd suggest looking into transformer coupling to the cable, using either an Ethernet transformer (manufacturers: Pulse, HALO, the former available through Digi-Key) or an RF transformer (http://www.mini-circuits.com/).

What kind of cable length are you planning to drive ?

JDB.

 

[Rochey]

Anything up to 30 meters I guess.

my main concern is to do with jitter. However,the recieving end doens't need amazing quality audio - the dacs will most likely be 105dB.

The reason I want high bandwidth is to cover the square wave nature of the signal. ADat's signal is around 6MBit... so, my fundamental will be at 6MHz, plus harmonics to make up the square wave.

I figure that to make a decent square wave, I'll need at least the 5th harmonic.

/R

 

[jdbakker]

[quote author="Rochey"]my main concern is to do with jitter. However,the recieving end doens't need amazing quality audio - the dacs will most likely be 105dB.[/quote]
The AL1402 ADAT receiver has a specced jitter of '1.5ns typical', whatever that might mean in real life. Your UTP needs to be pretty crappy to account for a similar amount of jitter over 30 meters. If you really care about jitter, I'd suggest retiming at the receiver.
[quote author="Rochey"]The reason I want high bandwidth is to cover the square wave nature of the signal. ADat's signal is around 6MBit... so, my fundamental will be at 6MHz, plus harmonics to make up the square wave.[/quote]
First, I couldn't find an open reference on ADAT, so I have no clue wrt its modulation; but 6MBit does not automatically mean that your fundamental will be at 6MHz.

Second, jitter is exactly why you do want to pre-filter your signal before sending it onto the cable. Google for 'group delay'; the cable forms a filter, wherein not all frequencies have the same transit time. This is a common phenomenon; imagine that fifth harmonic arriving a few ns later than the fundamental... This will shift your sampling point, thus introducing jitter.

JDB.

 

[Boswell]

[quote author="Rochey"]The reason I want high bandwidth is to cover the square wave nature of the signal. ADat's signal is around 6MBit... so, my fundamental will be at 6MHz, plus harmonics to make up the square wave.[/quote]
There are 256 bits in each ADAT frame, and frames are sent at the audio sampling rate. So 48KHz audio generates an ADAT bit rate of 12.288MHz. The entire ADAT lightpipe structure was designed round the fastest commonly available (at the time) optical TX and RX devices, which limit at around 13-15MHz. Working at the bandwidth limit of the transmission medium, the received waveform is almost sinusoidal, and is put through a Schmitt tigger circuit to square it up again.

For a cable version up to 30m, RS485 could be made to work. LVDS is nice, but an overkill. I don't think jitter would be a problem.

 

[Rochey]

let me double check something here...

with cat5 and RS-485, you terminate at the source (a 100ohm resistor accross the pins) and you terminate at the reciever - assuming a one to one relationship?

 

[Rochey]

So...

after a little a research... it looks like I could couple a 4ch RS485 transmitter with a 4ch RS485 receiver (which has schmitt triggers built in).

http://focus.ti.com/lit/ds/symlink/sn75lbc175a.pdf - 4ch 50Mb/S Reciever
http://focus.ti.com/lit/ds/symlink/sn75lbc174a.pdf - 4ch 30Mb/S Transmitter

that would technically allow me to use 4 adat chips in parallel over a Cat5 cable... for 32 channels of audio@48KHz... That's more than what I was planning on, I only wanted 8 channels.. but still.

 

[Boswell]

[quote author="Rochey"]with cat5 and RS-485, you terminate at the source (a 100ohm resistor accross the pins) and you terminate at the reciever - assuming a one to one relationship?[/quote]
No, not unless it's bidirectional (or multi-drop, where only the two ends are terminated). With point-to-point, terminate with the cable characteristic impedance at the receiver only. For CAT5, the value is 100 Ohms. The source sees 100 Ohms when looking into its end of the cable.

PS beware the funny pairing in a CAT5 cable.