Digital audio line driver

[Dirk Bruere at NeoPax]

I want to drive a digital audio signal over CAT5.
I intend to use the DA103 transformer but need to check impedance matching.

Will 100R each end of the CAT5 be suitable?
Should the 100R on the transmitting end be before, or after, the xformer?

 

[JosephKK]

I want to drive a digital audio signal over CAT5.
I intend to use the DA103 transformer but need to check impedance matching.

Will 100R each end of the CAT5 be suitable?
Should the 100R on the transmitting end be before, or after, the xformer?

The terminating restance should be at the far end of the cable. I
gotta ask though, in there any reasonable way that you could send the
signal digitally? like manchester encoded? The whole thing can work
much better that way.

 

[Dirk Bruere at NeoPax]

The terminating restance should be at the far end of the cable. I
gotta ask though, in there any reasonable way that you could send the
signal digitally? like manchester encoded? The whole thing can work
much better that way.

Forgot to add...
The input signal is standard SPDIF.
I'm going to bump it up to TTL level and use that to feed into a am26C31
line driver and then into the transformer. The other end is an SPFIF
receiver, so I have to do an impedance match (and voltage reduction to
around 1V).

 

[JosephKK]

Forgot to add...
The input signal is standard SPDIF. OK.
I'm going to bump it up to TTL level and use that to feed into a am26C31
line driver and then into the transformer. The other end is an SPFIF
receiver, so I have to do an impedance match (and voltage reduction to
around 1V).

Actually, i think all you really need is the impedance match (both
ends and let the voltage scale). Unless you are going a lot farther
than SPDIF is rated for, in which case the driver may be called for.

 

[Dirk Bruere at NeoPax]

Actually, i think all you really need is the impedance match (both
ends and let the voltage scale). Unless you are going a lot farther
than SPDIF is rated for, in which case the driver may be called for.

The only worry is that the receiving chip is 3V3 and I do not want to
risk damaging it by feeding it some overvoltage. Can bad impedance match
bump up the voltage through reflections?

I checked the specs for the am26c31 and the output is around 3V (5V
Vcc), so after the xformer and resistors it will certainly be less.

 

[JosephKK]

The only worry is that the receiving chip is 3V3 and I do not want to
risk damaging it by feeding it some overvoltage. Can bad impedance match
bump up the voltage through reflections?

Yes it can. With this configuration and guessed line lengths it may
show up as increased overshoot / undershoot, more ringing or leading
pedistals on the pulses at the receiver.

I still recomment just Z matching both ends and just using Cat 5 / Cat
5E cable. It really is nice cable for fast signals. I know, i done
got me a copy of TIA-568 set.

 

[Dirk Bruere at NeoPax]

Yes it can. With this configuration and guessed line lengths it may
show up as increased overshoot / undershoot, more ringing or leading
pedistals on the pulses at the receiver.

I still recomment just Z matching both ends and just using Cat 5 / Cat
5E cable. It really is nice cable for fast signals. I know, i done
got me a copy of TIA-568 set.

Given that the SPDIF input is 75R, a couple of 12R resistors in each
signal path at the SPDIF end should be sufficient?

 

[Dirk Bruere at NeoPax]

Given that the SPDIF input is 75R, a couple of 12R resistors in each
signal path at the SPDIF end should be sufficient?

On a related issue, how important is termination over short runs with
low(ish) frequency? Audio is around 10Mhz and I doubt we will ever go
beyond 30m in length. In this case the wavelength is around the same
lenth as the cable. Of course, this is square wave and obviously there
are higher frequency components.

 

[Paul Keinanen]

On a related issue, how important is termination over short runs with
low(ish) frequency? Audio is around 10Mhz and I doubt we will ever go
beyond 30m in length. In this case the wavelength is around the same
lenth as the cable.

1/4 wavelength is an impedance inverter (an open 1/4 stub acts a short
circuit, a shorted 1/4 wavelength stub acts as an open circuit).

With a typical velocity factor for CAT5 cables of 0.67, at 10 MHz, the
wavelength is about 20 m. When operating below 1/10 wavelengths, the
transmission line issues are not very critical. Thus, with 10 MHz
signals, termination should be used for cables longer than 2 m.

 

[JosephKK]

Given that the SPDIF input is 75R, a couple of 12R resistors in each
signal path at the SPDIF end should be sufficient?

Might be. You may also try 15R or 18R to see what works best, then a
1:1 transformer right into the receiver. Resistors at both ends, it
will cost some signal amplitude, maybe 40% in voltage. (105/75)^1/2
~= 1.18 so 1.2:1 transformers are likely to work better, 1.2 side on
the Cat5 cable side.

 

[JosephKK]

On a related issue, how important is termination over short runs with
low(ish) frequency? Audio is around 10Mhz and I doubt we will ever go
beyond 30m in length. In this case the wavelength is around the same
lenth as the cable. Of course, this is square wave and obviously there
are higher frequency components.

It mostly depends on the edge speeds you need to maintain. They
determine the highest frequency components with signifcant (spectral)
power. So if you need 10 ns edges you have stuff up in the 100 MHz
range involved. Or if you use another approximating rule 35 MHz.

 

[Dirk Bruere at NeoPax]

Might be. You may also try 15R or 18R to see what works best, then a
1:1 transformer right into the receiver. Resistors at both ends, it
will cost some signal amplitude, maybe 40% in voltage. (105/75)^1/2
~= 1.18 so 1.2:1 transformers are likely to work better, 1.2 side on
the Cat5 cable side.

The SPDIF input is already xformer isolated, 1:1
However, there should be plenty of signal amplitude

 

[JosephKK]

The SPDIF input is already xformer isolated, 1:1
However, there should be plenty of signal amplitude

So let us know how it all works out, ok?

 

[JosephKK]

The only worry is that the receiving chip is 3V3 and I do not want to
risk damaging it by feeding it some overvoltage. Can bad impedance match
bump up the voltage through reflections?

Yes it can. Depending on line length and signaling rate it may be in
the form of overshoot and undershoot or maybe ringing. Or simple
changes in volatge amplitude. Or a mixture of these.

I still recommend Z match and go for it, Cat 5/ Cat 5E is pretty nice
cable for fairly fast signals. I done bought me a copy of TIA-568
set.

 

[Dirk Bruere at NeoPax]

So let us know how it all works out, ok?

Well, feeding straight into CAT5e and AES/EBU i/f it all works fine.
SPDIF will have to wait until next week.