Transformers and digital signals

[shadewind]

How can isolation transformers work digital signals such as Ethernet? Doesn't a transformer generate current on the secondary side when the magnetic flux in the core changes which only changes when the current on the primary side changes? If so, this would mean that the there would only be current spikes generated on the secondary side when the signal switches?

There is something I am missing obviously...

[deephaven]

Ethernet uses Manchester encoding which doesn't require a DC component to work because the data is carried as transitions.

See: http://en.wikipedia.org/wiki/Manchester_code

[shadewind]

Yes, I realize that, but between the switchings, di/dt is ideally zero which would give the spikes I described above. I'm still missing something I believe

[deephaven]

It's all about frequency response. On the transformer, if you plot secondary voltage against primary voltage across a frequency range from DC to 100 MHz (say), then you will indeed see a signal level of zero at DC. But as the frequency rises, the response will increase until you get a close to 1:1 transfer (if the turns ration is 1:1). This response should remain relatively flat over the frequcny range the transformer is designed for. The Manchester encoding scheme guarantees a minimum frequncy component which means that the signal will pass through the transformer with little loss or distortion.

[shadewind]

Ah, so the transfer improves as the frequency increases? Also, I interpret your comments about transformer design that this also depends on the way the transformer is designed, i.e. number of turns, core material, core shape et.c.

This makes sense to me since coming from an audio background (not engineering, music production), I've understood that better quality audio transformers primarily have better low end response.

[scrat]

Think of the transformer as a band-pass filter and you get it. If the signal you're interested in is lying mainly in the band passed by the transformer, then the digital transmission through it works.
Ethernet Manchester signal is at very high frequency with respect to audio, around 10MHz for the 10Mbit/s, IIRC.

[jahonen]

Transformer lower frequency limit is set by amount of volt-seconds that transformer core can take before the core saturates, and the primary inductance is lost, leading to "droop" of output signal. That means if transformer is used, there can't be any kind of DC component. It is common practice in telecommunications to apply such an encoding scheme so that DC-component vanishes, so that there could be any form of AC coupling in the signal path.

Faster ethernet uses much different encoding scheme, 100BASE-T uses MLT-3. Also, Gigabit ethernet has same symbol rate (125 Msyms/sec) than 100 Mbit/s ethernet, but symbols are bigger and all four pairs are used to transmit data.

Regards,
Janne