Electronic – Purpose of anti-parallel diodes in isolated gate driving circuit

Ethernet uses \$\pm0.85\;\mathrm{V}\$ signaling so a signal will only be able to "activate" one diode pair. So any transmission only has a chance to reach the two immediate neighbors (no current will flow more than 1 "hop" round the loop).

Now you can notice the trick which will work only with 3 devices: both of computer A TX lines' have RX lines from computers B and C in immediate neighborhood and it's own RX lines on the far side of the loop. You really don't want to hear what you are sending since it would trigger the collision detection algorithm.

Two interesting points:

1. The diodes will heavily attenuate the signal so it probably won't work over longer distances.

2. You cannot use resistors since they would linearly attenuate the signal and after going all the way around the loop it would finally reach your own RX line. It would be attenuated but the receiver circuit is very sensitive so it would still be able to detect is as a collision. You need a nonlinear element (like a diode) that provides a sharp cutoff.

PS. It is a really clever circuit. I cannot imagine inventing something like that on my own. :)

I think you may be exceeding what is known as the volt-microsecond limit for the transformer. Basically this figure tells you how long you can apply a pulse of a certain voltage before the voltage reverses. Real transformers will saturate and the V.\$\mu s\$ limit tells you, in practical terms how far you can push the transformer.

The waveform initially looks good on the rising edge - it attains a peak of 4 or 5 volts (it looks that way from your pictures) but after about 10 us it goes wrong - this means your transformer is probably rated as having a volt-microsecond product of about 50.

Other problems with the circuit are: -

• You need more gate drive voltage to adequately and efficiently turn the mosfet on
• As you vary mark-space ratio, the peak voltage seen by the gate (from the transformer) will vary considerably - you need to have a clamp circuit on the secondary that keeps the gate drive at a decent level for varying mark-space.
• I don't think that transformer is man-enough to provide the current needed by the gate to efficiently switch the mosfet without incurring heavy losses.
• The transformer drive circuitry also isn't powerful enough to drive a transformer should you have a transformer (and clamp) that would fit the bill efficiency-wise.

Here is a helpful (in parts) article on the subject. And here is a helpful guide on designing transformer circuits for driving mosfets: -

You can probably see in one of the diagrams (fig 3) the problem you are getting.