3
\$\begingroup\$

Can someone tell me what is the meaning of Open circuit inductance and why we need to take care while choosing magnetics?

Also I have seen that, the turns ratio for selecting magnetics is 1:1 and must be 350uH. Can someone tell why is this value of 350uH is recommended? Can we increase or decrease this inductance value?

\$\endgroup\$

3 Answers 3

5
\$\begingroup\$

Ethernet is a digital signaling standard. Pulses must be sent and received, over a modest distance (~100m), without consuming excessive power, avoiding random errors, and tolerating ambient noise (and emitting little in turn), etc.

To meet these goals, many choices can be made: coaxial or twisted pair cable; pulse, amplitude, frequency, or other modulation or coding schemes; framing can be synchronous or asynchronous, with fixed timing or clock recovery, etc.

For 10BASE-T, clearly they chose twisted pair, FM/Manchester encoding, clock recovery (I think?) (which is fairly trivial with Manchester), and about a 2V signal level; and because Manchester is DC balanced (mostly, anyway), it can be transformer coupled, affording isolation and significant noise immunity (common mode rejection (CMRR)).

As it turns out, it's not very well DC balanced -- or 100BASE-T is; I'm not real clear on the exact details of how these work -- there are certain bit patterns that result in fairly significant low-frequency components, i.e. the DC balance acts slowly, over say hundreds or thousands of bits. This causes the baseline to shift, making it harder to detect the level and timing of the received bitstream. When a single flat pulse passes through such a channel, the waveform is distorted* and what's supposed to be a flat top, "droops" over time. Consider a waveform like:

*In the geometric sense. Electrically, this is still a linear process.

enter image description here

(From: Reason behind loss of low frequencies in square waveforms)

particularly the bottom-right waveform. If the droop is small like this, the pulse width and level is received with good fidelity; middle or left waveforms have an ambiguous or indeterminate level (voltage near zero) before the next edge comes in, losing information. Furthermore, consider a waveform like this, with finer pulses superimposed upon it: the fine pulses represent the bulk of the data transferred, while the slow varying (drooping) low frequency components correspond to "unlucky" trends of too many high or low bits, that momentarily shift the DC balance.

Instead of an RC highpass filter, we have an LR highpass filter, the L being the transformer's magnetizing inductance, and R being the line impedance and termination resistors. A transformer is used at each end, so they act in parallel, and we must have \$ \frac{R}{2 \pi L} < F_{min}\$ for some minimum frequency (corresponding to the DC balancing action of the digital code). Ethernet uses 350µH and 100Ω, evidently needing bandwidth down to about 90kHz. This 90kHz figure is determined by the line coding (and thus expected DC balancing action).

\$\endgroup\$
1
3
\$\begingroup\$

Can someone tell me what is the meaning of Open circuit inductance

That's the inductance of a winding when all the other winding(s) are left open (Similarly, the inductance measured when all the other winding(s) are shorted is called leakage inductance). It's basically the magnetising inductance and the one that you should be checking when selecting ethernet magnetics.

Can someone tell why is this value of 350uH is recommended? Can we increase or decrease this inductance value?

It's what the standard defines. For 100 Base-T the open circuit inductance (OCL) of ethernet magnetics should be at least 350µH with any DC bias between 0 and 8mA.

NOTE: That requirement may change soon (or maybe has already changed). The revised standard may allow the designer to use lower inductances just to meet the performance needs.

\$\endgroup\$
5
  • \$\begingroup\$ Could you tell me why is it important to measure OCL? \$\endgroup\$
    – Newbie
    Jun 21 at 8:14
  • \$\begingroup\$ Could you also answer on what happens if we increase or decrease the OCL of Ethernet magnetics? \$\endgroup\$
    – Newbie
    Jun 21 at 8:15
  • 2
    \$\begingroup\$ @Newbie It's almost all about the droop requirement of the physical layer (The standard defines an allowed percentage on the voltages). About increasing the OCL, I'd like to say that as long as you meet BER (bit error rate) requirement the inductance should not be important, IMO. \$\endgroup\$ Jun 21 at 8:38
  • \$\begingroup\$ Can you share me the link on where we can read the magnetics standard so that I can also study that? I've been trying to locate the standard but I am unable to get it? Is it free? \$\endgroup\$
    – Newbie
    Jun 21 at 8:40
  • \$\begingroup\$ @Newbie It's IEEE 802.3 but I don't know if it's free. \$\endgroup\$ Jun 21 at 8:44
2
\$\begingroup\$

In transformers, we distinguish between excitation current and magnetizing current. The excitation current represents energy passed from primary to secondary. The magnetizing current represents energy retained in the transformer's magnetic field. In a signal transformer, the magnetizing current loads down the driver without doing anything useful.

If the open circuit inductance is small, the magnetizing current will be large. The inductance should be adequate to limit the magnetizing current to a tolerable level.

\$\endgroup\$
2
  • \$\begingroup\$ Thank you very much for the answer. Could you please tell me on how the magnetizing current will be large when the OCL is small? Is there any formula relation or any intuition behind it? \$\endgroup\$
    – Newbie
    Jun 23 at 3:20
  • \$\begingroup\$ It's like any inductor. The rate of change of the magnetizing current is simply the applied voltage divided by the OCL. \$\endgroup\$
    – John Doty
    Jun 23 at 12:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.