Why Choose Schematic A Over Schematic B for Transformers?

[es]

Hello All,

I understand this is more of a question for electrical engineers however I have received only baffling replies from electrical engineers so I was hoping a physicist, or someone a little better versed in E&M than I am, could assist in finding an answer to this question.

Long story short, I am trying to understand the tradeoffs involved in selecting the different schematics listed in the attached data sheets. (I'll call the S558-5999-M8 schematic A and the S558-5999-P3 schematic B) Now one would think this would be a trivial question for the supplier of this part. However this is the answer I received from a, not-belfuse, vendor, (paraphrased) "Schematics of type B have 12 cores while the other has 8. Therefore we can build type A parts smaller and cheaper." To which I replied, "But they come in the same package and you quote the same price, so then why would anyone select A over B or vice versa?" "Because B is the recommended schematic." "Really. Why?" "Schematics of type B have 12 cores while the other two have 8." From here the conversation just circles. I tried to rephrase the question as, "What physcial property does B posses that A lacks which would make B the recommended schematic." The reply was related to its physical size...

I have done a little homework so I can state what little I know in the next post.

 

[es]

Background Information

Devices of this type are used to isolate and filter electrical signals which leave one circuit board, travel through a pair of cables, and arrive at another circuit board (for example a typical ethernet circuit). One will find a schematic of type A or B at both ends of the cable. The electrical signal travels down this cable pair differentially which means one cable carries negative current and another carried positive each of equal magnitude in the ideal case. So in our device of interest we can assume a positive current flows from pins 2 to 23, I1, and a negative current flows from pins 3 to 22, -I2, and the magnitude of these currents is as close as we can practically make it. In a typical configuration pin 1 is connected to power (usually +2.5V) directly and pin 24 is connected to a low pass circuit (usually R=75ohm, C=1uF for ethernet) designed to match the impedance of the cable at high frequencies.

In schematic B moving from pins 2 and 3 to pins 23 and 22 the first pair of coils is called "the transformer" (original I know), the next pair of coils is called the "common mode choke", and the third is called an "auto transformer."

 

[Doc Al]

I understand this is more of a question for electrical engineers however I have received only baffling replies from electrical engineers ...

Let's see if our crack team of PF electrical engineers can do better.

 

[es]

What I know about the device operation

One function of this part is to allow both ends of the cable to have different references (i.e. isolates DC) as you would expect from a transformer.

However another critical function is to remove the common mode component, defined as the average of the two currents, of the signal.

In a perfect world the magnitude of I1 equals I2 so they sum to zero meaning there is no common mode component but alas, we do not live in a perfect world. The small variations in magnitude lead to a finite common mode current which in turns generates RF noise as it travel down the unshielded cable. The magnetic fields of the differential signal cancel so the differential signals do not contribute RF noise. If to much RF noise is radiated by the circuit one will violate FCC requirements which will result in not being able to legally distribute the circuit. Therefore preventing common mode current from entering the cable is vitally important.

The transformer leaks common mode current though the capacitance across the primary and secondary coils (Cww), so it attenuates but not by enough.

The common mode choke attenuates the common mode current by a lot. The differential mode current, flowing in opposite directions through the choke windings, creates equal and opposite magnetic fields which cancel each other out. The result is the choke presents low impedance, or little attenuation on the differential signal. The common mode current, flowing in the same direction through the choke windings, creates equal and in phase magnetic fields which add together resulting in the choke presenting high impendence to the common mode signal, i.e. the common mode signal is highly attenuated.

A common mode signal sees the two halves of the center tapped auto transformer in anti-phase. Therefore this current causes equal and opposite magnetic fields to be generated which cancel each other out meaning the common mode signal sees a low impedance path to the center tap which is basically grounded meaning the common mode current is sent to ground not the cable. Vice versa for the differential signal, i.e. the differential signal sees a high impedance path to the center tap, so most of this differential current goes to the cable.

 

[es]

Finally my question!

I am trying to understand the tradeoff between selecting a magnetics package that includes an auto-transformer and one that does not. I would prefer to pick the part without the auto transformer because they are cheaper and smaller. (I am extremely limited in available board area) However, as stated earlier, this is "not recommended."

In your opinion what purpose does the auto-transformer provide? Why should it be recommended while the other is not? Does it improve the common-to-common-mode-rejection-ratio (CCMR) while the common mode choke coils improve the differential to common mode rejection ratio (DCMR)? (These are both spelled out in the IEEE specs) If this is the case then it seems these last coils may be necessary to meet the transmit CCMR IEEE spec. But then the CMR spec (unfortunately it is not clear which they mean in the various data sheets I have read) rarely changes between vendors or the varying configurations. To further confuse me, in the Belfuse data sheet for example, CMR doesn't change at all between configurations and they are both advertised as IEEE compliant!

If the specs don't change why offer two configurations? There must be a difference. (There is actually a third and forth that I haven't mentioned and their specs are the same as well).

I have to admit this whole transformer thing is black magic to us mere EEs. There have been reports of measured RF noise during design verification changing by as much as 15dB with no obvious change in any of the data sheet parameters. The vendor's reply was "there was a minor change in the magnetics." What ever that means...

ZapperZ's so you want to be a physicist never seems more appealing. :)

 

[faust9]

In response to your first question, I'd use whatever module your phy xceiver calls for.

http://www.alldatasheet.com/datasheet-pdf/pdf/NSC/DP83861.html
The xceiver from national spells out which magnetic module(s) to use.

Now, from the looks of things, the autotransformer is stepping up a filtered signal which may or may not improve your SNR while the 8 core unit taps the primary coil so the voltages on the primary are amplified more than with the 12 coil unit thus unchoked noise also gets amplified. Just a guess. Not my field of expertise. I have found though that the recommended part is recommended for a reason be it stability, SNR, or EOF considerations. Go with what is recommended unless you absolutely need that extra board space and the few cents that gets saved.

Good luck.

 

[es]

An Idea...

I think I have one idea on what additional purpose the auto-transformer serves.

In the schematic lacking the auto-transformer, if the cable picks up common mode noise when this noise hits the common mode choke it will see a high impedance and thus get reflected back onto the cable (with some loss so eventually this noise will reflect out) causing RF noise to be radiated. In the case with the auto-transformer between the choke and cable this common mode noise will have path with the characteristic impedance to ground, and thus will leave the cable.

But this is a case of the environment effecting the cable, not the cable effecting the environment, so perhaps this is why both can be IEEE compliant since the testing is done with the environment quiet.

 

[es]

Thanks for the feedback faust9.

The magnetic data sheets I am looking at specify a 1:1 turns ratio on the coils so I don't think anything is getting stepped up or down. I may be wrong though, this is not my area of expertise either.

Unfortunately I am having a hard time placing components even using the smaller part. If I am forced to use the larger, which I am afraid I will be due to "recommendations," things will be even worse. Then again, if nothing fits, there will be no board at all and therefore no problem ;)

 

[berkeman]

>I understand this is more of a question for electrical engineers however I have
>received only baffling replies from electrical engineers so I was hoping a physicist, or
>someone a little better versed in E&M than I am, could assist in finding an answer to >this question.

LOL. Keep in mind that there are lots of EEs here that love physics. With a good problem description, you should be able to get a good EE answer (allowing for work schedules interfering with forum checks, of course).

The main difference between the different xfmr models seems to be the availability of the center tap, at least from my quick look. You would generally use the center-tap-available xfmr for applications where you need to manage the common-mode impedance versus frequency of the transformer transfer function. For example, if you have an EMI problem meeting FCC or EN radiated EMI limits, you can often put a HF cap in the range of 100-200pF on the CT tap of the comm xfmr to notch out the problem. Or if you have a problem meeting the EN 61000-something-6 CM RF immunity tests, you can tie off the CT tap harder with something (the -6 tests run from 150kHz to ~80MHz I think).

If you don't care about the CM Z of your device, then you can pick either transformer type based on availabilty, IMO. -Mike-