EMI-resistant input transformer design

Sure! That is a time-honored way to do it, input transformers on sound boards have had this for many years. You can probably buy them.

Jon

If it's a toroidal core, make sure that the winding is non-progressive (so you don't end up with a single-turn loop antenna wrapped around the core).

Screening is good, double-screening may be better - so one screen is grounded to the remote earth and the other to the local earth.

Ralph Morrison's "Grounding and Shielding: Circuits and Interference"

may provide more detail. Amazon is now selling the sixth edition - I cut my teeth on the first edition, published in 1967, and I've got a copy of the fourth edition. He's good on this sort of stuff.

--
Bill Sloman, Sydney

** Having a grounded CT tap on a mic input transformer is rare. Normally th e primary is simply left floating and the CT is left available for connecti on to a 48V, phantom supply.

The problem with grounding the CT is that the signal source must be either fully floating or *precisely* balanced so the signals at each end of the pr imary are identical in amplitude otherwise any imbalance in the two signal results in serious loading of the source.

.... Phil

Do you have a common-mode choke in the input line?

--

-TV

Yes, and note that the primary should be tightly coupled (adjacent layers, or bifilar), otherwise imbalance and leakage between the two halves manifests as poor CMRR.

Excellent point: CMC can make up for shortcomings of the transformer. The CM inductance is in series with the split winding's leakage, thus making the leakage that many times less significant.

It's part of how Ethernet achieves its magic!

You can take this farther, as bandwidth permits, by adding more stages of CMCs and split windings:

This filter network (in effect, a CM-only bias tee) uses a tee section (series CMC, grounding CT, series CMC) to provide lots of isolation between source and load sides. I built one, using short enough CT windings not to matter (to the diff mode, they manifest as transmission line stubs, so the winding lengths need to be as short as possible), and if nothing else, it transmits 100BASE-T successfully (though with what eye diagram, I don't know).

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

If all the tricks fail, you can use active EMC to cancel out the EMI

Like this for DM noise:

Or like this for CM noise:

Cheers

Klaus

GNDs are bad news for EMI and ESD discharges, all practioners , OTT and Morrison, advise against them. You defeat the purpose of an isolation xfmr by GNDing a center tap.

Yeah- always recommended if the environment is severe, kills the capacitive coupling into your receiver.

Can you do synchronous (lockin) detection? (I know little of transformers.)

George H.

I should have deleted item a) in Spehro's post, I only meant that item b) was the way to go.

Jon

Amen. Microphone coupling has this problem very well studied and the hardware support is out there.

Yup, this is the front end for that. The full scale is only a handful of nV.

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

Thanks, Phil. there's a lot to the analogy with mic inputs, despite the many orders of magnitude frequency difference.

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

How about installing an amplifier close to the signal source ?

The signal levels would be larger and less sensitive to interference. Use phantom power feed if the number of wires is limited.

Thanks for the suggestion- yes, that would be ideal, and I've considered it in some detail, but the environment where the source is located is constrained and rather hostile to conventional electronics so unfortunately it's a no-go.

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

** Your " shielded twisted pair " is what exactly ?

Cat5 with a foil screen ?

Cheap twin core with a copper braid ?

A signal in sub mV range at a few hundred kHz might best be transferred and protected from EMI by good quality coax cable. Foil shielding preferred - and long as the foil is properly RF grounded the receiving end, should be almost immune to EMI.

FYI: Co-ax cable rejects line frequency, magnetic interference just as well as twisted pair, sometimes better. Balanced inputs are only needed when there is a known source of common mode signal to deal with - like ground loops.

So if your signal source is floating, foil shielded co-ax is best. Terminating it with the correct impedance eliminates capacitive loading too.

.... Phil

Similar enough to shielded Cat 6 from an EMI pov. There's a bunch of complexity to the system which I can't go into here. If you're curious I could go into more offline.

There are ground loops for sure. It's sub-microvolts- any small voltage from SMPS capacitive coupling or pickup from cell phones or whatever will cause currents in the shields.

A resistive termination would greatly compromise the SNR even with zero EMI.

Is there a standard reference book for high-end and professional audio work? Grounding, balanced line level analog input/output and so on?

--sp

--
Best regards,  
Spehro Pefhany 
Amazon link for AoE 3rd Edition:            http://tinyurl.com/ntrpwu8

What is the impedance level of the signal source ? Does it differ much from the CAT6 100 ohms ?

If there are temperature restrictions at the signal sourse and active amplifiers not possible, how about a transformer at the signal source, to increase the impedance to something like 100 ohms ?

-----

One traditional method is driving the common shield aroud a twisted pair is to drive the shiled with the common mode voltage. Is this possible ?

----

Since you seem to be afraid of both SMPS as well as cellular EMC, have you really thought about your gain distribution ?

With your initial design with a less than 1 MHz signal and a step-up transformer, killing the cellular EMC before the transmformer should be quite easy.

The SMPS EMC (and AM brodcasts at least in the US) might be a bit harder, since it might be close to your frequencies. However a passive HPF between your transformer and your _first_ amplifier would help alot and maintain the SNR. A more selective (and lossy) bandpass filter between your first and second active stage should protect the second stage from overloading.

The point is that you need to reduce the bandwidth each time you are applying active gain to protect saturation of the next stage.

** Really?

So you claim to have a balanced, TP line operating from a high impedance, sub microvolt RF source of a with a STEP-UP transformer the receiver end ??

I give up.

This is bollocks.

..... Phil

I have been wondering about this too.

At room temperature the thermal noise is at -174 dBm/Hz. With some stellar antennas (e.g. cold sky with satellites and radio astronomy) you can go below this.

Since the OP was talking about transformer upconversion, the original impdance must have been well below 100 ohms. You really need to do a proper power match.