Noise-Free Cable Transmission Of VLF Signals

[Melt-O-Tronic]

I'm trying to help a close friend solve a problem in seismic research (academic, not petroleum).  He needs to get the output of a function generator through several hundred feet cable to a shaking truck that renders the waveforms as physical vibrations in the earth.  However, they are plagued by electrical noise -- often a 60 Hz hum picked up by the cables which the trucks dutifully turn into actual ground vibrations.  This tends to only happen in electrically noisy environments.

The cables, as I understand it, are shielded untwisted pair and the frequencies they're working with are 0.5 Hz - 100 Hz with a peak amplitude of 2V.  They may be sinusoidal at times, triangular or ramp at other times or arbitrary waveforms.  I'm interpreting the problem as common-mode pickup, so my first thought is to use differential signalling.

Due to my amateur radio focus, I considered transformer coupling, but at 100 Hz max, maybe not.   Then I thought of mixing the VLF signal with a 2 MHz or so RF carrier and demodulating it on the other end.  I simulated that on the bench with a pair of Mini Circuits ZFM-2 mixers (IF = DC to 1 GHz) and it's not working out so well.  Perhaps this won't work for such low frequencies with a diode ring mixer?

I'm guessing there's a much simpler way to tackle the problem.  I'd be happy to share my test setup if you'd like to point out what I'm doing wrong and appreciate any tips or pointers.   


EDIT:  Within minutes of posting, I got it to work!  I'm recovering accurate waveforms on the other end -- I just needed to adjust my levels because I was overdriving the IF port on the transmit side just a bit and getting distortion.  Anyway, I'm still thinking there is probably a much better / simpler way that I'm missing.  What do you think?

[chuckb]

Jensen Transformers are designed to solve this problem.
http://www.jensen-transformers.com/

See the attached files. These models work down to 0.2Hz with -1dB loss. They have over 100 dB of isolation at 60 Hz. The transformers are also half the cost of the packaged solution.

Lots of app notes on the site for this kind of problem.

[Melt-O-Tronic]

Nice!  I didn't know such a thing was out there for such low frequencies!  The PO-2XX looks very promising.

As I continue thinking through this, I'm wondering if a 4 - 20 mA current loop would also be appropriate for the scenario.  The relationship between the time and amplitude domains are the important factors and I'm thinking current loop covers that.

[David Hess]

The common mode difference may be considerable over 100s of feet unless special precautions are taken so I would use a solution which includes galvanic isolation.

Modulation to a higher frequency so an RF transformer can be used for galvanic isolation is one way.  I would probably use a linear optocoupler and isolated power supply on the receiver side.  Another way would be to convert to pulse width modulation, differentiate the output so it can go through a transformer, and then drive a set/reset flip-flop to recover the pulse width modulation on the receiver side.

[chuckb]

With two hundred feed of excitation cable the capacitive load on your driver stage may cause phase shift issues. The capacitance per foot varies with the type of cable but I usually use 30 pf / foot to estimate it. The driver stage should have a minimum or at least a know phase shift when driving a 6000pf load.

I'm not familiar with a current loop setup but some part of the cable driver and receiver design needs to have a good CMMR at 60 Hz.

[David Hess]

With two hundred feed of excitation cable the capacitive load on your driver stage may cause phase shift issues. The capacitance per foot varies with the type of cable but I usually use 30 pf / foot to estimate it. The driver stage should have a minimum or at least a know phase shift when driving a 6000pf load.

Source and load terminate the cable and the capacitance becomes irrelevant.  In practice only source termination which will lower the bandwidth depending on cable length do to capacitance as you describe will work fine at low frequencies and isolate the driver; the cable capacitance becomes part of the noise filter.

[Melt-O-Tronic]

I'm ordering an assortment of loop drivers, isolators and current sense amplifiers from DigiKey to experiment with.  This will be my first time tinkering with current loops, but I have a some books, the internet and two Fluke 789 Process Meters to work with.  The principles look pretty straightforward.

I'll be starting with some XTR117 transmitters, RCV420 receivers, linear optoisolators, an isolated DC-DC converter for loop power and several other parts.  Other suggestions are welcomed.

Should be fun! . . .

[BrianHG]

Here is flat to 10hz:
http://www.lundahl.se/input-and-line-output-application/
and here:
http://www.lundahl.se/pro-audio/

You have a variety of taps on each one so you can configure signal gain and convert back and forth from balanced to unbalanced.

[BrianHG]

Jensen Transformers are designed to solve this problem.
http://www.jensen-transformers.com/

See the attached files. These models work down to 0.2Hz with -1dB loss. They have over 100 dB of isolation at 60 Hz. The transformers are also half the cost of the packaged solution.

Lots of app notes on the site for this kind of problem.

Nice for bridging a high end balanced DAC's output to drive a vacuum tube gate, for a super Tube-DAC.

[ruairi]

I think you might indeed be overcomplicating this a little?  In the audio world we are are sending signal over hundreds of feet of cable all day every day, even at very low levels and with negligible hum pickup.

First, how much interference/hum/distortion is acceptable?  Without a target spec we are shooting in the dark (this matters for phase and frequency response too, see below).

Some thoughts, roughly in order of signal flow...

Sending the signal over a differential pair is definitely the way to go.  At the transmit end you will need to ensure that the output impedance of the single ended to balanced amp is matched as closely as possible on each leg (stability of output amp into capitative loads will be a consideration, see more below).

You might consider a Star Quad cable for interconnect,  see this video for an example of it's superior hum rejection properties - https://benchmarkmedia.com/blogs/application_notes/117842759-star-quad-cable-demonstration-video

Obviously you don't need an esoteric audio cable, but a reputable performer will be required, look at Belden.

Is cable durability a concern? 

Note we would typically be concerned about the higher capacitance of Star Quad in a long audio line causing high frequency rolloff but with your 100Hz upper limit you can ignore that. You do however need to consider the capacitative load of the cable and how it will effect the stability of your balanced line driver amp.  As this will be a dedicated one function rig you can easily optimize that circuit for stability.

At the truck / receiving end you will need an input amplifier with a high impedance and very high common mode rejection ratio (CMRR).  Again in audio we are usually faced with a noise vs CMRR tradeoff but you will have an easier time with that.  The differential to single ended receiver will need to be optimized to prove the truck with a suitable signal.  THAT Corporation make a line of line receivers with a clever bootstrap circuit to give high CMRR - http://www.thatcorp.com/Balanced_Line_Receivers.shtml

Transformers can offer some very useful benefits like galvanic isolation and high CMMR but distortion rises considerably at low frequencies due to various mechanisms, some of which you can optimize, some not.  This may or may not be an issue in your situation.  Jensen do indeed make some of the lowest distortion audio transformers on the market. Even the best transformers have a natural 6dB per octave high pass function that will have an effect on phase and frequency response.  Again we are back to needing a minimum spec.

In truly hostile electrical environments you might need to consider a simple low pass filter on the truck side at chassis input to filter high frequency (anything above signal of interest) noise before it demodulates.

There are smarter audio people here than me so I'm sure you'll get some useful input - paging Richard Crowley...

Cool project by the way!

[dmills]

Concur, this is a trivial audio problem solved every day by every concert audio company in the world.

Balanced line, screened twisted pair (Or quad!), with the screen connected directly to chassis at both ends (See the AES-48 standard for why!), and either transformers or the THAT line rx chips (Which really are very good, and probably more linear then the transformers at very low frequency).

While you can skin this with off the shelf isolating transformers, be a little careful as most do not do that well at very low frequency, you have to spend proper money to get the good iron (Jensen or Lars Lundall being the names to conjure with, but a sand based solution should probably work and will be cheaper and flatter).

Regards, Dan.

[David Hess]

Concur, this is a trivial audio problem solved every day by every concert audio company in the world.

It may not be as simple in an industrial application.  We do not know what is going on here but the possibility exists that there is a large difference in common mode voltage between the source and destination.  Galvanic isolation neatly solves this however an integrated differential receiver or instrumentation amplifier has a limited common mode range which may not be enough here.

The various THAT line receivers are limited to common mode input voltages roughly within their supply voltages although there are line receivers which do not have this limitation and they might be suitable like the LT1990 which has a common mode input voltage range of +/-250 volts on +/-15 volt supplies.