Differential Inputs and Noise Rejection---How simple?

Hello Folks,

Here is one for circuit connoisseurs and creative types: What is the simplest possible way, barring the use of transformers, to reject noise due to a noisy source "ground"? Let's restrict the bandwidth of the signal and the noise to the nominal audio frequency range, and suppose that the voltage relative to our local circuit common is fairly small, say less than a few hundred millivolts rms. So what we are looking at as a source is something nominally referenced to our system common, with the whole thing bouncing on the noisy ground local to the source (or from the viewpoint of the source, seeing our noisy ground). The impedances of the lines are anything but balanced, i.e., this is a single-ended, nominally "grounded" source. Let's say the source device has an highish output impedance, in the hundreds of ohms range (but not known precisely). Suppose there is maybe an ohm or two and less than a microhenry, at most, between the source common and our common.

How would you do it? As yet undetermined prizes, at least extending to my appreciation and respect, will be awarded for the solution with the best performance/minimal cost. This should be something that does not require adjustment in manufacturing.

Brad

> What is the simplest possible way, barring the use of transformers, to reject noise due to a noisy source "ground"?

Very often: float the source. Don't tie it to local ground at all. This may require battery operation to avoid stray capacitance to power-lines (though sometimes not).

After that: the classic way is to run a 3-conductor cable, two hots and a shield. The shield is connected only at the destination, and goes to chassis, not some clean ground. The two hots go to a high-Z differential amplifier input (whatcha call an Instrumentation Amp). Their source end goes to the two sides of the output jack.

In scenarios similar to your highly-defined thought-problem, such schemes work well. They even work fine with terrible imitations of instrumentation inputs, such as the ubiquitous 1-opamp "differential input".

> does not require adjustment in manufacturing.

Ah, well, if this is a production problem, you should offer a kickback for each unit produced, not free advice.

But the simplest way to handle practical situations where the conditions are not so constrained is to un-bar transformers. My upstairs booth has two PA inputs: one a sweet diff-amp, the other a hunk of iron.

Thanks PRR---all duly noted and acknowledged.

Of course, I have a solution in mind, and I confess to shamelessly trolling to see if anyone out there has a similar one, or a different and superior one. Nothing is being manufactured as yet, and may never be. But...

A decent In Amp is not anywhere remotely in the budget, and furthermore the cabling and connectors are nearly as bad as they come. Battery operation as well is out of the question, although I may demonstrate the efficacy of the solution by operating one end of the system from batteries, just to make the point.

But you are right---If the solution offered is used and money is made I will compensate the originator in the coin of the realm. I don't stand to make much myself per unit, but there could be a lot of them in time. (insert disclaimer here in the legalese of your choice)

Brad

[quote author="bcarso"]This should be something that does not require adjustment in manufacturing. Brad[/quote]

I am interpreting this to be a very diffcult, if not impossible, to satisfy requirement. What constitutes to adjustment to manufacturing, and what doesn't? Why are transformers barred? What means are acceptable? How about opamps, transistors, cheese or peanut butter?

Think cheap. If the tranformers (2 channels are needed) are less than a few cents each then that's a possibility. But we have to pass 50Hz at least so I doubt they are in the running.

When I say no adjustment I just mean no trimpots or other tweaky things.

If cheese or peanut butter work I'd run with them too

Brad

Can a third wire be added for ground sensing between the input and output of devices?

If so an opamp (plus a few resistors) could add the difference of the two grounds to the signal.

Tamas

yes, it's a three wire cable to a powered speaker, and the third wire can function as a combined shield and ground sense. Only problem is even an op amp is too much money for this app. I know that sounds insane...

PS: You can figure transistors at the far east price of about 1.3 cents. "Lyitc caps of relatively small size are less than a penny, and resistors about free (5%'s anyway---1%'s atad more).

Is there budget or allowance per unit? (This process of trying to coax requirements out of users/customers reminds me of work. :sad: )

Given the small signal one could make a decent diff stage with a few transistors to substitute for the opamp.
What sort of power is available?

The power is dirty 5V d.c. There was no specific budget proposed other than whining about how much money the guy was already going to lose ...of course that got louder when I discussed my potential compensation.

The trickiest part imo is that the source Z can vary, depending on a given configuration, so the effectiveness of most any scheme will be compromised by a need to tolerate source Z variation.

At least the input voltage is only about 1V rms or so in most cases.

Brad,

Has this worked out with somehow? You don't have to disclose technical details, but I am curious if you found a solition that the customer accepted.

Cheers,
Tamas

Yeah the customer has accepted it so far. The funny thing is, the computer soundcard and dirty power problems are intermittent on what was to be the proof-of-design host machine, so now they are looking for a consistently bad machine for the A/B demo to their customer.

The solution was two bipolar Q, 9 R (5% !), 4 ~small 'lytics. Performance is better than 50 dB CMR across the band of interest. I got a verbal agreement from a kicking and screaming company owner to get a fixed few cents a unit shipped if they get the biz. I am sure I left money on the table---I am not good at haggling But if the volumes are as large as they may be, and they get the biz, and they honor their agreement...I could make out well in time.

But one recalls what Sam Goldwyn said about verbal contracts---they aren't worth the paper they're printed on.

Brad