Signal Transformer Bug

If the signal is from the current shunt of a large DC motor, is there any chance of using a Hall effect sensor to get this signal?

If you still wanted to try opto-isolators, there might be a way to amplify the signal into suitable range, right?

Example: say the incoming shunt signal is 150 mV. If you put a factor of 20 amp on that, you get a boost to 3.00 V. Surely the isolator would transmit that signal faithfully.

If the signal has to be put back to original scale a divider will do that handily.

I tried a really nice opto-isolator circuit that worked beautifully for DC and sinusoidal signals of some 10-12 V rms, but it just would not work for the shunt voltage levels.

I trying to understand what is wrong with the submitted picture. That yellow trace has a peculiar shape where it is distorted slightly at the same time the blue trace is flat. I'm trying to determine if I've encountered saturation of the magnetic core or if I'm leaking stored energy and thus not re-creating the waveform faithfully. I think it looks like two problems instead of one because of the difference in the distortion.

It almost looks like current is being limited due to core saturation. Or it could be a drive problem where the back EMF is suppressing current before the square wave half cycle. A couple of decades ago, I might have remembered how all of that works. I'd get back to basics if I could just remember where I left them.

Amplifying the original signal requires a ground connection which is at a different potential than the motor. That's the beauty of the opto isolator. It doesn't care about different ground potentials.

I tried using a dual pair of HCNR200-000E optoisolators in the circuit provided by the manufacturer, but the shunt voltage seems to be too small. I will contact the manufacturer for suggestions if I can't figure this out. For reasonable signals that opto isolator circuit really did a great job. But my hands are somewhat tied when it comes to changing shunts, etc. It's kind of frustrating.

This project was the first time I have ever been required to make my own signal transformers. I managed to get the ratio of windings about right, but I'm trying to decide if I need a different magnetic core. That waveshape should have told me the direction in which I need to work, but those brain cells haven't been activated for at least 40 years. Sinusoidal signals through the transformer coils don't exhibit that dropout.

A DC input causes a level shift, not a phase shift, in the original circuit. The square wave remains square and there is no ringing or slump in the square wave of the original design. And it seems senseless to have to stick with the original concept, but that is part of the requirement. It literally is giving me headaches.

Have you considered stray inductance, capacitance contribution to those waveforms?

Impedance mis-match is something else to peer into, and see if there is tuning to be done between output and input of scope.

Good point, you need to calibrate the scope probe with the square wave calibration signal on the scope. A miscalibrated probe will often result in overshoot that looks a lot like the square wave shown.

Any chance you could bias the signal high enough to run the optoisolator?

The only other thing I could suggest would be to put an amplifying transistor or op amp ahead of the optoisolator.

A picture is worth a thousand words. If you could post a picture of the schematic and the points you are measuring the waveforms, it would help us to help you.

I'll second that.

A capacitance & resistor timing circuit was my first thought. Then I thought the resistance would be in the form of an inductor. A picture of the circuit would be a help. Lets know the final outcome please.

Would the circuit by any chance be one of the Reliance CVTB circuits?

If it is Reliance I have a lot of the manuals and engineering notes.

Burr Brown used to make a linear opto-isolator "ISO 100AP" - it isolated shunts beautifully.

Not sure if there is an equivalent circuit today. E bay has some available.

If a picture is worth 1000 words, I guess I was a little short in my original description, so here is the problem in a picture. What happened to my signal?

That sure turned out fuzzy. In any case, there would be no signal if the shunt had no voltage on it. In this test case, the simulated shunt voltage was 0.2 VDC and the more voltage that is added, the more the output changes shape. On my 40 year old circuit, the output looks like a full wave bridge output before filtering. That is to say that the output is a pretty decent reproduction of the input even if it is DC or a low frequency AC signal.

Lastly, I may have swapped the Source for the Drain on the output side (on the right). I'll have to look, but that might cause this. The 40 year old version produces a nice DC signal with a little bit of ripple for the same 0.2 VDC input.

It looks like there are two effects going on at the same time.

Just checked. FETs are installed as shown. That does not preclude the possibility that I wound one of the coils backwards, but I should have realized it if I did. Here is the prototype. It has two of these circuits on it.

P6 and P7 keep my square wave at a duty cycle of 50%. Transformers are hand wound.

The circuit takes the DC input voltage and modulates with the two input FET's. The output is a demodulator using the FET's. The gates are synchronized so input and output FET's should be synchronized.

Using a dual channel scope with isolated probes you can see which FET is not switching.

The final Output OA is amplification and some filtering.

It seems so obvious when you put it that way. I'll let you know what I find.

This circuit is known as a "chopper" in its most basic form. In the days of vacuum bottles, they did not make drift-free DC amps. One solution was to "chop" the DC into AC with a vibrating switch (with gold or platinum contacts for millivolts) and follow with an AC amplifier. Anyhow, the point is that the circuit will work with a "half wave" switch. It also works without the synchronous output switch, but without the benefit that the detector responds only to signals at/near the switching frequency and its harmonics, ignoring power supply frequencies.

Imagine it with one switch on the input side of the transformer and a synchronous switch on its output side - the input switch is on for half a cycle. If you get the synchronous output switch operating on the wrong polarity, then instead of selecting the desired copy of the transfo input, you select whatever ringing or pickup is on the transfo when the input switch is off.

So the polarities, shown by the dots on the transfo windings matter and swapping over the winding polarity at one pair of FET gates may fix the problem. All the dot ends of the windings should be positive to the other end of their own windings at the same time. Otherwise, the detector FETs are turned on when the drain is negative to source!

All the commons could be joined [bypassing the isolation] for test purposes with the DC input from a battery & resistive divider - but the "production" solution is to wind all the transfos in the same way with the start/dot ends all identified and checked, then connect them consistently to circuit boards.

I would check that the wave form at the secondary winding of the second transfo is a faithful copy of that at the primary of the first transfo, to be sure there is no saturation occurring. If not, go back to the secondary of the first transfo, to be sure there is not a problem with the first transfo.

A point on layout - the 470k [looks like k, but fuzzy] gate resistors should be as close as possible to FET gates, else it being very high impedance, long wires connected direct to the FET gates can easily pick-up noise.

Well, the part that was causing the trouble was far from obvious. I did replace the two FETs on the right side of the circuit, but very little changed.

As it turns out, my noise suppression devices on the plus and minus 12 VDC supply consisted of a pair of Common Mode Choke followed by a 0.47uF capacitor. When I added a pair of 220 uF capacitors in parallel with the noise suppression capacitors, the problem went away completely. Now the goofy signal is gone. It works just like the old tech.

I have to tell you that I didn't expect that. Without the CM chokes the old machine would have fried all of my transistors. The CM noise consisted of 150 V spikes due to the age of the design. Today's systems have filters and other things to keep that kind of trash out of everything else.

Conclusion: If you use CM chokes, make sure you have sufficient capacitors on the far side so the choke will not starve your main circuit for DC power.

still curious - who's design was it originally? - the topology is so close to the '70's Reliance Electric isolators for their drive controls.

Well, the company is still in business but they have no electrical engineering people any more. Their business track has turned as happens with marketing demand. I will tell you that the circuit was from a company in Europe. It really is a common sense circuit but my error was due to not considering the effect of the series inductance of the common mode choke. With that frequency and the square wave drive circuit hammering the DC buss, that choke was reacting to the current draw on the plus and minus power supplies 180° out of phase. Adding large electrolytic capacitors next to my chokes (inside my circuit) the interference seemed to disappear.

I suspect the technique was fairly well known. The use of signal transformers has diminished substantially since the 1970's. The input signanls came from current shunts in a DC motor drive.

Nice to hear the end of the story...
I was expecting Jessica Fletcher to have to turn up an solve it
Del