Ammeter giving almost random readings

[jaba-who]

I have finally got around to addressing a problem with my ammeter present from when I initially installed my D-180 2 years ago.  

Basically the ammeter is little more than a random number generator. The pattern is for it to start about - or + 1 or 2 then to fluctuate up and down in sort of sinusoidal pattern sometimes dipping down to -6 or -7 though not often peaking up over +3 or 4.

It does this no matter what the connections are. I have tried it with the sensor wires disconnected from the shunt, shorted to each other, with and without the main 12v wires connected to the shunt and even with all the rear looms/connectors (and thus all wires) removed from the dynon itself.

With the sensor wires shorted to each other the spread was all - amps but still in a sort of waveform.  

It doesn't really seem to be altered by real loads too much - perhaps at most a tendency to run in the - range.
Any hints?

 

[jakej]

PM me, I may be able to help

Jake J

 

[G3WGV]

I've just started my Rotax 912ULS engine with the D120 EMS and have this fluctuating ammeter problem, but only when the engine is running AND the alternator is charging. I can turn the alternator off with the engine running and then the ammeter behaves perfectly, showing a steady discharge of reasonable proportions. I DO have the 22,000uF capacitor fitted across the alternator output, per latest Rotax instructions.

I'd be grateful for information on any fixes that may have been discovered.

 

[Roger_Lee]

A poor ground will cause this. We have this same problem on Flight Design CT's. Go back and tighten all grounds and make sure they are solid and no paint under them. Put a screwdriver or wrench on each ground whether you think it is snug or not. I have found several with this problem and it has usually been a bad ground.

 

[G3WGV]

Thanks Roger.

I am as certain as I can be that grounding is not the issue but I will check it out one more time. I suspect the problem is that the ammeter logic samples the voltage across the shunt at a very short instant in time, rather than integrating it over time like an analogue meter would do. Therefore any noise (which has to be coming from the alternator, because the problem is absent when the alternator is turned off) gets counted as current.

A further curiosity is that the voltage indication is steady and reasonable. So that tends to rule out noise on the +12V line itself. The shunt is 1mV/A, which means that the ammeter circuit is measuring very small voltages in the handful of mV range. It doesn't need a lot of noise to be induced into the unscreened wires to the D120 for the readings to go haywire.

I think the answer is some sort of integrating circuit or filter. Does anyone know what the input impedance is on the D120 ammeter lines? It shouldn't be too hard to make something up with a few resistors and capacitors if the impedance is fairly high. I will measure the DC resistance of the ammeter inputs today, which will give me a clue.

 

[pbennett]

I'm not familiar with Rotax but Lycons usually have a switched regulator. This turns the alternator field supply on when the bus falls below a given voltage, and switches it off when the voltage rises above the threshold. The lag in the electromagnetics means that the bus voltage has a low frequency "wave" on it, imposed by the alternator. This may only be a few tenths of a volt but could cause significant current fluctuations into a partially charged battery, which has a very low dynamic resistance. Like you I'd have thought the EMS would integrate over several seconds, and I'd imagine it would take a sufficiently large number of samples to average out the "wave". It's a long shot but to see if the problem might be to do with this, try charging the battery to full charge then run the engine and check the current under different loads. You might have to wait while the alternator replaces the starting charge.
DS advises that the input resistance is a couple of K ohms. See http://dynonavionics.com/cgi-bin/yabb2/YaBB.pl?num=1266633320

Peter

 

[G3WGV]

Thank you for that Peter. I think the Rotax regulator is a bit less brutal - it is electronic, contactless. The alternator itself is a rotating magnet type, so there is no field coil. I don't have the regulator circuit diagram to hand but I expect it's a solid state constant voltage regulator, since that is the correct charging profile for a lead acid battery.

A couple of K input resistance eh? I reckon a 100R in each lead and a few uF across the leads on the D120 side should show some sort of improvement if the problem really is current noise. It's a pity the input impedance isn't higher, although that would create its own set of problems, no doubt. As it is the 100R's will make the meter under-read current by about 10%, which I could live with.

 

[dynonsupport]

The amps inputs are about 33K to ground.

But the resistor / capacitor trick doesn't work. We've tried it before. The EMS doesn't load the two lines identically, so when you add a resistance between the source and the EMS, the 100 ohms causes so much difference in voltage that the ammeter just pegs at 99 amps.

We've run into this before, and it's always on Rotax and Jabiru engines. The other stuff never seems to have a problem. We're not 100% sure why it happens, and we can't just up our filtering in the EMS since part of the issue is that the variation is fast enough that it's faster than our sample rate and causes aliasing.

 

[G3WGV]

The amps inputs are about 33K to ground.

But the resistor / capacitor trick doesn't work. We've tried it before. The EMS doesn't load the two lines identically, so when you add a resistance between the source and the EMS, the 100 ohms causes so much difference in voltage that the ammeter just pegs at 99 amps.

Ahha, I did wonder whether that might be the case. Nothing is ever easy in avionics, eh!

We've run into this before, and it's always on Rotax and Jabiru engines. The other stuff never seems to have a problem. We're not 100% sure why it happens, and we can't just up our filtering in the EMS since part of the issue is that the variation is fast enough that it's faster than our sample rate and causes aliasing.

Well I am an electronics engineer with good process control/ computer systems knowledge. Now I know it's not just a simple filtering issue I can investigate more thoroughly. I feel a bit of 'scope probing coming on. I'll probably be back asking questions about the interface once I have some more test results. It has to be possible to fix this!

One question already - what level of integration does the D120 do? Does it simply sample the instantaneous voltage (== current across the shunt) every now and again, or does it attempt to integrate a number of samples to arrive at one reading?

 

[dynonsupport]

There is some filtering done, it is not just one sample. I think it's a 2 bit digital filter with samples at 4Hz.

A scope trace of the shunt voltage would be nice to see. We've never been able to get one.

If you really are into it, the best thing to do would be to build a filter into a buffer and then feed us a single ended voltage.

 

[G3WGV]

On getting a scope trace of the shunt voltage, are the ammeter inputs referenced to ground/some other reference, or are they truly floating? This has relevance for how I attach the scope to the shunt. Can I assume that all the Dynon is looking for is differential voltage between the two lines, at the rate of 1mV per Amp?

Please amplify how I would supply a single ended voltage to the Dynon, given that it has, I am assuming, a differential input. What would the voltage be referenced to?

I am certainly willing to put some effort into this. If you would be willing to let me have sight of the sensor input circuitry then that might help.

 

[dynonsupport]

They are truly floating when using a shunt, but we also need a common mode at least 6V or higher.

We also support single ended as a "hall effect" sensor, You can hook a single ended sensor to the AMPS+ line (pin 24). When you do this, we expect a rate of -16.3mV per amp, with zero A at 2.5V. Note the amps are negative against voltage. We do support an offset in software in case your zero point is not dead on.

 

[G3WGV]

Excellent. I can work with that, thank you. I'll be using a 100MHz PicoScope and will be able to record many minutes of data with a sample rate of 1ms or better. Unlikely I'll be able to do this until the weekend but I WILL get back to you with some results.

The single-ended approach will be my fallback option but it ought to be easy enough to engineer with an op amp.

Technology... don't ya just love it? ;D

 

[G3WGV]

They are truly floating when using a shunt, but we also need a common mode at least 6V or higher.

Sudden thought... might the problem be poor common mode rejection? You need a common mode of >=+6V but what if that voltage is wandering around? Something you said about not loading both ammeter lines the same just makes me nervous of possible common mode issues.

 

[Bo_Hannington]

I think what John says about common mode is quite likely to be right. Trying to measure a voltage difference of a few millivolts accurately (i.e to better than 1mV) between 2 wires both at around 13,700 millivolts (say) is not that easy, and common mode drift may well be part of the problem.

Bo Hannington

 

[G3WGV]

Indeed, Bo. I suspect that the reason for requiring a >6V common mode voltage is that the two ammeter inputs feed into an op amp in differential fashion. Achieving a common mode rejection ratio in the order of 13700:1 (about 84dB!) is quite a tall order. But why is it only apparent when the engine is running? That's the conundrum!

 

[G3WGV]

I can't get to the airport to put the 'scope onto my ammeter shunt until tomorrow.

Meanwhile, I thought I'd investigate the GRT CS-01 hall-effect transducer. Google is unusually silent about this device. Does anyone know where it can be obtained, ideally in the UK but if necessary shipped from the US? Apparently they cost about US$60, I've found that much out. If it fixed the problem then I'd probably be willing to spring for that. The ammeter shunt is so 1900's after all!

EDIT:
I've now worked out that GRT is Grand Rapids and with that info was able to find a few places in the USA that stock it at $60. But why use a CS-01 (±100A)? I can't ever imagine my aircraft load/charge exceeding ±30A or so, so can the CS-02 (±50A), as stocked by Aircraft Spruce, be used instead?

 

[G3WGV]

I wasn't able to get the aircraft out of the hangar over the weekend and so was unable to do an engine run. However, I did connect up my digital 'scope and experiment with getting current readings, measured as mV across the ammeter shunt.

A problem that immediately became apparent is that dealing with such small voltages is very hard. The 'scope itself has a noise floor in the 100-200uV region and I cannot zero the probe offset to better than 4mV. I still think the results are instructive and may yet yield some useful data when I can run the engine.

Meanwhile, here's some static analysis.

In the above trace, the first 3 seconds are with all aircraft power off. You can see the 'scope noise floor and also the fixed offset of about 4.5mV. Negative excursions are battery discharge, positive are the battery being charged.

At 3 seconds, I turn on the aircraft master and this in turn brings on the Dynon D120 EMS. There is a step down of about 1A, which I interpret as the master solenoid coil current. Shortly afterwards, a further load of about 1A is visible. This is the D120 coming on. Note two things:

1. There is a substantial increase in current noise
2. There is a period of about 1 second when the current drops back to about 0.5 A and the noise returns to background level

At second 5, the Dynon screen is showing and there are noise spikes all over the place. Not on this trace, is a test where I turned off the D120 which proved that this current noise is produced by the D120.

I realised that although I could not run the engine, and therefore the alternator would not operate, I could arrange a charge of about 5A using my battery charger, applied at the output of the regulator. In the second image, you can clearly see the step up when the charging current is applied (note that the vertical scale has changed). The D120 correctly showed this charge current. There are no new noise spikes and the D120 showed a steady charge.

Does this get us anywhere? I think it does. We now know the static (engine not running) current profile and we also know that the D120 puts quite a lot of current noise on the power line. We also know that a charge applied from a battery charger shows correctly and is stable, both on the 'scope trace and also on the D120 ammeter display.

I will try to run the engine with this test configuration later this week. If there are any other tests I can usefully do, please let me know. I can expand any part of the display, if you want finer detail. The sampling period was 1ms.

 

[G3WGV]

I haven't forgotten about doing these tests with the engine running! Unfortunately the weather has been poor recently, definitely not conducive to getting the aeroplane out of the hangar and running up the engine. Hopefully we'll get some better weather soon.

 

[G3WGV]

We also support single ended as a "hall effect" sensor, You can hook a single ended sensor to the AMPS+ line (pin 24). When you do this, we expect a rate of -16.3mV per amp, with zero A at 2.5V. Note the amps are negative against voltage. We do support an offset in software in case your zero point is not dead on.

Although I haven't given up on trying to find the reason for fluctuating amps using the shunt, I'm now looking into getting the CS-01 Hall effect device.

I see from the specification that it requires a supply of +4.8V. Is it intended that we should use the Dynon 5V excitation circuit (pin 18 on the D120 37 pin connector)?