AMC1200: Output noise question

Hi Hans,

I think this is more noise than we'd expect from the AMC1200.

In your schematic it doesn't appear you tie VINN to GND1. This should be done to establish the common mode input voltage. Try adding this first and check the results. 

What does your input look like? Have you probed the input pins of the device? You might add more filtering to the inputs to suppress high-frequency noise. You could also consider adding filtering to the output to dampen ringing effects. 

Please let me know what you see. 

Best regards,

Eva

Hi Eva

The VINN and GND1 are connected together, you cannot see it because I did not show the shunt resistor. You can only see it in the layout above at the shunt left from the amc1200. There is a net tie at the kelvin connection of the shunt, that connects the two together at one point. Am I wrong with that? Should I maybe use bigger traces for the kelvin connection?

The filtering i can of course adjust, but it is still strange, that it produces so excessive noise, so there must be something fundamentally wrong

I will do a remeasurement tomorrow of the input

Hi Hans,

Okay, I understand. You did tie VINN to GND1 correctly, but I think you should still try soldering a direct connection between GND1 and VINN when you debug next. The ground plane could be causing some of the noise. Let me know what you see on the input tomorrow and I'll look into what else could be causing the noise. Thank you!

Best regards,
Eva

Hi Eva

What do you mean with a direct connection between VINN and GND1? No Kelvin connection for VINN, but directly solder VINN and GND1 together at the IC input pins? 

In the datasheet, there is a layout recommendation, but it seems not like GND1 and VINN are connected close to the IC and since there are vias, there must be more than one layer used.

I have to do the input measurements on monday, I am out of office today.

Thanks,

Hans

Hi Eva,

I could do again the input noise measurements, but there is no input noise at all (we are directly on the noise floor. The only thing I could see is the supply noise on the VDD1, which has a ripple f +-5mV and some HF switching from the PSU:

50us resolution:

2us resolution:

From the datasheet, the 20kHz ripple should be attenuated by around 73dB, which should be deglictable:

About the higher frequency switching noise, I don't know if this would be filtered out by the AMC1200. I changed the the power supply to the external linear one and had less ripple and no HF parts, but the output noise waveforms look exactly like in the pictures above. So, I don't think it is an input noise issue and it really looks like the AMC1200 noise is like that.

I also shorted the input leads out, but there was no change at all in the output noise. When I disconnect VDD1 from the input stage, the noise is gone.

It would be amazing, if you could test the noise floor internally on a AMC1200.

Hi Hans,

Thank you for getting these figures for me.

As far as connecting VINN to GND1 directly, we usually do this after the capacitor by drawing a line directly from VINN to GND1. I had thought there could be noise on the input because of the ground plane, but this screenshot you got without that trace looks relatively clean. 

To confirm, after changing your input power supply, this cleaned up your input further? And after disconnecting your input power supply, the output noise is gone? What does the output look like without power on the input? If you found that removing your power supply eliminated the output noise, possibly we can consider more options for VDD1. Have you looked at the power supply going into the amplifier? Make sure this is stable and free from excessive noise as this can impact output noise. Parallel decoupling capacitors can help.

The noise floor of the device itself shouldn't be the source, but if you shorted out your input and saw noise that is the noise floor. To reduce the noise floor created, we would still suggest considering your power supply design and grounding strategies. 

Other considerations could be how long the traces are from your output to your filtering. This is a very long path, where are you measuring output from? The output on VOUTN is odd to me and almost looks like three events which I am curious about. 

If you think this could be an AMC1200 concern, you could always try replacing the unit with another AMC1200 to confirm if the error is internal to the device or to the system. 

Please let me know what you try and what results you see. 

 Best regards,

Eva

Hi Eva,

Thank you for your answer,

Eva Mason said:

As far as connecting VINN to GND1 directly, we usually do this after the capacitor by drawing a line directly from VINN to GND1. I had thought there could be noise on the input because of the ground plane, but this screenshot you got without that trace looks relatively clean. 

In the manual of the AMC1200 the VINN is connected to GND1 at the shunt, but not at the filter capacitor. Is this wrongly depicted in the datasheet?

I did not post a picture of the input noise, because it was on the noise floor (around 300uVrms) of my oscilloscope. The pictures from the last post are from the VDD1 power input.

Eva Mason said:

To confirm, after changing your input power supply, this cleaned up your input further? And after disconnecting your input power supply, the output noise is gone? What does the output look like without power on the input? If you found that removing your power supply eliminated the output noise, possibly we can consider more options for VDD1. Have you looked at the power supply going into the amplifier? Make sure this is stable and free from excessive noise as this can impact output noise. Parallel decoupling capacitors can help.

I just switched out the Mornsun DC/DC converter (5mVrms noise), depicted in the initial post, with a lower noise linear PSU, with around 1mVrms noise, but this did not change the output noise at all, also again when shorting the input leads VINP and VINN at the IC. When I disconnect the VDD1 pin ( so the input stage has no power), the output signal is on the noise floor of the oscilloscope. So either the AMC1200 disconnects the outputs when there is no input VDD1 voltage or the noise really somehow comes from the input.

Eva Mason said:

The noise floor of the device itself shouldn't be the source, but if you shorted out your input and saw noise that is the noise floor. To reduce the noise floor created, we would still suggest considering your power supply design and grounding strategies. 

The noise floor with 300uVrms should probably be low enough. This is only the noise floor of my oscillosope, in reality it is probably lower.

Eva Mason said:

Other considerations could be how long the traces are from your output to your filtering. This is a very long path, where are you measuring output from? The output on VOUTN is odd to me and almost looks like three events which I am curious about.

I would be really surprise if this trace length is already too much (maybe 3cm), especially since we have a differential transmission.

Eva Mason said:

If you think this could be an AMC1200 concern, you could always try replacing the unit with another AMC1200 to confirm if the error is internal to the device or to the system. 

I already did, I have 5 boards ready for testing, all with the exactly same behaviour

Edit: I ordered now also the AMC1200EVM board for verification. Do you have the Altium design files for it? Further, it looks like in the EVM they also use a GND plane on the bottom side (not sure if it is only a 2 layer PCB or not).

In the EVM board, it looks like GND is not connected to the VINN pin at all, only capacitively coupled:

But in the datasheet, it is connected together:

So, I got the EVM Board and did a reference measurement, it looks like here the AMC1200 produces even more noise than on my own evaulation board. So it definitely seems, that this sensor is as noisy as I measure now. I also used two different oscilloscopes (one handheld in a different environment) with exactly the same result.

VOUTP:

VOUTN:

Thank you for the measurements