Noise issue on T7 in differential mode

[MarkRivers]

I have found a strange problem when using the LabJack T7 analog inputs in Differential mode. My test configuration is the following:

• High-precision Intek power supply.
  • +Vout connected to AIN0 on T7
  • -Vout connected to AIN1 on T7
  • Voltage output set to 5.0532 volts
• T7 power from "SUNNY" USB power supply provided by LabJack.
• T7 connected to Ethernet
• No other connections to the T7
• EPICS software
  • Configures the T7, only AIN0 enabled, +-10V range, default resolution
  • Waveform digitizer for 4096 points, 10 kHz acquisition
  • Poll sleep=1 ms, ~300 samples/s
  • Analog input record averaging for 0.1 second, so about 30 samples/reading

This is the Instek power supply connected to AIN0 and AIN1:

This is the LabJack supply USB power supply:

This is the T7 in an enclosure. The power supply normally supplies the +5V to the T7 USB connector, but for the first part of this test it was disconnected, and the external SUNNY supply was used. The only connections to the T7 are USB power, Ethernet, and AIN0 & AIN1 from the Instek Power supply.

This is the input of the waveform digitizer over the full .4096 second acquisition. The maximum value is 5.05, but there are lots of negative spikes down to about 3.5 V.

This is the same data but only looking at the first 0.1 seconds. The spikes are at exactly 120 Hz, and are about 1.25 V in magnitude.

I then changed to use the USB power supply inside our enclosure, rather than the SUNNY external supply from LabJack,

This is the waveform digitizer using this supply. The range of the plot is only 0.01 V, and the noise is just a few mV.

Why are there 1.5V noise spikes at 120 Hz when using the LabJack USB power supply, but no such noise when using the power supply in our enclosure?

[MarkRivers]

These screenshots show the configuration details for the T7.

[MarkRivers]

I then tested using single-ended mode, rather than differential mode. I switched the -V output of the Intek power supply to be connected to GND on the T7, rather than AIN1.

This is the trace when using the LabJanu SUNNY external power supply. It is excellent, the noise is less than 1 mV.

In this case the noise with the internal power supply is several mV, and is actually larger than with the external SUNNY supply,

[MarkRivers]

I can make the problem go away in differential mode if I connect the GND from the power supply to the GND of the T7. This is a trace in differential mode using the SUNNY external USB supply. The large noise spikes are now gone, and then noise is less than 1 mV.

I don't understand this. If I am using differential mode why does the ground of the T7 care about the GND of the power supply? Furthermore, the Instek power supply and the SUNNY USB power supply are plugged into the same 115V outlet strip, so their grounds should already be the same, correct? Except the SUNNY supply is only 2-prong, so it is not actually connected to ground?

[MarkRivers]

This issue was originally raised because Jesse Smith at HP-CAT was observing noise on his T7 when reading the voltages from a system with capacitance sensors and an amplifier module. He observed a lot of noise when using the LabJack USB power supply, but not when using USB hub connected to a laptop.

This is the T7 ADC trace when using the LabJack power supply. There are lots of positive spikes.

After we realized that the problem could be that the sensor amplifier ground was not connected to the T7, he connected that ground. The trace then looks like this:

The T7 now works well for reading the capacitance sensors.

[MarkRivers]

The following is an explanation I received from the LabJack support engineer.

That is interesting. My best guess at this time is that it has to do with the power supply's isolation. The combination of power supplies and analog input configuration creates four scenarios:

Differential + LJ (isolated) supply - In this case the gnd of the LabJack is floating, there is nothing stabilizing it relative to the signal. Any charge getting onto the line will cause the LabJack's ground to drift relative to the sensor's gnd. From the graph, it looks like some charge is getting quickly induced then decaying out. My guess is that the charge is draining back through the differential signal, which must have some sizable impedance.

Differential + Alternate Supply - If this supply is not isolated or is connected to the signal generator's power supply, then the GND connection provides a path for quick equalization of the charge imbalance.

Single-Ended + LJ (isolated) supply - The SE connection provides a good ground connection which keeps the T7's and the sensor's gnd equalized.

Single-Ended + Alternate Supply - The alternate supply combined with the SE configuration providing two ground returns. That creates a ground loop. The loop will cause GND to flow through unexpected paths and can even alternate between the two. That usually causes increased noise, but can be much worse.

Would you test with the Differential + LJ (isolated) supply configuration with the addition of a GND connection between the sensor's power supply and the LabJack?

Best regards,
Steve

[kurtgoetze]

Mark, Re: "If I am using differential mode why does the ground of the T7 care about the GND of the power supply?"

https://support.labjack.com/docs/differential-readings-app-note

See "Differential inputs must have a reference"

[kurtgoetze]

" Except the SUNNY supply is only 2-prong, so it is not actually connected to ground?"

Yes