Different Noise from different computers; how?

[Tall_Michael]

NI cards:

6052E

6733

I have a test system; MatLab runs to output a signal through the NI PCI-6733, feeds into a High Voltage Amplifier, a Signal Monitor reads the Current and Voltage across my test piece, and the Current and Voltage signal is sent back in through the 6052E.

I have 3 computers; all have the same version of NI Drivers and Software, MatLab, and operating system. All have well over the specs needed to run the cards. Program files are the same.

However, the clarity of the signal (especially the output signal) is substantially different from one computer to the other. I am using the exact same cards (swapping back and forth for determining noise) and cables on each computer. One has a little noise that tracks the AC line signal from the wall socket. The other two computers have varying degrees of additional noise ontop of this. I am testing at 100Hz, so a good amount of this noise can't just be fitlered or averaged out. I found that even leaning on the case on one of the computers will offset my signal by a few millivolts.

All systems are plugged into the same socket. This is not a regular noise (pulses and changes), but seems to be set at around 60Hz and/or 120Hz.

I ran the test on an identical computer to my best performer and it acted very similar as its brother, as expected.

Any ideas on what inside a computer would be causing noise on the card? Any fixes to this? I am going to be setting up a few systems, and I can't have each one performing differently. Again, software and testing hardware is all the same. Internal hardware to the computer (mother board, power supply, etc.) are different.

-Michael

[Tall_Michael]

Added note: I am outputting a 100mV peak to peak 100Hz sine wave at 80,000 samples a second.

[johnsold]

Michael,

Computers were never designed to be data acquistions systems.  The way some of them are put together makes it very difficult to do any high performance A/D conversion with internal boards.

Most likely the problem has to do with grounding.  The computer is (at least should be) grounded in such a way as to meet the electrical safety codes.  Unfortunately the easiest (= cheapest) ways to do this are often very poor ways to ground from the perspective of signal integrity.  The fact that you ae seeing 60 Hz in the noise largely rulse out the switching power supplies as the noise source.

You mention a High Voltage Amplifier: How high is the voltage? How is the output of the amplifier connected to ground? Is the high voltage DC, 100 Hz, or something else?  How much current flows in the test piece?  How are the measured current and voltage signals obtained from the test piece? Are these signals ground referenced, differential, or something else? How long are the cables?

Do you see noise if you do not excite the test piece with the high voltage? What if you disconnect the cables at the amplifier or at the PCI-6733?

Is there anything else in the vicinity running off the power lines, such as fans, pumps, lights,...?

Lynn

[Tall_Michael]

Lynn,

The test system runs at 100V and at 100Hz, sine wave. We have a custom circuit that monitors the current with an in line resistor. The max voltage off of that which we are recodring through the DAQ is around 60mV. As I am recording a sine wave and concerned about phase shift, presicion is very important, and the spikes at the 60Hz and random inbetween are nearly impossible to filter or average out.

I still see the noise if there is no signal going to the High Voltage Amplifier (with all systems on). The noise goes away if I disconnect the cable at the PCI-6733. Noise goes away if I sever the lines coming from the HV Amp, disconnect the test device, or generally anything to break the circuit. The issue is that the noise from the DAQ is amplified through the High Voltage Amplifier even if no signal is coming through (1V = 1000V). The amplifier is grounded through the socket plug into the same socket as the computer and the rest of the system, and the ground of the HV Output connects to the test device, which in turn has the ground connected to the DAQ (connected into computer, which AC Plug is connected to the common socket).

I have accepted that I will be getting nosie from all of the test systems I have running, but I want it to be consistent between the computers. Once that problem is solved I can work through the non-NI systems.

As for other machines in the vicinity, we do have other test systems going. No other systems are running off of the specific wall socket that my test system is plugged into. The main problem is if I run a different computer in the exact same place with the exact same cards (pull out of one computer and place into another) then the noise level is different. Of the 4 computers I tested, 2 were the same exact model and looked similar as far as noise goes. The other two models looked really bad and different.

You mentioned that computers were never designed to be data acqusition systems; can you suggest a computer (make and/or model) that has good grounding that works well for doing data acquisition?

I would have to check with our EE, but I believe we are set up Differential.

When I can get to it, I will post a picture of the noise I am seeing with details. Thank you for any thoughts.

[Tall_Michael]

Or, perhaps is there a way I can fix the grounding issues within the computer? I have no issues voiding warrenties as it were. I may have an answer (poorly done computer grounds), but what can be my solution?

[johnsold]

Michael,

It sounds like a ground loop problem.  As you described it, the HV Amp has at least 3 different ground connections: The power line cable, the AO ground at its input, and the output ground connected to the test device.  The computer also has at least 3 ground connections. If any of those grounded conductors has current flowing in it, then the impedance of that conductor will produce a voltage drop.  Depending on the exact current path and wiring configuration, that voltage drop may get added to your signal. Since one side of everything is grounded, differential does not give you any advantage.

The ideal grounding system has every device which is to be grounded connected exactly once to ground in a star configuration.  Then if any currents flow in the ground conductors, no votlages are "shared" between devices because none of the impedances are common. That is not possible in your system either because of the internal ground connections in the computer and in the amplifier.

That is the generic explanation (one possible explanation, anyway).

Now, what can you do about it? That may require some trial and error.  I do not like disabling the power line ground connections. If there are serious grounding issues, there may also be safety issues. A laptop on battery would eliminate the computer ground issues but the PCI cards rule that out.  A PXI system or cRIO would get the measurement instrumentation out of the computer and if you communicated the measurement data wirelessly, you would break the ground connection between the computer and the test system.  You might try to create a star grounding system with much lower impedance than the internal wiring.  Select one point to be your ground reference.  The grounded end of the test device has some advantages if it is feasible to use that.  Run a separate heavy wire from the gorund reference point to each of the devices in the system.  "Heavy" is a relative term.  If your signals are on RG-58 coaxial cable or 26 gauge ribbon cable then 14 or 12 gauge wire is "heavy." If your test device draws 50 A, then you probably cannot get away with trying to bypass grounds like this. You can try with and without a ground wire from the ground reference point to the power line ground. Sometimes that helps and sometimes it makes things worse.  This extra grounding might work but usually does not have much of an effect.  It is relatively easy to try and much less expensive than completely redesigning the system.

Have you looked at the current signal with an oscilloscope?  Does the noise on the scope look different with the various computers? Do you see noise if the 6052 is disconnected but the rest of the system is still running?  If you use a line powered scope try plugging it in to the same circuit or a different one.

Lynn

[Tall_Michael]

I will talk to our EE when he is in regarding the ground loops. However, as the problem is fine with one computer and not with the other, I stil lam more concerned regarding the computer guts than the rest of the system.

As far as looknig at the current trace with an oscilliscope, that is how I have been tracking the noise. Disconnecting the 6052 makes no difference, disconnecting the 6733 does remove the noise.

I have attached 2 pictures of the oscilloscope screen; one with the output DAQ card disconnected (showing a 60Hz sine wave with spikes at 60Hz intervals) and then with the DAQ output card connected, showing the same spikes but amplified along with a lot of other crud. I am disconnecting the large shielded cable, not removing the DAQ card from the computer. This is from the better computer, the worse one has large spike when all connected and has noise that does not track with an AC trigger.

[johnsold]

I would not call those signals sine waves.  It looks to me like the current ramps up (as through an inductor) when the power supply rectifiers are conducting for a low duty cycle (like a capacitor input filter).  I would guess that the inductor is in the interference filter at the power line input and the rectifiers and capacitors follow that inside the computer power supply.  Why or how it couples to ground is still a mystery.  The noise which does not rtack on the other computer is probably from the switching power supply.

If you (or your EE) have a clamp ammeter which can read low levels of current (down to a few mA and up to the rated line currents for your devices), it might help you figure out where the current is going.

Lynn

[Tall_Michael]

After some poking and prodding, it seems a lot of the issues are coming from the computer's power supply. Some of the connections were not good, so that got rid of a lot of the junk noise. There is still a consistent blip that comes triggered off of the AC line. Swapped the powert supply and it reduced but is still there. I am going to look into the wireless option.