NI PXIe 6124 strange noise problem.

[Zeus99Knight]

I'm working on a high performance scientific instrument for my laboratory.

I've build 4 identical silicon photodiode amplifier and I tested them using a NI PXI 5922 board. The test consist in powering the circuit and connecting it to the board without shining any light on the photodiode. The signal in this configuration is only the background noise generated by the circuit (resistance thermal noise and opamp noise), the so called "dark noise" of the circuit. This noise is analysed by computing its Power Spectral Density. The same is done with the noise generated by the board itself, by putting its input in short circuit.

Using the 5922, the circuit performed as expected. But when I did the same test using my NI PXIe 6124, I observed very strange behaviour, and I dont know what's happening. It looks like the DAQ noise receive some strange gain, or interact in a strange way with the circuit.

Configuration when using the 5922 :

 - circuit powered by a low noise linear benchtop power supply

 - board configured in high impedance input mode (1 MOhm in parallel to 60 pF, acording to the spec sheet)

 - circuit output connected to board BNC input using a shielded, twisted pair cable 1 meter in lenght

 - I dont remember the PXI chassis used, but I dont know if it's important. In all cases, in this configuration everything worked well

The result from this measurement is given in the attached figure "new circuit compare.png". The acquisition rate was 15 Ms/s.

Configuration when using the 6124 :

 - circuit powered as before

 - circuit connection to the DAQ described in the attached figure "capture shielded twisted pair"

 - The DAQ is the only board in a NI PXI 1073 chassis. Connection is made through a NI SCB 68 A connector bloc and a 5 meter long SHC-68-68 EPM cable. The total cable lenght is thus 6 meters.

 - The DAQ noise was measured with this cabling, but by replacing the circuit with a short circuit wire (a wire between the IN+ and IN- analog input, and a second wire between IN- and GND). Note that contrarily to the 5922, the 6124 have differential input.

What I've obtained Is given in the figure "CIRCUIT TO DAQ COMPARE". In this figure, the 4 spectra at the bottom correspond to the DAQ noise, while the 4 spectra at the top are what I obtain when I connect the circuit. The acquisition rate is vary, but the input range is always +/- 1V.

The DAQ noise change when its acquisition rate is modified, as expected. The DAQ noise level correspond exactly to the ones specified in its spec sheet. However, when my circuit is connected, everything look like the DAQ noise is multiplied by 200 (100 at 4 Ms/s), and I dont have the same result I had with the 5922. Note that the circuit noise level correspond to the DAQ noise level at 4 Ms/s, at least between low frequency and 100 kHz.

The DAQ noise also change with input voltage range. But stangely, if I change the voltage range instaed of the acquisition rate, the noise level remain the same...

What I've tried so far :

 - change the cabling between the circuit and the connector bloc: no effect

 - change the length of the cable between my connector bloc and the DAQ (from 5m to 2m) : no effect

so it is not a cabling problem

 - use a different analog input (the 6124 have 4 AI) : no effect

 - use a different copie of my circuit : no effect. Instead of this strange noise, the circuit behaved well.

 - add resistor's in parallel between the DAQ IN+ and GND, and also between IN- and GND (recommanded if you connect a voltage source with a high output impedance to the DIFF input) : the continuous background noise didn't changed, but It added spurious sin peak in the spectra.

 - add a resistor's in serie between the DAQ IN+ and the circuit OUT : It does change the noise level, but I also change the circuit bandwidth by doing this, so this is not a definitive solution. By looking at the circuit new bandwidth with this added resistor, it look like the resistor is forming a pole with a capatance around 1 uF. I dont know where this capacitance can come from.

Additionnal information : the circuit diagram is attached. The Opamp I used is the Analog Device AD8067. It have a very low ouput impedance, this is why I connected my circuit this way.

Any hypothesis or explanation about what's happening here would be appreciated. I dont know why my circuit and my DAQ dont like each other. The main difference I see between the 5922 and the 6124 is the DIFF input of the 6124, but I dont know how it can bring me trouble. I you want additionnal information just ask.

Please Help me !

[johnsold]

Interesting question.

While I do not have the answer, I would add these comments.

1. The baseline noise floor of your amplifier as measured by the PXI 5922 is ~2E-15. Only the 4 MHz configuration of the PXIe 6124 is even slightly less, at ~1.5E-15. Thus it will be difficult to get meaningful measurements with the 6124. It is just not quiet enough.

2. The wiring for the 6124 includes a possible ground loop. Any currents flowing in that ground could affect the measurements. Try grounding the shield only at the Source end.

Please keep us infomred of what you find.

Lynn

[Zeus99Knight · Accepted solution]

Hi,

Thank you for the quick reply.

Has you noticed, The noise floor of the 6124 DAQ is almost at the same level than the circuit noise itself (measured with the 5922). In fact, I was expecting only a slight change in the noise of the whole system with my photodetector circuit connected, at least on the +/- 1V range at a sampling rate of 4 Ms/s. In my case, as I want to acheive le lowest noise level, seeing "nothing" when connecting the circuit was my goal, but this strange noise problem keep bothering me.

I tried a large variety of connection for the reference voltage of the DIFF input of the 6124, including not connecting anything at all, but it doesn't improved on this problem. And I don't understand this proportionnality between this noise and the 6124 DAQ noise, if it is really that of course.

I'll keep you informed if I correct the problem.

I know that you DAQ designer add a little bit of "dithering" random noise before (or during) the quantisation step of the ADC, can you tell me how it is acheived in the 6124? Maybe if I know the architecture of the PGIA at the AI of the 6124 I may be able to figure out what's happening.

Thank, Richard

[Zeus99Knight]

I eventually figured out what my problem was, and it was a pretty common one : It was simply aliasing.

My amplifier generated a lot of noise above the NI PXIe 6124 cutoff frequency, and the single pole low pass response of the DAQ after the Nyquist frequency was simply not enough to get rid of it. So this high frequency noise aliased back in all my noise spectra. Adding a simple RC filter between my circuit and the DAQ revealed the answer, and a design modification associated with the addition of an additional active low pass filter corrected it.

While I was trying to understand this problem, I've made two mistakes :

 - first, looking at the noise spectra without filter, I associated their very flat appearence with the DAQ's one. Aliasing explain why it was changing with acquisition frequency, but not with input range

 - second, I made a very bad beginner mistake : to believe that aliasing can be removed with a digital filter after the acquisition

step. It is simply impossible.

The 5922 embed its own high order antialiasing filter, and that's why I didn't had any trouble with it. The various resistor I added I serie with my circuit output most probably formed a single pole RC antialiasing filter in combination with the cable capacitance, hence the improving noise level with growing resistance value.