DAQ system issues: Unwanted peaks/spikes in the frequency spectrum and DC offset in the time signal

[Toros_Sorama]

Hi everyone,

In the past I have purchased a DAQ system from National Instruments to use for sound measurements with a microphone. My measurements are performed in the NI DAQExpress software and I use the NI MAX software for self-calibration of the system. Unfortunately, I have encountered issues during the sound measurements, which are limiting me from using the measurement results.

My setup consists of:

• USB-6289 MASS TERMINATION;
• BNC-2110 Noise Rejecting, Shielded BNC Connector Block
• SH68-68-EPM Shielded Cable, 68 D-Type to 68 D-Type, 1M.

One of the Analog Input channels of the BNC-2110 Connector Block is connected to a microphone via a power module, which powers the microphone using batteries. The Analog Input was set on FS, since the microphone is powered by a battery (However also a couple of measurements were performed using GS, simply for the sake of trying). The setup can be seen in ‘Figure 1 – Measurement setup with microphone’ in the attachments.

While performing measurements in NI DAQExpress with the microphone a strange response was seen. This response consisted of:

• A DC offset in the measured time signal. The measured voltage signal appears to be rising over time (even after waiting the requested warmup time).
• A noise floor of background noise measurements which would deviate over time, influencing the sound measurements (See ‘Figure 2 - Background noise floor deviations over time ' in attachments à The background noise measurements are displayed on blue)
• Strange peaks/spikes in the frequency spectrum, which are not actually present inside the measurement room (verified with other microphones). These peaks do not have a fixed position (fixed frequency), but move alongside the frequency axis in the spectrum. The peaks also sometimes disappear, only to return moments later. The behavior of the strange peaks over time in the frequency domain could be seen in ‘Figure 3 - Peak issue over time’ in the attachments.

To solve the issue, the microphone and the power module were checked at by the manufacturer. The manufacturer reported to us that the microphone and the power module were in perfect shape and could not cause these strange findings.

Subsequently, I focused on the DAQ system to possibly discover the underlying cause of the issue. This time, I directly connected a function generator to the Analog Input of the channels of the BNC-2110 Connector Block. The Analog Input was set on GS, since the function generator is a Ground Source (however tests were also conducted on FS, as an extra check).

Again, strange peaks/spikes were witnessed in the frequency spectrum, which should not be there (apart from the 20 kHz sine tone which was supplied by the function generator). The strange spikes in the frequency domain could be seen in ‘Figure 4 - 20 kHz sine directly for function generator’ in the attachments.

Subsequently, the USB-6289 MASS TERMINATION and BNC-2110 Noise Rejecting, Shielded BNC Connector Block were returned to National Instruments for a ‘Repair’. During the repair, the USB-6289 MASS TERMINATION was checked and returned without noticing me of any present malfunctions. The BNC-2110 Noise Rejecting, Shielded BNC Connector Block was replaced with a new Connector Block of the same type, since it was an item which could not be checked/repaired, meaning it was immediately replaced.

After the products were returned to me, I once again tested the DAQ system. I tested it with the microphone and also directly connecting a function generator (with an output of 1 kHz sine tone and a 60 kHz sine tone) to the Analog Input of the channels of the BNC-2110 Connector Block. In both cases, the peaks/spikes remained visible. In these tests I used different BNC-cables for my setup (also from different manufacturers). This did not solve the issue. See ‘Figure 5 - 60 kHz sine directly for function generator using 5 different BNC cables’ in the attachments for the results of a function generator generating a 60 kHz sine tone using 5 different BNC cables.

Thank you for your attention. I hope someone can guide me into the right direction to solve my issue.

[Kevin_Price]

Bottom line: I think the problem may be more in your expectations than in the measurements.

1. DC offset - one would normally want AC coupling for such measurements, using a device that supports it.  I'm pretty sure yours doesn't.  Barring that, you could do some highpass filtering.  Dunno what's available in DAQExpress, but you might need an analog filter, perhaps as simple as an RC.

2. The extra spikes you see when generating 20 kHz with the function generator sure look equally spaced, strongly implying harmonics.  Real-life measurements only rarely *don't* show harmonics.  Only a perfectly pure sine wave would have none, and those are hard to come by.

    That said, a slightly imperfect 20 kHz sine wave ought to be showing a major peak at 20 kHz and lesser peaks at multiples of 20 kHz.  Your graph shows something closer to multiples of 18 kHz where curiously, the fundamental freq of 18 kHz is *not* the highest amplitude.  *That's* the part that seems a bit off to me.  The rest of it looks pretty normal.

3. Note that your 20 kHz noise floor is way, *way* down in the sub -100 dB range.  Frankly, that's pretty freaking awesome!  However, even your peaks are very very low, so perhaps you can improve your measurement setup to help your mic pick up more of the sound?

4. The 60 kHz measurements look pretty great!   No prominent harmonics and you've got > 90dB of signal/noise ratio.   That's a winner!   The little spikes here and there are barely above the noise floor and probably don't deserve your attention.

-Kevin P

[Henrik_Volkers]

---

@Kevin_Price wrote:

Bottom line: I think the problem may be more in your expectations than in the measurements.

...

2. The extra spikes you see when generating 20 kHz with the function generator sure look equally spaced, strongly implying harmonics.  Real-life measurements only rarely *don't* show harmonics.  Only a perfectly pure sine wave would have none, and those are hard to come by.

    That said, a slightly imperfect 20 kHz sine wave ought to be showing a major peak at 20 kHz and lesser peaks at multiples of 20 kHz.  Your graph shows something closer to multiples of 18 kHz where curiously, the fundamental freq of 18 kHz is *not* the highest amplitude.  *That's* the part that seems a bit off to me.  The rest of it looks pretty normal.

...

-Kevin P

---

I fully agree with Kevin 🙂

Additional hint to real world measurements:  at these low levels you you have other sources of signals: Down mixing of higher frequency content with the sample frequency due to imperfect low pass filters and EMC sources like LED light or other switched power supplies, pulsed RF, other oddities ..  all in all combinations 😉   

Harmonic distortion will change with you source , so measure at 20, 21, 23 kHz and monitor the peaks.

down mixing peaks will shift if you change the sample rate (keep in mind that these only change in steps , see spec ..)

direct EMC peaks stay until you turn them off...

Peeking into my crystal ball and your fig. 2&3 : : Measured over different time length?   Common practise in switched power supplies to pass qualifications is a spread spectrum.. no peak on one place for a long time... bastards...

[Toros_Sorama]

Hi Kevin,

On closer inspection, I noticed that the frequencies of the unwanted peaks using the same setup are not the same, but they differ. To show this I have re-performed the measurement with the DAQ system directly connected to a function generator, with an output consisting of a 20 kHz sine tone. I have done three measurements over time (steps of roughly 5 minutes). The result can be found in Figure 1 in the new appendix ‘Measurements Kevin and Henrik’. As can be seen is that the expected peak at 20 kHz is the same for all three measurements. However the other peaks contain of different frequency components.

Regarding these new finding, do you have an idea on where the problem might be?

Best regards

[Toros_Sorama]

Hi Henrik,

Thank you for your response.

Are you referring to a difference in time length between Figure 2 and 3 or between the measurements performed in Figure 2 and the measurements performed within Figure 3?

Since the measurements performed in Figure 2 and the measurements performed within Figure 3 consist of the same time length. However comparing the two Figures with each other you are right, there is a difference in time length (Figure 2: 10 seconds and Figure 3: 1 second). Also Figure 2 and 3 consists of different measurements. Figure 2 of the background noise in the room and Figure 3 of the unwanted peaks. Considering that the microphone that was used has got a lower dynamic limit of 52 dB and the peaks appear roughly 50 dB above the noise floor, these peaks contain Sound Pressure Levels of roughly 102 dB, which does not make sense, since they do not exist in the room.

Both measurements (Figure 2 and 3) were performed with the combination microphone + DAQ system.

If you are still interested in seeing Figure 3 over the full 10 seconds, I have placed it in the new appendix ‘Measurements Kevin and Henrik’ and can be found in Figure 2.

Best regards

[Kevin_Price]

I don't have any fresh insight to offer.  On the whole, things still look more like something to celebrate rather than worry about.  I'm seeing ~100 db signal/noise and even the "unwanted peaks" are down about 75-80 db below.  Acoustic measurements are not my field, but that sure *seems* like pretty good data to my eye.

What kind of results do you get from other mics that leads you to expect better than this?  

-Kevin P