I see frequency spikes in the FFT that I suspect are unrelated to environmental magnetic noise?

[Vishal Shah]

The digital and the analog output is derived from the same source, so the FFT from the analog output should look the same digital output. If you see excess spikes on the analog output, something is not right. (Just a note: the digital output is in pT and the analog output is in volts. The analog output voltage to magnetic field conversion is 2.7V/nT is default gain mode and after calibration.) Spikes from unknown sources can also originate in QuSpin's user interface when plotting the digital output.

Here are some of the sources of spikes that we or other users have experienced in the past.

1. 77 Hz spike (only seen on the analog output)

The sinusoidal modulation field on which the QZFM relies for operation is at 923 Hz. After demodulation, we try and filter out residual 923 on the analog output as best as we can, but a tiny bit of 923 Hz modulation signal leaks out on the analog port. If you are sampling the analog output with ADC running at 1kHz (or 500 Hz), then the 1 kHz can alias with 923 Hz producing a 77 Hz spike in the FFT.

To avoid this problem, we recommend sampling at a minimum of 2 kHz when possible. If sampling at 2 kHz is not an option, we suggest sampling at a frequency slightly higher than 1 kHz (something like 1.1 kHz for example) such that the spike is outside the 0-100 Hz band which is critical for MEG and related applications.

2. USB Grounding

If the USB ground is different from the power supply ground, then ground loops can form which an introduce random and sometimes inconsistent spikes. This is a design oversight at our end and we will try to fix this problem in future electronics but for now, the best option is trying to plug the USB cable elsewhere and see if that helps. If you have found some nice tricks to solve the problem, please post here to benefit the rest of the community. To identify USB cable is a source of the noise spikes, do the following: Run field zeroing and calibration, and after that unplug the USB cable and measure the noise spectrum with analog output. If you see spikes that disappear, then USB ground loop is the source of the problem.

3. Interference spikes from adjacent sensors in multichannel operation

This is the most vexing problem at this time but it is critical to resolving the issue when operating in a multi-channel configuration. We expect that future generation of electronics  [V8 and beyond will be immune to interference issue (V8 is in internal testing stage right now as of April'19)]. The laser and the vapor cell in the sensor are temperature stabilized using AC currents @ roughly 400 kHz. Each sensor electronics generates its own 400 kHz which can be slightly-off due to variation in reference clock inside each electronics. The sensor cables radiate the 400 kHz and a slight variation in the heater frequencies in different sensors can produce a beat note visible in the DC-100 Hz operating band of the sensor. To avoid this problem, please follow the procedure provided here: link.

If the problem is not resolved, try the following process. Place all the sensors in close proximity of one another. Then turn-on just one sensor (Sensor-1, Master). Set Master to 400 kHz. Record the noise spectrum. Then turn-on sensor 2, and then tweak its heater frequency such that no spike is seen on the FFT in the 0-100 Hz band. Once sensor 1 & 2 are working fine, turn on sensor 3 and repeat the process while sensor 1 & 2 are also on. Continue doing this for all sensors in the chain. 

Again, if you have found a better process, please do post your results or suggestions here.