Lena/Nov 2016
November 2016
BACK/NEXT
11/02/2016
Magnetic connectors
It seems like many parts of our magnetometer setup are actually very magnetic. That includes the BNC cables, banana plugs and the circuit boards.
The circuit boards from OSH Park were plated with Ni, and that caused problems.
Replacement non-magnetic connectors:
- BNC plug http://www.mouser.com/ProductDetail/Amphenol-RF/112710/?qs=sPbJEtTr5b45g4lDF9tMFw%3D%3D
- BNC jack http://www.mouser.com/ProductDetail/Amphenol-RF/112740-13/?qs=ZnztrWCQEtE9xZQYo%2FWsjg%3D%3D
- Banana plug http://www.mouser.com/ProductDetail/Pomona-Electronics/3276/?qs=sGAEpiMZZMvdy8WAlGWLcPkjJE6AhYLi
- Banana jack (except the nut) http://www.mouser.com/ProductDetail/Johnson-Cinch-Connectivity-Solutions/108-0903-001/?qs=sGAEpiMZZMuIC3ROaEqRYUclhTyEUpND
- Pin headers http://www.mouser.com/ProductDetail/Molex/22-28-4363/?qs=sGAEpiMZZMs%252bGHln7q6pm%252bS0pk2Wo0XxTAhHStsXU8w%3d
11/07/2016
Pump modulation and noise
We are trying to verify whether the drifts in the pump laser are the reason why our measured magnetic field noise is so large. We have really low probe noise, around 1 fT/sqrt(Hz), and the measured magnetic field noise was consistently higher than 10 fT/sqrt(Hz), especially at the frequency below 10-20 Hz in WIMR. The SQUID measurements suggest that it should be ~5 fT/sqrt(Hz).
We modulated the pump current, which modulated both optical frequency and power at 10 Hz. The change in frequency was ~1 GHz, and the change in power was ~0.02%. It produced a 1.5 pT peak in the magnetic signal. That's 1.5 fT per 1 Mhz, or 2 ppm of optical power change. 2 ppm requires at least 19 bits, the feedback loop only has 16. The problem can be circumvented by averaging the AI data and creating a custom PID loop block.
Mike suggested that we measure how the magnetic field depends on the pump optical frequency.
11/08/2016
Stabilizing optical frequency
Built a setup for saturation absorption spectroscopy to stabilize the pump optical frequency in Chamberlain. The current control signal consisted of ~20 MHz sine modulation at 12345 Hz, and a DC feedback signal. The signal was demodulated by a lock-in amplifier and given to a 16-bit digital PID controller. The laser was locked to the brightest visible crossover resonance (~50% contrast). We measured the magnetic field with and without the frequency stabilization, and there was no change in the magnetic noise floor. The magnetic noise remained at 10 fT/sqrt(Hz).
Diff/sum board oscillations
It appears that the output filters on the diff/sum board oscillate at 4 MHz with about 100 mV amplitude. It might be happening because the opamp is directly driving the 820 pT feedback capacitor. We should replace the capacitor with a smaller value and see if that helps, or alternatively we could use a different opamp in the future frontend designs.
11/09/2016
Stabilizing optical power
Added a pickoff made from a microscope slide to the pump path in Chamberlain that picks off pump power and sends it to a photodiode. The photodiode is connected to an SRS 570 current preamplifier with no bias voltage. The pickoff is installed after the fiber output and the polarizer before the cell inside the magnetic shields. The feedback loop is keeping the photodiode current constant by feeding back on the laser current. I attempted to do it with an AOM instead, but it took to long to align it.
The feedback loop managed to stabilize the input signal up to 10^-5, and there were no changes in the observed magnetic field noise. The input reading was most likely incorrect and suffered from the optical interference fringes, because the feedback sign spontaneously changed to the opposite over a course of a few hours.
Mode hopping
We discovered that the DFB pump laser in Chamberlain is mode hopping at the frequency close to the optical transition. That made it difficult to measure how the field response depends on the pump optical frequency. We changed the laser temperature to move the mode hop further away from the transition.
Optical frequency dependence
Measured how the magnetic field change depends on the optical frequency of the pump laser. The pump current is modulated with a sum of two waveforms - 0.1 Hz ramp that sweeps through the optical frequency, and 38 Hz harmonic modulation that changed the optical frequency by 1 GHz. We connected the magnetometer signal to the lock-in to see how both phase and amplitude depend on the optical frequency.
We observed response in both X and Y channels of the lock-in. The X response remained positive, and changed by 20-50%. The Y response was an 10 times smaller than the X response, and changed sign as the optical frequency changed. The pump is parallel to the Z direction and orthogonal to the probe. It appears that the measured magnetic field noise has contributions both from the changes in atomic polarization and light shifts, but the light shifts contributions are an order of magnitude smaller.
11/10/2016
Optimizing magnetometer performance in WIMR
Tuned the probe and pump lasers to get the optimal magnetometer signal. The magnetometer appears to have residual sensitivity to the magnetic field when the pump is turned off. After tuning the probe further away from the resonance and decreasing the probe power, we could get nearly flat noise at 1 fT/sqrt(Hz) in Y mode.
Before (large probe power, close to the optical transition
TODO: add the plot
After (reduced probe power, detuned from the optical transition
TODO: add the plot
Popcorn noise in the diff/sum circuit
There are strange noise peaks in the time series in the probe noise. They appear only in the diff/sum mode of the polarimeter circuit, and go away if only the differential channel is used. They might be caused by a large digital gain (30), that results in the significant bits being lost and digitization noise appearing and passing through the filters.
Difference only
TODO: add the plot
Diff/sum
TODO: add the plot
fMCG live preview
Tested out the fMCG live preview in the FPGA magnetometer. The filter cuts out 60 Hz and 120 Hz and they are not visible in the preview. Then a large (almost of the size of the full dynamic range) 60 Hz disturbance is applied to the magnetic field, the filter eliminates it completely. For such a large disturbance, it takes several seconds until the 60 Hz exponentially disappears. The filter doesn't seem to distort the MCG signal significantly. The signal to noise on the MCG signal is poor, and the DC offset makes it hard to preview. I have replaced the 100 Hz low-pass filter with a 1-80 Hz band-pass filter to fix this problem, and make the preview signal match the pre-processed data in Saki as much as possible.
Digital current driver
Preliminary tests of the digital current driver circuit. Verified that the circuit works, but the components that Mike has found in Billy's can only run at 2 MHz (we need 20 or 40 MHz). We have also tried applying the digital modulation to the input of the existing current driver, and it works as expected. The atoms react to the applied field change and not to the the high-frequency noise. It is unclear whether the existing current driver is low-passing the digital signal, or the atoms.
Pump current stabilization
Wanted to check whether some part of the measured magnetic noise might actually be coming from the pump power noise. First we tried to stabilize the power by feeding back on the LD current, but that didn't work. The laser current was changing, but we didn't observe significant change of laser power after the tapered amplifier. The we locked the power on the tapered amplifier itself, and that didn't reduce the measured magnetic field noise. So either the room is too noisy, or the power stabilization wasn't enough.
11/14/2016
Pregnant patient
Measured a pregnant patient, the baby is 24 weeks old. The fetal signal is visible, but the S/N is really poor. The fetal signal is about 10% or less of the maternal signal. I spent rest of the day post-processing and trying to get the maternal and fetal separated, but I can barely identify the fetal QRS complexes.
11/16/2016
Pregnant patient data processing
More data processing. Checked out the signals from SQUIDs and SERFs that Ron's group have taken. The SQUID signals typically have much higher fetal signal amplitude, typically 1/3 of that of the mom for the same age. Some older babies even have the same heartbeat signal amplitude as the mom.
Meeting with Thad to discuss the results. Thad suggests that probably our results are getting worse over time because the room hasn't been degaussed in a long time. And SQUIDs are not affected that much because they are gradiometers. We compared the "golden baby" signals taken by Bob several years ago, and we seem to have similar maternal MCG S/N, but his baby amplitude is larger, and our magnetic noise floor is significantly larger, especially at the low frequencies.
Thad suggested that we take the ICA data (independent component analysis that Saki outputs) and look at the noise FFT to see how Bob's data compares to ours.
Degaussing
Zack contacted Ron, and it looks like the room hasn't been degaussed in 7 years. Thad said if the room is not degaussed, the shielding factor goes down. Zack contacted the company that installed the shielded room (ETS Lindgren) to order degaussing.
11/17/2016
OTS cells
Optimal operation parameters at T = 140°C are:
Probe
- Rb D2 2-3'
- Rtec = 12.561 kOhm
- Id = 160.2 mA
- Pwr = 1 mW out of the fiber, 50-240 uW at PD
Pump
- Rb 1-2' D1
- Rtec = 8.67 kOhm
- Id = 160 mA
- Pwr = 1.6 mW out of the fiber (signal saturation)