150
edits
Lena/Oct 2016: Difference between revisions
Jump to navigation
Jump to search
→Group meeting
ElenaZhivun (talk | contribs) |
|||
| (16 intermediate revisions by one other user not shown) | |||
|
[[File:MagRun2PSN 2016.09.27 R20N020304 Chan1Y.png|300px]]
Talked to Mike, he says it's because of the difference between the resolution and standard deviation of of that signal. If the noise is gaussian, it should be a factor of Sqrt(12)
=== Ripples ===
==== Relevant NI devices ====
Shot noise is ~1 fT/Sqrt(Hz)
{| class="wikitable"
|-
! Type !! Model !! Channels !! Bits !! Speed !! Expected noise (10 μV/fT gain) fT/sqrt(Hz) !! Price USD
|-
| FPGA || PXI-7851R || 8 AI, 8 AO || 16 || 500 ksps each || 0.58/Sqrt(12) = 0.16 || 2800
|-
| ADC || PXIe-6358 || 16 AI, 4 AO || 16 || 1250 ksps each || 0.28/Sqrt(12) = 0.08 || 5532
|-
| ADC || PXI-6281 || 16 AI, 4 AO || 18 || 500 ksps scanning || 0.58/Sqrt(12) = 0.16 || 1929
|-
| ADC || PXI-6289 || 32 AI, 4 AO || 18 || 500 ksps scanning || 0.82/Sqrt(12) = 0.23 || 2244
|-
| DAC || PXIe-6738 || 32 AO || 16 || 350 ksps each? || -- || 1530
== 10/13/2016 ==
Talked to Chad and
Other than the evaluation board or an engineer, we could try the Raspberry PI with external DAQs. We could either run the stock kernel and do chrt, or write a kernel-level application, or find a real-time system for it, or run our application without the OS, but then we would need the basic drivers.
Jes also mentioned that 8 AI, 8 AO is a very small FPGA, so maybe there are FPGAs that can drive all sensors at once.
We can make DACs out of the digital FPGA pins with PWM http://www.allaboutcircuits.com/technical-articles/turn-your-pwm-into-a-dac/
== 10/20/2016 ==
=== Testing the new differential polarimeter ===
inally completely debugged the differential polarimeter v0.1. Prototyping in the electronics shop turned out to be a bad decision because of the low board quality. Through-hole pads kept falling off, and most of the vias were busted.
The device itself performs as good as the SRS current preamplifier. We took several magnetic field measurements with both current amplifiers:
==== SR570 ====
[[File:MagRun2PSN 2016.10.20 R02N000102 Chan1Y.png|300px]]
==== Differential polarimeter v0.1 ====
[[File:MagRun2PSN 2016.10.20 R08N000102 Chan1Y.png|300px]]
==== Design problems ====
There were several issues with the polarimeter design. The biggest issue is that it picks up and amplifies the heater noise (150kHz). The noise is picked up by the cable that connects the photodiodes to the amplifier. The photoamplifier has a single-pole the filter on the input gain that rolls off at 20kHz. 150kHz signal is only 3 octaves away from 20kHz, so the amplitude attenuation is only 18 dB.
The photodiodes cable has shielding, which works much better when it's connected to the amplifier's ground. When the shielding is connected to the ground, the 150kHz component amplitude drops by an additional factor of 5 (14 dB). The next review of the board will have grounded pins in the photodiode header to connect the cable shielding.
There will be additional 2-pole filters at the circuit output to deal with this problem.
Expected attenuation with fc=20kHz (shielding + input filter + output filter) is 14 + 3 * (6 + 12) = 68 dB
Expected attenuation with fc=10kHz (shielding + input filter + output filter) is 14 + 4 * (6 + 12) = 86 dB
This noise will alias when the ADC samples at 20kHz, with a peak at 10 kHz, the digital low-pass filter should be able to take care of that.
Low-frequency gain of the differential polarimeter seems to be larger than the of SR570. It's unclear if it's a feature or a bug or a poor connection.
== 10/24/2016 ==
=== Testing the difference/sum mode ===
Tested how the difference only mode compares to the difference/sum mode in terms of field sensitivity. Both measurements were taken with the same polarimeter board.
When the polarimeter is balanced, the probe sensitivity in both modes is the same. When the polarimeter is misaligned, the noise in 0-10 Hz band increases for the difference-only polarimeter, but not for the difference/sum polarimeter. In overall, the noise floor for difference/sum polarimeter is really good and the probe noise level drops rapidly in 0-0.5 Hz band.
=== Small polarimeter imbalance ===
Difference between channels' light powers is 1.4 μW.
[[File:20161024SmallImbalance.png|300px]]
=== Large polarimeter imbalance ===
Difference between channels' light powers is 6 μW.
[[File:20161024LargeImbalance.png|300px]]
== 10/26/2016 ==
=== Pregnant patient in WIMR ===
Measured a pregnant patient in WIMR. Although all channels worked properly and probe sensitivity was 1 fT/sqrt(Hz), we didn't get any good data, and couldn't see the fMCG signal at all. Maybe there were large field gradients that we didn't properly get rid of, like from wires in a bra or other clothes. The patient returns on November, 14th.
Added a real-time fMCG preview into the FPGA so that we don't end up in this situation again.
== 10/27/2016 ==
=== Pulse-density modulation ===
Implemented delta-sigma PDM (pulse-density modulation) for the FPGA. Now the FPGA can output "analog" control signals from the digital outputs, which our FPGA has plenty of. This would allow us to control every sensor shell coil on all three axis, and permit us to feedback on the field to keep all the sensors in average zero field and thus extend the dynamic range of our measurements. The signal to noise ratio in 0-1 kHz range is surprisingly good, and is around 100 dB. The VI is configured to output the "analog" levels with 16 bit precision. The switching rate is 20 MHz, with the CLK output.
We want to drive the atoms with the high-frequency signal and have them act as the low-pass filter. Mike built a current supply for the coils that reverses the polarity of the coil current for 0 and 1 logical levels.
It would be great to replace the analog inputs with digital signals as well. This would permit us to use a bigger FPGA (NI PXIe-7971R) that could drive as many as 25 sensors. Texas Instruments has Verified Designs section, that explains how to design and layout ADC circuits. [[File:TI ADC design ECG.pdf]]
=== Group meeting ===
Thad is concerned that the excessive noise in our measurement could have been caused by the probe power noise. We need to modulate the pump power and see how in affects the measured field.
== BACK/NEXT ==
[[Lena/Sep 2016|<< BACK]] [[Lena/Nov 2016|NEXT >>]]
| |||