Journal notes, Moving the electronics indoors
GELLER Labs "Backyard Science"
Thoughts on a proton precession magnetometer design - a Proton Magnetometer Project
The goal of this project is a low cost high performance proton magnetometer (a digital magnetometer) kit for amateur scientists to be able to accurately measure and monitor changes in the Earth's total magnetic F field and to observe geomagnetic storms. There is a regular daily (diurnal) variation in the Earth's magnetic field. During events related to solar activity, there can be sudden changes in the field (such as a sudden impulse) as well as large excursions in the field which can be more than ten times the regular diurnal variation caused by magnetic storms.
(be sure to hit refresh to pick up our latest changes and entries)
Saturday, March 20, 2010
I started to reconfigure the test stand today to move everything but the counter-wound coils into the lab. I am using one of the existing twisted pair shielded cables to bring back both leads of both coils (two pairs and a shield). One of my best finds last year was a hp 4265B RLC bridge in super nice condition. Initial readings in the lab (end of the long cable) at 1 kHz are 10.0 mH and 9.7 mH. If those numbers are correct, my coil balancing efforts are still very far off (although noise cancelling is certainly working ... at least to some extent). It's 20 mH for the two in series, but they are within an inch of each other in the sensor stand, so who knows what kind of interaction leads to the series measurement with interacting magnetic fields. Shield to one wire capacitance appears to be around 8 nF (measured with the Agilent 34410A). (I also tried to operate the bridge at 2,290 Hz using the 3581C as the null meter with mixed results. The Q values did not seem right. Probably a project for another day.) I hope I am still well clear of self resonance at 2.3 kHz with the long cable, something to look at tomorrow.
If all goes well, tomorrow I will look at resonating the coils and re-connecting the same test setup, except now with all of the electronics in the lab.
There is more design effort to follow, especially to eliminate the PARC 113 preamp in the lab (running just over gain 10 now). Everything should get easier ... maybe ... now that all will be in the lab. I made some baseline measurements of dual coil - amplifier noise floor levels with the 3581C before taking apart the test setup. 60 Hz harmonics (35th to 40th harmonics) at the final amplified output ranged from about 1.5 to 9 mV today.
Sunday, March 21, 2010
I pulled the sensor box from the sensor stand today and confirmed the different inductances. I am thinking that I probably shorted out several turns during the over-heat situation some weeks ago when I accidentally stopped the program with the polarization current on (leaving constant current for several hours). This makes sense, since only one coil is powered and one coil reads low now. I did it again today, but fortunately only for ten minutes or so. It is probably time to wind new front end coils. (See March 23: Turns out the powered coil was not damaged. Also, the low L coil was the unpowered coil).
The electronics is now in the lab and the NI USB 6008 is digitizing directly the output of what was the outdoors front end amplifier. It needs another stage of amplification, but for now is working okay (under 200 mV, so not at present making good use of the USB 6008 +1/-1V scale.
As expected, grounding techniques are going to be a challenge. Best results today were had by dropping a short 1 foot copper ground rod at that sensor coils. There, earth ground is connected to the center tap of the two counter-wound coils. Back in the lab, the analog ground is connected through the sensor cable shield to earth ground out at the sensor. The data is a little noisy, however the new curve seems to be drawing in fine with FDM easily achieving figure of merit (FOM) values of 2e-6 to 7e-8 during the first hour of run or so.
Also I need to check the center frequency adjustment of the front end analog second order filter. It ran fine outdoors from about 60 degrees F to -15 degress F, so it can't be too far off. Should have a new plot to publish soon ...
Monday, March 22, 2010
The sensor stands alone in the yard with the shield of the cable to the sensor (two side-by-side counter-wound coils) connected to a short ground rod at the sensor stand. The shield is also connected to the amplifier common (which is also ultimately the center tap of the two series coils during the digitizing part of the cycle. The amplified signal (now without the PARC 113 preamp) is down about 80% (so only using about 8 bits). I need to add a gain of 5 (another opamp stage) to get back to about 1V peak to peak at the beginning of the precession signal to fully use the 12 bits of the NI USB 6008 +1/-1 Volt scale. That said, here is the overnight run: PDF, TXT, USGS PDF. The large swings were caused by garbage trucks. Here is a sample FDM spectrum taken with the new setup: PDF.
For our fixed base magnetic observatory, it might make sense to begin by providing a path to earth ground at the sensor. There might also be an intrinsic safety advantage to hard tying the sensor to earth ground.
In the present test setup, there is galvanic isolation between the polarization circuits and the low-level AC circuits by the relay (FET-relay hypbrid as described earlier). There is a differential input (but, not a galvanic isolation) at the input to the first amplifier following an input common-mode filter. There is a second differential input (but, not a galvanic isolation) at the NI USB 6008 internal differential amplifier input. The second differential input might be important, since the USB 6008 is hard tied to the PC by the USB common.
There are likely other grounding solutions, including a combination of ground breaks (shielding at one side only, especially to prevent ground loops), isolation (e.g. transformers), and/or floating some or all of the electronics. Optimal grounding might vary site to site.
Tuesday, March 23, 2010
Added gain to bring the raw signal up to about 2 V peak to peak to better match the USB 6008 +1V / -1V scale. Overnight run: PDF, TXT, USGS PDF.
Some of the added noise pickup might be related to the powered coil (turns out not to be damaged). I will re-wind that coil next. The inexpensive coil winder motor was not reversable, so last time I had take the winder apart and mechanically reverse the position of the motor. Last night I realized that a series wound motor is reversable, but only by reversing the polarity of the rotor winding with respect to the stator winding, so I had to take the motor apart to bring out the rotor winding and the stator winding connections so I can now add a reversing switch. That will make winding counter-wound coils a lot easier now.
The powered coil was unwound (446 Turns, #24 wire). There was no evidence of overheating or any other damage. It looks like it simply was never correctly matched to the un-powered coil.
Also, I want to revisit the way FDM sends data back into LabView. I might send back several frequency values instead of just one pre-selected value. That will give the ability to display an artifact / noise pick up indication. I probably need to think about how to have FDM tell LabView how many frequency values it sends in any given measurement so that Labview can know the array size.
FDM continues do a remarkably good job of pulling out the field value to seven digits in all sorts of adverse conditions. Also, LabView continues to be a very easy and user friendly develoment environment.
Re-wound the powered coil to match the inductance of the unpowered coil using the hp impedance bridge. Also, improved matching of the bias and filter resistors at the input of the front end amplifier. All seems well, should have a new plot in the morning.
Wednesday, March 24, 2010
Overnight run: PDF, TXT, USGS PDF. A couple of spectra from this morning: spectra 1, spectra 2. Note that spectra are produced using a different copy of FDM which is set to view a far wider portion of the spectrum as well as to report many signals including those of relatively low amplitude (takes several seconds to run), whereas the measurement version of FDM is set to view a relatively narrow portion of the spectrum and, using an amplitude filter, to report only the fundamental precession frequency (<1 second on an old Pentium 4).
Recap: The counter-wound sensor coils now stand alone at the outdoors test stand. A four wire shielded cable >50 feet doubles for polarization current (running about 1.3 A for 1.8 seconds) and bringing back the raw < 1uV precession signal (only one coil with a sample (presently RainX windshield DeIcer) is pulsed). The shield of this cable is grounded to a short ground rod at the sensor stand. The FET-relay hybrid switching board (using zero-current switching) and amplifier are in the lab. A NI USB 6008 provides the two control signals as well as 12 bit digitization using a +1V / -1V scale. The system is running on a LabView 2009 VI program. The frequency estimator is based on Professor Mandelshtam's FDM harmonic inversion technique (UC Irvine, 1997) as Windows executable compiled Fortran code (See: January 23, 2010 entry). Presently a measurement is made every 2 minutes. If the figure of merit (FOM) for a given measurement is reported by FDM as over 2e-6, Auto-Retry continues to take measurements at about 8 second intervals until an acceptable measurement is recorded. The Auto-Retry feature largely filters out perturbations causes by moving vehicles and measurements disturbed by short term interfering signals. Vehicles within several hundred feet that become stationary during a digitization cycle cause step changes in the field which are recorded. Vehicles having large ferrous metal content, such as garbage trucks, the village front loader, dump trucks, and school buses perturb the field at larger distances of over 300 feet.
For the intended GeoMagnetic Observatory application, it is anticipated that that one data point every two minutes is sufficient. It is still possible, however to take data at 20 second intervals with moving averages. On the other hand, FDM is such a strong filter in and of itself, the two minute intervals with Auto-Retry may prove sufficient.
Thursday, March 25, 2010
Overnight run: PDF, TXT, USGS PDF, a spectra from this am PDF. As before, step changes are due to stopped or parked vehicles coming and going. One large excursion (single point, 4:10:48, 2274.896 Hz, 53447.55 nT) appears to be noise (e.g. EMI or RFI) pickup or a calculation error. It is so interesting to see the diurnal variation in the Earth's magnetic field. It is surprising that there are not more plots on the web showing the field changes over several days.
Friday, March 26, 2010
Overnight run (includes past days): PDF, TXT, USGS PDF, am sample spectra PDF. There was a lot of nearby vehicle activity, however the overall cycle still shows clearly. Ideally the sensor should be about 300 feet away from all activity, yet a credible record can still be made with 75 to 100 feet of clearance.
The overall noise is way down in amplitude and so is the computed fundamental frequency. It is unclear if there has been a change in the amplifier, ambient noise, or some other change. The computed waveform envelope is similarly much cleaner. A mystery for now.
I probably need to make some careful amplifier gain measurements to be able to track down such changes. The amplifier has been re-worked many times in developement, I suppose there could be an intermittent problem somewhere on the board.
The polarizing current looks fine, so I doubt it is a relay issue, yet I should probably re-check the system timing too. All easier tasks now that the electronics is in the lab.
I started to look into the changing amplitude this evening and noticed that some of the problem is that I set the outdoor box next to the computer monitor. The amplifier is in a shielded box, but the hybrid FET-Relay board is not. Moving that board causes changes in the amplitude, so it probably is related to noise pickup and/or relay orientation. More problems to look into before this can move towards a production experiment. Ah well, better to deal with all this now.
Sunday, March 28, 2010
Overnight run: PDF, TXT, USGS PDF, sample spectra PDF. Confirmed that the odd changes in amplitude are caused by the position of the electronics box (the former outdoor container). Also, the orientation of the hybrid FET-relay board is significant. I need to look into how much is simple EMI coupling and/or relay orientation. Also, I have some new ideas for a relatively simple non-relay method, ordered some parts, will try it out sometime in the next couple of weeks.
Tuesday, March 31, 2010
Last ~24 hours as of 9 pm today: PDF, TXT, USGS PDF, sample spectra PDF. An interesting increase in the total field variation shows in the USGS FRD, VA plot. Also includes a morning truck and the usual coming and going of nearby vehicles. The slope of the field changes just before 12 pm with the regular diurnal variation.
Sunday, April 11, 2010
I finally got back to the prototype today and worked a little on the all solid state design. A very simple non-relay solution is difficult because of the isolation needed between the polarizing system and the ultra-high gain amplifier. I'm not there yet, however, I do believe there are some relatively simple low parts count solutions, but for now, it is not a priority. At 20 million cycles a small signal relay switching only at zero current, the hybrid-FET solution remains a very attractive approach for an inexpensive observatory application. The "free" galvanic isolation with low capacitive coupling offered by a high quality small signal relay is hard to beat. Mounted in a socket, at ~$3 retail, less surplus, it is easy enough and inexpensive enough to change out the relay as needed. We don't think twice about changing spark plugs in a car, so what is the big deal with plugging in a new tiny inexpensive relay once in a great while?
I think for now, I will continue to refine the FET-relay hybrid approach. The next job is to rebuild the hybrid FET-relay circuit in a shielded box and with a relay socket. Also, I will see if I can dissassemble the present relay (a tiny high quality small signal relay) to check for any noticable contact wear. It was in service since about November or December of last year, have to check back in my records. We are getting into the first of the G2, G3, and G4 events, I would like to observe some of them with the FDM prototype. Probably a couple to few days before the old prototype is back up and running.
Monday, April 12, 2010
Small signal relay autopsy results: Apparently at least for a time during development, I did not achieve "zero current relay switching". The tiny relay contacts do show signs of arcing on the powered side and no arcing (as expected) on the small signal amplifier side. pic 1. Will need to check the timing again, perhaps make more measurements, and then autopsy another relay. This time, the relay goes in a socket.
Interesting, The "K" index (NOAA Space Weather Prediction Center site) has been relatively high at times lately at Boulder, CO: JPG , USGS PDF.
Project Documentation (very early stages)
QUESTIONS/COMMENTS/notice of typos, etc. send email to joegeller @ gellerlabs dot com
COPYRIGHT © 2009, 2010 JOSEPH M. GELLER, All rights reserved.