Journal notes, FOM filter, Polarizing Power Supply
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, May 22, 2010
Overnight: PDF, TXT, USGS PDF, 7 hour view of recent data PDF, new fast spectra PDF.
The geomagnetic field has been relatively quiet the last couple of days. The 7 hour view shows the consistency and relatively low noise of the present configuration.
Note that these are "single-shot" measurements, with no averaging. When the figure of merit (FOM) > 2e-6, the FDM magnetometer continues to take new measurements until an acceptable measurment is made. In this fast auto-retry mode (several seconds as compared to the normal 2 minute data interval), each new measurement is an independent measurement, not influenced by the previous failed measurement.
A new log spectral display uses FDM settings closer to what the operating algorithm is using and a bar graph seems to a better job representing the data. Each spectral data point is above a certain amplitude threshold, set low for the spectral display and relatively high as a filter for the working FDM magnetometer data point.
Sunday, May 23, 2010
Overnight: PDF, TXT, USGS PDF, fast spectra log bar graph PDF, more traditional view of fast spectra PDF (same data set).
Each successive line of the line spectra graph is the next recorded frequency (the x-axis spacing is not constant). The traditional spectra (from the same measurement data) is an x-y plot, and the points along the x-axis are equally spaced. Also, the vertical amplitude axis of the line spectra graph is a log scale.
I notice there appear to be a few more auto-retries when some of the various lawn machines are in operation within about 500 feet. I suppose with HV coil and spark plug operation, this sort of interference is inevitable. On the other hand, perhaps these auto-retries just show the FOM filter doing its job.
Regarding the overall FDM magnetometer project, I suppose I should start to draw some flow charts for the instrument software. One possibility for this project, might be to offer the analog and/or digital sections as hardware kits, with a compiled LabView program (user does not need LabView) with the FDM module as a called Windows executable (the present configuration). Then amateur builders could a) duplicate the instrument as later described in more detail, b) provide their own control software and call the FDM executable, c) roll their own FDM routine, such as from the MIT harminv in LINUX, or d) use a more conventional ppm frequency estimator (although the FDM method combined with the FOM filter is what provides the high resolution and accuracy). The present instrument with the USB 6008 data acquisition module could be copied, or an experimenter might use another similar module (there are several available), or roll their own control and data acquisition module using other digital I/O techniques and an ADC. Also, experimenters could opt to use my zero-current hybrid FET-Relay switching scheme, or roll their own. More to think about.
Monday, May 24, 2010
Overnight: PDF, TXT, USGS PDF. The geomagnetic field has been relatively quiet. There is currently one sun spot PDF, (pdf printed to preserve this morning's page at www.solarmonitor.org), 11072, but so far very low magnetic activity here on Earth. (All of the usual morning vehicles show up in the record (newspaper delivery (~3am), garabage trucks, and school buses). It appears that one error point somehow got through the software filters at 6:51 am, a 53,726 nT point with a FOM of 9e-7. I believe this point with a -75 nT swing to be non-physical, probably some sort of interference. I suppose one error point in a tens of thousands of measurements is tolerable.
It is interesting to look at the unselected text file data (includes failed measuremnets with FOM > 2e-6) in amplitude and FOM plotted vs. measurement number. Something to study more closely in the future. From an equipment point of view, so far I do not see any failure trends.
Tuesday, May 25, 2010
Overnight: PDF, TXT, USGS PDF.
Interesting look at the FDM amplitude over the same period (using the TXT file, includes rejected measurements) PDF. Running about 1.2 A polarization current using a voltage slightly higher than needed and current limiting (constant current mode), so I do not think it is lower current due to increased cable and coil resistance with temperature. However, it does appear that the FDM voltage at the fundamental frequency might be falling off with increased sensor temperature. Not sure why.
On the other hand, here is a view of FDM FOM and amplitude for only measurements that were taken as acceptable (those plotted on the field plot): PDF. While only a couple of hundred measurements were dropped for amplitude below the FDM threshold amplitude (currently about .89 V), most were dropped for high FOM (about half of the measurements). Of those accepted, most had a FDM amplitude of well above the 0.89 V FDM amplitude threshold. So, perhaps there is not much to be gained by worrying about a dynamic threshold adjustment. Rather, it would be good, if possible, to increase the yield of acceptable measurements by achieving a higher percentage of measurements with an acceptable FOM.
Perhaps a slightly longer polarization time and/or slightly higher polarization current would be desireable. More to look at. The present polarization supply is limited to 1.2 A. I would like to try a slightly higher current of 1.5 A to 1.75 A.
Or, is just that for those failed measurements, no matter what percentage of protons are polarized, the magnetic gradient (db/d distance across a coil) or rate of change of the magnetic field (db/dt) at the sensor is just too high at a given measurement interval? I am guessing there is at least some small improvement in the ratio of acceptable to failed measurements still to be had.
evening: I broke out an old Kepco ATE 25-4M power supply from the equipment room (almost forgot I had one, the last time it was used was back around 1999 or 2000 during one of my earlier failed ppm experiments). The ATE 25-4M (25 V 4 A max) has "automatic cross-over", meaning you can set the voltage slightly higher than you think you will need, and let it automatically cross over to constant current mode during the pulse. This is just for testing, I am still thinking a small inexpensive open frame or bench supply (probably 12 to 15 V (depends on your feed and counter wound coil resistance), just voltage mode, with a 1.5 A to 2 A output current rating) will be fine for a working home FDM magnetometer. University experimenters will probably already have a small bench supply with a current mode. Afterthought: I also forgot about inexpensive imports with automatic voltage to current cross-over, such as the HY1803D power supply (LCD), or the HY1503C if you like to watch meters move, 15 V should be enough voltage for most setups (no affiliation with or knowledge of these sellers). Perfect! I have an 1803D that I use for battery charging. I will run overnight on the HY1803D. Note these are not the quality of an Agilent supply, but at well under $100 they are perfectly well suited for a relatively inexpensive version of the FDM magnetometer. Another possibility (but at 30 V 5A, more power than needed).
I am trying around 1.7 A, which should give a polarization field of around 250 Gauss. Will run overnight on the HY1803D at about 1.7 A. From an intial superficial look at a few data points (with the Kepco ATE), the envelop is cleaner and has more swing from the first displayed point of a 600 point moving average to the 11,000th point, however there are still auto-retries. The number of auto-retries goes back to my earlier question, if auto-retries are mostly due to a momentary db/dt or gradient at the sensor, perhaps more polarization current just gives a stronger impression (higher amplitude) of a still unacceptable "bad" measurement. Samples with the Kepco ATE power supply, (huge envelope amplitude) PDF , and a shot with a much lower amplitude, but a great FOM PDF .
Wednesday, May 26, 2010
Overnight: PDF, TXT, USGS PDF. There were a couple of small impulses just before 1am and after 2am EST. The USGS plot shows another one just before this record began. Ran about 10 hours on the inexpensive old HY1803D power supply overnight as the constant current polarizing power supply. (The "polarizing power supply" provides current to the powered coil, of the counter-wound coil sensor pair, to create a magnetic field across the sample volume (of the fluid in the sample bottle) to align a large number of proton spins in the sample fluid. After the polarziation time, the protons of the sample fluid precess in the Earth's magnetic field, causing the precession signal (the free induction decay or FID signal) that is then picked up by the coil and conveyed to the narrow band low noise amplifier. The fundamental frequency of the precession signal yields the exact value of the total magnetic field, the "F" scalar total magnetic field via the Larmor precession constant.) The HY1803D was set to about 12 V, which automatically reduced to about 10.6 V to provide a 1.7 A polarizing pulse. So, the total resistance of the powered coil circuit including the powered coil of the counter-wound coils combined with the feed cable (the pair for the powered coil) is presently about 6.24 ohms.
Graphs of selected field data (shown in the field plot PDF, from the TXT file): PDF. There were just over 700 data points recorded, so just over 1 a minute on average. 256 measurements were deemed acceptable (FOM > 2e6) and plotted on the field graph. Only 16 measurements failed for insufficient amplitude, 435 failed for high FOM. The graph looks great, there is no problem, however, it would seem desirable to have less rejected measurements. It appears that increasing the polarizing current from 1.2 A to 1.7 A, while considerably improving the amplitude of the fundamental FDM frequency, has not significantly reduced the number of auto-retries.
Recall that measurements are made every 2 minutes, with auto-retry taking repeat measurements as needed at about 8 seconds until an acceptable single-shot measurement (no averaging) is recorded. Accounting for auto-retries at 8 second intervals, the overall measurement rate was around 26 measurements per hour, versus 30 per hour for perfect data collection at 2 minute intervals (with no auto-retries). The two minute interval is simply a convenient data rate that gives a nice 12 or 24 hour plot. Were it important to take more measurements, the 2 minute interval could be significantly shortened.
Questions to look into are: Is the HY1803D ripple too high? Is there a distortion in the turn on or turn off waveforms? Or, are there just environmental RFI, EMI, or magnetic factors that would cause those measurments to be rejected with any polarizing power supply?
Here is an interesting NASA webpage the Costello Geomagnetic Activity Index that predicts upcomming geomagnetic activity. A website spaceweather.com, a great aggregator site of all solar, geomagnetic and related to auroras, is predicting activity in the next couple of days. I do not see similar predictions at the more formal NOAA, solarmonitor.org site, or at the Big Bear Solar Observatory (click on "operations", "activity report") yet, but stay tuned.
Please let us know if you know of a good solar or geomagnetic activity website for our upcoming list of website references. We very much appreciate reader feedback and input!
I re-coded slightly the amplitude threshold routine within the FDM executable. Before, I just looked through the FDM spectra for an FDM frequency above a certain threshold. The danger was that the search was done from lowest f to highest f of the spectra. If the threshold was too low, FDM could report a spurious f instead of the fundamental f, particularly in a noisy spectra during high noise conditions. However, even with amplitude variation of the whole spectra and varying signal to noise ratio (e.g. the ratio of the amplitude of the fundamental f to the amplitude of nearby spurious fs), (I think) the fundamental always has the highest amplitude. Now, the routine searches for the highest f, then double checks the f with the highest amplitude using a modest amplitude threshold and frequency range. This change should virtually eliminate the 0 Hz returns. It will have no effect on the number of auto-retries due to high FOM. The FOM filtering is done by the LabView program based on the returned FOM of the highest amplitude frequency. If all this seems cryptic, don't worry. There will be clear flow charts, and for most experimenters, FDM will be just a "black box" software module.
Thursday, May 27, 2010
Overnight: PDF, TXT, USGS PDF. A quiet geomagnetic period provided a nice field plot overnight. The small spikes are from trucks that stopped for a moment within some hundreds of feet of the sensor.
Polarizing Power Supply: Still running on the relatively inexpensive HY1803D import power supply set to run in constant current mode (voltage is set to 12 V, slightly higher the the about 10.6-10.9 V to provide the 1.7 A polarization current, the load resistance varies slightly with temperature). This is very encouraging. So, while some experimenters will opt for new or surplus relatively high end hp or Agilent supplies with a constant current auto cross-over mode (e.g. hp or Agilent U8001A, U8002A, E1315A, 6632B, etc), others will have an affordable option in the $50 to $75 range with the HY180* or HY150* type import power supplies (See May 25 entry) widely available from distributors and on eBay. The desired specs. are 15 V to 20 V range (depends on the resistance of your feed line and powered coil), and 2 A to 3 A rated output current, with the operating current between about 1.5 A and 1.75 A. (Sadly, the 1.2 A output of the hp 6237B triple (also provides the plus/minus analog power) is too low).
FDM Signal to Noise Ratio: I added a new parameter, the FDM S/N (next to the FDM FOM) PDF. This is a relatively narrow band calculation, so it may or may not turn out to be of any value in studying the operation of the FDM magnetometer. FDM, as presently configured, calculates over 100 (164) frequencies (the fundamental plus noise spurs) in a relatively narrow band (presently 300 Hz). The frequencies are not equidistant, but rather high resolution determinations of where the signal energy lies within the 300 Hz band. Since the raw digitized waveform has been modified by an exponential function (apodization), the frequencies of continous amplitude noise spurs are calculated as well as the fundamental frequency of the FID (free induction decay) signal. The FDM S/N is 20 LOG of the ratio of the amplitude of the fundamental f divided by the squareroot of the sum of the square of the amplitude of each of the noise spurious frequencies. I expect loose correlation at best, to noise viewed on the precession waveform filtered envelope, since that waveform reflects noise over a wider bandwidth.
Evening: There was an interesting sudden impulse recorded across the country: PDF, USGS PDF. A impulse records well with the FDM magnetometer since there is no moving average.
Friday, May 28, 2010
Overnight: PDF, TXT, USGS PDF. The NOAA / Space Weather Prediction Center reported the sudden impulse PDF . Clicking on the blue triangle summary symbol at the website (not in the PDF) reports, "Space Weather Message Code: SUMSUD, Serial Number: 108, Issue Time: 2010 May 28 0308 UTC, SUMMARY: Geomagnetic Sudden Impulse, Observed: 2010 May 28 0259 UTC, Deviation: 33 nT, Station: Boulder".
About 60% of all measurements are being automatically discarded. Since the auto-retry mode (which rejects measurements) is presently running at about 8 seconds and the measurement interval is 2 minutes, there is plenty of overhead, so this is not a problem, more of a curiosity. I might try running at a different current, say 1.5 A (~220 Gauss) versus 1.7 A (~250 Gauss) to see if it changes the measurement yield. I suspect that FDM is such a powerful filtering system (effectively providing ultra high Q for each returned frequency) that above some reasonable working level, FID amplitude (the free induction decay or precession signal amplitude) is less important.
The question remains, would a perfect instrument still have a 40% yield? If the discarded measurements are an accurate read of the sensor at that moment in time, as opposed a noisy reading caused by a difficulty in making a good measurement, the answer is yes.
Speaking of auto-retries and measurement overhead, it occurs to me that the two minute interval begins with an accepted measurement. So, the intervals are variable, depending on how many auto-retries occured (0 ... n) during a given measurement. I suppose measurement intervals could time more uniformly, with each auto-retry event decreasing the time to the next measurement. With eight second auto-retries, it is highly unlikely the two minute interval would be exceeded. On the other hand, the actual measurements would still be unequal in time, with some measurements closer to others (e.g. after say 10 auto-retries (highly unusual), it would be less than a minute to the next measurmement. something to think about, not sure which approach is better, the present graphs seem okay.
Evening: There was a slightly unsettled period in the field this afternoon: PDF . The "noise" here is actual noise in the geomagnetic field, not instrument noise; the offset is a parked vehicle).
Saturday, May 29, 2010
Very early morning: PDF, USGS PDF, GOES 13 satellite magnetometer PDF. Looks a minor to moderate geomagnetic disturbance, possibly a geomagnetic storm, is developing. Usually our plots most closely follow the Fredericksburg (Corbin), VA site, and ocassionally more closely some features of the Stennis, MS plot. However this morning, I also noticed that there is a close correlation between the shape of our graph (did not compare measurement times) and the GOES 13 real time magnetometer plots, interesting. One reason for the correlation might be that we are also very near the W75 degree Longitude line where the GOES 13 satellite is parked in a geostationary orbit (-4 hours from UTC). Afternote, other times there appears to be less similarity between our ground based measurments (which generally agree with the USGS VA and/or MS station plots) and the GOES 13 graphs.
Overnight: PDF, TXT, USGS PDF, GOES 13 (last two days) PDF. The local field is almost -100 nT from its peak last evening. NOAA is reporting a G1 geomagnetic storm. It is interesting that having one's own magnetometer, the onset of a magnetic storm can be detected before notice shows up at some of the official websites. I assume that critical sites, including the power infrustructure, have some sort of email, text message, and/or telephone notification system that is closer to real time. The aggregated information and predictions at the aurora site spaceweather.com have been quite good this month.
Ran overnight, still using the inexpensive HY1803D import polarization power supply at about 1.5 A. The number of auto-retries was down slightly (50% over about 1,200 measurements). I make no conclusions other than that the inexpensive supply does appear to be a viable option. I hope to try a couple of types of relatively high end Agilent power supplies (6632B, E3615A) next week. I do not expect to see much change, if any, in the measurements.
While relatively difficult to prototype and design, once up and running, the FDM magnetometer prototype is proving to be a very robust instrument. It is very insensitive to both the analog voltage (6 V to 15 V) and the polarizating power supply (about 10 V to 11 V, or as presently configured, having auto-crossover to a constant current mode at about 1.5 A to 1.7 A). With the HY1803D shown in the pic, we are not using the third output of the hp 6237B. The 6237B could be replaced by any suitable +/- 6 V to +/- 15 V supply, ranging from a home brew full wave rectifier ps to an open frame, bench supply, or even lead acid batteries. Also, recall from earlier description, the transformer circuit in the small box next to the NI USB 6008 moves into the larger analog box in the production experment, so there will probably be two aluminum boxes.
While we like the Prestone De-Icer, since it easily got us through an upstate NY winter, the instrument can run on many different types of fluids ranging from ordinary tap water to wine vinegar and organic cleaners! I still need to investigate the FDM amplitude variation with temperature. That might be a fluid temperature and/or fluid bottle pressure issue.
I will be revisiting some of the amplifier design over the next few weeks while I think about wrapping up the design. I will start with the input common mode filter and post some notes here. The main function of the input common mode filter is to block RF pickup, such as from nearby radios and cell phones. Such RF signals can be rectified by the small signal analog circuits causing problems ranging from distortion to input stage saturation.
Input Common Mode Filter: I free air wired something similar to my present input common mode filter JPG. The input signal was applied between the 50 ohm resistor and common. The output signal was measured from the junction of one of the 49 ohm and 10 k resistors to common using a hp 400EL AC voltmeter. As can be seen from the handwritten notes on the schematic, by 18 MHz there is a 30 dB attenuation (the reference level was about -10 dB). VHF to UHF transmissions from passing cars or cell phones should be sufficiently attenuated, as evidenced by good operation to date. (Here is a very handy downloadable (free) mini ring core calculator.) I pickup up some hundreds of high mu T-37 cores for the project. I need to more accuratly measure the mu at some point, however for now, the inductance and attenuation test is good enough. 10 turns gives an inductance of on the order of 0.7 mH. I simultaneously wound two wires to build the common mode transformer (see pic). The 10k ohm resistors are needed to bias the SSM2019 input stages. Afternote: The T-37 ferrite core has proven a little difficult to measure at first glance. At 1 kHz with 10 turns, I get somewhere between about 500 to 700 uH. At 500 kHz, more like 1.1 mH (not surprising, since the ferrite mu will vary with f). Also, I am not confident in the 18 MHz value in the test data, however I do believe that this circuit will be completely effective at VHF and UHF frequencies, the biggest concern. This common mode circuit is intended as an RF filter, and mostly ineffective at low frequencies, where the first line of defense is the counter-wound coil pair.
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.