Frequency Estimator Dead Zone

DEAD ZONE MYSTERY

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)

DEAD ZONE MYSTERY 12/28/09

Here are some pdfs and data files showing the dead zone. (I need to add frequency to the .txt file log.)

Using the "live" system, I disconnected the USB 6008 analog input from the preamplifier and plugged into the Agilent Synthesizer (GPS locked). I ramped the Agilent through the "dead zone", first at .001 Hz steps, then at .01 Hz steps. The dual slope ramp shows at the left of the pdf. Then I took actual Earth's field magnetic data for a while. Next, I manually set two frequencies (again with the synthesizer at about 500 mV p-p) to two frequencies (2291.810 Hz and 2291.840 Hz) in the dead zone. The two horizontal lines show proper reading of the two tones. Next, I resumed actual data taking. Here is the pdf.

Just for general info about the experiment, here is a plot showing the magnetic field changed caused by the car (a Subaru wagon, at about 33 feet from the sensor) going out for a while. Here is the pdf.

The overnight run gives a great example of the dead zone as the field just happened to be hovering near it. The pdf. The data. The USGS pdf plot. The long vertical line of dots is the garbage truck going down the street. The shorter lines are the snow plow (just a few inches and it was going fast). (about 30 seconds per point).

DEAD ZONE MYSTERY 12/29/09

I began today by testing the tone measuring express vi. While appearing promising for not exhibiting the same dead zone, the scatter of the data (short term instrument noise) is way too high. Here are two short runs of 100 points at about 5 seconds per point comparing Buneman to Tone Measure: In the Buneman PDF, (Buneman txt file), the sample rate was 6 kHz, number of samples per shot, 8192, 512 taps on the bandpass digital FIR filter, 3 second loop delay (so probably about 5 seconds total with the 2 second polarization time) the shot to shot variation is well under 1 nanoTesla (some of the variation is probably a correct measurement of noise in the local scalar field value). In the Tone Measure PDF, (TM txt file), the sample rate was 10 kHz, number of samples 12000, and there were 128 taps on the digital bandpass FIR filter. The point to point noise appears five to ten times higher than with Buneman and is believed to be computational error (i.e. not representing short term changes in the magentic field).

Then, I tried some of the express VI filters in place of the digital FIR filter with the Tone Measuring VI. With the 10 kHz oversampling, I preceeded the 22 order Bessel bandpass filter with a low pass filter. The data (txt) looked promising at first, (low noise shot to shot) but then a similar dead zone appeared. The shift is somewhat similar to that caused by nearby vehicle, however, a bistable dead zone shift has flickering back and forth between two values, something non-physical that does not occur with an offset caused by a vehicle.

I think all of this still shows promise for using LabView signal processing, however, of course the instrument is not viable with a bistable dead zone phenomena. Still early in the process though ...

DEAD ZONE MYSTERY 12/30/09

Tried Butterworth Express VI bandpass filter, 14th order, 2288 Hz to 2300 Hz. Looked promising overnight (30 seconds per sample): PDF, txt.

Since field happend to be low, was able to very slowly back car out from about 30 feet to about 100 feet or more (stopping about 3 seconds every few feet) to cause a positive change in the field (car depresses field artificially low in normal spot in garage). 6 seconds per sample: PDF, txt.

Apparent dead zone showed. Field happened to be moving into and through dead zone (6 seconds per sample), so plotted for a while, then started to inject sine waves (the straight horizontal lines) across range from Agilent 33120A synthesizer. PDF, txt. (the slight rectangular depression and odd points lower left is a vehicle arriving at an adjacent property, then leaving).

This time the dead zone is repeatable with the signal generator: PDF, Excel

DEAD ZONE MYSTERY 1/01/10

I switched back to the Keithley 16 bit digitizer (Keithley 194A High Speed Digitizer, an older but very flexible and easy to use up to 100 kHz at 16 bit digitizing instrument) to continue with the new focus on the frequency determining algorithm. A quick test this morning by moving the car out of the driveway shows one of the largest apparent dead zones to date. This was with 10,000 samples at 10 ksps and the Buneman frequency estimator following a 128 Tap FIR filter. Here is the PDF, txt file.

I am going to try some adaptive filtering techniques to move a second narrower FIR filter at each measurement to see if it has any impact on the Buneman frequency determination. The Buneman estimator remains attractive, sine sweeping a synthesizer shows that it can track a sinewave correctly to about +/- .002 Hz or better.

DEAD ZONE MYSTERY 1/02/10

It was easy to implement the tracking filter, after a first determination, loaded the new filter parameters, into the tracking filter +/- 0.3 band pass Hz +/- 1 Hz band stop 256 taps. The tracking filter then operated on the raw (analog filtered) data that was input into the first filter. Overnight paramaters were the Kiethley 194 high speed digitzer, 4096 pts, 166 us per point. Also, I removed more end points (to remove processing artifacts) which left 3839 points left going into the Buneman estimator. The Buneman estimator does not seem to need a 2toN input number of points. Whether it doesn't need it or it pads with zeros to 2toN, I do not know. Same problem ... more deadzone: the PDF and txt files. USGS, same period PDF. I need to look at the txt data more carefully. A first quick look shows quite a different dead zone for the first field determination versus the output of the tracking filter. Maybe there is a clue there?

DEAD ZONE MYSTERY 1/03/10

Tried going back to a single FIR filter overnight. Surprisingly 12 taps works pretty well with Buneman (scatter in shot-to-shot data seems reasonable). However the dead zone persists: PDF, txt, USGS, Histogram of data PDF.

Project Articles!

Project Documentation (very early stages)

Past Project Journal Notes

QUESTIONS/COMMENTS/notice of typos, etc. send email to joegeller @ gellerlabs dot com

Geller Labs	Geller Labs
Products	DEAD ZONE MYSTERY 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) DEAD ZONE MYSTERY 12/28/09 Here are some pdfs and data files showing the dead zone. (I need to add frequency to the .txt file log.) Using the "live" system, I disconnected the USB 6008 analog input from the preamplifier and plugged into the Agilent Synthesizer (GPS locked). I ramped the Agilent through the "dead zone", first at .001 Hz steps, then at .01 Hz steps. The dual slope ramp shows at the left of the pdf. Then I took actual Earth's field magnetic data for a while. Next, I manually set two frequencies (again with the synthesizer at about 500 mV p-p) to two frequencies (2291.810 Hz and 2291.840 Hz) in the dead zone. The two horizontal lines show proper reading of the two tones. Next, I resumed actual data taking. Here is the pdf. Just for general info about the experiment, here is a plot showing the magnetic field changed caused by the car (a Subaru wagon, at about 33 feet from the sensor) going out for a while. Here is the pdf. The overnight run gives a great example of the dead zone as the field just happened to be hovering near it. The pdf. The data. The USGS pdf plot. The long vertical line of dots is the garbage truck going down the street. The shorter lines are the snow plow (just a few inches and it was going fast). (about 30 seconds per point). DEAD ZONE MYSTERY 12/29/09 I began today by testing the tone measuring express vi. While appearing promising for not exhibiting the same dead zone, the scatter of the data (short term instrument noise) is way too high. Here are two short runs of 100 points at about 5 seconds per point comparing Buneman to Tone Measure: In the Buneman PDF, (Buneman txt file), the sample rate was 6 kHz, number of samples per shot, 8192, 512 taps on the bandpass digital FIR filter, 3 second loop delay (so probably about 5 seconds total with the 2 second polarization time) the shot to shot variation is well under 1 nanoTesla (some of the variation is probably a correct measurement of noise in the local scalar field value). In the Tone Measure PDF, (TM txt file), the sample rate was 10 kHz, number of samples 12000, and there were 128 taps on the digital bandpass FIR filter. The point to point noise appears five to ten times higher than with Buneman and is believed to be computational error (i.e. not representing short term changes in the magentic field). Then, I tried some of the express VI filters in place of the digital FIR filter with the Tone Measuring VI. With the 10 kHz oversampling, I preceeded the 22 order Bessel bandpass filter with a low pass filter. The data (txt) looked promising at first, (low noise shot to shot) but then a similar dead zone appeared. The shift is somewhat similar to that caused by nearby vehicle, however, a bistable dead zone shift has flickering back and forth between two values, something non-physical that does not occur with an offset caused by a vehicle. I think all of this still shows promise for using LabView signal processing, however, of course the instrument is not viable with a bistable dead zone phenomena. Still early in the process though ... DEAD ZONE MYSTERY 12/30/09 Tried Butterworth Express VI bandpass filter, 14th order, 2288 Hz to 2300 Hz. Looked promising overnight (30 seconds per sample): PDF, txt. Since field happend to be low, was able to very slowly back car out from about 30 feet to about 100 feet or more (stopping about 3 seconds every few feet) to cause a positive change in the field (car depresses field artificially low in normal spot in garage). 6 seconds per sample: PDF, txt. Apparent dead zone showed. Field happened to be moving into and through dead zone (6 seconds per sample), so plotted for a while, then started to inject sine waves (the straight horizontal lines) across range from Agilent 33120A synthesizer. PDF, txt. (the slight rectangular depression and odd points lower left is a vehicle arriving at an adjacent property, then leaving). This time the dead zone is repeatable with the signal generator: PDF, Excel DEAD ZONE MYSTERY 1/01/10 I switched back to the Keithley 16 bit digitizer (Keithley 194A High Speed Digitizer, an older but very flexible and easy to use up to 100 kHz at 16 bit digitizing instrument) to continue with the new focus on the frequency determining algorithm. A quick test this morning by moving the car out of the driveway shows one of the largest apparent dead zones to date. This was with 10,000 samples at 10 ksps and the Buneman frequency estimator following a 128 Tap FIR filter. Here is the PDF, txt file. I am going to try some adaptive filtering techniques to move a second narrower FIR filter at each measurement to see if it has any impact on the Buneman frequency determination. The Buneman estimator remains attractive, sine sweeping a synthesizer shows that it can track a sinewave correctly to about +/- .002 Hz or better. DEAD ZONE MYSTERY 1/02/10 It was easy to implement the tracking filter, after a first determination, loaded the new filter parameters, into the tracking filter +/- 0.3 band pass Hz +/- 1 Hz band stop 256 taps. The tracking filter then operated on the raw (analog filtered) data that was input into the first filter. Overnight paramaters were the Kiethley 194 high speed digitzer, 4096 pts, 166 us per point. Also, I removed more end points (to remove processing artifacts) which left 3839 points left going into the Buneman estimator. The Buneman estimator does not seem to need a 2toN input number of points. Whether it doesn't need it or it pads with zeros to 2toN, I do not know. Same problem ... more deadzone: the PDF and txt files. USGS, same period PDF. I need to look at the txt data more carefully. A first quick look shows quite a different dead zone for the first field determination versus the output of the tracking filter. Maybe there is a clue there? DEAD ZONE MYSTERY 1/03/10 Tried going back to a single FIR filter overnight. Surprisingly 12 taps works pretty well with Buneman (scatter in shot-to-shot data seems reasonable). However the dead zone persists: PDF, txt, USGS, Histogram of data PDF. Project Articles! Project Documentation (very early stages) Past Project Journal Notes QUESTIONS/COMMENTS/notice of typos, etc. send email to joegeller @ gellerlabs dot com COPYRIGHT © 2009 JOSEPH M. GELLER, All rights reserved.
Manuals
Tech Notes
About Geller Labs
Contacts
Links
Ordering