Skeptic wrote on Aug 24^th, 2003 at 6:10am:

Transducers used in the BP cuff produce relatively small signals and would be most sensitive. The failure mechanism is rectification or other non linear interations with jamming RF. Shutting down a control CPU is best done with wide spectrum impulse noise.

Jamming with EM fields in the area of interest (in band) would require huge B or E fields. Given the wavelength here, all effects are near field.  I would give B fields the biggest chance. Any effect would require B fields so large that they would vibrate metal of any significant cross section. In band E fields would not show since GSR would couple poorly (cap impedance is quite high compared to skin Z) 

-Marty

Food for thought. I know this is an MRI paper, hence the relationship to Tesla type fields, but anyone that has to go to this much trouble to prpoerly detect the human resistance response........got to be a fire in the somke somewhere......

Thus:

human skin conductance changes (SCRs)

Changes in human skin conductance are frequently used as an indirect measure of a subject’s cognitive effort, emotional arousal, or level of attention.  Recording skin conductance responses (SCRs) is useful during functional magnetic resonance imaging (fMRI) studies that hypothesize correlations between blood oxygen level dependent (BOLD) signals and anxiety levels or emotional responses.  Because clinical MR scanners do not come equipped to monitor SCRs, one must independently acquire the apparatus to do so, and then adapt it for use in an MR scanner.  The two main technical challenges are:  a) obtaining SCR signals that are not corrupted by interference from the MR scanner’s changing magnetic field gradients, and b) maintaining low background noise in the MR images.  Recent studies reported monitoring SCRs during fMRI (1, 2), but the technical issues above are not addressed in detail.  In addition, reference (1) reported that scanner-induced artifacts rendered SCR data useless in 4 out of 9 trials.  Both studies reported normalized SCR data rather than amplitudes measured in S (S=1/), making it difficult to compare their SCR data to values reported in the literature.  We therefore feel the need for an article that addresses the technical aspects of monitoring SCR on human subjects during fMRI.  In this article we describe the construction, calibration, and testing of a versatile, low-cost system for monitoring SCRs in a clinical MR scanner.  We show that the system suppressed scanner interference during fMRI to an acceptable level independently of the repetition time (TR) and slice orientation of the fMRI sequence.  In addition, the presence of the SCR system inside the MR scanner did not adversely affect fMRI images as measured by the image background noise level.   
SCR monitoring circuit:  The design of our system was based on the frequency and amplitude characteristics of SCRs.  The range of human SCR magnitudes is from SCRmin0.01 S to SCRmax 1 S  (3,4).  In order to ensure adequate resolution in measuring the smallest conductance responses, we required the SCR monitoring system to suppress interference during fMRI to a level ,  an order of magnitude below SCRmin 
                              =SCRmin/10=10-3 S   
We hypothesized this could be accomplished by using a low-pass filter with cutoff frequency of 1 Hz (5), because SCR signals change relatively slowly compared to interference generated by the MR scanner. .  Figure 2 shows the schematic diagram of the SCR monitoring circuit consisting of a Wheatstone bridge, a differential amplifier, and low-pass filter.  The design of the bridge circuit was taken from Ref. 3, and a complete description of its use is given there.  R1 is a 10-turn potentiometer, and all fixed resistors have 1% tolerance.  The bridge output Vout (between points C and D of Fig. 2) is proportional to the change in the subject’s skin conductance C4, as given by
                                (1)                                       
(see Appendix) where C3=1/R3=5000 S, and Vin=0.488 V is the voltage from A to B, regulated by the LM113H reference diode.  If one inserts the component values into Eqn. (1) and uses the lower limit of human SCR amplitudes C4=SCRmin, one finds an expected minimum bridge output voltage of Vmin=1 V.  We adopt as a rule of thumb for determining adequate amplifier gain G that a minimum of 10 voltage steps Vres must be present over the lower limit of amplified output voltage GVmin, or GVmin/Vres10, yielding for G the criterion 

                                   (2)

We therefore calculated that the gain should satisfy G  610.  We chose to use the differential amplifier chip AD624AD (Analog Devices, Norwood, MA) designed for a gain of 1000  (details are in Ref. 6).  The amplifier output was fed to a low-pass filter with a 3 dB cutoff frequency of 1 Hz.  We selected the 2-pole, Butterworth low-pass filter for its flat pass-band magnitude response and its moderately steep attenuation (-40 dB/decade) beyond the 3dB point.  The filter circuit of Fig. 2 was designed using FilterProTM software for the universal filter circuit UAF42AP (Burr-Brown, Tucson, AZ), where R6=1.58 M and C=100 nF.  NPO ceramic, silver mica, or metallized polycarbonate capacitors are recommended (6).  Voltage supplies ranging from 6 to 18 V may provide power to the amplifier and filter circuits. The resistor R5 limits current through the LM113H, which maintains a constant voltage of 1.22 V, and satisfies


Cable shielding:  Shielding the cable (shown schematically in Fig. 1) with braided copper sheath was another step we took to reduce scanner interference.  This established a ground connection between the chassis of the electronics box, the custom-built aluminum door, and the scan room shielding, and reduced interference considerably.   In our case shielding was particularly important because we put the 40 m data cable permanently in place over the scan room ceiling.  The cable extended from the rear scan room door to the console room in front of the scan room.  The length of the cable and its proximity to gradient power supplies made this step necessary.
Door:  To prevent the SCR cable from picking up outside radio frequency (RF) interference, transferring the RF interference into the scan room, and degrading the quality of the functional MR images, we filtered the cable as it passed through the penetration ports on the custom-built door (Fig. 3(b)).  The door itself was assembled from aluminum angle irons cut to the dimensions of the doorway, with allowance made on the sides and top for the addition of metal contact fingers (Lindgren, Glendale Heights, IL) compressed 70% of their width.  The lower half of the door was covered with an aluminum plate 0.64 cm (0.25 in) thick, and the upper half with aluminum screen.  Aluminum side handles were also added to make moving the door easier.  A twin BNC connector was mounted to the aluminum plate (Amphenol #31-225) to mate with the SCR electrode cable on the immobilizer.  Mounted to the aluminum plate was a filtered penetration port (Fig. 3(c)), consisting of an interior aluminum bulkhead through which two 400 Hz, low-pass, feed-through capacitors (Spectrum Control, Fairview, PA) were mounted.   
Calibration:    SCR amplitudes C4 are related to the amplified output voltage of the circuit by C4=-Vtotal  where = C3/(VinG) (see Appendix).  Given that C3=5.00x10-3 S, Vin=0.488 V, and G=1000, one expects theoretically that =10.2 S/V.  Further, because C3, Vin, and G are known to within 1% , one can show that  should have an uncertainty of  3%.   We also determined  experimentally by sampling Vtotal while R4 was switched to the internal resistors 1 M and 0.1 M.  Then =-C4/Vtotal was used to obtain
 
in good agreement with the expected value.   We used the experimentally obtained  value, and  programmed the data acquisition software to convert from volts to S.  As a check of the system accuracy, we connected a precision resistor and potentiometer in series across the SCR leads and measured the difference between the maximum and minimum conductance.  The difference in conductance was determined beforehand by digital multimeter to be 0.49 S.  The calibrated SCR system yielded 0.46 S, in good agreement. 

Skeptic wrote on Aug 24^th, 2003 at 8:20pm:


Skeptic and AH:

I agree the box would be EMI shielded. In a commercial product the incoming leads would be shielded as well to prevent internal RF from escaping.

Rectification comes from rf on the bipolar amplifier inputs (most low noise, high sensitivity amps extant). They casue what lookes like an out of spec input EOS or IOS. Since it's at the inner most point on the loop, the offset gets directly amplified by the loop gain. I've seen significant effects from local FM radio on Analog Devices modules.

Quote:

A rather simple way to look at it is this way. The capacitance between a 1m wire and a surface m^2 at a distance of 1m is going to be around 10^-12.  Impedance at 1Hz is about 10^11.  Assuming skin Z at 10^3, there would be a coupling reduction of the E field of approx 10^-8.  Assuming a 1V DC operating point, a 1 Million volt, 1Hz E field would produce about a 1% modulation in measured skin Z. For impulse noise, anything more than a 1 pole filter would attenuate faster than the increased coupling at higher Hz's.

Out of band RF is a much bigger problem but easily contained, both by shielding and good bypass designs on the front end. The latter is typical of modern commercial designs but ad hoc lab work is far more variable - as the MRI-GSR descr. shows.

Wheatstone bridges have an advantage in reducing sensitivity to power supply variation and even working well with old, direct drive centered galvanometers. This has long been a non issue as high stability, cheap, voltage sources have been around 30 years or so.

-Marty