in vitro exposure setup for elf magnetic fields · pliﬁers (tda 7294) working in parallel. a...

Chapter 9

In Vitro Exposure Setupfor ELF Magnetic Fields

123

124 CHAPTER 9. ELF IN VITRO SETUP

SUMMARY

For the conduction of in vitro studies on the effect of ex-tremely low frequency (ELF) magnetic field exposures in dif-ferent laboratories, a programmable high-precision exposuresystem enabling blinded exposures has been developed andfully characterized. It is based on two shielded 4-coil sys-tems, that fit inside a commercial incubator. The volumeof uniform B-field exposure (standard deviation: < 1%) is3500 cm3, which is considerably larger compared to a Merrit4-coil system with the same coil volume. The uncertaintiesfor the applied magnetic fields have been specified to be lessthan 5%. The computer-controlled setup allows signal wave-forms that are composed of several harmonics, blind proto-cols, monitoring of exposure and environmental conditionsand the application of B-fields up to 3.6mTrms. Sources of ar-tifacts have been fully characterized: sham isolation > 43 dB,parasitic incident E-fields < 1 V/m, no recognizable temper-ature differences in the media for exposure or sham state,and vibrations of the mechanically decoupled dish holder< 0.1m/s2 (= 0.01gGr) which is only twice the sham accel-eration background level produced by the incubator and fanvibrations.

9.1 Introduction

A considerable variety of exposure setups has been used to investigatethe biological effects due to the exposure of extremely low frequency(ELF) magnetic fields. Among them are Helmholtz coils, e.g., [29],4-coil systems [38], [126], [124], solenoids [53], and also ferromagneticcores [88]. Most of the systems employ two separate units in order toisolate the sham group from the field unit, although shielded configu-rations have also been in use.

The objective of this study was to develop a setup which is op-timized for a uniform B-field exposure and for minimum non-B-fielddifferences between the exposure and sham groups. The presented

9.2. REQUIREMENTS FOR THE EXPOSURE SYSTEM 125

system is applied within the “REFLEX” project1, which is part ofthe 5th European Framework program “Quality of Life”. REFLEXis focussed on the in vitro analysis of non-thermal and low level elec-tromagnetic field exposures with respect to possible genotoxic effects,effects on differentiation and function of embryonic stem cells, effectson gene expression and protein targeting, effects on the immune sys-tem and effects on transformation and apoptosis of cells.

9.2 Requirements for the Exposure Sys-tem

The requirements for ELF exposure setups in health risk research havebeen formulated by several authors [68], [119], [84]:

large loading volume with uniform exposure

high dynamic (µT - mT) and frequency ranges (subHz - kHz)

enabling complex signals and intermittent exposure

good isolation between exposure and sham

identical atmospheric parameters for exposed and sham cells(preferably placed in the same incubator)

blinded exposure protocols by a computer controlled randomdecision maker

continuous monitoring of all environmental and technical pa-rameters in order to detect any malfunctions

evaluation of possible artifacts such as parasitic E-fields, tem-perature loads, vibrations, etc.

9.3 Design of the ELF Setup

The setup is based on two identical coil systems (Figures 9.1 and9.2) that are placed beside each other inside a commercial incubator.

1REFLEX: Risk Evaluation of Possible Environmental Hazards from Low En-ergy Electromagnetic Field Exposures Using Sensitive In Vitro Methods.


µ

!

"#

" #

$!

Figure 9.1: Functional parts of the coil chamber in top view (a) andside view (b).

They are optimized to produce a homogeneous, linearly polarized B-field over the area of the Petri dishes in which the cells are located.The B-field vector is perpendicular to the plane of the Petri dishes.For the coils, a pair of electrically isolated copper wires (∅ = 2mm)is used. The wires are stabilized against vibrations by epoxy resin.Each coil system is composed of four quadratic subcoils (side length20 cm) which are arranged symmetrically to the 4-coil center. The

9.3. DESIGN OF THE ELF SETUP 127

Figure 9.2: (a) ELF coil system (lid and dish holders have been placedbeside the chamber); Pt100 temperature probe and E-field shield canbe seen. Two dish holders are shown with and without vibration-damped elastic feet; (b) Setup consisting of an exposure and a shamcoil installed in the incubator (one lid removed).

vertical distances of the center and outer subcoils are 6 cm and 18 cm,respectively, whereby the center subcoils consist of 50, compared to 56


windings for the outer subcoils. The entire 4-coil system is placed in-side a 25 cm x 25 cm x 25 cm µ-metal box (Amuneal ManufacturingCorporation, box consisting of 80% nickel-iron alloy with a perme-ability of This box serves first to shield the ambient incubator ELFfields, and secondly to isolate the sham group, so that both coils canbe kept inside the same incubator and do not influence each other.µr = 400′000). Removable lids on top of the boxes allow access to thePetri dishes, which are placed on a holder. Dish sizes up to a diam-eter of 12 cm fit inside. The dish holder as well as µ-metal box areplaced on elastically damped feet in order to minimize vibrations ofthe dishes (most partners of the REFLEX project used an older dishholder model without the damped feet (Figure 9.2a left); both holderswere analyzed for vibration load). The current in the coils produceheat that must be removed from the exposure chambers. For thatpurpose an air cooling system based on two fans (Pabst 612NGHH)per coil has been introduced (Figure 9.1b). The Petri dish holderis separated from the coils by an electrically grounded metal shield(dimensions: 16 cm x 16 cm x 23 cm). It is open at the top and bot-tom sides and connected at the side via a plastic spacer (Figure 9.1a).It fulfills two tasks: 1) shielding of the exposure area from parasiticelectric fields that are generated by the setup and 2) controlling theair flow (air cooling of the dishes from below - air cooling of the coilsfrom the top - warm air blown out over the fans). For the monitoringof the air temperature inside the chamber, Pt100 temperature probes(Labfacility) are fixed at the inner side of the shield.

9.4 Signal Generation, Monitoring and DataEvaluation

Figure 9.3 gives an overview of the electronic setup. A software-controlled arbitrary function generator (Agilent 33120A, 0.1 mHz -15 MHz, 16k points / 12 bit, 50mVpp−10Vpp) is used as signal sourcefor two identical current sources. They are designed as feedback-controlled, constant-current drives based on two monolithic audio am-plifiers (TDA 7294) working in parallel. A current of 10App can beachieved. Voltage compliance is 80Vpp (power supply: Lambda, Alpha

9.4. SIGNAL GENERATION, MONITORING AND DATA EVALUATION129

Current source 1

Exposurecoil

Current source 2

Shamcoil

Ucoil

UIB

Ucoil

DP Relays

Arbitrary functiongenerator

Fan

Fan

UIfan

Fan

Fan

UIfan

Fan Fan

DL,PC

DL,PC

DL,PC

T T

UIfan

UIB

UIfan

Figure 9.3: Signal unit consisting of an arbitrary function generatorand a data logger (DL), two identical current sources, double-pole re-lays (DP) and two coil systems. The paired wires forming the coils areconnected in parallel for exposure and anti-parallel for sham. Severalsignals are sensed by the data logger: UIfan

: driving currents of ven-tilators used for cooling, UIB

: coil currents, Ucoil: voltage drop overthe coils (exposure / sham indicator), T : air flow temperature.

600). To avoid artifacts due to the digitally generated signal or noise,a third-order butterworth low-pass filter with an edge frequency of1.25 kHz is introduced. The maximum unclipped magnetic field de-pends on the frequency content of the signal and is 3.6mTrms for a 50Hz sinusoidal signal, 2.3mTrms for the “power-line” signal (see below)and 0.6mTrms for pure 1000 Hz. The signals pass two double-polerelays that control the current flow in the two parallel wires of eachindividual coil: same direction for magnetic field and opposite forsham [55].

The relays are switched by a digital signal from the data logger(Agilent 34970A). The decision of which chamber becomes exposedand which one not is taken randomly by the PC. The currents in


the coils are acquired by the data logger as true-RMS measurement ofthe voltage drop on low-temperature-coefficient power resistors (ArcolHS50) mounted on the heat sink of the current source electronics. Thevalues are used for monitoring, as well as for feedback control of themagnetic field. Furthermore, the voltage drops over the entire coils aremeasured in order to check for exposure or sham. Finally, the drivingcurrents of the fans are monitored for the detection of malfunctions.All sensor signals are read out over a GPIB connection by the PC.

An arbitrary exposure signal can be synthesized and downloadedto the function generator. This allows to generate power line signals,consisting of up to 25 harmonics. The signal defined as standard forthe REFLEX experiments represents the maximum accepted distor-tion for low- to medium-voltage power systems by the InternationalElectrotechnical Commission (IEC) [48]. Figure 9.4 shows its spec-tral content as well as the time domain waveform. The given signalis an interesting alternative to the often used pure 50 Hz sinusoidal,because it is much closer to daily exposure. It is more relevant forhealth risk assessment since a wider spectrum is covered.

The spectral content of the currents in the coils was checked bya spectrum analyzer for exposure and sham states. For a 50 Hz si-nusoidal signal the third harmonic of 150 Hz is dominat and is 52dB below the 50 Hz signal. For the power line signal, the distancefor the lowest generated component (0.2% of 50 Hz amplitude) to thehighest unwanted harmonic is still more than 15 dB. Soft switchingof the fields is realized by a linear ramp of 3 s duration from zero tomaximum.

An exposure is initialized and controlled by MS-Windows-baseduser software. All exposure settings like signal type, frequency, dura-tion, etc. can be defined. During exposure, all sensor signals and com-mands are stored in 10 s intervals. The software is able to self-detectmalfunctions by tracing and handling of about 60 types of errors. Itgenerates warnings or abortions if required. After completion of theexperiment, all data is stored in an encoded file, which can be decodedonly by a dedicated program. Decoding is performed after biologicaldata evaluation in order to guarantee the blinded study design.

9.5. MAGNETIC FIELD EXPOSURE 131

Figure 9.4: Standard power line signal for REFLEX experiments;(a) amplitudes of the second to 25th 50 Hz harmonic in percent, asdefined by the IEC; (b) three periods of the power line signal in thetime domain.

9.5 Magnetic Field Exposure

The magnetic field that is produced by the coils has been optimized bynumerical simulation (Mathematica V4.1) based on the law of Biot-Savart. The numerical model of the setup consists of quadratic loopsfor each of the 4 coils, which carry the total current, correspondingto the respective number of windings. The resulting B-field is thesuperposition of the 4 coils. The Petri dishes with the cell mediumand cells do not need to be considered, because they have a relativemagnetic permeability of µr = 1, as long as non-magnetic tissues areunder investigation. However, the coil is placed inside a µ-metal box


Dimensions of µ-metal box 25cm× 25cm× 25cm

Box for air separation 16cm× 16cm× 23cm

Dish holder 12cm× 12cm× 24cm

Dimension of coils quadratic 20cm× 20cm

Number of windings 56 (outer) and 50 (center)

Distance of windings from center 3cm and 9cm (0.15D, 0.45D)

B-field efficiency Bavg/I = 1.06mTrms/Arms

5% deviation area, approx. 12cm x 12cm x 24cm

1% deviation area, approx. 10cm x 10cm x 20cm

Inductiveness and resistance at 60 Hz L = 9.3mH, R = 0.98Ω

Table 9.1: Geometrical and physical parameters of the coil system.

that has to be taken into account by introducing the correct boundaryconditions. This can be achieved by mirrored currents outside of thechamber that enforce a B = 0 condition for the box boundaries. The26 closest mirror currents were taken into consideration for the B-fieldcalculation. Since 16 current filaments form the rectangular 4-coilsystem, 16 · 27 = 432 Biot-Savart integrals have to be calculated inorder to get one sample of the magnetic field.

To optimize the field distribution, a trial-and-error approach wasused. The degrees of freedom were the number of windings and thepositions of the outer and the center loops of the coil. Based on theresults of the simulations, the volumes with field tolerances of 5% and1% relative to the coil center have been analyzed and optimized.

The parameters for the final design of the coils are listed in Ta-ble 9.1. Figure 9.5 compares the B-field distribution with the oftenused Merrit 4-coil configuration, which has no µ-metal shielding [82].Merrit coils with similar side lengths of 20 cm were analyzed usingthe same methods as described above. The grey areas of Figure 9.5represent the area of less than 1% field deviation from the center. B-field samples on a 5 mm equidistant grid have been extracted (74100points). The useable volume with 1% deviation are 41% of the to-tal coil volume for the shielded 4-coil and 27% for the Merrit coils.Therefore, in addition to better shielding of exposure and sham, theµ-metal casing also enlarges the 1% tolerance volume. However, if thetolerance is reduced to 0.1%, the Merrit system provides better uni-

9.5. MAGNETIC FIELD EXPOSURE 133

formity (Figure 9.5b). To obtain a measure of nonuniformity of theB-field at the cell location, the ratio between the standard deviationand the mean value of the magnetic field, analyzed over the volumeof the dish holder (12 x 12 x 24 cm3) was evaluated, and a value of0.99% was found. For a B-field of 1.06mTrms, a current of 1Arms isrequired.

Figure 9.5: (a) Simulated B-field distributions for the shielded 4-coiland a Merrit 4-coil system (axis dimensions are given in meters); (b)Uniform B-field volumes for both configurations as function of allowedtolerance in %.


9.6 Validation

The magnetic field of the coil system has been measured for expo-sure and sham conditions using a Gaussmeter (Magnet-Physik, FH49)equipped with a 3-axis, temperature-regulated magnetic field probe(HS-ZOA71-3208-05-T). The probe was handled by a positioner robot(SPEAG, DASY3 near-field scanner). In order to insert the probe, 10holes were punched into the wall and top of the the µ-metal box, thepositions of which can be seen in Figure 9.6a. The maximum avail-able current at 30 Hz (Irms = 3.534A ± 0.06%) was applied. Eachline defined by the holes was scanned with a 1 cm resolution. Thisprocedure led to 126 measurement points in the exposure area of thedishes. To compare the measurements to the simulations, the B-fieldat the positions of the field mapping was determined by simulations.

Figure 9.6b shows the results for simulation and measurement forfour selected lines. It can be seen that the simulations result in similarfield distributions, but in higher field values compared to the mea-surements. In average an overestimation of 2.9% of the simulationscompared to the measurements was found. This discrepancy can beexplained by the uncertainties of the measurement as well as of thesimulation model. The tolerance of the Gauss meter including thefield probe is specified at 2.9%, which is already in the order of thedeviation. The main sources of errors in the simulation model are theinfinite permeability of the box, current filament modelling coils andthe finite number of mirror currents.

The residual exposure inside the sham coil is determined by theimperfection of the field cancellation in the sham coil and by theshielding efficiency of the µ-metal enclosure. To quantify the fieldcancellation, the area of the dishes has been scanned with full shamcurrent. The highest sham field was 43 dB below the lowest exposurefield in the area of the dish holder. No B-field coupling from the activeexposure coil to the sham coil could be measured. The advantage ofequal temperature load in the sham chamber comes at the cost of asmall residual field due to non-ideal current cancellation caused bythe finite separation between the wires (d = 2 mm).

9.7. VARIATION AND UNCERTAINTY ANALYSIS 135

∅

Figure 9.6: (a) Locations of the holes for the field scanning; (b) Simu-lated (straight lines) and measured field values (symbols) for 4 selectedlines.

9.7 Variation and Uncertainty Analysis

It is important to quantify the uncertainties and variations for thegiven B-field values in order to specify reasonable B-field separations


Source of B-field variation Uncertainty SD

Nonuniformity of B-field 1.0%

Nonlinearity of B(I) relation 1.1%

Temperature dependency of resistors [∆T = 50K] 0.5%

Temporal instability [48 h exposure] 0.01%

Source of absolute B-field uncertainty Uncertainty SD

Tolerance of field probe 2.9%

Deviation between measurement and simulation 2.9%

Tolerance of data logger current measurement 1.4%

Combined relative uncertainty (RSS) 4.6%

Table 9.2: Standard deviations (SD) from specified B-field values dueto variation and absolute uncertainty.

for multilevel experiments (for investigations on dose-response rela-tions). Data for variation and absolute uncertainty are listed in Table9.2. Variations occur due to spatial and temporal changes of the expo-sure conditions, whereas absolute uncertainties consider the method ofcalibration. All given uncertainties are quantified according to [114],and ratios between standard deviation and the average value are givenin percent. The relative uncertainty of 1.9% due to all considered vari-ations and the absolute uncertainty of 4.3% lead to an estimation of4.6% overall uncertainty for the B-field.

9.8 Sources of Artifacts

9.8.1 Induced and Parasitic E-Fields

The electric field induced in the cell medium of Petri-dishes is:

Erms(r) = πr∑

f

fBf = crBrms

whereby r is the radial distance form the dish center and f the fre-quency of the magnetic field with an RMS value of Brms. c [V/m2/T]is a frequency-weighted parameter for signals composed of multiple

9.8. SOURCES OF ARTIFACTS 137

harmonics. c is 664 V/m2/T for the powerline signal, which is 4.2times higher than for the pure sinusoidal signal at 50 Hz. Experi-ments to distinguish effects of magnetic fields and induced E-fieldscan be performed by using large ring-like Petri dishes, since the in-duced E-field increases linearly from the dish center to its edge andis circularly oriented, whereas the B-field exposure is independent ofthe location.

Despite the built-in E-field shield residual parasitic electric fieldsare produced by the imperfections of the setup. They mainly resultfrom the voltage drop over the highly inductive coils. A Wandel andGoltermann EFA-3 sensor system with a cubic 3-axis probe (8 cmside length) was used to measure the E-fields with sinusoidal 50 Hzexcitation. In the center, a dominant vertical E-field component ispresent. Closer to the coil wires, also horizontal components exist.The E-field never exceeded 1 V/m for the exposed and 0.4 V/m forthe sham condition at the maximum current of 3.5Arms and decreaseslinear with the current. The total electric field at the location of thecells is a superposition of the B-field induced and the parasitic fields.Assuming a vertically polarized incident parasitic E-field of 1 V/mresults into an E-fields of about 0.01 V/m inside the medium (εr =80). Parastic E-fields are therefore in the same order of magnitude asthe induced ones. Without the E-field shield fields up to 300 V/m aremeasured, which represents a strong exposure to nonwanted E-fields.

9.8.2 Temperature Load

The construction of the coil chambers was optimized to limit tem-perature and avoid differences between probes in the exposed andin the sham chamber: Energy due to Ohmic losses is dissipated byforced air flow; same currents are applied for both, exposure and shamchambers. Heat production in the cell medium can be neglected be-cause E-fields in the medium are very low (corresponding SAR values< 10−5W/kg).

Nevertheless, the temperature of the air flow as well as of themedium were characterized, since they are dependent on the expo-sure strengths. A flexible T1V3 thermistor probe (SPEAG) with adiameter of 1 mm was fixed in the center of the cell medium within a90 mm Petri dish. Its absolute accuracy is specified to be better than


Figure 9.7: (a) Temperature rise due to 2.5Arms coil current, mea-sured with Pt100 probes in the air at the top and bottom of the cham-ber, and a flexible T1V3 probe inside the cell medium; (b) Typicalair flow temperature history for exposure and sham chambers (T1,T2)with a 1 mT field, applied with a 5/5 minute on/off cycle scheme.Lower trace shows the difference.

0.2C and the noise level is less than 2 mK. The measurements in airwere performed using two Pt100 resistive thermometers (Labfacility)

9.8. SOURCES OF ARTIFACTS 139

with an accuracy of better than 0.1C. One was fixed close to thePetri dish at the top of the chamber; the second was positioned atthe bottom of the box. All measurements were performed inside aHeraeus BB6620 incubator at a temperature of 37C and more than95% humidity.

The temperature over time was measured for 4 different currents:2 A, 2.5 A, 3 A and 3.5 A. Figure 9.7a shows the results for I = 2.5A,B = 2.6mT . The deviations between all three sensors are less than0.05C, which indicates that temperature gradients between the topand bottom of the chamber and the cell medium are negligible. Dueto the isolation of the Petri dish, the medium temperature increaseswith a lower time constant than in air. An incubator temperature riseof 0.1C was measured for 2 A, 0.25C for the shown 2.5 A and 0.3Cfor 3 A. For I = 3.5A the increase is about 10C, indicating that theincubator’s cooling capabilities are exceeded.

Temperature tracking between exposure and sham states was checkedby measuring medium and air temperatures at maximum B-field. Dueto the forced air flow, the temperature differences between both cham-bers can be kept < 0.1C. A typical data set for an experiment con-taining a 1 mT 5/5 min on/off cycle scheme over 16 h can be seen inFigure 9.7b.

9.8.3 Vibrations

Vibrations are critical for the described setup, if they differ for theexposed and the sham cell probes, because they could possibly leadto biological responses. While vibrations by external causes and fansmay be assumed equal for exposure and sham, magnetic forces aredifferent due to the field cancellation.

To assess the acceleration for a Petri dish with and without elasti-cally damped dish holder, a Wilcoxon Research accelerometer Model728T equipped with an amplifier unit P704T was applied. The sensorprobe’s mass of 45 g and attached cable result in a weight comparableto a filled 90 mm Petri dish (proposed volume 35 ml). Therefore, asensor fixed on the dish holder plate approximates the vibrations ofa filled Petri dish. The sensitivity of the non-magnetic sensor to theB-field has been assessed by hanging the probe at the incubator topin vertical and horizontal orientation avoiding mechanical contact to


Figure 9.8: (a) RMS acceleration with non-damped holder as a func-tion of the B-field at a 50 Hz resonance (two measurements to indicatereproducibility); (b) Acceleration spectrum for vertical orientation atresonant 50 Hz B-field excitation (data for nondamped holder shown).

the dish holder or coil chamber. A linear dependency with B-fieldamplitude was found with a constant of 0.56mgGr/mT at 50Hz and a

9.9. CONCLUSIONS 141

frequency sensitivity at 3.6 mT according to 1.5mgGr+0.01mgGr/Hz.For the sham state all vibrations were below the background level

of 5-13mgGr. In horizontal direction, no acceleration could be mea-sured on the damped dish holder with magnetic field applied. Theother dish holder showed a 56.4 Hz resonance with a peak of 53mgGr

at 3.6 mT. The maximum vertical vibration was 8mgGr at 69.7 Hzfor the damped and 92mgGr at 25.4 Hz for the non-damped holder(background level of 5mgGr and 7mgGr, respectively).

The vibration as function of the magnetic field strength was deter-mined for a setup with a resonance at 50 Hz and a 50 Hz sinusoidalB-field excitation, see Figure 9.8a for the non-damped dish holder.The behaviour is parabolic as expected since the Lorentz force is pro-portional to both, current and B-field (resulting in an accelerationproportional to B2). The repeated measurement 2 shows, that thereproducibility is limited due to the large sensitivity of the resonanceto sligth mechanical changes. The spectral distribution is shown inFigure 9.8b.

9.9 Conclusions

An exposure system that allows flexible signal and intermittent ex-posure schemes has been developed and characterized. It is easy tohandle due to automated software control. Coil currents, chambertemperatures and fan speed are continuously monitored and allowthe experimental history to be traced with 10 s resolution. B-fieldand E-field distributions were characterized. The B-field shielding ofthe 4-coil configuration considerably enhances the uniformity of thefield distribution, and a highly efficient E-field shielding inhibits strongparasitic electric fields generated by the coils. Temperature differencesbetween exposed and sham-exposed cells are kept below 0.1C. Thevibrational load on the exposed Petri dishes is sensitive to mechanicalresonances; however a mechanically isolated and elastically dampeddish holder limits this effect to less than 10mgGr, which is no morethan twice the background vibration of the sham setup.

in vitro exposure setup for elf magnetic fields · pliﬁers (tda 7294) working in parallel. a...

Documents