(
Biopotential amplifier for simultaneous operation with biomagnetic instruments J. V i r t a n e n 1
L Parkkonen ~ R. J, I l m o n i e m i z E. P e k k o n e n 1 R. N&~it~nen 1 aCognitive Brain Research Unit, Department of Psychology, PO Box 1t, 00014 University of Helalnki, Finland ~Neuromag Ltd, Helsinki P O Box 68, 00511 Helsinki, Finland nBIoMag Laboratory, Medical Engineering Centre, Helsinki University Central Hospital, 00290 Helsinki, Finland
J\
AbstractwA multichannel biopotential amplifier for simultaneous use with biomagnetic measurements in a magnetically shielded room is designed and evaluated. Particular care is taken to make the amplifier electromagnetically compatible with the biomagnetic instruments over the whole frequency spectrum, from DC to RF. The electromagnetically quiet environment allows the use of high electrode impedances; the preamplifier has been designed accordingly. Special care is taken to analyse the coupling mechanisms of mains interference to the amplifier. Over 170 simultaneous electric and magnetic recordings have been performed using the system in a hospital environment. Keywords--Bio-electric recordings, Interference, Noise, Amplifier, EEG, MEG, ECG, MCG Mad. Biol. Eng. Comput., 1997, 35, 402-408
1 Introduction
ELECTRO-ENCEPHALOGRAPHY(EEG) and magneto-encephatography (MEG) provide complementary information about sources in the brain (WIKSWO et al., 1993; NAXT~,NEN et al., 1994). Likewise, electrocardiography (ECG) and magnetocardiography (MCG) provide complementary information about the operation of the heart (MAKIJARW eta/., 1993). Here, a 64channel easily expandable biopotential recording system is described. The system is designed to be used simultaneously with the MEG and MCG in the magnetically shielded room of the BioMag Laboratory of the Helsinki University Central Hospital, where a 122-charmel MEG system~: and a 68channel MCG system have recently been installed. In an MEG measurement, the helmet-shaped dewar containing the 122 gradiometers and superconducting quantum interference device (SQUID) sensors is placed above the head of the patient. In an MCG measurement, a flat dewar containing the 68 sensors is lowered over the chest of the patient lying on a bed. The pickup coils of the biomagnetic instruments are typically first-order gradiometers that detect only local spatial variations in the magnetic field, the sensitivity to distant sources being thus reduced ~ ' E N et aL, 1993). Typically, the peak values of the measured flux density gradients are 100 ft era-t for evoked magnetic fields, 500 ft o m - t for spontaneous MEG, and 10 pT cm-1 for MCG. The frequency range is typically from direct current (DC) to I kHz.
S Biomagnetic measurements are sensitive to a wide spectrum of electromagnetic interference. The instruments are usually installed in shielded rooms (HAMALs et al., 1993), where typically no other electronic equipment is used. The shielded room in the Helsinki University Hospital* attenuates electromagnetic fields by 40 dB at DC and by 100 dB at frequencies above 100 Hz. When placed inside the shielded room, commercially available EEG and ECG recording systems interfere with the .MEG and MCG instruments. The most important noise sources are power supplies and digital electronic circuits. In some systemst only the isolated patient interface can be taken into the shielded room. Patient safety regulations (OLSON, 1978) make it impossible to filter the isolated signals against ground; thus significant radio-frequency (R_F) power may be transmitted into the room via these lines. Here, an amplifier system is described that is electromagnetically compatible with the MEG/MCG, maintains the integrity of the shielded room and meets the safety regulations. 2 Compatibility w i t h b i o m a g n e t i c m e a s u r e m e n t s
Alternating currents in electronic instruments inside the shielded room may cause artefacts in biomagnetic measurements. The magnetic flux density B(r), at point r, generated by a small current loop at the origin, with its axis along the z-axis, is given by the equation ~xa 2
B(r) ~ --~-fi-(ur2cos0 + u0 sin0) First received 20 September 1995 and in final form 30 May 1996 Correspondence should be addressed to J. Virtanen. email:
[email protected]
r IFMBE:1997
402
a << r
(1)
*Euroshield Ltd, Eura, Finland l"Synamps, Neuroscan, Inc., Herndon, VA, USA Spectrum 32, Cadwell Laboratories, Kennewick, WA, USA :~Neuromag Ltd., Helsinki, Finland Medical & Biological Engineering & Computing
July 1997
w h e r e a is the radius of the loop, I is the current in the loop, 8 is the angle between r and the z.axis, and u, and u0 are the
radial and tangential unit vectors. The rate of change in the radial direction of the tangential component of the flux density can be calculated from eq. I dB0 dr-
3~Ia ~ 4ra sin0
(2)
Likewise, rate of change of the radial field component is dB, d-T =
Uo/'~ 9 -~-sm(1/r)
(3)
where I is the distance measured along a tangential path at distance r from the origin_ If a = 5 ram, r = l m and I = I 0 mA, the maximum flux density gradients to both directions are about 2 ft cm - I , which shows that currents driving optocoupiers, for example, may interfere with the magnetic measurements. Even when the interference is well outside the measurement band, problems can occur. For example, RF fields can decrease the SQUTD gain and increase the noise level (I-I~.M~LMNEN et al., 1993). The maximum acceptable level of high-frequency interference is extremely difficult to estimate, because it depends on unspecified component properties and on stray impedances between different parts of the magnetometer. For example, no digital data transfer should be allowed inside the shielded room during data acquisition. Magnetically shielded rooms are capable of providing shielding factors better than 100dB at radio frequencies (HAMALMNEN et al., 1993). To maintain the integrity of the shielding, the signal feedthrough filters should have similar attenuation factors. Even tiny ferromagnetic objects attached to the patient disturb the biomagnetic measurements. This is especially true for EEG electrodes, which, during the measurement, are at a distance of about 2 era from the magnetometer sensors. All components attached to the patient must be tested for remanant magnetism. As a summary, the following features are required for the biopotential amplifier not to interfere with the magnetic measurements. 9 no digital signal processing or data transmission inside the shielded room during data collection 9 all the signal lines through the wall of the shielded room must include low-pass filtering, attenuation better than 100 dB at RF frequencies 9 no significant AC loops inside the magnetically shielded room 9 no ferromagnetic components that move with the patient or vibrate as the patient breathes 9 no large ferromagnetic components in those parts of the amplifier that can be moved inside the magnetically shielded room.
are less stable in long-term recordings (CoOPEa et al., 1969), they are perfectly suitable for event-related potential (ERP) recordings and other applications where long-term DC stability is not required. See Fig. 1 for a photograph of a typical EEG/MEG measurement set-up. All the materials, including solder, wires, insulators etc., were tested for their magnetic properties by moving the material samples under the magnetometer array. The material was accepted if it was not possible to distinguish the signals induced by the movements from the background noise of the magnetometer at the frequency range from 0.03 to 30 Hz.
4 The amplifier The amplifier system is designed to be easily expandable up to 128 channels, which has resulted in quite a large overall size. The amplifier system is divided, so that the isolation amplifier, filters and power supplies are placed inside the RFshielded cabinet but outside the magnetically shielded room, to minimise the coupling of low-frequency magnetic fields to the magnetometer. The pre-amplifier unit is placed inside the shielded room to minimise the cable length between the electrodes and the pre-amplifier, which is necessary to keep the coupling between adjacent channels within reasonable limits. See Fig. 2 for a block diagram of the amplifier. As it is not possible completely to avoid ferromagnetic components in the pre-amplifier (component leads made of ferromagnetic alloys, nickel layers in gold-plated connectors etc.), the amplifier must be located about I m away from the magnetometer. The electrode leads are connected to a passive, non-magnetic connector box, which is connected to the preamplifier unit by a 1 m long shielded cable. The power supplies for both the isolated and non-isolated sections are implemented using line transformers and linear regulators. A specially designed transformer is used for the isolated section. (See Section 5 for more details.) The I3(2 component is removed from the signal between the first and second amplifier stages with a first-order high-pass filter. The high-pass comer frequency is 0.03 Hz for both electric and magnetic signals. Both electric and magnetic signals are digitised by the same high speed AD-converters,
3 Electrodes To avoid artefacts in magnetic recordings, the electrodes can contain no ferromagnetic materials. This applies, not only to the metal parts, but also to the colour pigment in the plastic insulator over the electrode leads, for example. To facilitate the preparation of the subject for multi-channel EEG recording, a cap-like fixture for the electrodes was made of elastic fabric. Custom-made low-profile electrodes of gold-plated silver were developed to fit the cap. Gold plating was chosen because it is easier to maintain than the commonly used silver/silver-chloride coating. Although gold electrodes
Medical & Biological Engineering & Computing
July 1997
Fig. 1
Typical measurement situation. The magnetometer is slightly raised from the measurement position to show the electrode hat
403
!_
=
I
/
-
I=
~-~s RF shielded cabinet
Fig. 2 Block diagram of biopotential amplifier HP3852A ~ with 16-bit resolution and 200 kHz aggregate sampling frequency. The maximum sampling frequency with 122 -MEG and 64 EEG channels is currently 600 Hz. The gain of the amplifier chain can be selected to be either 20 000 (for EEG) or 1000 (for ECG). The system includes signal generators for electrode impedance testing and calibration. The filters and the pre-amplifiers have several softwareselectable options. The filters are controlled by a dedicated computer via optical fibres, and there is no digital data transmission during data collection.
DG441 9 102 u
osce
*""'q~, ~"~ R101
R105 1K
DG441
C101 ,,2p
di(femnti=l
**Linear Technology Corp., Milpitas, CA, USA ffAnalogue Devices Inc., Norwood, MA, USA ~Hewlett-Packard, Santa Clara, CA, USA
404
311~ I~
;u,02v~
~R104 x~. t0M
AI2620
P~, P,~
4.1 Preamplifier To take full advantage of the fixed number of isolation and filter channels, the pre-amplifier unit has several softwarecontrolled options. In the present version of the amplifier, there are 64 single-ended input pre-amplifiers, eight differential input pre-amplifiers and three dedicated limb electrode amplifiers for ECG recordings. 64 of the pre-amplifier outputs are connected to the isolation amplifier and to the data acquisition unit at time with analogue switches ('pre-amplifier selector' in Fig. 2). The augmented leads aVR, aVL and aVF are calculated in the pre-amplifier hardware. The simplified circuit diagrams of the differential and single-ended pre-amplifier stages are shown in Fig. 3. All the single-ended chan~els share a common buffered reference (U301, LTI012) ~ In the first stage, the signal is amplified against the reference signal (mr') with gain of about 50, using the low-noise operational, amplifier LT1012 (U201). The signal is AC-coupled to the second, adjustable differential gain stage, which is implemented using the instrumentation amplifier AD620 (U202)**. Except for the first stage, the differential stage is identical to the single-ended one. Although a slightly better noise performance would have been achieved by building the first differential stage from two LTI012 components, AD620 was chosen because of the limited printed circuit board space. To improve the coupling between the isolated ground and the patient, a driven-right-leg scheme is used (WINTER and
R107
~1 ,,
R
U2~
P301
~
R206 2K64
I
l~ig. 3 Circuit diagrams of di.fferential and singte-ended pre-amplitiers (input protects'on circuits and some other details are omitted for clarity)
WEBSTER, 1983). In EEG recordings, a single electrode is used as a reference, whereas, in ECG, the average of the three buffered limb electrodes acts as a reference (not shown in Fig. 3). No shielding or guarding (METTrNGVAN RI3N et al., 1991) around the electrode leads is used as the mains interference proved to be no problem in the well-shielded environment. For the same reason, it is not necessary to increase the I0 M ~ input impedance. (See section 5 for a more detailed analysis. For electrode impedance testing, 10 Hz test signals (osc0, oscb osr can be connected to the imput terminals of the amplifier with laigh-impedance analogue switches (15102, U103, U203). The signals are coupled through 1 MQ resistors (RI01, R102, R201) to feed constant current to the electrodes. The polarity of the osc signal ca~a be independently selected Medical & Biological Engineering & Computing
July 1997
for each channel. If the impedances of all the single-ended channels are measured simultaneously and the polarity of every other channel is inverted, the net current fed to the patient is close to zero, and the driven-fight-leg circuit can easily sink the residual current. Likewise, the inputs of the differential channels are driven with test signals of opposite polarities, OSCoand osc t- No means are provided for testing the reference and neutral electrode.s.
Table t Ground capacitances of patient and pre-amplifiex (Fig. 5) meast~ed with an impedance meter Escort ELC-131D ~
Component Patient C=~y Patient CMeO Patient CMcG Pre-amplifier C=r.p
Both the isolation amplifier and the filter unit have 64 identical channels. The analogue signals are transmitted from the isolated side to the non-isolated side by optoeouplers that operate in analogue mode. The schematic diagram of one isolation amplifier is shown in Fig. 4. Two optocouplers are used for each channet, so that the non-linearities of the two components compensate for each other. The circuit is thoroughly described by TEDER (1994). The capacitance between the isolated and non-isolated sides was measured to be less than 1 pF per channel, and so the total capacitance remains tolerable even with 128 channels. In the filter bank, there are two second-order digitally controlled state-variable filters for each channel. Both the comer frequency (30-1000 Hz) and the type of filter (Butterworth, Bessel, Chebyshev) can be selected by software. The filters are identical to the ones used for magnetic data (AHONEN et al., 1993). There are identical feedthrough filters for all signal and neutral wires. One filter consists of three pi-type LC-filters in series. The middle one of the pi-filters is implemented with a Murata Erie 1207-0250 filter component*, which alone has an attenuation of better than 85 dB at 100 kHz. The low-pass comer frequency of the filter is 60 kHz, and the theoretical slope is 60 dB per decade.
5 Mains interference
The coupling mechanisms of mains interference to biopotential amplifiers in normal conditions have been thoroughly analysed by ME't'rINGVAN RIJN et al. (1990), PALLAS-ARENY (1988), HUHTAand WEBSTER(1973) and WOOD et al. (1995). The capacitively coupled currents to the body and to the electrode leads are considered to be most important sources of interference. These currents can be reduced by minimising the ground capacitance of the isolated amplifier. It is recommended by MET'rINGVANRIJN et al. (1991) that the amplifier is made small and battery-powered. Table I lists the measured ground capacitances of the patient and the isolated pro-amplifier. According to PALLASARENY (1988), the ground capacitance Cp,n between patient and ground is about I00 pF, which is in good agreement with the data in Table 1. The total capacitance from patient to ground is exceptionally large (280--400 pF) because of the
~,~,~
s,..
~"
~"-~I"-
1
'
RIO6 1K
33tl
1K69 1K69
P102
1K69 1K69
II I"':,'.~x= l v.~w it ~
80 pF 200 pF 320 pF 800 pF
Transformer type Standard Custom made Custom made Custom made
Explanation
V=,~
(pre-amplifieralone) 15 Vma= (pre-amplifieralone) 0.10 Vrms (pre-amplifierand patient sitting) 0.09 V.,= (pre-amplifier and patient under MEG) 0.07Vrm=
vicinity of grounded metallic parts in the biomagnetic instruments. Fig. 5 illustrates the situation. The problem caused by the large ground capacitance Cp,t is well demonstrated by the measurements that were made using an ordinary 50 W medical grade isolation transformer~'~;. The capacitance C~n,f between primary and secondary windings was 60 pF. The ground capacitance of the amplifier Ca,,p was 800 pF, and the potential of the amplifier V=,,~ was 15 Vm~s (Table 2). The equivalent source voltage Vp,~,,~which couples from the transformer primary to the secondary, can be calculated from these data C~,~,i
V,~p = 2 15 Vrms
(4)
The ground capacitance of the patient Cpat = C~sroy+ CMeG was 280 pF in the MEG measurement set-up (Table I). The potential difference Vet over the neutral electrode with electrode impedance R, = I 0 k.Q can be calculated, if it is assumed that Ctror~: << C~t. C,r << C , ~ and 1/coCp,t > R,
G=
V~ ~ R~ x C:at + C=,,p x (c~C~'~'~f)-l = 10 mVpp
(5)
The amount of potential difference Vet transformed to differential input voltage of the amplifier depends on the true input impedance of the amplifier and on the imbalance between the electrode impedances (I-IUHTA and WEBSTER, 1973). One metre of shielded cable with 50 pF m-~ capacitance gives an impedance o f ~ 6 4 Mf~ to amplifier ground in parallel with the 10 Mxq input resistance at 50 Hz. Thus the total input impedance I Zo, I = 8 M~. The current let through the electrodes generates a differential input voltage Vt, for the amplifier, if the impedance difference in electrodes ARel = 10 k.Q V~, = I a x AR,z ~ ~
o,L :,C mo~ . ~
(stray capacitance, sitting) (tkrough =MEGinstrument) (through MCG instrument) (total)
Table 2 Potential of isolated amplifier at 50 Hz measured with HP54603A oscilloscope and 10 MI'Zprobe HPlOO71A
CtOl
33o ~,
Ground capacitance
*Escort Instrmnents Corp., Taipei, Taiwan
4.2 Isolation amplifier and fllters
v,Io~3
Coupling mechanism
x ARa = 12.5 mVpv
(6)
Eq. 5 suggests that the interference can be attenuated either by increasing the ground capacitance of the amplifier Ca,v, or by reducing the coupling between the primary and secondary windings of the transformer. Increasing C,,~, is not possible because of safety regulations that indirectly require Co,,p to be less than 800 pF, by stating the maximum allowed leakage
u~uu
*Murata Electronics N. America Ltd,, Smyrna, Georgia Fig. 4
Circuit diagram of isolation amplifier
Medical & Biological Engineering & Computing
~r
July 1997
Ltd, Helsinki, Finland
4O5
,r MEG
t---
,.t..-
"r ~'
-r'-
~
~
-It'-"
,t,-,
,
Fig. 5 Equivalent circuit illustrating coupling mechanisms o f mains interference to patient and isolated amplifier
- ~ current to ground to be 50 btA,-,~, if the 220 V ~ mains voltage is applied to the patient (OLSON, 1978). Because of the sensitivity of the system to mains interference coupling to the amplifier, a special toroidial transformer was designed. A grounded metal sheet was placed between the transformer core and the windings. The primary and secondary did not overlap, but were wound in different segments of the core to minimise the mutual capacitance. With this transformer, the potential of V~,~pwas only 0.1 V ~ . and Vin was well below microvolt level. V,,,p was also measured with a patient connected to the amplifier, when he was just sitting and when he was in MEG measurement position (Table 2). The attenuation of Vamp in both cases was proportional to the capacitance added (Table 1), which indicates that there are no other sources of capacitively coupled line interference inside the shielded room. The custom-made transformer proved to be as good as battery power, in the sense that it did not couple significant mains interference to the amplifier. In some other operating environments, the ground capacitance of the transformer might appear to be significant. In the absence of other sources of capacitively coupled line interference inside the shielded room, the relatively large ground capacitance of the amplifier appears to be harmless. It is even beneficial in the sense that it attenuates interference coupling directly to the amplifier. For example, it may well be possible that, if C,,,~p were significantly smaller, battery powering would perform better than the custom transformer.
6 Discussion A 64-channel biopotential amplifier for simultaneous use with biomagnetic measurements was designed and evaluated. The main motive for designing and building the amplifier especially for this application was to develop a system that is fully compatible with the Neuromag MEG/MCG systems. In addition, the electric data are filtered and digitised in exactly the same way as the magnetic data. Also, the amplifier can be easily and economically expanded up to 128 channels. A 32-channel version of the amplifier was installed into a hospital environment in March 1995, and nine months later it was expanded to 64 channels. As of May 1996, over 50 simultaneous multi-channel EEG/MEG and 120 ECG/MCG recordings have been made on volunteers and patients. An example of auditory evoked responses is presented in Fig. 6. EEG and ECG instruments are often operated in RFshielded rooms in research laboratories. The analysis of the coupling mechanisms of mains interference suggests that an ordinary R.F-shielded room could provide significant shielding 406
r--
r-
r-- t
EEG
Zc- f f . , . t c ~ , t r - - t r
9" ~ . e . . . e _ . e .
"t~
Fig. 6 Event-related magnetic and electric responses for 50ms, 700Hz tone pips given to the left ear, with 2500ms stimulus onset interval, 100 averages, filtered from 1 to 30 Hz. The horizontal scale bar is 100 ms, and the vertical scale bar is 5 I.tVfor electric data and l OOf T cm- t for magnetic data
against line-voltage interference. For best results, no mainsoperated equipment, including room lights, are allowed inside the room. In the absence of mains noise coupling to the patient, a relatively large capacitance from the isolated amplifier to ground appears to be advantageous in attenuating interference coupling directly to the amplifier. The maximum value of the capacitance is limited by safety regulations. It has been shown that completely artefact-free electric and magnetic measurements can be made simultaneously. Also, it has been shown that electrode impedances as high as 100 k.Q can be used in EEG experiments, provided that the preamplifier's noise performance is optimised for the high impedance level and that the level of environmental interference is attenuated with a shielded room.
Acknowledgements--The authors thank Jukka Honkanen for this help
with the regulatory issues, Dr Wolfang Teder for the isolationamplifier design, and Minna Huotilainen for her comments on the manuscript. This work was supported by the Academy of Finland. Medical & Biological Engineering & Computing
July 1997
References
AHONE~. A. 1., H/~M/~t.~tNEN,M. S., KAJOLAM. J., K,.ba~Irt,.,~, 3. E., LAINE, P. P., LOUNASMAA,O. V., PARKKONEN,L. T., SIMOLA, J. T., and TESCHE,C. D (1993): '122-channel SQUID instrument for investigating the magnetic signals from the human brain', Pysica Scr/pta, T49, pp. 198--205 COOPgR, tL, OSSELTON, J. W., and SHAW, J. C. (1969): 'EEO technology" (Butterworth, London) pp. 25-28. I-IXM~LXINEN,M., HARI, R., |LMO,';IENfl,IL J., KNUUTILA,J., and LOONASMAA, O. (1993): 'Magneto~ucephalography - theory, insmam~ntation, and applications to non---invasive studies of the working human brain', Rev. Mod. Phys., 65, pp. 413--497 HurrrA, J. C., and WEBSTER, J. G. (1973): '60-Hz interference in electro-cardiography', IEEE Trans., BME-20, pp. 91-101 METTINOVANRL~ A. C., PEPER, A., and GR1MBERGEN,C. A. (1990): 'High-qualtiy recording of bioeleetric events. Part 1 Interference reduction, theory and practice', lVled. Biol. Eng. Comput., 28, laP. 389-397 METTING VAN RI.IN, A. C., PEFER, A., and GRIM.BERGEN,C. A. (1991): 'The isolation mode rejection in bioelectric amplifiers', IEEE Trans., BME-38, pp. 1154-1157 MAKIJ.&RVI,M., MONTONEN,J., TOPr L., SILTANEN,P., NIEMINEN, M., LEINIO,M., and KATILA,T. (1993): ' Identification of patients with ventricular tachycardia after myocardial infaetion by high-resolution magnetocardiography and electrocardiography', 3.. ElectrocardioL 26, pp. 117-124. NA~,T~L~'N, tLM., ILMONIEMI,R.J., and ALHO, IC (1994: 'Magnetoencephalography in studies of human cognitive brain function', Trends N~rosci., 17, pp. 389-395 OLSO~, W. H. (1978): "Electrical safety' in WEBSTER, J. G. (Ed.): 'Medical instrumentation: application and design' (Houghton Mifflin Co., Boston) pp. 667-706 PAt.LAS-AREI,nf, R. (1988): 'Interference-rejection characteristics of biopotential amplifiers: a comparative analysis', IEEE Trans., BME-35, pp. 953--959 TEDER, W. (1994): 'An analog optical link in surface mount technology for multichannel biomedical data acquisition', Behav. Res. Methods lnstrum. Comput., 26, pp. 416--420 WIKSWOJ. P., GEVINSA.. and WILLIAMSONS. J. (1993): 'The future of the MEG and EEG', Electroenceph. Clin. Neurophysiol., 87, pp. 1-9 WINTER, B. B., and WEBSTER, J. G. (1983): 'Driven-right-leg circuit design', IEEE Trans., BME-30, pp. 62--66 WOOD, D. E., EWINS, D. J., and BELACHANDRAN, W. (1995): 'Comparative analysis of power-line interference between twoand three-electrode biopotential amplifiers', Med. BioL Eng. Comput., 33, pp. 63-68 Appendix
C M R R u = . ~ . Even if Gt~o~ and Gv~ol significantly differ from each other, CMRR~ is always better than the l o w ~ of the open-loop gains. For LTI012, the open,loop gain is specified to be better than 80 dB at 50 Hz. Note that the tolerances of the resistors R2o3 and R2o, do not affect CMRR~. Ftmhermore, the common.mode rejection ratios of U~ol and U~ot affect the common-mode rejection of the whole circuit (PALLAS-ARENY, 1988). AS the CMRR of the LTI012 operational amplifier is specified to be as large as 130 dB at 50 Hz, this effect can be ignored. The CMRR at 50 Hz was measured by connecting a sinewave generator between terminals Pgadl and Pga,t/2 and setting R s = 0 (Fig. 2). For all 64 single-ended c "harmels using LTt012 operational amplifiers, the CMP, R was better than 86 dB and, for all 8 differential channels, it was better than 96 clB. The specified CMRR for AD620 is better than 110 dB at gains from 10 to 100. The common-mode rejection of the network formed by the electrode impedances and the amplifiers' input impedances is denoted CMRR,I. Assuming the same imbalance in the dect-rode impedances ARet = 10 k ~ and the same input impedance IZ~.I = l0 M r as previously, the CMRR~ can be estimated from eq. 6 CMRR't ~ ~ a = ~
Pre-amplifer noise The pre-amplifier components were chosen to operate well from high input impedances. For example, such widely used components as OP27 and AMP01* were rejected because of high current noise, and INAI 11t was rejected because of high voltage noise at low frequencies. Fig. 7 shows the calculated total root-mean-square (RMS) noise of the pre-amplifiers as a function of source impedance with two different bandwidths. The curves were obtained by combining the data from voltage- and current-noise curves from the component manufacturers' data books. As the fre~.s
The common-mode rejection of the differential input stage is expected to be about equal to that specified for the instrumentation amplifier. The common-mode rejection ratio (CMRR) of the single-ended first stage, built of two operational amplifiers U2ol and U3o~ and resistors R2o3 and R2o4, can be derived by writing the nodal equations for the circuit and solving for the CMRR~, (Fig. 2)
]
>= ~.0. . . . . . . . . . .
CMRRse = lff--~dl = l~ (l + l/Gu3~ Gu201G3ol "~ - ~ --~'--
---'G~3o~ + )
0,1
('#U301
USA
& Biological Engineering
& Computing
July 1997
i~ il i lt~,~
i i
i ":
!
,
i " [tllsl ,- : : : = = ~ :
i
l
I
i
l/,,Ll~i
]
:
'.i ii
! i ili~l] :
am , v'~u201 >> 1, Gu3oi >> 1
where fl = Rzo4](R203 + R2o4), Aa = lift, Ac,~ = VoqVc,,,, and Gv2o~ and Gt~o~ are the open-loop gains of the operational amplifiers U~ot and U3oi (Fig. 2). If Gmol = Gv3ob then
Medical
0.5'
~
"~ ".i"~, 't~ . . . . . . . . . . . ~ " ' "
I
i
[
(7)
*Analogue Devices Inc. *Burr Brown Corp., Tucson,
t--"~f',~"~
i
t=
--
(8)
This calculation shows that, in practice, the CMRR of the system is not limited by the amplifier components. The drivenfight-leg circuit does not literally affect the CMRR of the amplifier circuit. Instead, it reduces the common-mode voltage by effectively dividing the impedance of the reference electrode with the loop gain of the driven-right-leg circuit.
Common-mode rejection
(JU201
= - 6 0 dB
~
i
i
. ' ~ " ~ ! ' " "'~'P~" . . . . . . . .
f "-'I-"~"
I
I
i
iii l l , l i
i
.1r
i iit!
il~
9 ili:Lli
i
~, i ~.~
!
:
~
i :~il
i;
i. iI li~i , I
I :
!
! ill
!
i~
' !!
ii
+
...&;,,,""F.~L,'I'I~
~ f i i ; T / . i ' f ~ i i .... r - ~ ~ i l i .......
?''":
:
ili
1
il
10
~muree~ r , e ~ ,
l
i
i
!~iii
100
tan
Fig. 7 IL~IS noise curves of the pre-amplifiers as a function of source resistance R,; (a) single-ended, 0.1-100 Hz; (b) single-ended, 1.0--100 Hz; (c) differential, 0.1-100 Itz; (d) differential, 1.0-t00 1-1,.; (e) thermal noise of the source resistors only. The curves are derived from the data in manufacturers' data sheets 407
Table 3 Measured noise from preamplifiers over frequency range from 0.03 to t00 Hz
Single-ended channel Differential channel
R,=27 k.Q
Rs= 127 I ~
0.38 # V ~ 0.37 pVr,,~
0.63/~V,,,~ 0.68/~Vnm
queney axis was extended from 1 to 0.1 Hz, the l/f noise was extrapolated from the original data. The total noise Vn as a function of the source resistance Rs is ,
,
89
Fig. 7 demonstrates the dominance of current noise at impedances over 10 kfl and the significance of the Vf noise. In ERP experiments, the inter-stimulus interval (ISI) is typically in the order of Is, This means that the data cannot include frequency components lower than 1/IS/, I Hz with 1 ISI; therefore most of the l / f noise can be filtered away. In typical ERP studies, more than 100 responses are averaged; which reduces the amplifier noise by a factor of I0. The RiMS noise at 127 k~ (0.68 #Vnm ~ 4.1/z Vp.p) is reduced to 0.41 /~Vv_P These are adequate values, considering the typical signal amplitudes of a few microvotts in ERP measly:merits (COOPER et al., 1969). These figures are comparable, for example, with the specification of 0.5 #Vrms at 0.1-i00 1-[z used by METTtNGv~'q RIJN et al., (1991).
(9) where e ~ , is the voltage noise density of the main amplifier, e ~ is the voltage noise density of the buffer amplifier (for the differential input stage e,,a~= 0), i~, and i,,a., are the current noise densities, enR~is the thermal noise voltage density of the input resistor, and j~ and 3q are the limits of the frequency band. To verify the calculation, the pre-amplifier noise was measured using two different values of input resistance R, and with the terminals Pgna/l and Pg,~/2 shorted (Fig. 2). One eharmel was precisely measured, and the others were inspected qualitatively. The results in Table 3 show that the measured amplifier components were slightly better than specified.
408
Author's biography Juha Virtanen was born in 1961. After receiving the M.Sc. degree in 1986, from the Helsinki University of Technology, he joined the Research and Development department of K.one Corp. After that he worked for Picker Nordstar, Inc. as a project manager in MRI scanner development project. In 1990, he received his Lic.Tech. degree from the Helsinki University of Technology. Since 1994, he has worked with the Cognitive Brain Research Unit at the University of Helsinki. Currently, his main interest is to advance the simultaneous use of different functional brain imaging modalities.
Medical & Biological Engineering & Computing
July 1997