Proceedingsof the 25" Annual International Conference of the IEEE EMBS

Cancun, Mexico * September 17-21,2003

Precision Constant Current Source for Electrical Impedance Tomography
J eong Whan Lee, Tong In Oh, Sang Min Paek, Jae Sang Lee, and Eung Je Woo
College of Electronics and Information, Kyuny Hee University, Korea
Abstract- I n Electrical Impedance Tomography (EIT),
weinject currents through electrodes placed on the surface of
a subject and try to reconstruct cross-sectional resistivity
images. This requires a precision constant current source. I n
this paper, wepresent a design and performance analysis of a
constant current sonrce using an FPGA-based waveform
generator and improved Howland current pump circuit. The
designed current source was implemented and tested to verify
the total harmonic distortion of 0.0907% and amplitude
stability of 0.2361%. The output impedance of the current
source was bigger than 64 Mfl at de and smaller than 500 KR
at 50 KHz. We need to further improve the output impedance
at the frequency of 50 KHz or higher employing a better
method of stray capacitance cancellation.
Keywords- EIT, current source, waveform generator
1. INTRODUCIION
In Electrical Impedance Tomography (EIT), we use one
or multiple constant current sources to inject currents
through electrodes placed on the surface of a subject,
Measured boundary voltage data are used to reconstruct
cross-sectional resistivity or conductivity images. In most
EIT systems, the performance of the current source is one of
the major limiting factors determining its accuracy.
When multiple current sources are used, it is most
important to accurately calibrate all of them so that the sum
of all currents is always zero. This requires a very accurate
control over amplitudes and phases of multiple current
sources [ l , 21. In this paper, we propose using one current
source with a switching circuit to minimize errors caused by
multiple current sources. In order to reduce the data
acquisition time, we suggest fast multi-channel voltmeters.
The current source usually consists of a waveform
generator and voltage-to-current converter. In this paper, we
describe the design and implementation of the current
source using a digital waveform generator and improved
Howland current pump circuit with stray capacitance
cancellation capability [3].
11. METHODOLOGY
Table 1 shows the desired specifications of the current
source to be used in our EIT system. We determined the
specifications to achieve the 16-bit accuracy at least within
the current source. The output impedance of 64 Mi2 is
needed assuming the maximal load impedance of 1 KR. Fig.
1 shows the block diagram of the proposed current source
design.
TABLE I . conrtant Current Bourcc
Feature Specification
Amplitude ofcurrent -2.5 -+2.5 mb P
Frequency of Current 1 KHz - 1 MHz
Output impedance t 64 MC2
Calibration procedure Automatic
Digital wavefom data 16-bit
L1
E+
Fig.1. Block diagramofthe current source
A. Digitd Waveform G,?nerntor
In order to obtain a very low total harmonic distortion
(THD) and high stability, we designed a digital waveform
generator using an FPCiA and DAC shown in Fig. 1. This
also enables us to change the waveform and its frequency.
In the design, we used the hardware description language
(HDL) PI .
waveform
Address
Generator
Waveform
Data
Decoder
e process JDistance
Fig. 2. Block diagram of HDL code in the FPGA
0-7803-7789-3/03/$17.00 a2003 IEEE 1066
Fig. 2 shows the block diagram of the HDL code in the
FPGA. Digitized samples of a waveform are stored in a
ROM implemented within the FPGA. When it is enabled,
the stored binary value of the waveform data is sequentially
read and sent to a 16-bit DAC at 30 Msps. The current
output of the DAC is then transformed to a voltage signal
using an op amp circuit. In order to smooth the signal, we
used a passive band-pass filter with zero degree phase shift
at the frequency of the signal. We could change the
frequency of the signal by an appropriate division of the 40
MHz main clock of the FPGA. For the synchronization with
voltmeters using the phase-sensitive demodulation
technique, the FPGA provides a timing pulse indicating the
beginning of a new period of the signal.
B. AmpIitude Control
The amplitude of the signal needs to be changed so that
we can adjust the amount of injection current. This can be
done in two methods. First, the FPGA can scale the
waveform data with a shifter. Second, we added a
multiplying DAC (MDAC) to control the amplitude. When
we use a MDAC, we should be careful about additional
phase shift. Since we use one current source, this MDAC
does not have to provide four-quadrant multiplications.
C. Voltage-to-Current Converter
We used the improved Howland current pump circuit
shown in Fig. 3 to convert the voltage signal into a current
signal [2, 31. Assuming the perfect resistance matching
condition among the resistors used in the circuit, the ideal
output impedance is infinite. However, in practical
implementation, the output impedance becomes finite due to
errors in resistances.
R3 R4
R26
R2A
m
*
Fi g3 lmpmved Howland cumnt pump circuit
From the circuit shown in Fig. 3, the load current IL is
given by (1).
1 1
I -4--\
L - R 4
where V, is the voltage signal from the digital waveform
generator, VL is the voltage across the load,
When R. is finite, the load current IL varies depending on
the load since the voltage V, changes with the load. Since
this could severely deteriorate the performance of the
constant current source, we implemented a calibration
circuit.
D. Calibration Circuit
In order to maximize output impedance, we must first
match the resistors in the improved Howland current pump
circuit. Then, we should compensate the non-zero stray
capacitance at the output of the current source. Therefore,
we designed a calibration circuit using digitally
programmable potentiometers (digi-pot).
The output resistance is tuned to be a maximum using a
digi-pot at the place of R3 in Fig. 3. The stray capacitance is
eliminated with a negative capacitance circuit whose value
is controlled by another digi-pot as shown in Fig. 4.
FPGA for
Calibration
Neg. cap.
Fig. 4. Block diagram ofcalibration circuit
111. RESULTS
A. Digital Waveform Generator
Table 2 summarizes the performance of two waveform
generators. The analog waveform generator was constructed
based on the Wien-bridge oscillator circuit with an
automatic gain control [3]. From Table 2, we can see that
the digital waveform generator is about 5.2-13.7 times
more stabile. In terms of THD, the digital waveform
generator is also significantly better (about 54.5-87 times
smaller).
1067
TABLE 2. Comparison of two waveform generators
Waveform generator
FPlhlW . -
, Analog(%) Digital (%j
Stabilhy 0.84-2.2 0.1607
. .
" THD' 1.5-2.4 ' 0:0275 ' . '
',Told Harmonic Distortion
I " " " ' I
(4
3 . 5 , , , , , , ,
0 10 m . . a 40 54 Ea 10 10
.!I/ , , , , , , , I
Cmby, . .
(b)
Fig. 5. (a) Sinusaldal and (b) rectangular waveform
. .
. .
. .
. .
. ..
Fig. 6 Amplihlde linemiry ofthe variable mplihtde waveform generator
As described before, we can easily change the frequency
and amplitude of the voltage. signal fiom the digital
waveform generator. Throughout the frequency range of 1
KHz to 1 MHz, wefound the similar performance as shown
in Table 2. Fig. 5 shows typical voltage signals from the
digital waveform generator. Fig. 6 shows the linearity of the
amplitude as we controlled the amplitude.using the MDAC.
B. Voltage-to-Current Converter '
,' . . .
L
, . . .
. . , . :
In.order.to ohtain'an output resistance of 64 MQ, we
must match the resistors. in- Fig. 3 within 0.039%. Using a
10 KR digi-pot with 128 steps and a' 3 KR resistor in.
parallel, we could actjust the resistance value with a
resolution of 17.3 Q. Fig. 7(a) shows the change of the
output resistance R, as a function of digi-pot steps. In theory;'
we may get R, as high as 35 GQ. The measured value at dc
was 3.3 GO.
Fig. 7(b) shows 'the change, of the capacitance in the
negative, capacitance ciicuit as a functiodof digi-pot steps.
In'the current:design,'il,viyies in the range of
10 pF. Combining th!: effect of both Rq.and negative
capacitance circuit, we found that the implemented. current
source has an output impedance of 3.3 GQ at dc. However,
it was less than 500 KR at 50'kz.
. . , .
18.7 to -
Fig. 7. (a) Output mistance ofthe voltage-tomrent converter and @)
negative capacitanci: of the negative capacitance circuit
1068
IV. DISCUSSION AND CONCLUSION
The performance of the constant current source is one of
the major limiting factors determining the accuracy of an
EIT system. Based on our speculation that using one current
source may outperform a system using multiple current
sources, we designed and implemented the constant current
source including the FPGA-based waveform generator,
voltage-to-current converter, and automatic calibration
circuit.
We found that the digital waveform generator must be
used instead of the analog oscillator since it provides much
better stability and low total harmonic distortion. It allows
us to easily change the operating frequency and, therefore,
is also a more effective solution for an EIT system using
multiple frequencies.
In order to maximize the output impedance of the
current source, it is essential to use digi-pots or any similar
means for fine tunings. We found that it is little more
difficult to compensate the stray capacitance using a
negative capacitance circuit rather than finding the
matching condition of resistors. Without any proper
compensation of the stray capacitance, we found that it is
not possible to achieve an output impedance greater than
500 Kn at 50 KHz or higher. Considering the needs for
multiple frequency operation up to around 1 MHz at least,
an improved technique for the stray capacitance
cancellation should be studies in our future works.
ACKNOWLEDGMENT
This work was supported by the grant Rll-2002-103
from Korea Science and Engineering Foundation.
REFERENCES
[I] R. D. Cook, G. 1. Saulnicr, D. G. Gisser, 1. C. Goble, 1. C.
Newell, and D. Isaacsan, "ACT3: a high-speed, high-precision
electrical impedance tomography", IEEE Tmns. Biomed Eng.,
vol.41,no:8,pp.713-722, 1994.
[21 A. S. Ross, G. J . Saulnier, I. C. Newell, and D. Isaacson,
"Current source design for electrical impedance toimography,"
Physiol. Meas, vol. 24, pp. 509-516,2003.
(31 S. Franco, Design wifh Operafional AmpIWiers and Analog
IntegrafedCireuirs, 2"'ed., New York, McGmw-Hill, 1998.
(41 D. I. Smith, HDL Chip Design, Doone Publications, 1996.
1069