1.2 V sub-nanoampere A=D converter
G. Rachmuth, Y.-S. Yang and C.-S. Poon
CMOS-based integrated biochemical sensors generate photo currents at sub-nanoampere (nA) levels, which present a challenge for digital data acquisition. Proposed is a MOS current-mode analogue-to-digital converter with sub-nA sensitivity. Its inherent low-voltage low-power and small-size capabilities are ideal for portable chemi- or biosensor applications.
Introduction: High-performance analogue-to-digital converters (ADCs) are generally optimised for conversion speed and resolution with given size and power budget [1]. In CMOS-based mobile biochemical sensor applications[2, 3], however, speed and resolution are immaterial because of the slow reaction rates ( > seconds) and inherent experimental errors (10%) typical of most biochemical reactions. Instead, ADC sensitivity, power consumption and size are of major concern [4]. To convert sub-nA-level photo currents into voltage, [3] amplified the signal using large-gain (106) current
mirrors, increasing power and area requirements and the current mirror’s susceptibility to mismatch errors for a given chip area. In
[5], increased ADC sensitivity was achieved by successive capacitive integration and voltage-to-frequency conversion at the expense of increased power consumption and long conversion time (seconds). In this Letter we propose a novel current-mode ADC (IADC) capable of digitising the photo currents in[2, 3]at a speed and resolution commen-surate with such applications. The IADC operates at a supply voltage (VDD) as low as 1.2 V, contains no capacitors or clocks, and can be directly
integrated alongside the CMOS photodiode in any fabrication process.
Fig. 1 Subthreshold IADC
a Circuit design for unit cell with VDD¼1.2 V
n cell: input node, 1; output node, 3, m cells, input node, 1; output node, 4 b n-cell and m-cell IADC architectures
IADC design: Previous IADC designs allow high-speed conversion down to the mA range by biasing the transistors in the strong-inversion (above-threshold) regime[1, 6, 7]. To achieve sub-nA sensitivity and low VDD, we used an algorithmic IADC cell architecture similar to[7]
but with a modified circuit design that is amenable to subthreshold operations (Fig. 1a). Cells n ¼ 1, 2, . . . , N are cascaded with the analogue output of one cell connected to the analogue input of the following cell (Fig. 1b). The analogue input of cell n, IIN(n), is scaled
using a 1:kncurrent mirror (Q1). The comparator (Q5–Q8) output,
DO(n), is a HI if knIIN(n) is greater than a user-defined reference
current IREF, and is an LO otherwise. To operate in the sub-nA range,
a transmission switch gated by DO(n) in [7]is omitted in order to
avoid large switching current artifacts that may disrupt the signal conversion process. As a consequence, the input current into the next
cell (IIN(n þ 1) ¼ knIIN(n) IREF) via the current subtraction
transis-tors Q2 and Q3 is no longer dependent on DO(n). This design reduces
the IADC to a flash architecture, which allows the fastest conversion speed for any given operating conditions.
Traditional flash architecture generates a thermometer code according to a voltage divider sequence. Here, the 1:knscaling of input current in
each cell provides an equivalent current divider sequence [1 (k1k2 kn)1], n ¼ 1, 2, . . . , N, for the range IREF=k1
IIN(1) IREF, where IIN(1) is the current input to the IADC, and k1is
the mirror ratio of n ¼ 1. When knIIN(n) < IREFfor cell n, the inputs to
all subsequent cells are zero so that DO(n : N) is LO. Thus the analogue
input is quantised by the largest value of n such that DO(n) is HI.
A drawback of the cell design in [7]is that in the critical region knIIN(n) ’ IREFwhere the currents in transistors Q2 and Q3 of cell n are
almost balanced, the drain current of transistor Q1 in cell n þ 1 may cause a significant error. This difficulty is circumvented in the present design by redefining the input–output relationship of each cell as follows. A set of cells m ¼ 1, 2, . . . , M are cascaded as before but with IIN(m þ 1) ¼
kmIIN(m) via transistor Q4 (Fig. 1a). This results in a current divider
sequence (k1k2 km)1for the range 0 IIN(1) IREF=kM, such that
when IIN(m) > IREF=km, DO(m:M) is HI. In this case, the analogue input
is quantised by the smallest m such that DO(m) is LO. Both the n-cell and
m-cell IADCs have a variable dynamic range that is set by IREF. IREF
values in or below the nA range bias the transistors in the subthreshold regime, allowing the conversion of sub-nA currents.
Fig. 2 Simulation results for prototype (with M ¼ N ¼ 4, IREF¼160 pA,
VDD¼1.2 V) showing input currents at which each m and n cell begins to
switch
——
corresponding theoretical valuesResults: The above IADC design was simulated on T-Spice and a prototype chip was fabricated using an AMI 1.5 mm process. For convenience the results for 4 m-cells and 4 n-cells with kn¼km¼2 are
presented here although the design can be extended to any number of cells with arbitrary current divider ratios. Simulations showed that the m-cells generally had higher sensitivity and accuracy than the n-cells (Fig. 2). Measurements of the fabricated IDAC with a Keithley 6485 picoammeter showed that the m-cells had an input current sensitivity of <100 picoampere (pA).
The IADC response bandwidth is determined by the conversion delay, which is given by the maximum rise or fall time (whichever is longer) of all cell responses when switching on or off, respectively. For each cell, this value is determined primarily by the corresponding comparator’s switching time, t / (CLVDD)=(IINIREF), where CL is
the load capacitance. Simulations showed that for VDD¼1.2 V and a
square-wave current input with amplitude IˆIN(1) ¼ 0.1 nA and
IREF¼1.6 nA, the conversion delay was < 500 ms. At IˆIN(1) ¼ 1 nA
and IREF¼16 nA, the conversion delay was < 15 ms.
In Fig. 3a, the fabricated IADC chip responded faithfully to a triangular-wave input with peak current IˆIN(1) ¼ 30 nA at a frequency
of 1 Hz. Similar results were obtained for IˆIN(1) down to 1 nA. At even
lower input currents or higher frequencies, measurements were limited by the response of the testbed, which was not designed to operate at such low current levels. Simulations showed that the IADC faithfully converted the signal with IˆIN(1) ¼ 50 pA at a frequency of 1 Hz
(Fig. 3b). In practice, the IADC can interface directly to the CMOS photodiode on chip, preserving the IADC response as predicted by simulations.
Fig. 3 Fabricated and simulated IADC response
a Response of fabricated m and n cells to 30 nA triangular-wave input, (VDD¼1.2 V)
b Simulation results of m and n cells’ response to 50 pA triangular-wave input, (VDD¼1.2 V)
Because of inevitable current–mirror mismatch, the resulting current divider ratios may differ from the designed value of 2. InFig. 3, a simple calibration procedure yielded the actual ratios of 2.44, 2.36 and 2.6 for m4=m3, m3=m2and m2=m1, respectively. Once calibrated, proper
IADC operation was achieved. In practice, mismatch errors may be further minimised by increasing transistor sizes or decreasing process dimensions for a given size.
Discussion: The salient features of this IADC are its high input sensitivity, a low VDD and a programmable dynamic range, with a
design that is simple, small, and power efficient. A conversion cell with IREF¼1 nA uses < 10 nW of static power.
The resultant IDAC conversion speed is adequate for most biosensor applications. For example, the sub-nA level currents from the HRP-luminal-H2O2system in[2, 3]can be digitised in <1 s, allowing ample
temporal resolution for the measurement of initial reaction rates that are important to enzyme kinetics.
This new IADC can be readily integrated with any portable CMOS sensor at reduced overall power, size, and cost. Its input sensitivity, speed and resolution can be further enhanced by employing sub-pA circuit[8]and low-voltage wide-input comparator[9]design techniques and with increased number of conversion cells.
Acknowledgments: The VLSI chips were fabricated with the support of the MOSIS Education Program. G. Rachmuth was a recipient of the US National Defense Science and Engineering Graduate Fellowship.
#IEE 2005 28 December 2004
Electronics Letters online no: 20058259 doi: 10.1049/el:20058259
G. Rachmuth (Division of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA)
Y.-S. Yang (Department of Biological Science and Technology, National Chiao Tung University, Hsinchu, Taiwan)
C.-S. Poon (Harvard-MIT Division of Health Sciences and Technol-ogy, MIT, Cambridge, MA 02139, USA)
E-mail: cpoon@mit.edu
G. Rachmuth: Also at Harvard-MIT Division of Health Sciences and Technology, MIT, Cambridge, MA 02139, USA
References
1 Razavi, B.: ‘Principles of data conversion system design’ (IEEE, New York, 1995)
2 Lu, U., et al.: ‘CMOS chip as luminescent sensor for biochemical reactions’, IEEE Sens. J., 2003, 3, pp. 310–316
3 Lu, U., et al.: ‘The design of a novel complementary metal oxide semiconductor detection system for biochemical luminescence’, Biosens. Bioelectron., 19, 2004, pp. 1185–1191
4 Yotter, R.A., and Wilson, D.M.: ‘A review of photodetectors for sensing light-emitting reporters in biological systems’, IEEE Sens. J., 2003, 3, pp. 288–303
5 Simpson, M., et al.: ‘An integrated CMOS microluminometer for low-level luminescence sensing in the bioluminescent bioreporter integrated circuit’, Sens. Actuators B, 2002, 72, pp. 134–140
6 Ravezzi, L., Stoppa, D., and Della Bettam, G.-F.: ‘Current-mode A=D converter’, Electron. Lett., 1998, 34, pp. 615–616
7 Nairn, D.G., and Salama, A.: ‘Current-mode algorithmic analog to digital converter’, IEEE J. Solid State Circuits, 1991, 25, pp. 997–1004 8 Linares-Barranco, B., et al.: ‘Current mode techniques for
sub-pico-ampere circuit design’, Analog Integr. Circuits Signal Process., 2004, 38, pp. 103–119
9 Hung, Y-C., and Liu, B-D.: ‘A low-voltage wide-input CMOS comparator for sensor applications using back-gate technique’, Biosens. Bioelectron., 2004, 20, pp. 53–49
