You are on page 1of 2

IDEAS FOR DESIGN

Circle 520 skin (fingertip, ear lobe, etc.) fur-


nished with a good blood supply, al-

Build Your Own Optical ternately expand and contract in time


with the heartbeat. An ordinary in-
frared LED/phototransistor pair can

Heart-Rate Sensor sense this rhythmic change as small


but detectable variations in skin con-
trast (Fig. 1, upper half).
W. STEPHEN WOODWARD When gently held against the skin
Venable Hall, CB3290, University of North Carolina, (too much pressure will flatten the
Chapel Hill, NC 27599-3290; e-mail: woodward@net.chem.unc.edu. surface capillaries and suppress the
pulsation effect), some of the radia-

P
lenty of noninvasive methods ex- manometer), electrically (EKG), and tion from D1 reflects back into Q1.
ist for electronically sensing the optically. One handy optical tech- Q1s photocurrent produces an ac sig-
human heartbeat. The job can be nique presented here exploits the nal across Q2 and Q3 of 500-V p-p
done acoustically (stethoscope or fact that tiny subcutaneous blood for every 1% change in skin re-
Doppler), mechanically (sphygmo- vessels (capillaries) in any patch of flectance. This logarithmic relation-

+5 V
+
100 mF
R1 6V
Q2 300
200 C3
1 mF
Heartbeat
Q3 sensor
Q4
C5
1 mF
R2 C6 0.001 mF
1M

Q5
2 A1
Retrosensor V2
V1 3 + 11
1
3.9k
6 4
A2
Q1 Honeywell Digital heartbeat output
HLC1395-001 5 + 7 (TTL/CMOS compatible)
or similar
1M
+5 V 680k
F1 D1 S2 10 +5 V 11
R3 1M* S3
Finger 1 12
2k 13 Analog heart-rate output
2 (1 V = 100 bpm)
cal. 16
Q6 15 14 C1
300 30k 1 mF
3.40k* (see text)

A1,,A4 = LMC6484 V4
Q7 C4 1 mF
S1,S2,S3 = HC4053 549k* 13
562*
Q2,,Q5 = 2N3904 HC4053 14
Q6,Q7,Q8 = 2N3906 C2
S1 9
* = 1% 5 1 mF 12 + LMC6484
3 A4
9
ELECTRONIC DESIGN / DECEMBER 15, 1997

6+7 8 8 V3
4 10 + A3 Fast-settling
frequency-to-voltage
100 + converter
100 mF
6V
+ C7 Q8
15 mF
6V

1. The upper half of this optical heart-rate sensor contains an infrared LED /phototransistor that senses rhythmic change as small variations in skin
contrast. The lower half constitutes a zero-ripple frequency-to-voltage converter thats optimized for human pulse-rate measurement.

104
IDEAS FOR DESIGN

resulting transfer of charge into C2


Period to frequency approximation
4V causes A3s output to slew positive un-
A B til clamped by Q6. Adjusting R3 trims
the clamp voltage and thereby sets
full-scale calibration for the circuit.
C2 immediately begins to discharge
back toward zero, but it doesnt do so
+5% (V3 = 10.5 mV/bpm) linearly. Instead, R6, R7, and Q7 syn-
Volts

C thesize a composite exponential curve


2V V3 = 10 mV/bpm = 0% error
(200 bpm) V3 (Fig. 2) that, from 285 (210 bpm) to
5% (V3 = 9.5 mV/bpm) 1250 ms (48 bpm), is a good approxi-
mation (within 5%) of the reciprocal
1V relationship between pulse period and
(100 bpm) pulse rate. Thus, each time the digital
0.5 V signal from A2 returns high, V3 will
(50 bpm) % error
equal the reciprocal of the time
0V elapsed since the previous transition
0 0.30 0.60 1.2 and, therefore, the actual instanta-
Inter-heartbeat period (sec)
neous heart rate. S2, S3, and C1 then
Curve A: C2 discharge curve = composite exponential = V3 Curve C: V3/exact = error curve of approximation
Curve B: Exact reciprocal 10 mV/bpm curve transfer V3 to sample/hold A4 for con-
tinuous output as V4 = 10 mV/bpm. If
2. C2s discharge is nonlinear. R6, R7, and Q7 synthesize a composite exponential curve V3 that C1 = C4, only one charge transfer is
approximates the reciprocal relationship between pulse period and pulse rate. needed for instant convergence.
However, normal hearts have sig-
ship is constant over many orders of need an unpleasantly long ( >10 sec- nificant beat-to-beat aperiodicity that
magnitude of photocurrent. Conse- ond) output settling time. Some con- sometimes causes V3 to jump around
quently, reliable circuit operation is verters avoid this limitation but tend quite a bit. If desired, choose C1 < C4
possible despite wide variations in to be complex (ELECTRONIC DESIGN, to provide a degree of signal averag-
skin contrast and light level. A1 and Feb. 21, 1994, p. 115). ing and thus smooth this effect out. Q8
the surrounding discrete components The relatively simple instant-set- provides Vbe and temperature com-
comprise a high-gain adaptive filter tling FVC in Figure 1 employs a pe- pensation for the Q6 and Q7 break-
that rejects ambient optical and elec- riod-to-rate approximation trick that point voltages. Thus, these transistors
trical noise (mostly 60-Hz pickup) and works well in this application. To un- should be thermally coupled for good
presents a cleaned-up signal to com- derstand the idea, consider S1, which tracking. Notice that with an appro-
parator A2 so that it can extract a is arranged to alternately switch C7 priate change of time constants, this
digital pulse-rate signal. between ground and A3s summing circuit also has general potential as an
A2s TTL/CMOS-compatible out- point. When a rising edge from A2 at optical tachometer for other difficult
put is suitable for direct input to a dig- S1 pin 9 connects C7 to A3 and C2, the low-contrast applications.
ital period-measurement circuit, and
for such applications, thats all thats
needed. But for some simple heart-
rate display situations, an analog rep-
resentation of pulse rate is convenient.
So the lower half of Figure 1 illus-
trates an unusual, zero-ripple, fre-
quency-to-voltage converter (FVC)
uniquely optimized for human pulse-
rate measurement.
Most FVCs are characterized by an
unavoidable trade-off between re-
ELECTRONIC DESIGN / DECEMBER 15, 1997

sponse time and output ripple. Usu-


ally, in order to have an acceptable
output ripple of the order of a few per-
cent, the settling time of the converter
must be at least ten periods of the low-
est expected input frequency. Normal
human hearts usually beat at rates
within the 4:1 range of 50 to 200 beats
per minute (bpm) 0.83 Hz to 3.3 Hz.
A conventional frequency-to-voltage
converter (FVC) would therefore

106

You might also like