Pulse detection with light

I decided I wanted to try shining light through my earlobe and detecting my pulse optically.

The hardest part of this was coming up with a way to hold the light-emitting diode and phototransistor in alignment on opposite sides of my ear. My first thought was to make a clip out of a clothespin.  I tried putting a clothespin on my ear, but it squeezed painfully hard (the spring in the clothespin is much to stiff for this application).  I reduced the pain by filing of the bumps at the tip of the clothespin, adjusting the gap until it squeezed just hard enough to keep the clothespin on my ear.  This was still harder than I liked, but was tolerable for the experiment.  A lighter weight clip would not have to squeeze so hard. (I have an old pulse meter that uses an earclip—it is a bit uncomfortable also, though not nearly as bad as the clothespin).

Side view of the clothespin clip. Note that the bumps at the tip of the clothespin have been filed flat to reduce how hard the clothespin squeezes.

I needed to choose what wavelength of LED I would use for this project. We need a color that hemoglobin absorbs, but not too strongly (or there won’t be enough signal for the phototransistor to see). I found some absorption spectra for hemoglobin online at http://omlc.ogi.edu/spectra/hemoglobin/, which not only has pictures of the spectra, but tabulations and comparisons between different sources for the spectra.

Hemoglobin absorption spectra copied from http://omlc.ogi.edu/spectra/hemoglobin/summary.gif

The difference in absorption between hemoglobin and oxyhemoglobin around 686nm is used in pulse oximeters to determine the oxygenation of the hemoglobin.  For my purpose today, I probably want to stay away from that wavelength.  I decided to try using a WP710A10F3C infrared emitter that peaks around 940nm, since that is the maximum sensitivity of the phototransistor.  (The spec sheet for the WP3DP3BT phototransistor even says “Mechanically and spectrally matched to the infrared emitting LED lamp.”)

To line up the LED and phototransistor, I drilled a ⅛” hole, the closest drill bit I had to the 3mm diameter of the devices. The LED and phototransistor were a tiny bit loose in the holes (as expected), so I glued them on the back with “cool-melt” glue (a low-temperature hot-melt glue).

The IR emitter peeking through the hole in the clothespin.

I hooked up the IR emitter with a 180Ω series resistor to my function generator (whose output is is always ≧ GND), and the phototransistor with a 330Ω series resistor (on the emitter) to a battery, and looked at the signal on the scope. With nothing blocking the gap between the emitter and the phototransistor, this circuit worked quite well, with a voltage gain of about 3 for small signals (large signals resulted in clipping, most likely from the IR emitter having non-linear response when the input voltage is near threshold).

Clothespin on earlobe

When clipped onto my earlobe the signal from the signal generator still went through, but the gain was reduced to about 0.2 (100mV sine wave in, 20mV sine wave out, up to about 128kHHz—after that the gain decreased, probably because of the slow response of the phototransistor). I did an x-y plot on the scope and got an ellipse rather than a straight line above about 2.5 kHz, showing the phase change due to the slow response. The phase change is 90° (a circle on the x-y plot) at about 170kHz.

I could not see any fluctuations in the output signal due to my pulse, so I disconnected the function generator and hooked the IR emitter up to the 5v battery (still with the 180Ωseries resistor, to limit the current). When I turned the gain of the scope all the way up, I could see my pulse as a 1 mV change in level. This is about half the size of the signal I got from the EKG! Perhaps I should hook it up to the EKG amplifier circuit, if I can get all the DC biases worked out.  (It may be easier to use a simple op amp circuit, since this is not a differential signal.)