Opacity Meter
This device is for monitoring the rate of flow of powder used in a rapid prototyping machine.
The sensor and LED are mounted about 2.5 inches apart. The sensor and LED have working apertures of about a quarter inch. The powder stream is one or two millimeters across. The first major problem is that only a small fraction of the sensor aperture is actually used so the maximum signal is only a small fraction of the zero opacity signal. Also, the actual opacity of the powder stream is not great. In addition to an inappropriate sensor, there is also the constraints that only a single (inappropriate) five volt power supply is allowed and that there is almost no money available.
Since a great deal of stability and insensitivity to optical noise are necessary, AC processing is obviously needed. The red LED is driven with a one kilohertz square wave and the resulting photocurrent is amplified and passed through a bandpass amplifier before being rectified, offset, and greatly amplified. LM2900 quad Norton opamps were chosen because they work well at five volts and their current inputs simplify the sensor interface. An output voltage range of +1 volt or a bit less to +4 volts or a bit more is realistic. The user wants 0 to +5, but that's not attainable with real devices and the single 5 volt supply. (The prototype actually produced a minimum output of 160 millivolts.) Note: while some of the vendor documents are for the LM3900, LM2900s were actually used - the difference is mostly in the allowed temperature range.
Shift click here to get the original Visio Pro 2002 .vsd file.
Oscillator/Driver/LED
There is little of interest in this circuit, except that the special characteristic of the LM2900N (14 pin plastic DIP) that the output pulldown is a 1.3 millamp current sink allows driving the PNP driver transistor directly (this works as long as the collector (LED) current (in ma) is less than 1.3 times the current gain. The oscillator circuit is straight out of the National Semiconductor AN-72 application note with the component values recalculated to match available components. The opamp generates a 1 kHz square wave that gates Q1 on when low and off when high. This results in the LED producing a pulsed light output.
Photocurrent and BandPass Amplifiers
Since the entire object of this design is to isolate the top of the peak of the photocurrent, it's OK to let the bottoms be nonlinear. The first amplifier on this sheet is biased by the photocurent and produces a more or less square wave output of variable amplitude. This amplifier's + input bias requirements are satisified if there is some phototransistor current during the off phase. The transfer function lies between ~0.005 V/μA and ~0.025 V/μA, depending on the setting of R7.
That square wave is passed through the second amplifier which is a 1 kHz bandpass amplifier with a Q of 10. This circuit is explaned in the AN-72 document referenced above. Output of this amplifier is a sine wave (more or less) in the 1-2 volt peak-to-peak range - all DC information from the phototransistor has been removed so the peak-to-peak value reflects the original peak value.
Rectifier, and Low Pass and Output Amplifier
The rectifier is simply an unbiased amplifier - only positive halves (approximately) of the input sine wave are passed with a gain of 2. This presents a half wave rectified signal to the top of the span pot - the low side is held at one diode drop above ground so that all of the signal is at or above the minimum input level of the following stage (one base-emitter drop).
The rectified output is applied to the span pot which is biased at its low end to ensure that even at low values of span there is still enough DC component to bias the following amplifier input stage on.
The last stage is a low pass filter with a DC gain of about fifty and a DC coupled input, so if a pulsed input is present, there is still DC bias to the inverting input through the feedback resistor. An offsetting voltage is applied to the current mirror input to offset all but the very top of the unput pulses. This is a bit weak, but seemed a reasonable solution given the severe constraint of the single 5 volt power supply and the limited number of usable amplifiers (I wanted to avoid using the extra amplifiers in the oscillator package.
In theory, the output increases for decreasing light transmission, but in practice, it may decrease. This effect may be due to excessive photocurrent causing some stage to fold over, or it may be due to the light level actually increasing due to refraction in the translucent powder. Since any change is useful, this becomes a minor matter of inversion during data reduction. The rather large amount of noise apparently due to mechanical vibration requires extensive post processing anyway.
Setup and Adjustment
TP (Test Point) 1 is simply for checking that the oscillator is operating and that the LED is being driven - there should be a square wave of approximately 1 kHz and 2.2 V.
TP2 allows setting the gain of the photocurrent amplifier - with the LED and phototransistor in there mounts, adjust R7 for approximately 1 volt peak-to-peak at TP2.
TP3 should present a reasonable sine wave of about two volts peak-to-peak.
TP4 should present a roughly half-wave rectified version of the sine wave at TP3 - there are no adjustments.
TP5 is the point to monitor when setting up the span. Begin with R17 set so that TP5 has about 1 volt peak.
TP6 allows monitoring the offset voltage - it should be adjusted to approximately 1 volt initially (about half of that is due to the bias diode at the bottom of R17.
TP7 is the output. Adjust R18 about its initial point until the unattenuated output (no powder) is about 2.5 volts. This should be a reasonable setting for use, but if it doesn't provide enough resolution, increase the voltages at TP5 and TP6 until they approximately match at a higher voltage and retrim R18 to bring the unattenuated signal to a convenient point. Span and offset can be reduced to provide a wider woprking range at the cost of reduced resolution. The offset can also trimmed to move the unattenuated signal to the top or bottom of the working range (roughly 1 to 4 volts) as needed tp provide the full working range for attenuation - the direction would depend on which way the output was changing with increased powder flow. Note: the offset voltage should always be slightly above the average value of the rectified input to the output stage.
T.E.D. 22 April 2004
How I would do it if the constraints of a single five-volt power source and an existing LED/Phototransistor pair were removed.
I'll work on this as time allows: Opacity Meter Mark II