Implementing LM339 comparator to count analog peaks

Question

Updates after applying Trevor's advice: I have attenuated the output of the CPS to between 2 and 2.5 volts, as shown in the scope pic: , channel 2 is the constant "high" output from the schmitt trigger, which is not connected to any input in that screenshot.

the issue now is that when I connect the trigger input it drops the voltage way down and the scope looks like this:

I have absolutely no idea why it is dropping the voltage so much, any advice on what to change? see Trevor's answer for schematic.

---

original question below:

I am an ME lost in a EE world... Overall problem is need RPM input from an engine to an arduino for PID control. Went to the shack and picked up some stuff to experiment with. Seems that the LM339 should be able to pickup the peaks of the analog signal from the crank position sensor and output a logic signal that the arduino can count and convert to rpm.

I have referenced the TI datasheet and application guide [see figure 6] in my efforts so far. See attached schematic for my starting point. Also attached the scope reading of the analog signal from the cps.

My question is basically can this work or am I missing some fundamental EE reasoning? Input/advice on resistor values, decoupling caps, and the proper Vref value would also be greatly appreciated.

would not let me post links to the application guide or my search results, found some info on decoupling caps for IC's but not sure how to apply it to my circuit.

Answer 1

The circuit below should do the trick.

I A.C. coupled the input signal and biased it to 2.5V using C1 and R1, R2.

Shotky diodes D1, D2, with R7 afford you protection in case the signal ventures over 2.5V peak to peak.

C2 provides some noise filtering. You may need to adjust this size to best suit your application

The reference is set on the plus pin at 2.5V with about +/- .44V hysteresis. If the signal has a ripple or noise larger than that decrease the size of R5.

R6 provides the final pull-up.

^{simulate this circuit – Schematic created using CircuitLab}

EDIT: Since your input signal is a lot larger than you originally intimated, You could also use the following circuit to simply half-wave rectify it and attenuate it a little.

Benefit of this method is you pretty much guarantee full swing on the input of the comparator. That allowed me to decrease R5 giving you more hysteresis and lowered C2.

^{simulate this circuit}

Answer 2

The LM339 won't like inputs that drop below 0 volts. The absolute maximum negative input voltage is -0.3 volts below the negative power supply rail.

I would consider biasing the input signal to somewhere between the power rails. It might be an issue if the RPM is low but I don't think you have much of an option. Some comparators (like the MAX999) are designed to be used with a negative input signal but this still has to be restricted and I would consider using a potential divider to reduce the input signal to +/- 75 mV.

Your pull up resistor is also far too low and implies an amp of current into the device when driving a low output signal. The MAX999 (and others) have a push pull output and don't require this resistor. Looking at the hysteresis resistors I think you have too much hysteresis, so much that the input signal won't be able to switch the output from one state to the other.

Answer 3

Suitable for signals 2Vpp with 500 Ohm load to reduce stray noise coupling. Use twisted pair , pref shielded from Arduino gnd side only.

^{simulate this circuit – Schematic created using CircuitLab}

Answer 4

If you feed a voltage below about -0.6V to LM339 inputs, then a high current will flow out (limited by your sensor) and LM339 input stage may not survive.

Simplest solution is to stick a resistor in the input, like 10k, and let the input protection diodes shunt the current to GND.

A much better solution, especially near motors which may create lots of noise, is:

• Use hysteresis
• Filter input signal

Here's a quick example. Here, I set hysteresis high enough to reject the noise on the 1kHz waveform. You can use LM339, I used LT1018 here because the model comes with LTSpice.

See how comparator ignores multiple zero crossings due to hysteresis?

A small RC filter on the input signal to get rid of HF noise is also a good idea.

Answer 5

can this work...

the basic concept is sound, aside from excessive positive feedback.

in cases like that, the challenge often is a signal with varying degrees of "average". so you cannot "pre-set" the reference voltage.

if your application calls for that, you may think about how to take that on.