Amplifying weak pulse signal in the presence of noise

Jeff311

Joined Jul 21, 2019
4
Hi guys

I want to be able to amplify a weak periodic pulse signal in the presence of noise. What are the key figure of merits I should be looking out for when selecting an op amp? e.g noise figure, CMRR, etc.

Cheers

Sensacell

Joined Jun 19, 2012
3,396
It totally depends on the technique used to recover the signal.

Tell us more about the problem? what signal? what purpose?

crutschow

Joined Mar 14, 2008
33,985
What signal characteristics do you need to measure?

Jeff311

Joined Jul 21, 2019
4
It totally depends on the technique used to recover the signal.

Tell us more about the problem? what signal? what purpose?
The signal is a low duty cycle (~0.1%) periodic pulse at 10 Hz. Using a non-inverting amplifier with op amp, the gain should be (1+Rf/R). Ideally, the pulse would get amplified no matter how weak it is, but realistically there's always some noise, and once the pulse becomes too weak it doesn't get amplified (no amplified pulse at output). Which parameters of an op amp should I be checking for in order to tolerate as low a pulse amplitude as possible?

Hymie

Joined Mar 30, 2018
1,271
Depending on the frequency of the noise, you might get away with a simple high pass filter.

Although the pulse is based on a fundamental 10Hz, at a 0.1% this pulse will be equivalent to a 10kHz signal. I suggest you start with a classic high pass filter made from an op-amp and two resistors & capacitors with a cut off frequency of 10kHz.

crutschow

Joined Mar 14, 2008
33,985
If you answer my question in post #3 we can give you better answers.

BR-549

Joined Sep 22, 2013
4,928
For an open, general question, you'll get a general answer. And a general answer might not work for you. We need details.

Analog Ground

Joined Apr 24, 2019
460
Some questions:
1. Your 0.1% duty cycle at 10 Hz infers a pulse width of 100 usec. Is this correct?
2. What are the rise and fall times of the pulse?
3. Do you care about the shape and amplitude or just want to know if a pulse is present?
4. Is your main concern noise produced by the circuit or getting rid of noise which comes into the circuit with the signal? Or both?
5. If noise coming in with the signal, what type of noise is it? Power line interference (50 or 60 Hz), RF pickup, etc.
6. Can you say what type of sensor you are using? What is the source of the pulse?

Analog Ground

Joined Apr 24, 2019
460
The signal is a low duty cycle (~0.1%) periodic pulse at 10 Hz. Using a non-inverting amplifier with op amp, the gain should be (1+Rf/R). Ideally, the pulse would get amplified no matter how weak it is, but realistically there's always some noise, and once the pulse becomes too weak it doesn't get amplified (no amplified pulse at output). Which parameters of an op amp should I be checking for in order to tolerate as low a pulse amplitude as possible?
A small correction on your view of amplification. Signal of any size will be amplified but it can be too small to use because of the noise. The correct way to think about it is the ratio of the size of the signal to the size of the noise (S/N). You want to make this ratio as large as possible. Too small a ratio and the pulse cannot be seen but it is there at the output.

Analog Ground

Joined Apr 24, 2019
460
Hi guys

I want to be able to amplify a weak periodic pulse signal in the presence of noise. What are the key figure of merits I should be looking out for when selecting an op amp? e.g noise figure, CMRR, etc.

Cheers
Noise figure is not an op amp specification. CMRR would only apply if you use an Instrumentation Amplifier instead of a non-inverting op amp circuit. The op amp specs are Input Voltage Noise and Input Current Noise. TI and Analog Devices have many application notes about op amp noise. An example, of noise analysis and use of op amp noise specs for a non-inverting op amp circuit is this article.

Jeff311

Joined Jul 21, 2019
4
Some questions:
1. Your 0.1% duty cycle at 10 Hz infers a pulse width of 100 usec. Is this correct?
2. What are the rise and fall times of the pulse?
3. Do you care about the shape and amplitude or just want to know if a pulse is present?
4. Is your main concern noise produced by the circuit or getting rid of noise which comes into the circuit with the signal? Or both?
5. If noise coming in with the signal, what type of noise is it? Power line interference (50 or 60 Hz), RF pickup, etc.
6. Can you say what type of sensor you are using? What is the source of the pulse?
1. Small correction sorry: it's closer to 0.2% -> pulse width of 200 usec
2. Rise time ~20 usec; fall time ~ 40 usec
3. Just knowing the pulse is present is enough, but ideally if the pulse transitions can be can be recovered and amplified I can then use that to trigger something else
4.5.6. The pulses come from pulsing IR LEDs. The sensor I'm using is a phototransistor, and the pulse I described is the output of a photodarlington. Sources of noise/interference include the sun (this is an outdoor application) and cloud movements etc. and component noise. The challenge is I can't freely increase my sensor sensitivity because of saturation from the sun, and the pulses become weak at long distances.

MrChips

Joined Oct 2, 2009
30,440
Now we're getting somewhere.
You are using a photo-transistor. Is the photo-transistor shielded from ambient light and has a filter that allows IR only?

It is common practice to modulate the IR LED at 38kHz or 40kHz. Have you considered this option?

There are IR receiver modules that are readily available that will detect the 38kHz or 40kHz IR signal for you and give you a logic output.

Jeff311

Joined Jul 21, 2019
4
Now we're getting somewhere.
You are using a photo-transistor. Is the photo-transistor shielded from ambient light and has a filter that allows IR only?

It is common practice to modulate the IR LED at 38kHz or 40kHz. Have you considered this option?

There are IR receiver modules that are readily available that will detect the 38kHz or 40kHz IR signal for you and give you a logic output.
Yes, the phototransistor is shielded and has an optical bandpass filter in front of its active region. Modulating the LEDs is not an option, unfortunately - I'm working on extending the range of detection solely from the receiving end.

Hymie

Joined Mar 30, 2018
1,271
I think my idea of a high pass filter with a cut off at around 10kHz might very well work, given the further information you have supplied.

The circuit below is the classic high pass filter with the cut off frequency given by:-

f = √2 / 4π RC where C1 = C2 = C and R1 = R and R2 = 2R

If my calculations are correct, with the capacitor values of 10nF and R1 is 10kΩ and R2 is 20kΩ will give a cut off at around 10kHz.

Attachments

• 39.3 KB Views: 7