SNR measurements at the output of opamp

[afr123]

Hi,

I have a very basic question.

I want to calculate SNR at the output of the opamp. First I measured the rms value of the noise without any input using a digital oscilloscope. Then I input a sinusoidal input of 40kHz to the opamp. The voltage at the output was 0-5V.

So to measure the SNR I divided 5V by sqrt of 2 to get the rms voltage. Then I used 20*log (rms voltage / rms noise) to get the SNR.

My question is do the need to remove the offset (2.5V) from the signal first and then calculate the rms voltage? And secondly, is it the proper method to calculate SNR?

Regards

 

[KlausST]

Hi,

SNR = Signal to Noise Ratio

Signal = max undistorted AC signal in RMS Volts
Noise = RMS noise voltage over the bandwidth of desire. (Mayb with additional weighting option)

**
Maybe a digital scope is not useful to calculate RMS noise:
* It usually has low ADC resolution introducing additional quantization noise.
* The analog input stage may introduce additional noise
* ..also cabling and GND loops
* there usually is no selectable bandwidth.

If you don´t have any other method I recommend to use a low_noise, high_gain amplifier with bandpass filter for to adjust bandwidth and weighting.

Offset = DC is not included in any noise calculation.

Klaus

 

[afr123]

The opamp is configured as a 4-stage bandpass filter (10kHz - 100kHz) with a gain of 500. What I need is to measure the SNR at the output.

My Oscilloscope settings are
BW limit: 20MHz
AC coupling
5mV/div

I only have a high-end oscilloscope from Agilent Technologies to record the data.

What I am not sure is my method. I just measure the rms noise 'without' any signal. I guess this is how noise is measured?

Then I use the rms value of the signal and noise to calculate the SNR. So I just want to know if my method is correct?

Thanks.

 

[KlausST]

Hi,

Impossible to say yes or no.

Your method is correct if:
* your scope introduces no quantisation errors.
* your scope internal amplifier is noise free
* your desired bandwidth for the noise is 0Hz...20MHz

***

The opamp is configured as a 4-stage bandpass filter (10kHz - 100kHz) with a gain of 500.

Is this the DUT?
Or is this the amplifier and filter for noise measurement?


Please give us your values.
* max RMS amplitude
* your measured RMS noise
* for comparision: Leave the scope_GND_terminal connected at the DUT. Connect the scope tip to the same GND connection. What RMS noise value do you measure now?

Another method to verify your measured data:
Run an FFT over the measured DUT noise signal. It should be relatively flat without high peaks.

Klaus

 

[afr123]

This is DUT. I want to measure the noise at the output of this filter and amplifier. This output of this opamp is connected to 16-bit ADC.

I put the gnd and tip at the gnd terminal of DUT.

Measured AC-RMS noise = 384uV (duration = 20ms)
Vpp = 6.8mV (duration = 20ms)


Regards

 

[KlausST]

Hi,

This is DUT.

--> but you need a band pass filter for noise measurement - at least I recommend it.
The problem is: I anyone else measures noise with it´s own measurement device ... the result may differ.
My opinion: The quality of a measured value is only as good as it´s specification.

One good example for this is with audio amplifiers. The manufacturer says it is a 200W amplifier, but if one measures with the specification: 1% distortion, 8 Ohms load, 1minute then the rusult is 20W only.
So the 200W is meaningless - without specification. In this case it´s more a marketing value.

You also need to specify your "amplifier and filter" input source for noise measurement: Open, short, or shorted with an specified resistor.

This output of this opamp is connected to 16-bit ADC.

I don´t recommend to measure noise with the ADC connected. The ADC may introduce noise caused by a switched capacitor sample and hold circuit. This may harm your measurement, while it does not harm ADC output quality.

I put the gnd and tip at the gnd terminal of DUT.

Just to verify:
384uV is the RMS noise of the shorted input?
And 6mVpp is the PP noise of the shorted input?

But what about the other two requested values? (I recommend to connect a resistor at the DUT input that about equals the expectable operational input sourceimpedance)

Klaus

 

[afr123]

okay. I just need to mention that I used this oscilloscope with these settings then the results should match (without considering the internal noise of the oscilloscope itself).

The measured values above are at input of the opamp while the gnd of the scope and the tip were shorted.

The measured values at the output of the amplifier:
rms noise at output of the amplifier = 5.06mV
peak-to-peak noise at the output = 56.3mV

What do you mean by max rms amplitude?

Regards

 

[KlausST]

Hi,

What do you mean by max rms amplitude?

--> post#2: "Signal"

peak-to-peak noise at the output = 56.3mV

--> next time you know you don´t need a 16 bit ADC. An 8 bit ADC gives almost the same information. ;-)

5mV RMS noise is huge. Either your amplifier input is left open, or your circuit has some noise problems.
A well designed amplifier/filter circuit should have 0.5mV noise RMS or less. (50 Ohms input, 1k...100kHz, linear weighting)

Klaus