op-amp output noise

Mitch,

Here is a brief answer to a question with a few twists and turns:

Op amps are specified by their noise density, a number in nV/rt-Hz. A spec is also provided for noise current, often in pA/rt-Hz or fA/rt-Hz. The noise current density is multiplied by the source resistance to get the noise voltage density created by noise current and combined with op amp's voltage noise by the root of the sum of the squares.

To get a voltage, you multiply this total noise density by the square-root of the measurement bandwidth. If you do not limit the bandwidth with a filter, the bandwidth will be approximately the unity-gain bandwidth of the op amp (because you are operating in G=1). Be careful when comparing noise of different op amps with different unity gain bandwidths as you are measuring noise over different bandwidths. Actually, the noise bandwidth is about 1.6 times the op amp bandwidth because the roll-off is only 20dB/decade. If you limit the bandwidth with 2-pole filter (or more) you can consider the noise bandwidth to be equal to the filter bandwidth. Noise of audio circuits is often measured with a 20kHz, 2-pole low-pass filter.

This gets you the noise voltage produced by the op amp. It does not include the thermal noise from the source resistance.

There are some other possible variables but these are the basics. If you would like to provide some circuit details, I'll check for other possible factors.

Regards, Bruce.

Bruce:

 Thanks for getting back to me so fast.

  I was not asking how to measure noise but rather how I can determine the noise of your various op-amps at unity gain without having to procure parts and test them. I was hoping to find a correlation between the specs you provide and low noise at unity gain.

Thanks

Mitch

Mitch,

The advice remains the same, minus the comments on measurement.

The noise in unity gain is the combination of voltage noise and current noise times source impedance. "Adding" the voltage and current noise by the root of the sum of the squares gives you the total noise.

If you can provide a circuit or application details I can work some numbers by example and make recommendations on the low noise op amp choices. The key information needed is the source impedance or type of input signal source.

Regards, Bruce.

Bruce:

 Let's make the following assumptions:

 The amp is connected as a unity gain follower, output connected directly to the inverting input. The non-inverting input has a 10 Ohm source to ground. The measurement bandwidth is 20KHz.

 When you use, say a ne5532 in this configuration, the output noise measures about 3uV RMS. (I believe that this translates to about 22nV/Hz^.5)

Now the equivalent input noise is spec'ed at 5nV / Hz^.5 for voltage. ( The Voltage noise is far more significant than the current noise due to the low input impedance of our example so I am ignoring it for now).

There is a noise floor in this case that is unrelated to the Equivalent input noise spec, So how do I choose the lowest noise op-amp for this configuration based on the specs you provide?

Thanks

Mitch

Mitch,

Your thinking is correct. With a 10 ohm source resistance, the noise of the resistor and current noise times the resistor value are insignificant. Measured over 20kHz BW, you should be getting closer to 0.7uV noise. There is no additional noise floor in the op amp that is related to operation in unity gain.

This may be a measurement issue. In unity gain, your measurement device must have a noise floor that this lower than the op amp you are measuring. When you short the input of your measurement instrument, what do you measure? You  should also look at the output with an oscilloscope. It is very possible that you are measuring extraneous 60Hz hum or other noise.

It can be difficult to measure noise performance in unity gain because the noise level is so low. We generally connect the op amp in a gain configuration to increase the output noise to a more easily measured quantity.

Regards, Bruce.

Bruce:

 We use an Audio Precision SYS-2722 and I monitor the measurement with an oscilloscope. I generally don't trust the analyzer measurements without seeing the spectra and character of the reading output. In this case, the AP reads 10dB lower when its' input is reading the grounded reference for the op-amp and the 22-20KHz reading output  spectra is basicly white in character.

  The main reason for my question is the fact that in my experience I have never measured an op-amp 's output noise as low as its' Equivalent input noise when running low or unity gain. Is it typical of your op-amps' unity gain noise to measure below 1uV RMS?

Thanks

Mitch

Mitch,

Sounds like your measurement setup is pretty clean and your procedure is commendable. Good job.

Still, 10dB margin is not good enough to make a measurement down to the specified level of the NE5532. !0dB margin is only 1/3 the noise of your measurement value. But you should be getting better numbers with an NE5532. Have you tested more than one unit? You could be measuring a damaged op amp. Bipolar-input op amps can be damaged in a manner to cause excess noise but otherwise function normally. This is especially true of older designs that have minimal or no ESD protection.

To be honest, I can't say how our op amps noise look in unity gain as we rarely measure them in this configuration. Furthermore, they are rarely used in this configuration if low noise is an important consideration because it passes the noise problem on to the next stage in the signal chain. I am, however, quite confident of the expected behavior of an op amp in this configuration.

Regards, Bruce.