Learn how to measure noise using LTspice for op amp circuits with practical examples.
In previous articles, we introduced modeling noise with LTspice and simulated noise sources in LTspice. This article shows how to use LTspice to measure the noise of an op amp and op amp combined with a dual JFET input stage. How accurate are the measurements? They are useful? To find out, the results are compared to the noise specifications on the data sheets.
An excellent summary of op amp noise and how noise sources are combined can be found in this article and in the Analog Devices application note. An excellent explanation of how to convert an RMS value to peak to peak is in this video from Analog Devices.
LTspice noise simulation example: low noise op amp
The first component to test is a low noise op amp. The circuit is a standard non-inverting amplifier with a resistor in series with the non-inverting input to measure bias current noise.
The values for the power supplies and input resistance have been parameterized for ease of change and use in the .STEP directives. More on this later. Why is “no noise” added to the resistance values? When adding this undocumented attribute to a resistor, LTspice ignores the resistor as a source of noise. This feature is very useful because the additional noise from the resistors does not have to be subtracted from the measurement. The silent attribute can be added using the Component Attribute Editor. Launch this editor by holding down the control key and right-clicking on the body of the resistor. Add the word “no noise” as an additional value. With this feature, the ONLY source of noise is the op amp. Perfect!
The op-amp is an ADA4627 analog device. Not much work was done on choosing this part. The ADA4627 caught my eye after a quick scan of Analog Devices’ low noise op amp selection chart. It has low noise and supports supply voltages from ± 5V to ± 15V. Then a check of the datasheet showed that the noise is well specified. It was not chosen because the specifications in the datasheet did or did not match the LTspice tests.
Here’s the noise section of the ADA4627 datasheet. The left column corresponds to the “Rating B” parts and the right to the “Rating A” parts. Specifications are used in the comparison table below.
The table is from the Analog Devices ADA4627 datasheet
Here are the LTspice directives used in the simulation.
The measurement conditions are set with statements on the left side. The .PARAM statements provide values to the variables in the schema. These are the names of the variables in braces, for example, {RINP}. .STEP statements run multiple simulations with different variable values. For example, two runs are performed with different values of input resistance. There is a small value for the input voltage noise measurement (Step # 1) and a large value for the input current noise measurement (Step # 2).
Measurements are specified with .MEAS statements on the right hand side. No additional calculations are required to compare with the datasheet. I’m not going to go into the details of every .MEAS statement. However, let’s look at one to get the noise from the input voltage at 1kHz.
.meas NOISE en1_1k_RMS FIND V (inoise) @ 1 AT 1K
NOISE – Apply the measurement to a noise simulation.
en1_1k_RMS – Only one name for the result. Used in the log file.
FIND – Specify the measure, which in this case is only getting a data value
V (inoise) @ 1 – The data set to be used in the measurement. Details are below.
IN 1K – Select the frequency of the data.
Now for the magic. Click on the running engineer.
Here is a graph showing the results of the two-step simulation. The green trace is Step # 1 with an input resistance of 1Ω. The purple trace is Step # 2 with an input resistance of 10MΩ. The red trace shows the difference between the two races.
“V (inoise)” needs an explanation. “V (inoise)” is the output voltage noise of the entire circuit that refers to the input. LTspice calculates this for each noise simulation. This is explained below. Remember that noise sources are added and subtracted RMS. As one source of noise grows larger than another, it begins to dominate completely. No problem! LTspice takes care of the calculation for us!
The data for a particular execution of .STEP is selected by adding “@ 1” or “@ 2” after a font. For example, “@ 1” is the execution with the small input resistor.
The data for the .MEAS statements are in the Spice Error Log. Here is a table comparing the data sheet values with the measured values. Not bad!
| Voltage noise density 10 Hz | 16.5 nV / √Hz | 16.6 nV / √Hz |
| Voltage noise density 100 Hz | 7.9 nV / √Hz | 6.9 nV / √Hz |
| Voltage noise density 1 kHz | 6.1 nV / √Hz | 5.0 nV / √Hz |
| Voltage noise density 10 kHz | 4.8 nV / √Hz | 4.8 nV / √Hz |
| Voltage noise 0.1 to 10 Hz | 0.7 μV pp | 0.55 μV pp |
| Current noise density 100 Hz | 1.6 fA / √Hz | 2.2 fA / √Hz |
| Current noise 0.1 to 10 Hz | 30 fA pp | 35 fA pp |
The datasheet provides peak-to-peak values for noise from 0.1 to 10 Hz (1 / ƒ region) but LTspice calculates RMS. I applied a factor of 5 to convert RMS peak to peak. The voltage noise is the same for grades A and B. However, the current noise is different. Grade B is shown in the table.
The measurement results can be found in the “Spice Error Log”. Open this file by selecting the plot window and select View-> Spice Error Log. Scroll down to the results for .MEAS. Here is the input for the voltage noise measurement explained above.
Scheme Directive:
.meas NOISE en1_1K_RMS FIND V (inoise) @ 1 AT 1K
Corresponding entry in the spice error log:
Medida: en1_1k_RMS paso v (inoise) @ 1 en 5.03632e-009 1000 5.03632e-009 1000
Measurement results are displayed for each step. However, in these runs, the value from Step # 2 is replaced with the value from Step # 1 since the data is specified with the “@ 1” specification.
What is the “V (inoise)” font in .MEAS directives? From the LTspice Help, V (inoise) is a calculation of “input referred noise voltage density”. This noise source is the output noise referred back to the input for ALL noise sources in the circuit. However, since all resistors are “quiet”, the noise is just the input noise from the op amp. Again, perfect!
Here are graphs to explain “V (inoise)” or “Voltage density of noise referred to input”. V (inoise), V (onoise) and gain are displayed in separate windows. The measurement cursors for V (inoise) and V (onoise) are set at 1 kHz. Dividing the output noise V (onoise) by V (inoise) is the circuit gain of 50. V (inoise) is the input noise of the op amp, since all resistors are “quiet”.
LTspice 2 noise simulation example: low noise JFET op amp
The next component to test is a very low noise dual JFET. Here is a diagram of the circuit.
Here are key points that make the noise measurement almost the same as that of the operational amplification circuit.
- The differential pair gain is large enough that the op-amp noise is very small compared to the amplified JFET noise and can be ignored.
- Current source noise that bypasses the JFETs (I1) is common to the two op amp inputs and is eliminated by op amp common mode rejection.
- The noise of the drain resistors (RD1, RD2) is eliminated with the attribute “no noise” LTspice.
The dual JFET is the LSK489 from Linear Integrated Systems. This part was chosen because it is in the LTspice standard library and is a very low noise device. It was not chosen because the specifications in the datasheet did or did not match the LTspice tests. Here is the complete schematic.
This circuit is adapted from an application note on linear integrated circuits.
There are additional parameterized parts. R3 sets the drain current from the JFETs and, together with R1 and R2, sets the drain voltage to the source. The circuit is adapted from a Linear Integrated Systems application note. I liked using an LED to set the operating point of Q1. This was new to me!
The measurements are very similar to the op amp simulation. One difference is that the circuit has two JFETs and they both contribute to the input voltage noise. The datasheet gives the noise for a JFET. When necessary, the combined noise measurement is multiplied by 1 / √2 to obtain the noise from a single device. This setting is not necessary for the input current noise measurement because the large series resistor is only on one input. The noise specifications in the datasheet assume a drain source voltage of 15 VDC. However, the test circuit uses 9.5 VDC to stay within the common mode voltage range of the op amp. For good measure, the tests were performed on a drain source voltage range and the results indicated that the drain source voltage does not significantly affect noise. Can you see the .STEP for this test?
Once again, a flick of the magic wand to the running engineer. How do the LTspice measurements compare to the datasheet?
| Noise voltage (10 Hz) | 3.5 nV / √Hz | 3.8 nV / √Hz |
| Noise voltage (1 kHz) | 2.0 nV / √Hz | 1.8 nV / √Hz |
| Voltage noise 0.1 to 10 Hz | Not specified | 0.12 μV pp |
| Current noise density (100 Hz) | Not specified | 1.8 fA / √Hz |
| Current noise 0.1 to 10 Hz | Not specified | 28 fA pp |
Not bad! Additional measurements are shown which are not in the datasheet but are useful for comparison with the op amp circuit.
conclusion
LTspice was used to measure voltage and current noise for a low noise op amp and the same op amp with low noise dual JFETs added to the input. There is a good agreement between the simulated values and the values in the datasheet for both devices. The .STEP and .MEAS directives are used to calculate values for direct comparison with datasheets.
Bonus tip
Low noise op amp datasheets often have a figure showing a 10 second oscilloscope noise trace in the 0.1 Hz to 10 Hz band. With the available tools, you can digitize this trace and convert it to a CSV file. Then the CSV file is used to specify a PWL voltage source. The figures show a data sheet plot and the corresponding plot of a Transients analysis with a PWL source. The site I used to convert the graph is WebPlotDigitizer (it’s free). Make sure to use the option to rearrange the file entries in order. Can you think of something useful for this waveform?
The graphic is from the Analog Devices ADA4627 datasheet.
Related Posts
LTspiceAmpNoise.zip
