Noise amplifier principle and knowledge questions and answers

There are two types of IC noise:

One is external noise, which comes from outside the IC;

The other is internal noise, which comes from the device itself.

External noise

Some engineers believe that external noise should not be called noise because it is not randomly generated. It may be more appropriate to use the term "interference". First, let's briefly talk about the three main sources of external noise:

RFI coupling

The environment is full of various electromagnetic waves. Although these RF interference signals are usually outside the target bandwidth, the nonlinearity of the device can sometimes adjust these signals and bring them into the target area. Especially when the leads connecting the sensor are long, noise generally enters the circuit from the input leads.

Ways to suppress RF interference include: input filtering, shielding, and twisted pair input.

Power supply noise

Electronic circuits have limited ability to suppress power line signals, especially at high frequencies, so high-frequency interference on the power line must be eliminated first so that it cannot reach the low-noise circuit. This can be achieved by properly filtering the power supply and taking good bypass measures on the IC itself. Sensitive analog circuits and digital logic should use different power supplies, or at least deeply filtered.

Ground loop

We often see a lot of ground symbols on schematics, but it must be noted that in actual circuits, the potential of any two points cannot be completely equal. Current will flow through the ground wire, resulting in a potential difference. It is important to consider how current flows and isolate high current paths from sensitive circuits. For example, a utility ground configuration or connecting the analog ground plane to the digital ground plane at the same point.

Internal noise

Internal noise originates from circuit elements in the signal chain, and the performance specifications in the IC data sheet are for this noise. Typical internal noise sources include sensors, resistors, amplifiers, and analog-to-digital converters.

Resistor noise

Resistor noise is divided into two categories: one is internal thermal noise, which is independent of the resistor construction and depends only on the total resistance, temperature, and bandwidth. It is independent of the applied signal; the other is additional current noise, often called excess noise, which depends on the construction of the resistor. Unlike thermal noise, resistor current noise is related to the applied voltage. Thin film resistors and wirewound resistors have excellent current noise performance, and their noise is mainly internal thermal noise. Carbon core resistors are not the same. They are generally considered to have poor noise performance. In the following discussion, we will assume that high-quality thin film resistors are used in low-noise designs, so current noise can be ignored and only thermal noise can be focused on.

The thermal noise formula of an ideal resistor is:



It can be seen that thermal noise depends on temperature, resistance, bandwidth, and Boltzmann constant. But in actual design, you don't have to remember this formula because we have a very convenient quick calculation method.

The square root symbol will appear again and again when discussing noise. The formula contains a constant term, which is the Boltzmann constant k. The second term is temperature. Note that noise increases with increasing temperature. The unit of this temperature is k, so the effect of temperature on noise may not be as great as you think. Most engineers ignore the effect of temperature on noise. Remember that the noise specifications you see are only valid at room temperature. The third term is the resistance value, and the last term is the bandwidth.

You should remember this formula. The thermal noise of a 1kΩ resistor at room temperature is , which is

No matter what noise-related work you do, this formula will always benefit you. This quick calculation formula can be easily applied to other resistor values.

Amplifier Noise

Figure 1 shows the amplifier noise model. Amplifier noise is divided into two categories: one is voltage noise (VX) and the other is current noise (IX). In real circuits, amplifiers are composed of many transistors, all of which have noise. Fortunately, the noise of all transistors can be referred to the input of the amplifier.


Figure 1 Amplifier
Noise Model ... The noise characteristics of instrumentation amplifiers are slightly different from those of op amps. For op amps, all internal transistor noise can be referred to the input. In other words, all noise sources scale with gain. This is not the case with instrumentation amplifiers. Some noise in the circuit scales with gain, while other noise is independent of gain. The amount of noise related to gain is shown here as eNI, and the amount of noise independent of gain is shown as eNO. The data sheet has a formula to show the relationship between the two. In addition to voltage noise, amplifiers also have current noise. If there is resistance at the input, the current noise will interact with it to produce voltage noise. For example, most source voltages have some resistance. After all, converting a high impedance signal source to a low impedance signal source is one of the reasons for using an op amp. Current noise flows through the resistor connected to the amplifier, producing voltage noise. Generally speaking, the higher the input bias current of the amplifier, the higher the current noise. Figure 2 shows a voltage follower configuration with some source resistance. The current noise of the op amp will interact with the signal source resistance to produce some additional noise at the output. Figure 3 shows how the resistors in the feedback path interact with the current noise, which flows through the parallel combination of the feedback resistors to create an additional noise source at the input, which is then amplified by the amplifier to the output. Figure 2 Voltage follower configuration with a certain source resistance Figure 3 Interaction of resistors in the feedback path with current noise Analog-to-digital converter (ADC) noise Sometimes analog-to-digital converter (ADC) data sheets provide noise characteristics in the form of Vrms or VP-P, but most of the time, the characteristics are expressed in terms of noise relative to the maximum full-scale range of the ADC, specified as the signal-to-noise ratio (SNR). The noise specifications in the data sheet occasionally include distortion characteristics and signal-to-noise ratio. In a pinch, the ideal formula provided in the article can be used, but this is a theoretical limit and is always better than the actual value. The formula here shows the conversion relationship between the ADC's SNR number and the Vrms number to compare the ADC and amplifier noise. One important thing to note is to make sure to use the rms noise within the maximum input range of the ADC. Peak-to-peak noise and RMS noise Peak-to-peak noise Vrms is the distance between the peak and the trough of a waveform. It depends on only two points, which has both advantages and disadvantages. The advantage is that it is very easy to calculate, just subtract the minimum point from the maximum point; the disadvantage is that it is not very reproducible and not very accurate. Noise is a random process, so this measurement actually relies on the extreme values ​​of the noise waveform. The longer the data is collected, the more likely it is to obtain extreme values. RMS noise uses all points in the waveform and is much more accurate than peak-to-peak noise. The more points are measured, the more accurate the RMS value. The disadvantage is that since all points are used, the calculation time is longer.
























One thing to note about peak-to-peak and RMS measurements is that they vary greatly with bandwidth, with lower bandwidths resulting in lower noise for the same amplifier. Figure 4 shows this clearly. In our lab, we measured the noise of the AD8222 instrumentation amplifier at several different bandwidths, and we can clearly see how bandwidth affects noise. For every tenfold increase in bandwidth, the noise increases threefold. Since these measurements are bandwidth dependent, there are a few things to be aware of: First, you need to understand the bandwidth characteristics of your circuit, and you need to make sure that the bandwidth of the instrument you are measuring is higher than the bandwidth of your circuit to get an accurate reading. Also, when using a DMM, when specifying RMS or peak-to-peak noise, you must also specify a specific bandwidth. For most data sheets, the bandwidth is the 0.1Hz to 10Hz band.


Figure 4 AD8222 Noise at Several Different Bandwidths

Spectral density plots take RMS measurements a step further, and actually break the noise measurement into different bins, so you can see which frequencies have more noise content. Figure 5, from the AD8295 data sheet, shows the averaged combined value of many measurements. Because the spectral density plot divides the measurement into many intervals, a large amount of data is required to obtain a clear plot.

Figure 5 Spectral density plot of AD8295

At lower frequencies, the noise curve of most amplifiers will slope up, and the noise density is inversely proportional to the frequency, so it is called 1/f noise. If you draw a straight line along the 1/f slope and intersect the horizontal noise line, you can get the 1/f corner frequency.

Noise calculation

The addition rule of noise is the sum of the squares of the noise. Assuming that the noise sources are uncorrelated, this assumption is valid in most cases. The multiplication and division rules of noise are the same as general signals.

First, there are a few things to note when calculating noise: At room temperature, a 1kΩ resistor corresponds to The noise, this quick calculation formula can be easily applied to other resistance values, just multiply by the square root of the resistance.

Second, when summing the signal sources, smaller terms can be ignored. The noise addition rule is the sum of squares. If a noise signal is only 1/5 of the dominant noise signal, then its contribution to the additional noise is only 1/25.

The third point is an extension of the first point. If the gain of the first gain stage is large enough, all the noise after it can be ignored.

Tips for designing low-noise systems

The first trick to designing low-noise systems is to apply as much gain as possible in the previous stage. Figure 6 shows two examples of an amplifier front end with a gain of 10. It can be seen that applying all the gain to the first stage is much better than distributing the gain between the two stages. Please note that sometimes the requirements for best bandwidth performance may conflict with the requirements for best noise performance. For bandwidth, we want each gain stage to have similar gain, while for noise, we want the first stage to have all the gain. Figure


6 Amplifier front end

The second trick is to pay attention to the source impedance. There are two reasons for this: first, the larger the source impedance, the greater the system noise; second, the amplifier must be well matched to the source impedance. If the source impedance is high, the current noise noise characteristics may be more important than the voltage noise characteristics. The

third trick is to pay attention to the feedback resistor. If you choose an ultra-low noise op amp but use a large feedback resistor, it is impossible to achieve a low noise circuit. In the non-inverting (Figure 7) or inverting configuration, pay attention to the feedback resistor as a noise source referred to the output. The other resistors are equivalent to the voltage source at the input, or more precisely, the voltage source at the input of the inverting configuration. As mentioned above, when designing a low-noise system, the first-stage application has a high gain, in which case the Rg noise is dominant.


Figure 7 Noise model of a non-inverting operational amplifier

For more details, please visit: http://seminar.eccn.com/100325/jchf_adi.asp

Q&A Selection

Q: How to accurately measure the internal noise of an amplifier? What factors are related to it? What issues should be paid attention to during testing?

A: Generally speaking, to measure the noise of an amplifier, the input of the amplifier is connected to 0, the output passes through a low-pass filter, and then a high-precision ADC is used to sample and perform FFT, or an oscilloscope is used to view the output.

Q: When judging the performance of an amplifier, which noise parameter should be mainly referred to?

A: The noise parameters of sensors, resistors, amplifiers, and ADCs should be considered.

Q: When designing an amplifier using an op amp, how to estimate its input and output impedance?

A: Usually, for op amp devices, we consider their input impedance to be infinite and their output impedance to be 0 (you can refer to the data sheet of the specific model to check the specific value). So the input and output impedance of the circuit can be calculated based on this condition.

Q: How to reduce the internal noise of the device and weaken the external noise?

A: The internal noise of the device cannot be changed, and the external noise can be limited by selecting the external bandwidth.

Q: What are the characteristics of LC circuit filtering and operational amplifier circuit filtering, and in what occasions are they used?

A: LC filtering is simple, but the filtering effect is not as ideal as active filtering. Moreover, active filtering can amplify the signal at the same time, while passive filtering cannot do this.

Q: How to consider the quantization noise of ADC?

A: Quantization noise exists in theory and cannot be removed. This is also the source of the theoretical signal-to-noise ratio of 6.02N+1.76.

Q: How to measure noise most accurately without introducing measurement noise?

A: If you want to get the most accurate noise, you should use the root mean square value measurement method. This method will take all the noise into account, but the disadvantage is that the measurement time is long and the data volume is large.

Q: How to eliminate noise by single-point grounding or multi-point grounding? What is the difference between them?

A: Single-point grounding means connecting the ground only at the chip power pin. This is to prevent the ground return of the digital power supply from affecting the ground of the analog circuit. It is also used when analog and digital chips are on the same board. Because the two grounds must be connected together in the end, they are generally selected at the junction of the analog and digital grounds. Multi-point grounding means that the ground pin of the chip should be grounded nearby, and there is no need to lead a long wire to the ground.

Q: How to divide the analog ground and digital ground of the A/D converter to better reduce noise?

A: There is no consensus in the industry on whether the analog ground and digital ground need to be divided. Some are just one ground plane, while others are divided into two areas connected by short wires under the ADC. There are various methods. It is necessary to keep the analog and digital parts as separate as possible and keep a certain distance. The analog signal and the digital signal should not cross the routing, and the filter capacitor of the power supply should be as close to the chip as possible.