Mobile phone receiving channel noise figure test

In order to test the noise coefficient of mobile phone receivers, this article proposes two simple and practical methods, and discusses their advantages and disadvantages. One method is to use a separate spectrum analyzer for testing, which has low accuracy; the other method is to use the noise source of the noise tester for testing, and use the principle of cold and hot load testing noise coefficient to obtain more accurate measurement results.

Problem statement

The figure below is the receiving circuit of the reference design of the TD-SCDMA mobile phone radio frequency unit of MAXIM. The voltage gain of this channel is greater than 100dB, and the interface with the baseband unit is an analog I/Q signal. We need to measure the noise coefficient of this channel. Our existing noise test instrument is HP8970B, the lowest frequency that this instrument can measure is 10MHz, while the highest useful frequency component of the TD-SCDMA baseband I/Q signal is 640KHz. Obviously, this instrument cannot meet our measurement needs.

Below we will introduce two test solutions, discuss their test accuracy, and finally give actual test data for comparison.



The instrument connection for directly measuring the noise figure using a spectrum analyzer is shown in Figure 2, where the point-frequency signal source is used to calibrate the entire channel gain. The attenuator has two functions: one is to improve the front-end matching; the other is to calibrate the channel gain. Because the receiver gain is often very high, greater than 100dB, and some signal sources cannot output very weak signals, this function can be accomplished with the attenuator.

Measurement step 1, first use the signal source to generate a point frequency signal (usually we are interested in the noise coefficient of the receiver when the signal is small, so the point frequency signal level should be close to the sensitivity level at this time), the frequency point and the local oscillator signal are offset a little, so that a point frequency signal can be obtained at the baseband I/Q port, and the receiver channel gain is adjusted to make the amplitude of the I/Q end point frequency signal moderate. The channel gain at this time can be obtained by measuring the size of the point frequency signal at the input and output of the receiver, which is recorded as G;

Measurement step 2, follow step 1, turn off the signal source, keep all the settings of the receiver unchanged, and use the spectrum analyzer to measure the noise power spectral density of the I/Q port at the point frequency just now, the I port is recorded as Pncdensity (dBm/Hz), the Q port is recorded as Pnsdensity (dBm/Hz), then the receiving channel noise coefficient is given by the following formula:



In the above formula, kb represents the Boltzmann constant, F is the true value of the noise factor, we use NF to represent the logarithmic value of the noise factor, NF = 10lg (F), G represents the gain of the entire channel, T1 is the current thermodynamic temperature, T0 is equal to 290K. Assuming T1 = T0, it is easy to obtain the explicit expression of NF as follows:



Regarding the correctness of equations 2 and 3, we can make the following simple deduction. First consider the point frequency situation, assuming that the point frequency signal at the receiver input is:



Now consider the noise problem. To simplify the calculation, the current temperature is set to 290K, which is the standard temperature for defining the noise coefficient. According to the definition of the noise coefficient, we can equate the noise generated by the system to the input port. The sum of this noise and the available noise power should be equal to F times the available noise power. Next, we use a narrowband stationary Gaussian process to describe the sum of these two parts of noise. Assuming the noise bandwidth is 2B, the following equation gives some characteristics of the noise:



Comparing Equation 4 with Equation 7, and referring to Equations 5 and 6, we can obtain the noise expression at the output of the receiver:



Combining equation 8 with equation 7, we can directly get equation 2. Combining equation 9 with equation 7, we can directly get equation 3. Note that the noise bandwidth of the I and Q ports is B, which is half of the RF noise bandwidth. The following figure shows the noise transformation process in a more vivid way:



Measurement step 1: first connect the receiver to the point frequency signal source side, use the signal source to generate a point frequency signal with a sensitivity level (because we are usually interested in the noise coefficient of the receiver when the signal is small), and the frequency point is slightly offset from the local oscillator signal, so that a point frequency signal can be obtained at the baseband I/Q port. Adjust the receiver channel gain to make the I/Q end point frequency signal amplitude moderate;

measurement step 2: follow step 1, keep all the settings of the receiver unchanged, connect the receiver to the noise source side, set the noise source to cold state, set the cold state noise temperature to T1, and use a spectrum analyzer to measure the I port noise power spectrum density (I and Q have the same properties, so only the I port is mentioned here), recorded as Poc (dBm/Hz);

measurement step 3: follow step 2, keep the receiver settings unchanged, set the noise source to hot state, set the noise temperature to T2, and use a spectrum analyzer to measure the I port noise power spectrum density, recorded as Poh (dBm/Hz);

the Y in the so-called Y coefficient method is the ratio of the two measured values ​​in measurement step 3 and measurement step 2:



Assume the receiver equivalent noise temperature is Te. We can use the cold source noise temperature, hot source noise temperature, and receiver equivalent noise temperature to express the coefficient Y as follows:



Assuming the noise head excess noise ratio is ENR and the standard noise temperature is T0 (290K), the following equation can be obtained according to the definition of excess noise ratio:



According to the definition of noise coefficient and equivalent noise temperature, the following formula can be obtained:



By combining equations 11, 12, and 13, we can easily obtain the functional relationship of the noise coefficient with respect to ENR, Y, T1, and T0. Its logarithmic expression is as follows:



Generally, the cold noise temperature is close to the standard noise temperature. When the accuracy requirement is not high, it can be considered that T1 = T0, and the above formula can be simplified to:



In the above formula, Y is given by equation 10, which is an indirect measurement value, and ENR is given by the noise head. According to this equation, the receiver noise figure can be easily calculated.

Comparison of the advantages and disadvantages of the two test methods

Using method 1 to test the noise figure of the receiving channel of MAXIM's TD-SCDMA mobile phone, first use the point frequency signal to measure the channel gain. The input point frequency signal is -105.6dBm, the frequency is 2015.95MHz, the LNA and mixer of MAX2392 are set to high gain and high linearity, the VGC voltage is adjusted to 2.63V, and the local oscillator frequency is set to 2015.8MHz. At this time, we measured a 150KHz point frequency signal of -3.5dBm at the I output end, and calculated that the entire channel gain is 102.1dB. Now turn off the input frequency signal and use a spectrum analyzer to measure the noise power spectrum density of the I port at the 150KHz frequency point. The spectrum analyzer we use is the RS FSEA. To make the noise measurement result accurate, the detection mode is set to "SAMPLE", and then the "Maker Noise" function is used for testing. We measured the noise power spectrum density to be -63.5dBm/Hz. According to equation 2, the noise coefficient of the entire channel can be easily calculated as:



Method 2 is used to test the noise coefficient of the receiving channel of TD-SCDMA mobile phone of MAXIM Company. The above measurement is followed, and the working state of MAX2392 is kept unchanged. The noise power spectrum density at the 150KHz frequency point of the I port obtained in the above test is the noise power spectrum density of the cold noise source. Now we only need to measure the noise power spectrum density at this frequency point in the hot state. Here we use the NC346A noise head of Noise/Com Company, which has an ENR of 5.91dB at the 2G frequency point. Using the same test method as in Method 1, we measured that the noise power spectrum density at 150KHz in the hot state is 60.4dBm. According to equation 10, the Y coefficient can be calculated to be 3.1dB, and then according to equation 15, we can calculate the noise coefficient of the entire channel as:



Comparing the measurement results obtained by the above two methods, there is only a 0.3dB difference, and the test results are relatively ideal. Of the two methods, the second test method is more accurate because the spectrum analyzer may have errors when measuring the noise power spectrum density. The signal bandwidth and noise bandwidth of the intermediate frequency filter of the spectrum analyzer are generally not equal. Some spectrum analyzers will give a correction value, while others do not. If we do not consider the correction value, or the instrument does not make corrections in the readings, the noise power spectrum density we measured may have a deviation of about 1dB, resulting in a final noise factor deviation of about 1dB. If the second method is used for testing, because we only need to know the power spectrum density ratio of the cold and hot noise sources, even if the power spectrum density measured by the cold and hot noise sources has deviations, the ratio is still correct, thereby improving the noise measurement accuracy.