RoadTest: Agilent N9322C Spectrum Analyzer & TI CC11XLDK-868
Author: ciorga
Creation date:
Evaluation Type: Independent Products
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: Anritsu MS2713E and MS2830A
What were the biggest problems encountered?: I haven’t encountered any major problem during this road test; everything went on smoothly and both the spectrum analyzer and the RF evaluation kit worked as expected. So, if I have to pick something for this question I would pick the measurement of phase noise, which is not supported as built-in function in N9322C spectrum analyzer. It took me a while to setup a procedure for direct phase noise measurement using the available measurements and features of N9322C and I was happy to see that the procedure worked.
Detailed Review:
Overview
First I would like to express thanks to Keysight/Agilent and Texas Instruments for selecting me to review the N9322C spectrum analyzer and CC11XLDK-868-915 Value Line Development Kit. I feel fortunate and honored to be one of the three people selected for this roadtest. I would also like to thank Element14 for organizing this roadtest.
N9322C belongs to the product category of basic spectrum analyzers and covers frequencies from 9kHz up to 7GHz. Besides addressing the primary frequency domain measurements, with built-in options N9322C can be used for signal modulation analysis like AM demodulation and listening through a built-in speaker, FM demodulation and time-domain display of demodulated signal, and ASK/FSK modulation analysis that can display the symbols, waveform, error, and eye-diagram. With the reflection measurement option, N9322C can be used to characterize transmission line cables and antennas by measuring reflections, one-port cable loss, and distance to fault.
The Texas Instruments CC11XLDK-868-915 development kit comes with two SmartRF transceiver boards on which users can plug various modules that can be used to evaluate TI’s sub-1GHz products: CC110L, CC113L, CC115L (included in the kit), and more others like: transceivers (CC1121, CC1200, CC1120), transmitters (CC1150, CC1175), and system-on-chip (CC430, CC1110, CC1111, ). The accompanying SmartRF Studio 7 software (that can be downloaded from TI website) can be used with the CC11XLDK-868-915 development kit to control and test the RF performance of the TI’s sub-1GHz products mounted on the modules. The SmartRF Studio software works also with 2.4GHz system-on-chip products.
First Impression when Receiving the Roadtest Products
TI CC11XLDK-868-915 Value Line Development Kit
My TI Kit CC11XLDK-868-915 came first packaged in a small box (270x180x100mm as written on a label). Enthusiastic to start working on this roadtest, I opened the box and I took out the contents. Here is a picture of what was inside:
Included in the kit there was this instructions document on how to access the Texas Instruments website for technical support and downloading software:
Another useful document was a quick start guide, shown below, which helped me setup a transmitter/receiver link smoothly and in a quite short time.
Next step was to examine the boards, so I started with the two SmartRF TRXEB. The top side has a TI MSP430 microcontroller, a debug interface connector for the MSP430, an LCD display, four push buttons to navigate through the control menu displayed on the LCD, a USB connector, four LEDs, and multiple switches for selecting the functionality modes. On the upper right corner there are two connectors for plugging one of the evaluation modules (CC110L, CC113L, CC115L). Here is a picture of the top side:
On the bottom side there are two battery holders for AA type batteries. Batteries are included in the kit, but the boars operate without batteries if connected through a USB cable to a computer. Here is a picture of the bottom side of the board:
The kit comes also with a debug interface module for the MSP430 microcontroller, which I am showing in the picture below:
The debug interface can be connected to the SmartRF board through a supplied cable.
The four modules included in the kit are shown I the figure below:
One module has mounted a CC115L transmitter, another module has a CC113L receiver, and the other two modules have CC110L transceivers. Each module has an open pad for mounting an SMA connector which can be connected to an external antenna (instead of using the built-in antenna on the module PCB).
Agilent N9322C Spectrum Analyzer
The spectrum analyzer came in a large box with padding for protection. Here are some pictures of the N9322C spectrum analyzer after I took it out of the package:
The N9322C spectrum analyzer came with a certificate of calibration:
In the box I have also found two bags with accessories, which I am showing in the pictures below.
Besides the documentation CD, cables and adapters there was a N9311X-201 OSL precision mechanic calibrator, and a Diamond RH799 70 to 1000MHz wide-band antenna.
Short Competitive Analysis
Here is a comparison chart of some of the functions between Agilent/Keysight N9322C and Anritsu MS2713E and MS2830A products.
There are multiple other models and manufactures, but I selected these ones since I considered them in the same category of performance versus price range.
Setting up the TI Evaluation Boards
After unpacking the contents of the boxes I continued with setting up the TI kit. I plugged the CC115L transmitter module in one of the SmartRF boards and the CC113L receiver in the other SmartRF board. Before connecting the boards to the computer through the provided USB cables I went on TI website and I downloaded the SmartRF Studio 7 software and the USB drivers. First step on TI website was to create an account so I could then get access to the support documents. When I got to the support website for the CC11XLDK-868-915 kit I was impressed of how many technical documents are available to users. Here is a screenshot of only part of what they have:
And here is a screenshot of the SmartRF Studio download page:
My first attempt to install the drivers was not successful, but that was just my fault. After poking around a few times I noticed a switch on the SmartRF board pointing to UART (this was the default). I switched it the other way towards “SmartRF”, I plugged the USB cable in the computer, and the USB drivers got installed.
Next I installed the SmartRF software without any difficulties and I started it. Here is a screenshot showing the features of this software:
In learning how to use this program I found very useful the document “SmartRF Studio 7 Hands-on User Guide and Tutorial”
The SmartRF Studio 7 starting screen shows all the evaluation modules that can be controlled with this software, and out of those the ones that are connected to the computer are highlighted and listed at the bottom window:
Next I double clicked on CC113L and CC115L, which opened the control panels for these two modules. There are two modes of operation: an easy mode that has predefined register values and packet date and an expert mode that provides full control of the RF devices. Here is a screenshot of the transmitter and receiver control panels in easy mode operation:
And below is a screenshot of the transmitter and receiver control panels in expert mode of operation.
In expert mode there is a full description of each register and users can read and write any register.
Starting Experiments and Measurements
For my first measurement I have chosen to transmit an unmodulated continuous signal and measure it at receiver using the SmartRF signal measurement function, as I am showing in the screenshot below. Looking at the displayed receiver waveform, the level was at –120dBm before I started the transmitter and moved up to –28dBm after the transmitter started emit the RF signal.
For next measurement I connected the Diamond RH799 wide band antenna to the RF input of the N9322C spectrum analyzer. The antenna has 50Ohms impedance that matches the N9322C input impedance, and matches the characteristic impedance of the interconnect cable. The antenna captured the transmitted signal and sent it to the RF input of the N9322C spectrum analyzer. Here is a picture of this measurement setup:
Next I used the peak search function of the N9322C spectrum analyzer to find the amplitude and frequency of the peak TX carrier frequency. Here is a picture of the N9322C display showing the peak search/measure function:
On this screen we can see that the received carrier signal has a level of –21.78dBm and the frequency is 867.3MHz. The –21.78dBm is higher than the –28dBm signal level at the CC113L receiver on SmartRF board displayed on the control panel of the SmartRF Studio that I have shown in a picture above. The carrier frequency measurement of 867.3MHz is close to the 867.999939MHz carrier frequency programmed in the CC115L transmitter control panel.
Next, I wanted to see how the N9322C spectrum analyzer can show a demodulated signal, so I turned on the modulation in the CC115L control panel of SmartRF software, and I setup the TX to send a random packet in GFSK modulation format. Here is a screenshot of the CC115L control panel in this experiment:
Next I setup the N9322C spectrum analyzer to modulation analysis by pressing MODE -> Modulation Analysis -> FSK. In modulation analysis mode, the Agilent N9322C spectrum analyzer displayed the demodulated signal, as I am showing in the picture below:
Direct Measurement of TX Carrier Signal
After these measurements I took one of the CC110L transceiver modules and I installed an SMA connector in the available open pad, which allowed me to connect it to the N9322C spectrum analyzer directly through a coaxial cable. Besides mounting the SMA connector I had to move a capacitor to disconnect the on-board antenna and connect the CC110L output to the SMA connector. Here is a picture of the modified board.
I then replaced the CC115L TX module on the SmartRF board with this modified CC110L module and I setup the TX mode in continuous unmodulated transmission mode. With this setting the N9322C spectrum analyzer measures the carrier signal of the transmitter without any modulation. Here is a picture of this test bench setup:
Then I have setup the Agilent N9322C spectrum analyzer to measure the transmitter power by pressing Meas -> Channel Power buttons on the front panel. Here is a picture of the display showing the measured power.
The measured channel power was –0.04 dBm which is close to the transmitter power set in the SmartRF control panel of 0dBm. The measured power spectral density was –63.05 dBm./Hz.
Temperature Stability of TX Carrier Frequency
The next measurement in this series of characterizing the transmitter was to see how temperature affects the TX carrier frequency. Agilent N9322C spectrum analyzer has a built-in function that tracks the signal drift. Here is a picture of the display showing the spectrum monitor on the upper part of the screen.
After setting up this measurement I took a hair dryer and I heated up the transmitter (oscillator crystal and CC110L transceiver integrated circuit). With this heat source turned on, the TX carrier frequency started to shift gradually as I am showing in the following picture.
Notice the bending of the spectrum monitor red line that shows the frequency drift in time as the temperature of the TX module increased. I have also captured this frequency drift in a video that I have linked blow:
Measuring Phase Noise
Phase noise is a figure of merit of oscillators that relates to frequency stability. In transmitters, like the CC115L and CC110L that are part of the TI development kit, the phase noise can be used to quantify the stability of the carrier frequency. To further analyze the phase noise of CC110L carrier frequency, I have setup the experiment shown in the following figure.
In this experiment I have modified the CC110L module by disconnecting the built-in antenna and mounting an SMA connector for an external antenna or for connecting directly to the N9322C spectrum analyzer, like I am showing in the picture. Next step, I setup the transmitter mode in SmartRF Studio control panel to send only the carrier frequency unmodulated. After setting up this experiment and preparing to measure phase noise, I discovered that N9322C does not do phase noise measurement. I searched the user manual and then I moved my search on the Internet until I found this comparison chart of Agilent/Keysight Spectrum Analyzer Series.
But I didn't get discouraged and I continued to look into a method of measuring phase noise. So next I tried to use the direct spectrum technique, since N9322C has all the features I need for this measurement method. To start this method I setup the N9322C to display the frequency spectrum of the carrier signal, as I am showing in the following figure.
I have annotated on this picture what an ideal (noiseless) signal would look like. The spectrum of an ideal signal has only one spectral component at the carrier frequency value and nothing else. So intuitively, phase noise should characterize all the “deviations” of the real carrier frequency spectrum from the ideal signal spectrum. Since the deviations are symmetrical around the carrier frequency, phase noise uses only one side, also called sideband phase noise. The following figure shows the components of the spectrum used in measuring the phase noise.
The sideband phase noise at a frequency offset from carrier can be calculated as area of 1Hz bandwidth at that offset frequency divided by the total area under the curve. This is a little hard to compute from the measured spectrum without exporting the data and post processing it in Excel or Matlab (or equivalent), so instead I found out that I can use a simplified techique. The simplified technique calculates phase noise as the ratio between the noise power in a 1Hz bandwidth at the measured frequency offset from the carrier frequency (marked as fc+fm in the figure above) and the power of the carrier signal (marked as P_fc in the figure above). So the phase noise at a frequency offset of fm can be calculated as:
phase_noise(fm)=P_fm(dBm/Hz)-P_fc(dBm)
P_fc(dBm) can be easily read from the frequency spectrum measured with the N9322C spectrum analyzer using the peak-detect function or just a marker. In this experiment P_fc is equal to –14.33dBm. The challenge comes now from determining the P_fm(dBm/Hz), since I needed to read the level and then integrate over 1Hz bandwidth. So next I setup the N9322C spectrum analyzer in average mode to make easier the reading of level values. The picture below shows the average version of the same frequency spectrum in the above picture.
Then I turned on all six markers available in N9322C spectrum analyzer and I placed them on one side of the frequency spectrum. The spacing can be adjusted to the desired frequency offset where we are interested to measure phase noise. Here is a picture showing the six markers and the markers table at the bottom of the screen. The markers table is a nice feature available in N9322C spectrum analyzer since it makes easier the reading of all marker values.
Alternately we can use the channel power measurement function of N9322C spectrum analyzer, as shown in the picture below.
We can adjust the integrated bandwidth to a smaller value and shift the center frequency to measure the noise power over a selected bandwidth region, as I am showing in the picture below.
We can narrow the integration bandwidth down to 100Hz (it doesn't go to 1Hz; 100Hz is minimum), as I am showing in the picture below.
Then we can measure the noise power at any desired offset from the carrier frequency by sweeping the center frequency, as I am showing in the following two pictures.
For each frequency offset from the carrier we compute the phase noise using the formula: Phase_noise(fm)=P_fm(dBm/Hz)-P_fc(dBm), as described above.
So the N9322C spectrum analyzer can be used to determine the carrier power. P_fc(dBm) and P_fm(dBm) at any offset frequency from the carrier. For simplicity I assumed that the marker read values are constant over 1Hz bandwidth, so this simplified the integration step; it may introduce some error but it allows me to calculate the phase noise easily only from values read directly on the Agilent/Keysight N9322C spectrum analyzer.
Measurement of TX Spectral Spurs
Ideal oscillators would have only one frequency spectral component; however, real oscillators have additional frequency components named spurs. A lot of effort is made during design to implement techniques to suppress spurs. The measured frequency spectrum of the CL110L on the Texas Instrument evaluation board does not show any visible spurs. However, I wanted to further analyze the spectrum and search for spurs. I started by first identifying the mechanisms of generating spurs from the schematic diagram of the evaluation board, and I selected to look into modulation type coupling of the local oscillator into the carrier frequency, and local oscillator feed-through. The following figure shows the measured frequency spectrum of the CL110L with my annotations of the locations where I could find the local oscillator feed through component (at 26MHz) and the modulation coupling spurious components (carrier frequency + 26MHz, and carrier frequency – 26MHz).
The picture below shows the local oscillator feed through into the output signal of the CL110L.
So indeed I found a frequency component at the local oscillator frequency value. The amplitude is very small, -100.6 dBm, (so small that it does not create any significant issue), but finding it matches my expectations that there is some feed-through from the 26MHz local oscillator. Also, I liked that the N9322C spectrum analyzer was enough sensitive to detect this component.
Next I looked for modulation type spurs from the local oscillator. I expected these spurs at two locations left and right from the carrier signal and separated by 26MHz offset. The picture below shows the spur at 893.97MHz, which is equal to the carrier frequency of 867.97MHz + 26MHz.
The spur amplitude is very small, -86.93dBm, but I was happy to find it where I expected based on my analysis of the evaluation board circuit. The spur on the other side of the carrier, at 867.97MHz - 26MHz = 841.97MHz, is shown in the following figure.
The amplitude of this spur is –86.79dBm,which is close to the other spur (we expect these two spurs to have the same amplitude).
Doppler Shift
Doppler shift is a phenomenon that results in frequency shift at the receiver when the transmitter moves. Some of you may have noticed or hear in a movie the whistle of a train approaching, passing by, and then moving away. The sound of the whistle has a higher frequency pitch when traveling towards us and a lower frequency pitch when moving away from us. This is due to the phenomenon called Doppler frequency shift or Doppler effect.
So I wanted to use the Agilent/Keysight N9322C spectrum analyzer to try to measure the Doppler frequency shift of the RF signal transmitted by the TI CC115L module mounted on the SmartRF evaluation board. This board is small and easy to manipulate, so it makes a good candidate for picking up in my hand and moving it while observing the transmitted frequency spectrum with the N9322C spectrum analyzer.
The experiment was successful in detecting the Doppler frequency shift, and I have captured it in the video below. As I move the CC115L transmitter towards the receiving antenna, the carrier spectral component moves towards the right side on the N9322C screen, which shows that the received frequency has increased. This is equivalent to the train whistle having higher pitch when approaching. Then, when I move the CC115L transmitter away from the receiving antenna, the carrier spectral component moves towards the left side of the N9322C screen, which shows that the received frequency has decreased. This would be similar to the train whistle lowering the pitch when moving away.
Analysis of GFSK Demodulated Data Packets
I first setup a communication link between one SmartRF board with CC115L transmitter module and one SmartRF board with CC113L receiver of the TI CC11XLDK-868-915 kit. In the SmartRF Studio control panel I selected GFSK modulation format, packet data size 30 (28+2), and packet count infinite. Next I connected the Diamond RH799 wide bandwidth antenna to the N9322C spectrum analyzer and I looked at the demodulated signal. The modulation analysis function built in the spectrum analyzer can display the demodulated waveform in time domain, in eye-diagram format, and also in demodulated packet data symbols. The picture below shows the demodulated waveform:
We can see in this picture the pulses corresponding to logic “1” and logic “0” symbols. They are shifted either down to –5.157kHz level on y-axis or up to +5.157kHz level. The same data can be represented in eye-diagram format, as I am showing in the picture below:
What bothered me was the fact that I saw demodulation errors reported at the bottom of the screen, like FSK Error and Mag Error. I decided to spend some time trying to understand why these errors are reported there. On the SmartRF Studio control panel all the packets were received correctly without any error. I tried to adjust the settings of the spectrum analyzer demodulation parameters, but I couldn’t make the errors disappear. However, from all these tries I noticed that the demodulated waveform sometimes stops or has glitches, and I assume that the transmitter does not send data continuously, which may “trick” the demodulator inside the N9322C spectrum analyzer. But why the SmartRF Studio showed that all packets have been received correctly without errors? My assumed explanation for this is that SmartRF Studio controls both transmitter and receiver boards, and knows when a packet is transmitted and when it stopped or when there are preambles or transitions that create demodulated glitches. Knowing this information SmartRF Studio may process only the valid packets and may ignore all the transient stop/start/glitches when checking for errors.
To further understand if my assumption is correct I setup the N9322C spectrum analyzer to display the demodulated packet data. This is a nice feature of the built-in modulation analysis function. So what I noticed was that when I looked at multiple data packets the symbol values were wrong, either all “1” or all “0”; however, if I looked at only one data packet at a time the bits were correct. The following picture shows one demodulated data packet of 30 symbols displayed by the N9322C spectrum analyzer.
I think this may happen because of the stop/start/glitches that I saw in the waveform between data packets, which may be taken care of in SmartRF Studio since it knows when the transmitter sends valid data, but not in the spectrum analyzer which does not know what the transmitter is controlled to do at each moment in time.
N9322C Reflection Measurement Calibration
In this experiment I mounted an SMA connector on one of the CC110L transceiver module and I connected an external antenna using three combinations of different RG58 cable lengths. The purpose was to check the effects of the cables and connectors on the frequency spectrum measured with Agilent/Keysight N9322C spectrum analyzer and the degradation of bit error rate (BER).
But first I wanted to measure the losses of these three cables, so I used the built-in tracking generator and reflection measurement function of the Agilent/Keysight N9322C spectrum analyzer. I turned on the spectrum analyzer and I set it in the reflection measurement operation mode. Then I selected the frequency range for these measurements around the carrier frequency of the transmitter, from 824MHz to 960MHz (the carrier frequency was 867MHz). Then I looked at the next screen, shown in the picture below, and I noticed the word “Uncalibrated” displayed in the upper left corner, which basically told me that I need to perform a calibration procedure before measuring the cable loss.
The N9322C spectrum analyzer was shipped with a calibration “T” shown in the picture above. This calibration device has three ports with three different input impedances. One port has infinite impedance “open”, another port has zero Ohm impedance “short”, and the other port has 50 Ohms impedance "load". So I started the calibration process by pressing the “Calibrate” button on the N9322C side buttons. Then I really liked how the N9322C spectrum analyzer guided me through the entire calibration procedure. First N9322C prompted me (and also showed me how to) connect the calibration device “open” port to the TG source output, as I am showing in the following picture.
Next I pressed “Enter” and the spectrum analyzer performed the calibration step in about 1-2 seconds. Then it prompted me to connect the calibration device “short” port.
I connected it and then I pressed “Enter” and the spectrum analyzer performed the “short” calibration step. After that it prompted me to connect the 50Ohms load port of the calibration device.
After this calibration step the spectrum analyzer went back to the reflection measurement screen, but this time it displayed “Calibrated” in the upper left corner and it also displayed the calibration frequency range.
Measurement of Return Loss of Cables and Antennas
I used three cables: cable1 was a single piece about 3ft long of RG58 coaxial cable, cable 2 was made out of two cable1 pieces connected with a coaxial adapter, and cable 3 was made out of three pieces of the same type of cables connected together. Besides these cables I also wanted to measure the return loss of the Diamond RH799 wideband antenna that came with the spectrum analyzer and another “home-made” antenna that I am showing in the figure below.
This antenna is made out of a simple wire sized at lambda/4 of the TX carrier frequency.
The cable loss measurements are shown in the following pictures.
In these measurements I used the marker function of N9322C to measure the return loss at the TX carrier frequency, as I have annotated in the figures above. Notice also more resonant peaks for cables 2 and 3 compared to cable 1. These additional resonance peaks are due the mismatch in characteristic impedance of coaxial adapters used to connect the individual cables together.
The following two pictures show the return loss measurements for the two antennas that I mentioned above, the wideband (Diamond RH799) and my home-made wire antenna.
Of course as I expected the Diamond RH799 performs better, but mine looked good too.
Measurement of VSWR (Voltage Standing Wave Ratio)
The Agilent/Keysight N9322C spectrum analyzer can also measure one-port insertion loss and Voltage Standing Wave Ratio (VSWR). VSWR is a parameter that measures how well the cable + antenna characteristic impedance is matched to the expected 50 Ohms of the transmitter output impedance. An ideal impedance match produces a VSWR = 1 and in non-ideal cases VSWR has values larger than 1. So first I measured the VSWR of the 50 Ohms load as I am showing in the picture below.
The measurement window shows the peak and valley values of the VSWR over the measurement frequency range, and we can see that both reported VSWR values equal to 1. Next I measured the VSWR of each of the three cables, and the results are shown in the pictures below.
From the measurement window at the bottom of the screen we can see that cable 1 has VSWR values between 1.05 and 1.34, cable 2 has VSWR between 1.02 and 1.37, and cable 3 has VSWR between 1.04 and 1.48. So the minimum values don’t vary significantly among these three cables, but the maximum values increase sequentially: 1.34, 1.37, 1.48. As shown annotated on the pictures, for each cable I used the built-in marker to measure the VSWR at the carrier frequency of 867MHz. The VSWR values at carrier frequency do not vary significantly among the three cables.
Connecting an External Antenna to the Transmitter
In the next experiment I connected my home-made antenna to the CC110L transceiver in four setups: first directly without any cable, second through cable 1, third through cable 2, and fourth through cable 3. In each case I used the SmartRF Studio software to setup a communication link that transmitted continuously data packets in FSK modulation mode, and I studied the effect of cables loss on the signal received by CC113L and on the reported bit error rate (BER).
First I connected my home-made antenna directly to the CC110L transceiver SMA connector. Here is a picture of the measurement setup followed by a screenshot of the SmartRF Studio control panels.
So the average Return Signal Strength Indication (RSSI) at the receiver is –73.4dBm and there are no bit error rate (BER) errors; all data packets are received correctly.
Next I connected my home-made antenna through cable 1. The picture below shows the experiment setup and N9322C measured spectrum followed by a screenshot of the SmartRF Studio control panels.
The average RSSI at the receiver is –54.5dBm, which surprisingly came higher than the antenna only case above. I suspect there might be some resonance created by the combination of SMA mounting parasitics and mismatch into antenna when I connect the antenna directly without cable. Before thinking more about this discrepancy I wanted to see what happens with the other two cables. The picture below shows the experiment setup of the antenna connected through cable 2 followed by a screenshot of the SmartRF Studio control panels.
The average RSSI dropped to –64.8dBm, which makes sense since I expect cable 2 to introduce more reflection losses than cable 1. Next I inserted cable 3. The following picture shows the setup and N9322C measured spectrum followed by a screenshot of SmartRF Studio control panels.
The average RSSI is –64.4dBm, almost at the same value as for cable 2. I expected cable 3 to introduce more losses than cable 2 but it did not happen. My only explanation is that there are multiple factors involved here and very possibly not the length of the cable is the dominant factor but the impedance discontinuity of connectors on the cable. Cable 1, which is just one piece, shows lower losses that cables 2 and 3, which are made out of multiple pieces connected together with coaxial adapters.
In addition to checking the average RSSI level I also checked the BER (bit error rate), and in all cases the BER was zero, which means that all the data packets came to the receiver without any error. Further more, I wanted to see if I could make BER to show errors, so I removed the antenna completely from the transmitter. The picture below shows this experiment and it is followed by a screenshot of SmartRF control panels.
The RSSI level dropped to –93.6dBm and I waited until 2200 data packets have been sent, but BER was still zero. So I couldn’t make BER fail even after I removed the antenna from the transmitter. My explanation is that the TX and RX SmartRF boards were quite close to each other on the test bench and the radiation level was enough to transmit a signal to the receiver. The receiver may also have a high sensitivity contributing to this. I expect that if I move the receiver in a different room I would be able to make the BER fail.
Distance to Fault Measurement
A nice feature built in the Agilent/Keysight N9322C spectrum analyzer is the distance to fault measurement, which basically measures the distance to characteristic impedance discontinuities along the transmission line cable (typically coaxial type but it should work with PCB traces and basically any type of transmission line). This is a great feature of N9322C since typically these types of measurements are done with Time Domain Reflectometry (TDR) functions of expensive oscilloscopes. The physical principle is to send a signal through the measured cable and analyze the time that it takes to receive back reflected parts of that signal. Based on this time and the velocity of signal through the cable the instrument can calculate the distance to the impedance discontinuity that generated the partial reflection. My next experiment uses the distance to fault feature of the Agilent/Keysight N9322C spectrum analyzer to measure a "compound" cable made of the short gray cable that came with the spectrum analyzer and two 3ft long RG58 cables connected together. But first I needed to recalibrate the spectrum analyzer for a wider range of frequencies, 5MHz to 7GHz, following the same procedure as for the reflections measurements. Next I connected the "compound" cable to the TG Source connector on the front panel of N9322C, and I selected the distance-to-fault measurement mode. Here is a picture of this experiment setup with my annotations in pink.
So my cable was made out of three segments: first the short cable that came with the spectrum analyzer and which measures about 24cm, second a piece of RG58 cable of 3 feet length (about 91cm), and the third another piece of RG58 cable. These cables have around 50Ohms characteristic impedance, but the adapters that connect them together introduce impedance discontinuities. These impedance discontinuities generate partial reflections as the signal travels through the cables. The partially reflected signals travel back to the spectrum analyzer, which computes the distance to the reflection point. If we look at the trace on N9322C screen starting from the left we see two spikes that I have encircled with pink color. The first spike represents the reflection from the first interconnect on the cable and the second spike represents the reflection from the second interconnect, as I have annotated with the two pink arrows. To measure the distance I then used the marker feature of the N9322C spectrum analyzer, and I placed one marker on the first spike and another marker on the second spike. Then at the bottom of the screen I could read the distance to first marker as 0.24m and the distance to the second marker as 1.15m. The picture below shows a magnified view of the N9322C screen.
The first measurement of 0.24m matches the length of the short gray cable of 24cm. The second measurement of 1.15m matches the sum of the lengths of the short cable and one piece of RG58 cable: 91cm + 24cm = 115cm. This was just an experiment where I knew the location of the impedance discontinuities, and the measurement results proved that the distance-to-fault measurement using N9322C can be used to determine the locations of the impedance discontinuities. The distance-to-fault function of N9322C becomes very useful in the design and troubleshooting of transmission lines since it can detect otherwise hard to find impedance discontinuities issues.
Measurement of Electromagnetic Interference
In this experiment I setup a communication link between CC115L transmitter and CC113L receiver and I injected electromagnetic interference from a cellular phone, since the operation frequencies are quite close. The test bench setup is shown in the picture below.
The CC115L transmitter is configured to send FSK modulated data packets and the N9322C spectrum analyzer measures the eye diagram. There were no visible changes in the eye diagram when I used the cell phone. Next I looked at the frequency spectrum with and without the cell phone activity. I captured the changes in frequency spectrum in the video below.
So before calling the cell phone the N9322C spectrum analyzer showed only the carrier frequency of the CC115L transmitter, and after calling the cell phone we could see another spectral component generated by the cell phone signal. I then tried to bring the cell phone closer to the transmitter and receiver to see if I get any BER errors, but in all settings the BER remained zero.
Effects of Power Supply Noise on Transmission Link
In this experiment I injected noise into the power supply of the CC115L transmitter and I observed the effects on the transmitted frequency spectrum, on the received RSSI at the receiver, and on BER (bit error rate). The following picture shows the 3.3V power supply voltage with injected single tone sinusoidal signal of 14.7kHz frequency.
The injected signal into power supply couples into the CC115L transmitter generating FM modulation type sidebands on the transmitted signal, as I am showing in the following picture.
Notice the carrier spectral component in the center at 867MHz and two sideband components equally spaced on the left and on the right. We can use markers to measure the spacing between the TX carrier and the sidebands, as I am showing in the following picture.
So we measure 16.5kHz spacing from the TX carrier to the sidebands. This is the frequency of the injected signal into the power supply (it has been measured as 14.7kHz with the oscilloscope not 16.5kHz, but close enough). Think now that we troubleshoot the TX signal in a system using a spectrum analyzer and we find “unwanted” sidebands like these. We don’t know where they come from and we want to find out, so using the spectrum analyzer we can measure the frequency spacing from carrier and then look at the schematics of the system and see what part of the circuit operates at that frequency. This would be the first step in troubleshooting when we identify the source of the problem. Further troubleshooting steps can be performed to identify the coupling mechanism by watching the level and frequency of sidebands on the spectrum analyzer while changing various functionality parameters of the system. The video below shows how the frequency offset of the CC115L TX sidebands changes while sweeping the frequency of the injected signal into the power supply.
The amplitude of the injected tone into the TX power supply in the experiment above was 50mV. For injected tones with higher amplitude or for lower noise coupling suppression paths other sidebands at twice the injected signal frequency appear. The picture below shows the spectral sidebands after increasing the amplitude of the supply-injected tone to 150mV.
The video below shows how sidebands appear and change location as I sweep the amplitude and frequency of the injected tone into the power supply of the CC115L transmitter.
Next I looked at the effects of supply noise on the quality of the transmission link between the CC115L transmitter and CC113L receiver on the SmartRF boards. So I have setup the experiment shown in the picture below.
I setup the SmartRF control panels to send GFSK data packets continuously from the transmitter to the receiver. Unfortunately the picture does not show the text on the SmartRF control panel, but in that corner we should see the bit error rate value. Notice on the spectrum analyzer screen multiple sidebands generated by the injected tone into the CC115L transmitter power supply. With this experiment setup I started to sweep the frequency of the injected tone while watching the bit error rate BER. So I found out that as I reduced the frequency of the injected tone and the sidebands came closer to the carrier frequency at some point I started to see BER failures. The picture below shows the frequency spectrum when I started to see BER errors.
Notice that the sidebands came so close to the carrier to the point where they overlapped. The screenshot below shows the BER on the SmartRF control panel when it started to show failures.
Not all the packets failed, some just went through while others failed. I have captured this experiment in the video below.
So in this video we could see the carrier spectral component “jumping” left and right due to the GFSK modulation, and we could also see how the side bands started to approach this carrier as I reduced the frequency of the tone signal injected into the power supply.
This concludes my roadtest review of the Agilent/Keysight N9322C Spectrum Analyzer and the Texas Instruments CC11XLDK-868-915 Value Line Development Kit. Both are great products that I recommend to anyone interested in the RF communication field. There is a lot of learning that can be done from all the experiments that can be configured on the TI development kit, and the N9322C spectrum analyzer can help in this learning and also in daily activities of design and troubleshooting of RF communication systems. I enjoyed all this work and I would like to express my thanks to Keysight/Agilent and Texas Instruments for selecting me for this roadtest.
Best Wishes,
Cosmin
Top Comments
Hi Ingit,
Thank you for your nice comments about my review,
Best Wishes,
Cosmin