Design and Construction of FM Transmitter and Receiver Final Report David Chen Abstract: Our FM transmitter and receiver are built with discrete analog components and integrated on two circuit boards. The modulation scheme uses a superheterodyne setup, in which the intermediate carrier is at 300-kHz and the transmission carrier is at 24.3-MHz. Using 711-mW of DC power, the transmitter outputs a 5-dBm signal centered at 24.3-MHz with 100-kHz bandwidth. Using about half as much power, around 342-mW, the receiver can detect incoming signals at powers as low as -110 dBm. This receptivity level translates to successful audio reception at distances of almost 2-km from the transmitting antenna atop Packard. Design Theory: Transmitter: The transmitter is designed to take a signal in the audio range (20-Hz – 20-kHz) and prepare it for transmission through the air. A signal path for the transmitter is depicted in Figure 1. Two modulation stages perform the up-conversion. The VCO (implemented through the LM566) converts the base-band signal into the frequency of a square wave. With only a DC input, the VCO is set to output at exactly 300-kHz. To control both the DC level and the maximum AC variation at the VCO input, an audio amplifier is placed between the VCO and the audio source. Following the VCO, the mixer up-converts the VCO output to the transmission frequency at 24.3-MHz. A crystal provides the mixer with its 24-MHz local oscillator reference. The mixer output can technically be used for transmission, but it is generally too weak to be sent far. A power amplifier that provides 20-dB of power gain is placed between the mixer and the antenna to boost the actual output. Since power is the primary concern at this stage, impedance matching between the mixer and the power amplifier is needed to minimize transfer loss. With adequate output power, this transmitter is able to send signals out to decent distances.
Figure 1. Stages of FM transmitter.
Receiver: The receiver is designed to work with the transmitter. Specifically, the receiver must pick up a noisy and attenuated transmission, selectively filter and amplify signals in the spectral regions around 24.3-MHz and 300-kHz, and ultimately demodulate the transmitted signal back to a recognizable audio signal. This process is illustrated by a block diagram in Figure 2. The signal off the antenna first enters the LNA. According to Friis’ Equation, noise through this first stage has the greatest impact on the system’s SNR. Thus, much care is taken to ensure that the LNA has the best SNR of all stages in the receiver, through a combination of high gain and low noise figure. The signal is then sent into the mixer, which down-converts the signal to an intermediate frequency of 300-kHz. Filtering and adding gain is much easier at this lower carrier frequency than at 24.3-MHz. At this point, a second-order Butterworth band-pass filter, centered at 300kHz with 100-kHz bandwidth, is used to attenuate noise far away from 300-kHz. Two noninverting amplifiers together provide 40-dB of gain. Unfortunately, the filter does not limit noise around 300-kHz and the amplifiers introduce noise of their own, so the IF stage will not boost the SNR as much as the LNA does. The gain provided by the IF stage, though, does help send a stronger signal to the PLL. If the PLL can lock onto the modulating signal embedded in the frequency of its input, the PLL will output the demodulated audio signal. With some additional gain and crude low-pass filtering from the audio amplifier, we should be able to hear the transmitted signal on a speaker.
Figure 2. Stages of FM receiver. Characterization of Transmitter: Overall Characteristics: System-level characteristics for the transmitter are summarized in Tables 1 and 2. The transmitter is able to output near 5-dBm power across a 50- Ω load. Since the antenna has approximately 50- Ω impedance (as measured on the network analyzer and SWR meter), almost all of the 5-dBm is transmitted. We must differentiate between the system’s input and output bandwidth, because the VCO translates voltages into frequency. The input bandwidth was found to be all of the audio range (20-Hz – 20-kHz); all signals within this range could be amplified by the audio amplifier and then converted by the VCO. The lower bound of 20-Hz is tight because DC blocking capacitors prevent operation closer to DC, but the upper bound of 20-kHz is much looser. The output bandwidth, on the other hand, is limited by the VCO. When the VCO is centered at 300-kHz, its square wave output can swing between 250-kHz and 350-kHz. Thus, the
2
output bandwidth is 100-kHz (listed as the system bandwidth in Table 1) and remains unchanged going through the mixer and power amplifier stages. In order to output at 5-dBm, the system consumes 711-mW of power. At this rate of usage, a 200- mA ⋅ hr 9-V battery can be exhausted in 2.5-hours. Power consumption for each stage is listed in Table 2. Clearly, the power amplifier uses the bulk of the power, a necessary but seemingly wasteful drain. Distortion never became a serious issue with the transmitter, as long as the audio source was limited to 200-mV amplitude and the VCO input was limited to 500-mV amplitude. These constraints kept the audio amplifier THD less than 5% and the VCO output stably within 100kHz of 300-kHz center frequency. Output Power 9-V Battery Life 2.5 hours = 150 minutes 5-dBm (across 50- Ω load) System Bandwidth Transmittable Audio Range 100-kHz (3-dB cutoffs) 20-Hz – 20-kHz Table 1. Key figures of merit for transmitter. Overall 711-mW VCO 54-mW
Audio Amplifier 4.5-V Regulator 27-mW 36-mW Mixer Power Amplifier 54-mW 540-mW Table 2. Power consumption of transmitter blocks.
Audio Amplifier: The function of the audio amplifier is two-fold: 1) to bias the DC level of the VCO input at 7-V so there is maximum frequency swing on both sides of 300-kHz, and 2) to limit the AC amplitude of the VCO input to 500-mV. Proper DC biasing is accomplished using a 10- kΩ potentiometer acting as a voltage divider. Since the largest audio signal we applied had a 200mV amplitude, we chose to set the gain of the amplifier at 3-dB. VCO:
Figure 3. VCO frequency vs. input voltage. 3
Figure 4. VCO output for 400-Hz input.
We used a timing capacitor of 100-pF and a potentiometer of 10- kΩ to shift the VCO center frequency to 300-kHz. Other values of capacitance and resistance could also be used to achieve the 300-kHz center, but they caused nonlinear frequency shifts at the VCO input in response to linear changes at the input voltage, as discovered in Lab 2. Figure 3 shows the frequency behavior of the VCO through a range of input voltages. Based on this test, we chose a DC bias of 7-V at the input to take full advantage of the linear region. The frequency-to-voltage sensitivity k of the VCO is measured to be 67-kHz/V, lower than predicted. A typical input (400Hz 400-mV amplitude sinusoid) produces the output depicted in Figure 4, where we can see the bandwidth is close to 100-kHz and the peak output power is -10-dBm. Mixer: The mixer correctly up-converted the 300-kHz signal from the VCO to 24.3-MHz (see depiction in Figure 5). Unfortunately, in the process, the mixer did not provide the 14-dB of conversion gain promised by the specification sheet. Instead, the mixer attenuated the signal by 2-dB. We were careful to select coupling capacitors at the input and output to represent low impedance at 300-kHz and 24.3-MHz, respectively, but beyond that, we could had no other design control over the mixer circuit. We accepted this unexpected loss and focused our energies on the power amplifier.
Figure 5. Mixer output for 300-kHz, -10-dBm input. Figure 6. Network joining mixer and PA. Power Amplifier: The power amplifier is designed to supply up to 20-dB of gain to compensate for the weak mixer output. Other than the basic Gali-5 amplifier, there is an RF choke branch consisting of a 30- µH inductance in series with a 67- Ω resistance (multiple discrete components had to be used to achieve these values). To ensure maximum power transfer between the mixer and the power amplifier, the matching network in Figure 6 was used. The network transforms the output impedance of the mixer (including a 1.8-nF coupling capacitor) into approximately 50- Ω . Additional matching was unnecessary, since both the input and output impedances of the Gali-5 were measured on the network analyzer to be very close to 50- Ω .
4
Characterization of Receiver: Overall Characteristics: System-level characteristics for the receiver are summarized in Tables 3 and 4. The system gain is 62-dB, consisting of 22-dB from the LNA and 20-dB from each of the IF amplifiers. Considering the exact system bandwidth is difficult because there are multiple blocks with differing frequency responses, such as the matching networks favoring 24.3-MHz versus the IF band-pass filter favoring 300-kHz. We choose the most unambiguous bandwidth constraint: the 200-kHz 3-dB bandwidth of the IF filter around 300-kHz center. The minimum detectable signal of -110-dBm was measured by alternating between two single-tone inputs (400-Hz and 1kHz), lowering the power gradually, and subsequently determining when we could no longer hear the alternating tones nor see their peaks on the spectrum analyzer. Unlike the transmitter, the receiver is a mobile device, so battery life becomes a greater concern. We were able to limit power consumption to 342-mW, translating to 5.3-hours of usage on a 200- mA ⋅ hr 9-V battery. The LNA uses a large fraction of the power, but this is a worthwhile investment to obtain high gain (and thus high SNR) through arguably the most important block on the receiver. System Gain 62-dB System Bandwidth 200-kHz
Overall 342-mW IF Amplifiers 72-mW Audio Amplifier 36-mW
9V Battery Life 5.3 hours = 318 minutes Minimum Detectable Signal -110-dBm Table 3. Key figures of merit for receiver. LNA 90-mW Mixer 36-mW
4.5-V Regulator 36-mW PLL 72-mW
Table 4. Power consumption of receiver blocks. LNA:
Figure 7. LNA using shunt-shunt feedback.
Figure 8. Matching networks for LNA and mixer.
5
Power Gain 22-dB Signal-to-Noise Ratio 15-dB IP3 Point -5-dBm (measured at input) Output Impedance (45-9j)- Ω (after matching)
Noise Figure 7-dB 1-dB Compression Point -15-dBm (measured at input) Input Impedance (43+6j)- Ω (after matching) Maximum Receivable Distance 2-km (west side of Escondido Village) Table 5. Key figures of merit for LNA.
The importance of the LNA to the overall functionality of our receiver cannot be overstated. This stage has the highest SNR of all stages and can potentially have the greatest impact on overall system noise according to Friis’ equation. Figure 7 shows that we used the shunt-shunt feedback topology presented in lecture. The collector and emitter resistances are designed to draw 9-mA of current through the transistor. In actuality, the entire LNA circuit is using close to 10-mA. Bias resistances of 10- kΩ are chosen to draw a current much smaller than the collector current and much larger than the base current. Originally, the feedback resistance was set at 100- kΩ , but we experimentally found that 2.2- MΩ moved the collector current and bias voltages closer to desired levels. The feedback capacitance we will talk about shortly. This circuit is able to provide 22-dB of gain with a noise figure of 7-dB, resulting in an SNR of 15dB, which is much higher than the SNR of subsequent stages. One of the most time-consuming and baffling challenges we overcame this quarter was the matching of LNA’s input and output impedances to 50- Ω . The problem was made difficult by the severe coupling between the LNA’s input and output. As soon as we matched the input close to 50- Ω and started matching the output, we destroyed the input match, and vice versa. We reached a point where we could match either port with accuracy but could not match both simultaneously. It was only when we discovered the main culprit for the strong coupling, the feedback capacitor, that we made progress. Our original feedback capacitance of 20-pF was too large. When the 20-pF was replaced with the current 5-pF, the coupling minimized significantly, to the extent that we could independently match the input and output. The final input and output impedances are (43+6j)- Ω and (45-9j)- Ω using the networks shown in Figure 8. These values are very sensitive to parasitic effects, so our wiring and construction had to follow stringent standards. Mixer: Again, the mixer failed to provide the expected 14-dB conversion gain, but fortunately it did not attenuate the signal passing through it. Both of the inputs are matched to 50- Ω as shown in Figure 8. The output of the mixer terminates into the first IF amplifier’s input, which has very large impedance, so matching in that case did not make much sense. The linearity characteristics of this stage are -10-dBm for 1-dB compression and -13-dBm for IP3. IF Filter and Amplifier: The passive filter in the IF section provides frequency selectivity of 200-kHz around 300kHz center, as depicted in Figure 9. Unfortunately, the filter cannot remove noise in the passband. Each amplifier in this stage contributes 20-dB gain at 300-kHz but also raises the noise
6
floor by 10-dB, resulting in an SNR of only 10-dB. This indicates that the IF stage is not as efficient as the LNA at amplifying the signal without introducing more noise.
Figure 9. Frequency response of IF band-pass filter. PLL: Minimum Detectable Signal Lock Range -50-dBm 60-kHz – 480-kHz Center Frequency Sensitivity 300-kHz 6-kHz/V Linearity 0.08% total harmonic distortion Table 6. Key figures of merit for PLL. After determining the correct bias components, the PLL did not present any problems in the latter integration steps. Its minimum detectable signal has -50-dBm power. Thus, given a weak -110-dBm on the antenna and assuming 62-dB gain prior to the PLL, the -110-dBm signal can be demodulated. Lock range was found to be 60-kHz – 480-kHz, slightly larger than predicted. The implication of this discrepancy is that the PLL will lock onto more signals around 300-kHz but will also introduce more noise into the final output. Nonlinearity of the PLL is negligible given the very small total harmonic distortion. Audio Amplifier: The audio amplifier between the PLL and the speaker is necessary for two reasons: 1) the PLL output may be too weak to be heard clearly on the speaker, and 2) the PLL output contains a leakage of the 300-kHz signal from the input, resulting in high-frequency perturbations on top of an otherwise clean audio signal. To the PLL output of -15-dBm power, we added 6-dB of gain through the audio amplifier. To eliminate the 300-kHz leakage, we used a low-pass filter with a 20-kHz cutoff in the feedback path of the amplifier. The low-pass filter resulted in remarkably clean audio outputs.
7
Field Tests: With the transmitter and receiver both functioning, we conducted three field tests, the last of which was the class field test on SPAM Day. First Walk: My partner went out in the evening, after the rain subsided. Our transmitter was connected to the roof antenna and sent out a simple music piece. He was able to get good signal reception as far as Encina Hall (less than 1-km). Along the way, he noticed that dead spots occurred most frequently in the presence of trees. Sometimes, standing in front of tall buildings helped reception, probably because some signals bounced off the building in the direction of the receiver. One problem that plagued the latter half of his tests was a faulty batter clip, causing the receiver to be powered only intermittently. Second Walk: Both my partner and I participated in this second field test. We went in a different direction this time, towards the lake. As the elevation beneath our feet increased, the reception improved because there was a clearer line of sight between our receiver and the antenna atop Packard. In particular, at the lake, we could pick up a decent signal. Trees often disrupted reception, while building faces again helped reception. In fact, one of the best long-range sites was in front of a music building which was even a few hundred meters further than the lake. Having lost reception in a forested region, we headed back towards the center of campus. By the time we reached White Plaza, we regained a clear audio signal. For the rest of the trip, we walked down Serra Street and had the good fortune of being able to receive audio as far as the Serra-El Camino intersection (about 2-km). There were dead spots along the way and the SNR degraded with distance, but the audio signal was still discernible on El Camino. Further on El Camino in the direction of California Avenue, we lost our signal but picked up some interesting excerpts from talk shows. These stray transmissions were especially strong near lamp posts and other large metal poles. Our audio signal returned briefly in front of a hotel near Starbucks, maybe because we had a better line of sight away from the trees. Third Walk (SPAM Day): SPAM Day was most enjoyable in that we were able to walk as a class and receive a transmission together. We were able to reach the gas station and still receive a clear signal. Further towards El Camino, the SNR degraded faster than it did on our previous field test, so we stopped halfway between the gas station and El Camino. Several factors could have been detrimental to reception: 1) the weather was not as favorable, 2) the newly created metal enclosure for the receiver actually hurt reception, or 3) the BNC output to a loudspeaker is not as optimal as a wire connection to a headphone. Conclusion: Block-by-block design and construction of the FM transmitter and receiver was the right approach. We become familiar with the benefits and limitations of each stage and could optimize them individually in the first few weeks. When integration occurred, the firsthand knowledge of the separate blocks became an invaluable part of debugging.
8
The main achievement of this project is the successful construction of two systems, shown in Figures 10 and 11. Our transmitter is able to output at 5-dBm, while our receiver is able to detect signals as weak as -110-dBm and as far away as 2-km. Our two circuits are by no means ideal. Given more time, we would fix some issues we have noticed during construction. Since the LNA provides the best SNR out of all stages on the receiver, we can increase its gain by drawing more collector current and try to lower the noise figure by experimenting with new resistance values. The IF amplifiers currently contribute too much noise in the region around 300-kHz. We can try a different op-amp and different values of bias resistors, as well as sharpen the cutoff of the passive band-pass filter by using a higher-order Butterworth topology. Neither mixer is providing any conversion gain, so we would try using a different mixer chip. There is certainly an impedance mismatch between the VCO output and mixer input, which we would solve by impedance matching one to the other or both to 50- Ω . Also, with more time, we would explore the transmission route through the Colpitts oscillator, to increase the transmission power without sacrificing stability.
Figure 10. Transmitter circuit.
Figure 11. Receiver circuit.
9
