FM broadcast receiver in function blocks

Created: Jun. 2020

Introduction

Many moons ago in 1999, I built a broadcast band FM radio from semiconductors manufactured by my (then) employer. The components were assembled on some scrap printed circuit boards (PCB). These PCBs were previously testboards for the automated testing of RF semiconductors, but retired from service after the contact areas have worn out. The objective for constructing the radio as distinct function blocks is educational - they clearly demarcated the various stages of a superheterodyne architecture. Additionally, the connectorized blocks eased the build process as they could be independently tested.

Overall architecture

Fig. 1: block diagram of a FM broadcast superheterodyne receiver constructed (almost) entirely from semiconductors made by my (then) employer

Fig. 2: top view

In constructing this radio, no component was bought - every part is re-cycled from somewhere. For example, the frequency tuning control came from an old TV game set (fig. 3). And the transparent plastic base came from an old notice board.

Fig. 3: perspective view.

The overall noise figure (NF) is dominated by the preselector's and the LNA's NF as a good design should be (fig. 4). Based on the 4.3 dB overall NF, the minimum discernable signal (MDS) is estimated at -117 dBm. The estimated sensitivity for 12 dB signal-to-noise ratio is -105 dBm.

The total gain, from RF input to limiter output, is ~105 dB and bulk of it comes from the two limiter sections. The large overall gain ensures that hard limiting occurs in order to reduce AM noise at the demodulator input and keep the demodulator's sensitivity constant.

Fig. 4: cascade analysis

Preselector, LNA & image reject filter

For simplicity, we opted for fixed tuned preselector and image reject filter (fig. 5). While this eliminated the complexity of LO-RF tracking, it has the trade-off of limiting the reception range to 91-105 MHz.

The low noise amplifier (LNA) used current shunt feedback to flatten its response inside the passband and to achieve 50 ohm terminating impedances. The quiescent current is ~25 mA and the gain is ~ 26 dB.

Fig. 5: Preselector, LNA & image reject filter

Fig. 6: PCB for preselector, LNA & image reject filter

The image reject filter consists of three stages of parallel LC. The stages are coupled capacitively. The inductors have ferrite slugs to permit tuning of the passband. The passband range is 91-105 MHz and insertion loss is around -2.5 dB.

Fig. 7: Image reject filter's close-in response. The filter has its passband fixed at 91-105 MHz.

Fig. 8: Image reject filter's far-off response. When receiving 98 MHz, the image frequency at 119.4 MHz is rejected by 28 dB

Mixer

Mixer is the classic doubly balanced diode ring mixer (fig. 9). The baluns are pseudo-Guanella because they were wound toroid rings instead of binocular cores. The Schottky diodes inside the HSMS-2829 are matched because they were picked from adjacent dies on a wafer. The optimum LO power is ~7dBm.

Fig. 9: mixer

Fig. 10: mixer PCB

Local oscillator

The local oscillator (LO) is a Clapp tuned by a varicap diode (fig. 11). A buffer (INA-021) amplifies the power to the 7 dBm level required by the mixer.

Fig. 11: Local oscillator & buffer

IF amplifier, channel-select filter & limiter

The intermediate frequency (IF) amplifier, MSA-0711, that immediately follows the mixer is not protected from adjacent channel interference (fig. 12). So, it is biased to double its nominal operating current (from 22mA to 44mA) in the hope of increasing its intermodulation resistance.

Selectivity is provided by two ceramic filters (Murata MS3 series). They have 180 kHz bandwidth each. The terminating impedance is 300 ohm. So, a L-section low pass network and a IFT are used to match the filters' impedance.

Four stages of limiters (INA-50311) provide ~88 dB of gain. The limiters' saturated output power is ~ 4 dBm. So, the minimum RF input signal required for limiting is approximately -101 dBm (Psat - G).

Fig. 12: IF amplifier, channel-select filter & limiter

FM demodulator

My employer didn't make FM demodulators. So, I re-purposed a mixer IC, IAM-81008 into a 10.7 MHz quadrature detector. The underlying principle is this: if the input signal is split and fed into both RF & IF ports, the IF port will produce a DC output proportional to the frequency change. A three-way resistive divider is used to split the input signal (fig. 13). In the LO path, a 12p capacitor provides ~90 degree phase shift. A slug-tuned parallel LC tank provides fine control of the phase shift and its secondary lowers the impedance to match the LO port's 50 ohm. The tank component is a common 10.7 MHz intermediate frequency transformer (IFT) - Sumida F35 type with a 9+9 turn primary and 4 turn secondary. The turn ratio is likely not optimum for this circuit, but it was what I had in the junk box. However, this component is not critical and other tunable 10.7 MHz IFTs should work too. The RC network at the output, in conjunction with the IF port 50 ohm impedance, provides de-emphasis (although the resultant 1 uS value is far short of the standard 50 uS)

Fig. 13: FM demodulator from re-purposed mixer

Fig. 14: demodulator PCB

FM demodulators work by converting the input frequency deviation into a varying DC voltage, thereby, producing the S-curve. The prototype produced a DC output voltage that is inversely proportional to the input frequency (fig. 15). These S-curves' slopes increase with input power level . The slope rates are 0.15 V/MHz, 0.31 V/MHz and 0.57V / MHz for -5 dBm, 0 dBm, and 5 dBm, respectively. The steeper slope produces a larger demodulated signal. The changing slope is probably inconsequential in actual applications as the input power is more or less held constant by the preceding limiter stage. The IFT is tuned so that the S-curve's top and bottom halves are approximately equal at 10.7 MHz. A minor niggle is the midpoint of the S-curve shifts slightly (tens of mV) with input power.

Fig. 15: The measured S-curve validates the improvised demodulator's suitability for detecting FM

Fig. 16: original, hand-plotted S-curve from Dec. 1998

As expected, the demodulator's Signal-to-Noise (S/N) ratio improves with input power (fig. 17). The test signal has 22.5 kHz deviation (m=0.3) and a 1 kHz audio. Presumably, better S/N would have been achieved using a 75 kHz deviation (m=1) test signal. Probably even better S/N could be achieved if the de-emphasis had been correctly dimensioned to 50 uS.

Fig. 17: Measured Signal-to-noise ratio vs. input power. Good S/N can be achieved at the target demodulator input power of 4 dBm

consent of the owner