Image, blocking or cross-modulation in a homebrew 14 MHz superhet
You built a single-conversion superhet for 14 MHz: a broadband input transformer (1–30 MHz), no preselector, a diode-ring mixer, a 455 kHz IF chain with a 2.4 kHz ceramic filter, a product detector and AGC. The local oscillator runs above the received frequency. On quiet winter evenings it works well. From about 18:00 local time, reception anywhere between 14.100 and 14.250 MHz is spoiled: a broadcast station that is S9+30 dB at your location is audible in the background, the S-meter sits high, the receiver sounds desensitised, and the AGC pumps with the modulation. The station's published schedule lists 15.110 MHz. Disconnecting the antenna removes the problem completely.
1. Calculate the image frequency and decide whether the 15.110 MHz station explains interference across the whole 14.100–14.250 MHz range.
Read the worked answer
Image calculation. With high-side injection the local oscillator sits 455 kHz above the signal. Tuning to 14.200 MHz puts the LO at 14.200 + 0.455 = 14.655 MHz. The image is the frequency that also converts to 455 kHz: 14.655 + 0.455 = 15.110 MHz.
But that is only true at one dial setting. The image frequency moves with the tuning: at 14.100 MHz the LO is 14.555 and the image is 15.010 MHz; at 14.250 MHz the image is 15.160 MHz. A single carrier on 15.110 MHz can therefore only appear as an image when you are tuned within the IF bandwidth of 14.200 MHz, i.e. about ±2 kHz. It cannot explain trouble across 150 kHz.
Conclusion. There are at least two mechanisms. The image response accounts for the broadcast programme appearing exactly at 14.200 MHz, because the 1–30 MHz broadband front end gives essentially no rejection 910 kHz away from the wanted signal. The wideband desensitisation and AGC pumping across the whole range come from a different effect - the same very strong carrier overloading the mixer and the first IF stage before any selectivity has been applied.
Confirming test. Tune slowly through 14.200 MHz. If the programme audio is present over a few kHz only and vanishes elsewhere, that is the image. If the receiver stays deaf and the S-meter stays high across the whole band, that is blocking. A 10 dB input attenuator will improve blocking but will not remove a genuine image.
2. Name the mechanisms that could be at work and say how you would tell them apart at the bench.
Read the worked answer
Image response. As calculated, 15.110 MHz converts to the same 455 kHz IF. It is a linear conversion, so it sounds exactly like a real station on frequency. It disappears if you raise the IF or add image rejection.
Blocking / desensitisation. S9+30 dB at HF is around −43 dBm, tens of millivolts at the mixer input. The mixer and IF amplifier compress, gain falls and the AGC backs off the whole receiver, so everything sounds weak and noisy. It is broadband and does not require any frequency relationship.
Cross-modulation. The modulation of the strong carrier is transferred onto whatever you are tuned to, because the front-end non-linearity acts as a modulator. The test is that you hear programme audio on a frequency where no station is scheduled and where there is no image relationship.
Reciprocal mixing. LO phase noise multiplied by the strong carrier spreads sideband noise across the band, raising the noise floor near it; the cure is a cleaner oscillator, not a filter.
How to separate them. Insert a switched step attenuator ahead of the mixer and watch the signal-to-interference ratio. If it improves markedly with attenuation, the front end is being overloaded (blocking or cross-modulation). If the ratio stays the same while both signals drop, the interfering signal is arriving through a linear path, i.e. an image or a mixer product. Repeating the test with a signal generator on 15.110 MHz and a known level makes the comparison quantitative and repeatable.
3. Propose the design changes and describe a measurement that would demonstrate the improvement.
Read the worked answer
Reduce the level before the mixer. Add a three-resonator band-pass preselector covering 14.0–14.35 MHz with 1–2 dB insertion loss and at least 40 dB rejection at 15.11 MHz. A simple double-tuned front end already gives 20–30 dB and will largely remove the image. A switchable 10 dB pad is the complement: it protects the mixer when the band is crowded.
Improve the mixer's handling. Terminate the diode ring properly on all ports, drive it with adequate LO level (around +7 to +10 dBm for a typical ring), and use matched Schottky quads. A badly terminated, under-driven ring has a poor third-order intercept and compresses early.
Change the frequency plan if you want a real fix. Moving the first IF to 9 or 10.7 MHz pushes the image several MHz away where the preselector easily rejects it; a double conversion with a high first IF achieves the same thing; an image-reject mixer is an option but needs careful balance.
Measurement. Feed a signal generator at 15.110 MHz into the antenna socket with the receiver tuned to 14.200 MHz, and use a step attenuator to find the generator level at which the broadcast-like signal becomes audible and the level at which the AGC starts to move. Repeat after fitting the preselector. The increase in those thresholds, in dB, is the improvement you have bought - and it tells you whether the remaining limitation is image rejection or mixer overload.