I'm currently learning about the Superheterodyne AM Receiver and I don't understand why does it use an envelope detector (as suggested here) rather than taking the signal back to its baseband frequency through a local oscillator. That seems more natural to me because it's already using one to take the signal to the intermediate frequency.

  • 1
    $\begingroup$ Simple engineering answer: it works and it's cheap. $\endgroup$
    – Hilmar
    Nov 13, 2021 at 13:52

1 Answer 1


The AM receiver could be done with a second Local Oscillator to baseband which would work fine if there was no frequency offset between the transmitter and receiver; otherwise the baseband signal would have an amplitude variation given by the AM in the signal, as well as phase rotation given by the frequency offset and relative delay (consider if the signal happened to be received with 90 degree delay- the result out of a single real mixer after low pass filtering would be 0!) So, the receiver would require two additional mixers driven in quadrature to be able to determine and remove these phase and frequency offsets. The phase and frequency offsets would need to be recovered and removed to intelligibly recover the desired amplitude variation; this is known as coherent demodulation. This can and is often done, at the increase in complexity but also an increase in performance with higher SNR achievable.

Such a frequency offset is inevitable given the transmitter and receiver operate on different reference clocks that can't possibly be at exactly the same frequency, and the frequency difference can be significant for low cost hardware with temperature variations etc. If there is motion between transmitter and receiver, this will add an additional Doppler offset. For software radios where the translation from IF to baseband together with carrier recovery can be done digitally this would likely be the favored approach at the benefit of the flexibility the implementation will provide (and all the other functions that motivated us to go down this path to begin with, certainly not for a fixed AM receiver). For a radio that only needs to demodulate the AM, a simple power detector on the IF output (or depending on the carrier frequency directly on the RF signal) is far simpler and therefore a lower cost, size and power solution.

  • $\begingroup$ also note that for a coherent receiver, you would still need filters after the final mixing stage – one to remove the LO leakage (i.e., a high pass filter) and one to remove unintended-order intermodulation products (typically, a low-pass filter). With an envelope detector, you also only need a high-pass to get rid of the DC, and a low-pass to get rid of the higher harmonics, and you don't actually get any worse signal! $\endgroup$ Nov 13, 2021 at 14:05
  • $\begingroup$ @MarcusMüller Right- the filtering is somewhat similar for each and not overly complex (series DC blocking cap and simple first order low pass filter)-- but I think you would get 3 dB better SNR since you can rotate the coherent receiver to I and ignore Q, thus getting rid of half your noise--- the detector is sensitive to the AM on both I and Q on the equivalent Analytic Signal. Am I correct? $\endgroup$ Nov 13, 2021 at 14:15
  • $\begingroup$ Ahaa. I assume then that such variations in frequency/phase would also affect the first local oscillator (used to take the signal to the intermediate frequency) but error there might be more tolerable or its just that we'd rather stay away from double trouble. $\endgroup$
    – Essam
    Nov 13, 2021 at 14:25
  • 2
    $\begingroup$ @Essam- No they have no impact which is the beauty of envelope detection. The signal at the IF frequency has an envelope that varies with the amplitude only and a phase that varies with the frequency offset. Any frequency offset within the bandwidth of the detector would have no impact! $\endgroup$ Nov 13, 2021 at 14:31
  • 2
    $\begingroup$ @DanBoschen hm, I hadn't looked at it that way. Your argument here is that half of the noise energy is "anti-symmetric in spectrum", i.e. lands on I, and can be ignored, because we know that our signal of interest is symmetric in spectrum. Hm... You're right. If I represent the passband noise as series of cos and sin, then all the sin are irrelevant to a coherent receiver, but would still lead to noise after demod in an envelope detector! $\endgroup$ Nov 13, 2021 at 14:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.