# Intermediate frequency

I'm a newbie to communication systems and circuitry. I was going through superhet receivers while trying to undertand AM/FM/DAB receiver architecture. For FM, IF is fixed to 10.7 MHz and AM 455 KHz. Once the carrier wave is shifted to this IF frequency, it is amplified and filtered. In DAB, it is mentioned that IF is 0 (Source). So, is it possible to set IF frequency as 0?

The 10.7 and 455 kHz IF frequencies are more "historic" choices for superheterodyne receivers. The choices for these frequencies might have to do with availability of components (filters etc.) and/or frequency of the LO signal.

To receive for example a 100 MHz FM station, when IF = 10.7 MHz the LO can be either 89.3 MHz or 110.7 MHz. If an IF of 1 MHz was chosen then the LO frequency would need to be much closer to the frequency you want to receive introducing all kinds of problems.

The main problem is that the LO signal can find its way back into the receiver (LO feedthrough) so the radio would be receiving its own LO oscillator. Also the signal you want to receive should have a bandwidth of less than the IF frequency. For an FM receiver a 1 MHz IF would affect sound quality as it cannot properly handle the a 1 MHz frequency deviation of the FM signal.

Yes but then when IF = 0 Hz the LO is at the same frequency as the signal you want to receive !

Yes that is correct ! But in more modern receivers which are mainly implemented on a chip, not using discrete components, the issues caused by the LO being at the same frequency as the RX signal can be solved using a more complex architecture (Quadrature LO, LO supressing mixers, digital DC calibration etc.)

Modern receivers like for DAB and the ones in your (smart)phone are almost without exception Zero-IF or same thing, different name: Direct conversion receivers. See this lecture for more information.

These type of receivers are easier to implement on-chip. The issues like DC-offset and LO suppression can be handled by calibration and digital signal processing.

There are also (on-chip) AM and FM receivers using zero-IF so that superheterodyne architecture with 10.7 MHz IF is not strictly needed. But in "classical" designs there was less choice to do things differently.

• Thanks for the explanation! You have mentioned that LO frequency when it is close to desired frequency introduces problems. Could you briefly explain what kind of problems? Thanks in advance!
– Gomu
Jul 6, 2017 at 13:35
• @Gomu I have added some to the answer Jul 6, 2017 at 14:01

AM radio receivers use simple envelope diode detectors to recover the audio signal. This means they need a version of the carrier wave and, of course, the IF output was perfectly suitable for this because (a) it was filtered to output just one particular channel and (b) the IF frequency was well above the audio frequencies making it easy for the simple diode detector to work effectively.

FM receiver have a similar story - the type of detector used relies on the presence of a carrier and although it is slightly more complex than an AM detector it is still very simple.

On the other hand, technology has moved on and now it is relatively easy to make a phase locked loop circuit lock into either the FM signal and directly demodulate OR lock into the carrier of an AM signal and do a bit of math (for the more esoteric AM methods) to recover the base-band signal: -

Using a PLL (or sampling the RF directly) directly converts RF to base-band and doesn't require an IF but, for a cheap AM/FM radio, the old traditional methods work just fine.

This wiki article on the direct-conversion receiver pretty much covers what I've said above in a little more detail such as: -

Thus direct demodulation of AM or FM signals (as used in broadcasting) requires phase locking the local oscillator to the carrier frequency, a much more demanding task compared to the more robust envelope detector or ratio detector at the output of an IF stage in a superheterodyne design.

One of the initial disadvantages of the original 1930s design is mentioned here: -

The design (homodyne or DCR) suffered from the thermal drift of the local oscillator which changed its frequency over time. To counteract this drift, the frequency of the local oscillator was compared with the broadcast input signal by a phase detector. This produced a correction voltage which would vary the local oscillator frequency keeping it in lock with the wanted signal. This type of feedback circuit evolved into what is now known as a phase-locked loop. While the method has existed for several decades, it had been difficult to implement due largely to component tolerances, which must be of small variation for this type of circuit to function successfully.

• This seems to be largely tangential: DAB is a digital modulation, and the demodulator is probably built to take a zero-centered IQ IF. Jul 6, 2017 at 15:13

Also note that hetrodyne receivers are also susceptible to receiving an "image" signal that's either higher or lower than the desired signal by 2x the IF frequency, because the mixing stage produces identical outputs for both the sum and difference of its input frequencies.

Thus, for historical FM receivers the image reception is 21.4 MHz offset from the desired signal, so outside the standard FM band. For historical AM receivers, the image reception was 910 KHz offset, so for most of the AM band the image frequencies were out-of-band. Suppressing image reception usually required a tuned RF input stage, which would knock down the amplitude of the undesirable signal.

As an earlier responder mentioned, for a zero-IF receiver both desired and image frequencies are mixed down to DC when the local oscillator is exactly correct, and to low-frequency signals that are indistinguishable from audio modulation when the LO drifts even slightly off. Thus, you'll only see such designs when the frequency of the LO can be controlled e.g. using a phase lock loop circuit.

• It's actually not true that zero IF only works when the frequency can be tightly controlled or that low frequency error is indistinguishable from audio modulation. This is because an IQ zero IF has not aliased positive and negative frequencies. Any frequency drift or error small enough to fit the conversion bandwidth which could be corrected before a traditional IF, can still be corrected subsequent to a zero IF, and arguably more easily, because you can apply complex algorithms to identifying the drift or error. Sep 17, 2019 at 19:36
• ie, if you decide that the zero IF is off by 50 Hz, you can readily remix it to another. If you want an ultra narrow carrier at the center of the zero IF to produce a 700 Hz audio tone, you can mix it up to that. And if you want to demodulate an SSB signal, you don't actually mix the suppressed carrier to zero, you mix the center of the passband, filter symmetrically around zero, and then mix it back to put the suppressed carrier at zero and make the audio frequency correct. Sep 17, 2019 at 19:38