1
\$\begingroup\$

In a superheterodyne receiver, what must one consider in order to chose an intermediate frequency?

Specifically, I'm working on a radio telescope project. It is basically a superheterodyne receiver where the IF signal is 1-bit sampled and then stored.

The bandwidth of the signal is 20 MHz. Now, to minimize aliasing, I'll be using a sample rate somewhat above the Nyquist rate. That means the minimum sample rate is ca 50 MHz. The highest I could sample, on the other hand, is about 120 MHz.

This would however mean an IF range from 0 Hz to 20 MHz. If we drop the lowest 1 MHz, this equals to 4.3 octaves.

The observing frequency is 1421 MHz, if that matters.

I'm mostly worried about the high fractional bandwidth of the signal. So, my questions are:

  • Would it be problematic to amplify the 1-20 MHz (or even 0-20 MHz) signal using a high-speed OP-Amp (considering that many electronic components' behaviour is (inversely) proportionally dependent on the frequency)?

  • Is the high fractional bandwidth a problem? i.e. Would the SNR improve significantly if I used a 40-60 MHz range?

  • Low IF receivers are used to evade brown noise (noise power constant per octave) zero IF systems inherently suffer from. How much does it practically matter for an application like this?

\$\endgroup\$
1
  • 3
    \$\begingroup\$ It might be helpful for you to draw a complete block diagram of your entire signal chain, starting at the antenna, and explicitly showing the places where amplification and filtering are done, right up to the one-bit quantizer. Keep track of where unwanted out-of-band signals are going to be attenuated, too. Doing this will help you focus on the issues such as what frequencies and what dynamic range each stage of amplification needs to deal with. You'll soon realize why "trivial" solutions rarely work. \$\endgroup\$ – Dave Tweed Sep 18 '14 at 21:52
1
\$\begingroup\$

OK you are basically proposing a zero I.F. in your question - I see that now!

  1. I don't see op-amps as a problem (not these days) - you can get ones that are over 1GHz GBP which means at 20MHz bandwidth you could get a gain of maybe 30 to 50 (op-amp choice variations). No need to lose the lower 1MHz though and, you ought to consider some real good low pass filters to remove out of band noise. Basically filter out whatever you can that's not required and that includes some good filters on the RF signal recieved - if your bandwidth is only 20MHz in 1.4GHz, removing as much as you can from outside this area is a good idea.
  2. Use the full base-bandwidth is my thought.
  3. Maybe but without a brain dump or a vulcan mind meld I can't anticipate anything LOL.
\$\endgroup\$
2
  • \$\begingroup\$ thanks for your reply. The reason I was worried about the lower 1 Mhz is that extending the range down to near-dc would incur a huge increase in fractional bandwidth. According to Wikipedia what I described is a link. In that article it also says that Low IF avoids brown noise which is constant with respect to log(f). I will edit the question to be more specific. \$\endgroup\$ – hasep Sep 18 '14 at 21:50
  • 1
    \$\begingroup\$ OK it's flicker noise that concerns you. If you don't need the low 1MHz then no problem - it's usually an artefact created by op-amps and if you don't require anything below 1MHz this is easily dealt with as you implied in your question. Clearly any band of frequencies around some nominally high RF doesn't carry i/f noise so down-converting that to baseband isn't generating anything except constant noise power per Hz. \$\endgroup\$ – Andy aka Sep 18 '14 at 22:26
1
\$\begingroup\$

The problem you might run into with a low IF receiver is that the image frequency is close in and hard to filter out. Seconding Dave Tweeds recomendation to draw a block diagram and figure out where the signals are, as well as where the image frequencies are.

\$\endgroup\$
1
\$\begingroup\$

An IF of zero has a few disadvantages:

  • the image overlaps your signal, which is fine for AM but not much else. You can use a second ADC for the Q signal, where the last downmix is done with an LO output at a 90 degree offset, but at this point you are effectively building a direct-conversion receiver anyway.
  • you get a DC offset in the analog-digital conversion, which you have to compensate for in the digital processing. With a nonzero IF, that shows up as a CW far outside the signal you want to track and can be filtered out easily.
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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