Tell me more ×
Electrical Engineering Stack Exchange is a question and answer site for electronics and electrical engineering professionals, students, and enthusiasts. It's 100% free, no registration required.
up vote 0 down vote favorite

Google is not helping me on this one.

I get that image rejection is important when demodulating an RF to an IF frequency.

But what about when demodulating an RF directly to an wideband signal?

I mean, then the image would be placed in the negative frequency domain, exactly as it was when the signal was generated in the transmitter.

Is it because of wasted energy of transmitting both the signal+image?

share|improve this question
2  
It is easy to construct a stable demodulator for a fixed IF frequency. It is reasonable easy to construct a stable and tunable mixer that allows you to tune any single station somewhere on in a wide frequency band onto the IF. On the other hand, it is extremely difficult to construct a receiver that tunes an RF signal directly to a baseband signal, without frequency shifting around the 0Hz marker. – jippie Jun 21 '12 at 8:04
Okay, thank you. But what about in relation to IQ-modulation/demodulation, most schemes I have seen demodulate the carrier directly to a baseband signal. I know my question was generalized, but can you help with this? – JakobJ Jun 21 '12 at 9:12

2 Answers

up vote 3 down vote

Yes, image rejection is a potential concern anytime you have a mixer, including in a "product detector" where you "demodulate" by mixing to baseband.

Such a mixer producing a single output will combine the positive and negative frequencies in a way which cannot be distinguished - a classic example would be tuning a steady carrier (or a modulated one, as in morse code) using a direct conversion receiver / product detector mixing to audio. As the local oscillator nears the transmit frequency, the audio tone will fall through the audio bandwidth towards zero. As the LO frequency passes the audio the transmit frequency, the audio tone starts rising in frequency again. With a single output and no other reference it is not possible to distinguish positive and negative frequencies from each other. An input spectrum which contains both could therefore become quite confused.

An alternative approach is IQ mixing. Here two local oscillator signals in quadrature are produced (ie, sine and cosine) and mixed with the same input signal to produce two outputs. The phase relationship of the outputs can then be examined to determine if (a given component of) the signal has positive or negative frequency. Many baseband signal processing algorithms and modulators/demodulators operate on IQ signals.

An additional option is to shift one of the I or Q signals by 90 degrees and subtract them. This will destructively cancel either the positive or negative frequencies, forming what is known as an Image Reject Mixer. However, generating a precise 90-degree phase shift can be tricky to do, especially over a wide bandwidth. The merits of doing this, vs using filters to eliminate one of the image possibilities depend on the frequency and bandwidth of the signals to be processed, performance requirements (allowable phase distortion and image leak through) and type of technology available for implementation (ie, DSP vs RLC, crystal, mechanical, SAW, cavity, etc filters)

share|improve this answer
up vote 2 down vote

I've been working through something similar, myself. Not having Matlab, I've had to do everything with [sigh] Excel. So my setup may take a little explaining...

Start with a column counting from 0 to 4095. Emulate a 50MHz ADC by filling the next column with time increments. After that, a few columns that make sine waves, based on the time. I selected 11.5MHz, 12.5MHz(desired signal), 13Mhz and 14MHz signals.

Sum up the sine waves, label that column FFT_IN. Use Tool to do a FFT Transform, plot the magnitude of the result: a nice little spectrum. That's your INPUT.

Next: Demodulate to get that 12.5MHz. Make "I" stream by multiplying FFT_IN by: 0,1,0,-1. "Q" is: 1,0,-1,0. Result: samples alternate between I and Q, with every second sample in a stream inverted. (Optional: I also run a little FIR on this data)

Merge the I and Q streams into complex samples, and do another FFT. Surprise: the signal starts at DC (was 12.5MHz) with signals present at 500KHz, 1MHz, and 1.5MHz.

Why? Because the 11MHz signal has been "reflected" back into the positive frequency domain.

share|improve this answer
1  
Thanks a lot for your reply. :-) Excel.. I have made my simulations using Octave, it is a free tool, and it works exactly like Matlab with script, plots and all. (Matlab scripts will run in Octave). – JakobJ Jun 21 '12 at 11:26
@Alan - you might also want to give Scilab a go, it's free and similar to Matlab. – Oli Glaser Jun 21 '12 at 16:38

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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