Forums

Image-reject IF downmixing

Started by Kevin Neilson April 19, 2004
When digitally mixing an IF down to baseband, one is left with a spectrum
that consists of the baseband (Fif - Fmix = 0Hz) and an image (Fif + Fmix).
If the IF is greater than the Nyquist freq, the image will wrap back into
the first Nyquist zone (0 to Fn).

Normally the next step in demodulation is decimation, which consists of
lowpass filtering out the image (often with CICs) and then dropping some of
the resultant samples to get a lower sample rate.

Is there a method by which an image-reject mixer could mix the IF down to
baseband while simultaneously cancelling the image?  If so, it seems like
there would be no filtering required for decimation, which would consist
entirely of throwing samples out.  Then it also seems that the mixer itself
could run at the slower decimated rate.

Or is that all just another way of saying "undersampling"?
-Kevin


Kevin Neilson wrote:

> When digitally mixing an IF down to baseband, one is left with a spectrum > that consists of the baseband (Fif - Fmix = 0Hz) and an image (Fif + Fmix). > If the IF is greater than the Nyquist freq, the image will wrap back into > the first Nyquist zone (0 to Fn). > > Normally the next step in demodulation is decimation, which consists of > lowpass filtering out the image (often with CICs) and then dropping some of > the resultant samples to get a lower sample rate. > > Is there a method by which an image-reject mixer could mix the IF down to > baseband while simultaneously cancelling the image? If so, it seems like > there would be no filtering required for decimation, which would consist > entirely of throwing samples out. Then it also seems that the mixer itself > could run at the slower decimated rate. > > Or is that all just another way of saying "undersampling"? > -Kevin
Look at sinrle-sidebamd receiver design. It's all there for you. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
I/Q demodulation - a complex mixer using the sine and cosine to deliver the
baseband without image - provides the single-sideband demodulation Jerry
mentions.

Decimation is a simple way of getting "processing gain" by taking a high
digital frequency IF with a lower number of bits per sample and - through
decimation - increasing the effective number of bits by filtering out the
system and quantization noise from the original A/D that is mixed to outside
the baseband.  If you don't need the processing gain, subsampling can be
used but realize that this effectively aliases the IF into several "folds"
such that any spurious signals or thermal or quantization noise gets added
to the desired IF.

Your description of the image comes off a little peculiar in my perception
suggesting you might not be getting your desired point across.  If the IF is
analog-filtered then subsampled, this aliased pseudo-IF can be I/Q mixed to
provide the baseband; it can even *be* the baseband depending on the
sampling rate.  If the IF is analog-filtered then sampled by at least twice
the IF, I/Q demodulation is still required to isolate the baseband from the
*mixed* image of Fif+Fmix.  The topology you're thinking about isn't clear.

Are your questions answered by the interpretation I took?


"Kevin Neilson" <kevin_neilson@removethiscomcast.net> wrote in message
news:HfYgc.173403$JO3.100798@attbi_s04...
> When digitally mixing an IF down to baseband, one is left with a spectrum > that consists of the baseband (Fif - Fmix = 0Hz) and an image (Fif +
Fmix).
> If the IF is greater than the Nyquist freq, the image will wrap back into > the first Nyquist zone (0 to Fn). > > Normally the next step in demodulation is decimation, which consists of > lowpass filtering out the image (often with CICs) and then dropping some
of
> the resultant samples to get a lower sample rate. > > Is there a method by which an image-reject mixer could mix the IF down to > baseband while simultaneously cancelling the image? If so, it seems like
> there would be no filtering required for decimation, which would consist > entirely of throwing samples out. Then it also seems that the mixer
itself
> could run at the slower decimated rate. > > Or is that all just another way of saying "undersampling"? > -Kevin > >
I'd forgotten about the decimation processing gain--that's a good reason to
stay with a high sample rate and conventional decimation process.
-Kevin

"John_H" <johnhandwork@mail.com> wrote in message
news:iaZgc.4$913.203@news-west.eli.net...
> I/Q demodulation - a complex mixer using the sine and cosine to deliver
the
> baseband without image - provides the single-sideband demodulation Jerry > mentions. > > Decimation is a simple way of getting "processing gain" by taking a high > digital frequency IF with a lower number of bits per sample and - through > decimation - increasing the effective number of bits by filtering out the > system and quantization noise from the original A/D that is mixed to
outside
> the baseband. If you don't need the processing gain, subsampling can be > used but realize that this effectively aliases the IF into several "folds" > such that any spurious signals or thermal or quantization noise gets added > to the desired IF. > > Your description of the image comes off a little peculiar in my perception > suggesting you might not be getting your desired point across. If the IF
is
> analog-filtered then subsampled, this aliased pseudo-IF can be I/Q mixed
to
> provide the baseband; it can even *be* the baseband depending on the > sampling rate. If the IF is analog-filtered then sampled by at least
twice
> the IF, I/Q demodulation is still required to isolate the baseband from
the
> *mixed* image of Fif+Fmix. The topology you're thinking about isn't
clear.
> > Are your questions answered by the interpretation I took? > > > "Kevin Neilson" <kevin_neilson@removethiscomcast.net> wrote in message > news:HfYgc.173403$JO3.100798@attbi_s04... > > When digitally mixing an IF down to baseband, one is left with a
spectrum
> > that consists of the baseband (Fif - Fmix = 0Hz) and an image (Fif + > Fmix). > > If the IF is greater than the Nyquist freq, the image will wrap back
into
> > the first Nyquist zone (0 to Fn). > > > > Normally the next step in demodulation is decimation, which consists of > > lowpass filtering out the image (often with CICs) and then dropping some > of > > the resultant samples to get a lower sample rate. > > > > Is there a method by which an image-reject mixer could mix the IF down
to
> > baseband while simultaneously cancelling the image? If so, it seems
like
> > > there would be no filtering required for decimation, which would consist > > entirely of throwing samples out. Then it also seems that the mixer > itself > > could run at the slower decimated rate. > > > > Or is that all just another way of saying "undersampling"? > > -Kevin > > > > > >
You need a filter in there (unless you are not downsampling and there is no
out-of-band signal), however you can play some optimization tricks with the
math to roll the mixer into the filter if the LO is a multiple of your input
sample rate.  What you end up with is a polyphase filter that has the mixer
convolved with it so that you get both mixing and filtering.   If you want to
run the mixer slower, that can be done by bandlimiting the signal before the
mix and subsampling.

Kevin Neilson wrote:

> When digitally mixing an IF down to baseband, one is left with a spectrum > that consists of the baseband (Fif - Fmix = 0Hz) and an image (Fif + Fmix). > If the IF is greater than the Nyquist freq, the image will wrap back into > the first Nyquist zone (0 to Fn). > > Normally the next step in demodulation is decimation, which consists of > lowpass filtering out the image (often with CICs) and then dropping some of > the resultant samples to get a lower sample rate. > > Is there a method by which an image-reject mixer could mix the IF down to > baseband while simultaneously cancelling the image? If so, it seems like > there would be no filtering required for decimation, which would consist > entirely of throwing samples out. Then it also seems that the mixer itself > could run at the slower decimated rate. > > Or is that all just another way of saying "undersampling"? > -Kevin
-- --Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com "They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759
Ray,
I've seen that idea presented before but I don't think I can combine
mixing/filtering in my case because I'm using the IF mixer to do carrier
synchronization so I need to be able to precisely control the LO with an
NCO.
-Kevin
"Ray Andraka" <ray@andraka.com> wrote in message
news:408466B4.85D6AF45@andraka.com...
> You need a filter in there (unless you are not downsampling and there is
no
> out-of-band signal), however you can play some optimization tricks with
the
> math to roll the mixer into the filter if the LO is a multiple of your
input
> sample rate. What you end up with is a polyphase filter that has the
mixer
> convolved with it so that you get both mixing and filtering. If you want
to
> run the mixer slower, that can be done by bandlimiting the signal before
the
> mix and subsampling. > > Kevin Neilson wrote: > > > When digitally mixing an IF down to baseband, one is left with a
spectrum
> > that consists of the baseband (Fif - Fmix = 0Hz) and an image (Fif +
Fmix).
> > If the IF is greater than the Nyquist freq, the image will wrap back
into
> > the first Nyquist zone (0 to Fn). > > > > Normally the next step in demodulation is decimation, which consists of > > lowpass filtering out the image (often with CICs) and then dropping some
of
> > the resultant samples to get a lower sample rate. > > > > Is there a method by which an image-reject mixer could mix the IF down
to
> > baseband while simultaneously cancelling the image? If so, it seems
like
> > there would be no filtering required for decimation, which would consist > > entirely of throwing samples out. Then it also seems that the mixer
itself
> > could run at the slower decimated rate. > > > > Or is that all just another way of saying "undersampling"? > > -Kevin > > -- > --Ray Andraka, P.E. > President, the Andraka Consulting Group, Inc. > 401/884-7930 Fax 401/884-7950 > email ray@andraka.com > http://www.andraka.com > > "They that give up essential liberty to obtain a little > temporary safety deserve neither liberty nor safety." > -Benjamin Franklin, 1759 > >
Which is why I stated that if the  LO is a multiple of your
input sample rate....   It should have read if the input sample rate is an
INTEGER multiple of the LO.

Kevin Neilson wrote:

> Ray, > I've seen that idea presented before but I don't think I can combine > mixing/filtering in my case because I'm using the IF mixer to do carrier > synchronization so I need to be able to precisely control the LO with an > NCO. > -Kevin
-- --Ray Andraka, P.E. President, the Andraka Consulting Group, Inc. 401/884-7930 Fax 401/884-7950 email ray@andraka.com http://www.andraka.com "They that give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -Benjamin Franklin, 1759