```Randy Yates wrote:
>
> praveen wrote:
> >
> > Hello,
> > I am doing IQ demodulation in DSP. Only one reference signal(sin) fo
> > is coming as reference. I have generate 2fo,4fo from it and also
> > generate orthogonal signal of each of it for IQ demodulation. All
> > these frequency should be phase looked with the reference signal.My
> > question is how to generate these.Any special techniques for
> > generation. Any reference or suggestion will be great.
>
> Here's one brute-force way. The down-side is that it requires a clock
> 4 times the reference frequency, but that may not be a problem if your
> rates are not that high.
>
> Generate a nice square 2fo signal by frequency synthesizing the reference
> signal up by 4, then dividing that by two using a D flip flop with Q and
> Q* outputs. Generate the in-phase signal, 1fo, by dividing the 2fo signal
> by 2. Then generate the quadrature 1fo signal, 1foq, by clocking the 1fo
> signal through a D-flip flop using 2fo* (the Q* output of the divider) as
> a clock. The Q output of this flip-flop is the 1foq signal. If you
> need nice sine waves, filter the 1fo and 1foq signals with a filter
> that has a cutoff just above 1fo.
>
> However, I've never done this before...
>
> A schematic is worth a thousand words, but hopefully you get the gist.
> --
I've done that (sort of), and for diode ring modulators, the square
signal is ideal.

With a 4x clock, feed the clock inputs of 2 D-type flip flops. Connect
the Q output of one to the D input of the other, and the ~Q output of
that one do the D input of the first.  The Qs will have a quarter of the
clock frequency and quadrature phase. With the ~Qs, all four phases are
available. There is no need for the 4x clock to be square; only one of
its edges is used.

Jerry
--
Engineering is the art of making what you want from things you can get.
&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
```
```Hello Praveen,

R-BJ mentioned my article about recursive osc. You can use one of these and
phase lock it to you incoming sine wave. A simple phase lock can be had if
you analytic extend your sine wave to include a cos component. You can use a
phase orthogonal filter pair to do this. Then use the osc (with quadrature
outputs) to heterodyne the incoming reference down to zero. This gets
lowpassed and is used to control the frequency of the quad osc. See my
article for details on dynamic tuning of oscs.  If your reference's
amplitude is all over the place, then you can add a simple arctan method to
get amplitude independence.  If you only need to freq. lock instead of phase
lock, you can used a derivative method (finds instantaneous freq.) and avoid
arctangents.  Once you have a quad osc locked to your signal, you can easily
create the square and fourth powers just by squaring the quad osc's output.
Treat the osc's output as a complex number. Just square it again to get the
4th power value. Another approach for locking is the Costas loop; however,
since it does real signal mixing, you get both the sum and difference
frequencies in two intermediate stages. The sum frequencies are knocked out
with a pair of lowpass filters. But since you don't want the sum frequency
to alias on top of your frequency of interest, you have to oversample
(relative to your final data bandwidth). The complex phase locked loop
doesn't suffer from this requirement - in fact you can let the aliasing just
wrap the signal around. For example, to demodulate BPSK, just phase lock the
square of the osc's output to the square of the signal to create the control
signal for the oscilator's frequency. Then let the original osc. output be
used to directly demodulate the unsquared signal.  I've implemented this and
it works well. The reason this works is a complex mix doesn't create both
images.

Clay

My paper can be found here:

http://personal.atl.bellsouth.net/p/h/physics/oscpaper.pdf

"praveen" <praveenkumar1979@rediffmail.com> wrote in message
> Hello,
> I am doing IQ demodulation in DSP. Only one reference signal(sin) fo
> is coming as reference. I have generate 2fo,4fo from it and also
> generate orthogonal signal of each of it for IQ demodulation. All
> these frequency should be phase looked with the reference signal.My
> question is how to generate these.Any special techniques for
> generation. Any reference or suggestion will be great.
>
>
> praveen

```
```In article 3EFDA062.4DF109B6@ieee.org, Randy Yates at yates@ieee.org wrote
on 06/28/2003 10:01:

> praveen wrote:
>>
>> Hello,
>> I am doing IQ demodulation in DSP. Only one reference signal(sin) fo
>> is coming as reference. I have generate 2fo,4fo from it and also
>> generate orthogonal signal of each of it for IQ demodulation. All
>> these frequency should be phase looked with the reference signal.My
>> question is how to generate these.Any special techniques for
>> generation. Any reference or suggestion will be great.
>
> Here's one brute-force way. The down-side is that it requires a clock
> 4 times the reference frequency, but that may not be a problem if your
> rates are not that high.
>
> Generate a nice square 2fo signal by frequency synthesizing the reference
> signal up by 4, then dividing that by two using a D flip flop with Q and
> Q* outputs. Generate the in-phase signal, 1fo, by dividing the 2fo signal
> by 2. Then generate the quadrature 1fo signal, 1foq, by clocking the 1fo
> signal through a D-flip flop using 2fo* (the Q* output of the divider) as
> a clock. The Q output of this flip-flop is the 1foq signal. If you
> need nice sine waves, filter the 1fo and 1foq signals with a filter
> that has a cutoff just above 1fo.
>
> However, I've never done this before...
>
> A schematic is worth a thousand words, but hopefully you get the gist.

another way, especially if the need is for nice round sine waves, is what
Clay Turner calls "Coupled-standard quadrature" in his "Recursive
Discrete-Time Sinusoidal Oscillators" paper he did last May for IEEE Sig
Proc magazine.

it is essentially:

cos(wt + phi) = cos(wt)cos(phi) - sin(wt)sin(phi)

sin(wt + phi) = cos(wt)sin(phi) + sin(wt)cos(phi)

where phi = w/Fs.  cos(phi) and sin(phi) should be bumped up by the smallest
amount possible (LSB) so that cos^2 + sin^2 will add to slightly bigger than
1.  that will keep the amplitude from decaying.  what keeps it from blowing
up is clipping.

and i *have* done this on the 56K long ago.  worked like a dream.

r b-j

```
```praveen wrote:
>
> Hello,
> I am doing IQ demodulation in DSP. Only one reference signal(sin) fo
> is coming as reference. I have generate 2fo,4fo from it and also
> generate orthogonal signal of each of it for IQ demodulation. All
> these frequency should be phase looked with the reference signal.My
> question is how to generate these.Any special techniques for
> generation. Any reference or suggestion will be great.

Here's one brute-force way. The down-side is that it requires a clock
4 times the reference frequency, but that may not be a problem if your
rates are not that high.

Generate a nice square 2fo signal by frequency synthesizing the reference
signal up by 4, then dividing that by two using a D flip flop with Q and
Q* outputs. Generate the in-phase signal, 1fo, by dividing the 2fo signal
by 2. Then generate the quadrature 1fo signal, 1foq, by clocking the 1fo
signal through a D-flip flop using 2fo* (the Q* output of the divider) as
a clock. The Q output of this flip-flop is the 1foq signal. If you
need nice sine waves, filter the 1fo and 1foq signals with a filter
that has a cutoff just above 1fo.

However, I've never done this before...

A schematic is worth a thousand words, but hopefully you get the gist.
--
%% Fuquay-Varina, NC            %       'cause no one knows which side
%%% 919-577-9882                %                   the coin will fall."
%%%% <yates@ieee.org>           %  'Big Wheels', *Out of the Blue*, ELO
```
```Hello,