Digital synthesis of Quadrature, please help

Greetings,

I have a nice little DSP radio working, see 
http://members.wideband.net.au/gzimmer/DSP_Rx/SimpleRX.html

It uses two mixers to develop the quadrature data streams.

I figured that I could eliminate the quadrature mixers,
if I could synthesise the quadrature streams in software.

I used a rolling buffer with taps 1&2 summed for one stream
and taps 2&3 summed for the other.

Trouble is I get no sideband separation at all.

I'd be gratefull if someone could steer me in the right direction.

Thanks in advance.

........................... Zim 



-- 
Posted via a free Usenet account from http://www.teranews.com

Graeme Zimmer wrote:
> Greetings,
>
> I have a nice little DSP radio working, see
> http://members.wideband.net.au/gzimmer/DSP_Rx/SimpleRX.html
>
> It uses two mixers to develop the quadrature data streams.
>
> I figured that I could eliminate the quadrature mixers,
> if I could synthesise the quadrature streams in software.
>
> I used a rolling buffer with taps 1&2 summed for one stream
> and taps 2&3 summed for the other.
>
> Trouble is I get no sideband separation at all.
>
> I'd be gratefull if someone could steer me in the right direction.
>
> Thanks in advance.
>
> ........................... Zim

Nice project, but I don't see how the VFO and mixers can be omitted
because it appears that the ADC you're using is intended only for audio
frequencies.

Are you undersampling the RF in the new configuration? If so then you
need to verify that the ADC has sufficient bandwidth to see the RF
frequency. Assuming that is the case, the signal will translate to a
digital IF that is a function of the RF frequency and sample rate. To
get to baseband you need to create sine and cosine of the digital IF
frequency, multiply the ADC samples with each, and then lowpass filter
to remove the images. If the RF frequency changes then you'll have to
adjust your mixing frequency. It seems to me that this could only work
over a fairly narrow tuning range.

John

Thanks john,

but no, my intention is to still use one mixer to convert the RF down to to 
audio.

I need to convert an audio stream of samples into a pair of quadrature 
streams.


Thanks ............... Zim



-- 
Posted via a free Usenet account from http://www.teranews.com

Graeme Zimmer wrote:
> Thanks john,
>
> but no, my intention is to still use one mixer to convert the RF down to to
> audio.
>
> I need to convert an audio stream of samples into a pair of quadrature
> streams.
>
>
> Thanks ............... Zim
>
>

I see. Then you might want to off-tune by Fs/4 (so the ADC input is
bandpass), mix down in DSP by sin and cos of Fs/4, and then apply
lowpass filters.

John

Thanks, john,

Yes, I had come to much the same conclusion....

I can push the sample rate up to 50KHz, so I have enough B/W to fit in 
another down conversion...

My first effort was to step a clock at sample rate and use it to steer 
samples and zeros to the filters in quadrature. It didn't work.

But the above sounds close to what you are suggesting. I'll have another go, 
but this time send samples and inverted samples alternatively.

Thanks ............... Zim



-- 
Posted via a free Usenet account from http://www.teranews.com

John,

I coded this up, but still no joy. No sideband supression at all.

I'm feeding the samples into a rolling buffer and taking them off in 
quadrature order..
(the same way that a commutating mixer would work)

CM    +1.0  IN7              ; Read chan 7 into A
SCA   +1.0  1000            ; Write A into rolling buffer

;----------- get I chan --------------------
CM       +1.0   1000       ; get sample
CMA     -1.0   1002       ; sum with inverted sample
SCA      +1.0   0            ; feed it to I chan

;----------- get Q chan --------------------
CM        +1.0   1001       get sample
CMA     -1.0   1003      ; sum with inverted sample
SCA      +1.0   495       ; feed it to Q chan


The assembler will look wierd, but you can probably see what is happenng 
from the comments.

Only thing I can thing I can of, is that the I&Q FIRs are still stepping at 
sample speed rather than one quarter (something I can't change).


Thanks for any help ................ Zim 



-- 
Posted via a free Usenet account from http://www.teranews.com

Graeme Zimmer wrote:
> John,
>
> I coded this up, but still no joy. No sideband supression at all.
>
> I'm feeding the samples into a rolling buffer and taking them off in
> quadrature order..
> (the same way that a commutating mixer would work)
>
> CM    +1.0  IN7              ; Read chan 7 into A
> SCA   +1.0  1000            ; Write A into rolling buffer
>
> ;----------- get I chan --------------------
> CM       +1.0   1000       ; get sample
> CMA     -1.0   1002       ; sum with inverted sample
> SCA      +1.0   0            ; feed it to I chan
>
> ;----------- get Q chan --------------------
> CM        +1.0   1001       get sample
> CMA     -1.0   1003      ; sum with inverted sample
> SCA      +1.0   495       ; feed it to Q chan
>
>
> The assembler will look wierd, but you can probably see what is happenng
> from the comments.
>
> Only thing I can thing I can of, is that the I&Q FIRs are still stepping at
> sample speed rather than one quarter (something I can't change).
>

Is this the same as multiplication by {1 0 -1 0 ... } and {0 1 0 -1
...} ? I don't understand what "sum with inverted sample" means.

John

Hi John,

> Is this the same as multiplication by {1 0 -1 0 ... } and {0 1 0 -1> ...}
> ? I don't understand what "sum with inverted sample" means.

Well, a simple mixer multiplies the incoming signal by +1 or -1 
alternatively.

Instead of a NCO and multiplier, I am using a shift register to simulate a 
commutating mixer,
that is, a 4-way switch which is rotating at NCO rate.

As the switch rotates it alternatively picks off +Signal and -Signal.
The other chanel does the same only rotated by 90 deg.

I think.

........... Zim



-- 
Posted via a free Usenet account from http://www.teranews.com

John,

After some thought, I decided my previous code was all wrong.

I'll have another go.


Tks ................ Zim 



-- 
Posted via a free Usenet account from http://www.teranews.com

Graeme Zimmer wrote:
> John,
>
> After some thought, I decided my previous code was all wrong.
>
> I'll have another go.
>
>
> Tks ................ Zim
> 
> 

You had me scratching my head.

Good luck,

John