Digital Down Conversion - Channel Selection

Hi,
I am facing problem in understanding the following.
I have a wideband signal 30 MHz bandwidth with 30 kHz channels on 21.4
MHz center frequency. I have sampled it on 90 MHz, so the band is
centered around 20 MHz due to sampling. I want to extract a single
channel from this signal. For that I need to mix this signal from the
carrier frequency of that particular channel. In that case, what I see
is that the positive and negative parts of the spectrum overlaps. I am
using gc5016 chip. Is there any way I can avoid this overlapping and
downconvert the signal.
I have a possible solution in mind. If i take hilbert transform of the
signal and remove the negative spectrum, would i be able to perform
this.

Aitezaz

On Jul 6, 1:21&#4294967295;am, Aitezaz <aitezaz....@gmail.com> wrote:
> Hi,
> I am facing problem in understanding the following.
> I have a wideband signal 30 MHz bandwidth with 30 kHz channels on 21.4
> MHz center frequency. I have sampled it on 90 MHz, so the band is
> centered around 20 MHz due to sampling. I want to extract a single
> channel from this signal. For that I need to mix this signal from the
> carrier frequency of that particular channel. In that case, what I see
> is that the positive and negative parts of the spectrum overlaps. I am
> using gc5016 chip. Is there any way I can avoid this overlapping and
> downconvert the signal.
> I have a possible solution in mind. If i take hilbert transform of the
> signal and remove the negative spectrum, would i be able to perform
> this.
>
> Aitezaz

If you sample 21.4 +/- 15 MHz at 90 Msps, why do you say the digital
signal is centered on 20 MHz?

John

On Jul 6, 1:30&#4294967295;pm, John <sampson...@gmail.com> wrote:
> On Jul 6, 1:21&#4294967295;am, Aitezaz <aitezaz....@gmail.com> wrote:
>
> > Hi,
> > I am facing problem in understanding the following.
> > I have a wideband signal 30 MHz bandwidth with 30 kHz channels on 21.4
> > MHz center frequency. I have sampled it on 90 MHz, so the band is
> > centered around 20 MHz due to sampling. I want to extract a single
> > channel from this signal. For that I need to mix this signal from the
> > carrier frequency of that particular channel. In that case, what I see
> > is that the positive and negative parts of the spectrum overlaps. I am
> > using gc5016 chip. Is there any way I can avoid this overlapping and
> > downconvert the signal.
> > I have a possible solution in mind. If i take hilbert transform of the
> > signal and remove the negative spectrum, would i be able to perform
> > this.
>
> > Aitezaz
>
> If you sample 21.4 +/- 15 MHz at 90 Msps, why do you say the digital
> signal is centered on 20 MHz?
>
> John

you are right... i was thinking of another application for 70 MHz IF.
in this case it will be on 21.4 MHz. Thanks for correction. Do you
have any comments on the question?

Aitezaz

On Jul 6, 5:59&#4294967295;am, Aitezaz <aitezaz....@gmail.com> wrote:
> On Jul 6, 1:30&#4294967295;pm, John <sampson...@gmail.com> wrote:
>
>
>
> > On Jul 6, 1:21&#4294967295;am, Aitezaz <aitezaz....@gmail.com> wrote:
>
> > > Hi,
> > > I am facing problem in understanding the following.
> > > I have a wideband signal 30 MHz bandwidth with 30 kHz channels on 21.4
> > > MHz center frequency. I have sampled it on 90 MHz, so the band is
> > > centered around 20 MHz due to sampling. I want to extract a single
> > > channel from this signal. For that I need to mix this signal from the
> > > carrier frequency of that particular channel. In that case, what I see
> > > is that the positive and negative parts of the spectrum overlaps. I am
> > > using gc5016 chip. Is there any way I can avoid this overlapping and
> > > downconvert the signal.
> > > I have a possible solution in mind. If i take hilbert transform of the
> > > signal and remove the negative spectrum, would i be able to perform
> > > this.
>
> > > Aitezaz
>
> > If you sample 21.4 +/- 15 MHz at 90 Msps, why do you say the digital
> > signal is centered on 20 MHz?
>
> > John
>
> you are right... i was thinking of another application for 70 MHz IF.
> in this case it will be on 21.4 MHz. Thanks for correction. Do you
> have any comments on the question?
>
> Aitezaz

I don't really understand the question. A Hilbert Transform should not
be required to perform the DDC operation; the GC5016 is doing that for
you using its complex mixer and LPFs. Please specify more detail --
tune frequency, decimation, etc.

John

Thanks for your reply john.

well you are right when the signal and my channel is centered on 21.4
MHz. I need to select the channel from the 30 MHz wide band signal as
well. Let's say my required channel is some where around 30 MHz and I
dont select it on the IF. Now to downconvert this, I need to set the
NCO frequency to -30 MHz. It will center my channel around 0 Hz. Now,
if i dont perform hilbert transform and dont remove the negative part,
there will be spectral overlap on the baseband. Isn't so?
This is the summary of my specifications.
IF carrier = 21.4 MHz.
Signal Bandwidth = 30 MHz
Sampling Rate = 90 MHz
Channel Bandwidth = 30 kHz
So No. of channels = 1000
On the front end, its only a 30 MHz wide bandpass filtering and not
the channel selection. I want to select any channel digitally.


Aitezaz

On Jul 6, 6:29&#4294967295;am, Aitezaz <aitezaz....@gmail.com> wrote:
> Thanks for your reply john.
>
> well you are right when the signal and my channel is centered on 21.4
> MHz. I need to select the channel from the 30 MHz wide band signal as
> well. Let's say my required channel is some where around 30 MHz and I
> dont select it on the IF. Now to downconvert this, I need to set the
> NCO frequency to -30 MHz. It will center my channel around 0 Hz. Now,
> if i dont perform hilbert transform and dont remove the negative part,
> there will be spectral overlap on the baseband. Isn't so?
> This is the summary of my specifications.
> IF carrier = 21.4 MHz.
> Signal Bandwidth = 30 MHz
> Sampling Rate = 90 MHz
> Channel Bandwidth = 30 kHz
> So No. of channels = 1000
> On the front end, its only a 30 MHz wide bandpass filtering and not
> the channel selection. I want to select any channel digitally.
>
> Aitezaz

Suppose your 30 kHz signal is centered at 21.4 MHz. You will set the
NCO frequency to 21.4 MHz and decimation to ~1000. Coming out of the
DDC you will have a single complex baseband channel covering the IF
frequencies 21.4 +/- .015 MHz. There is no "spectral overlap".

On Jul 6, 3:49&#4294967295;pm, John <sampson...@gmail.com> wrote:
> On Jul 6, 6:29&#4294967295;am, Aitezaz <aitezaz....@gmail.com> wrote:
>
>
>
>
>
> > Thanks for your reply john.
>
> > well you are right when the signal and my channel is centered on 21.4
> > MHz. I need to select the channel from the 30 MHz wide band signal as
> > well. Let's say my required channel is some where around 30 MHz and I
> > dont select it on the IF. Now to downconvert this, I need to set the
> > NCO frequency to -30 MHz. It will center my channel around 0 Hz. Now,
> > if i dont perform hilbert transform and dont remove the negative part,
> > there will be spectral overlap on the baseband. Isn't so?
> > This is the summary of my specifications.
> > IF carrier = 21.4 MHz.
> > Signal Bandwidth = 30 MHz
> > Sampling Rate = 90 MHz
> > Channel Bandwidth = 30 kHz
> > So No. of channels = 1000
> > On the front end, its only a 30 MHz wide bandpass filtering and not
> > the channel selection. I want to select any channel digitally.
>
> > Aitezaz
>
> Suppose your 30 kHz signal is centered at 21.4 MHz. You will set the
> NCO frequency to 21.4 MHz and decimation to ~1000. Coming out of the
> DDC you will have a single complex baseband channel covering the IF
> frequencies 21.4 +/- .015 MHz. There is no "spectral overlap".

but my 30 kHz channel can be anywhere in the 30 MHz. band. If it is
present on 21.4 MHz, there is definitely no problem. But problem
arises in the case when this channel is anywhere else specially far
away from 21.4.
Aitezaz

The datasheet states (page 15):
>>A programmable inversion is provided for each I or Q data source.
>>Programming Q x sin to be inverted corresponds to a mathematical view of
>>down-conversion (mix with negative frequency tone to get a positive
>>spectrum). Programming I x sin to be inverted corresponds to a radio
>>view (tune to a frequency to get the signal at that frequency). The
>>fields involved are mix_icos, mix_isin, mix_qcos, mix_qsin, and
>>mix_inv_icos, mix_inv_isin, mix_inv_qcos, and mix_inv_qsin. The cmd5016
>>software automatically programs these fields assuming a mathematical
>>view.

So, according to this description, if You have a signal, which is, for
example, +5 MHz apart from the IF, then You should set NCO to either 26.4
MHz (if I x sin mode is used) or -26.4 MHz (if Q x sin mode is used). And
the cmd5016 uses Q x sin mode.

--
Alexander

>On Jul 6, 3:49=A0pm, John <sampson...@gmail.com> wrote:
>> On Jul 6, 6:29=A0am, Aitezaz <aitezaz....@gmail.com> wrote:
>>
>>
>>
>>
>>
>> > Thanks for your reply john.
>>
>> > well you are right when the signal and my channel is centered on
21.4
>> > MHz. I need to select the channel from the 30 MHz wide band signal
as
>> > well. Let's say my required channel is some where around 30 MHz and
I
>> > dont select it on the IF. Now to downconvert this, I need to set the
>> > NCO frequency to -30 MHz. It will center my channel around 0 Hz.
Now,
>> > if i dont perform hilbert transform and dont remove the negative
part,
>> > there will be spectral overlap on the baseband. Isn't so?
>> > This is the summary of my specifications.
>> > IF carrier =3D 21.4 MHz.
>> > Signal Bandwidth =3D 30 MHz
>> > Sampling Rate =3D 90 MHz
>> > Channel Bandwidth =3D 30 kHz
>> > So No. of channels =3D 1000
>> > On the front end, its only a 30 MHz wide bandpass filtering and not
>> > the channel selection. I want to select any channel digitally.
>>
>> > Aitezaz
>>
>> Suppose your 30 kHz signal is centered at 21.4 MHz. You will set the
>> NCO frequency to 21.4 MHz and decimation to ~1000. Coming out of the
>> DDC you will have a single complex baseband channel covering the IF
>> frequencies 21.4 +/- .015 MHz. There is no "spectral overlap".
>
>but my 30 kHz channel can be anywhere in the 30 MHz. band. If it is
>present on 21.4 MHz, there is definitely no problem. But problem
>arises in the case when this channel is anywhere else specially far
>away from 21.4.
>Aitezaz
>

On Sun, 05 Jul 2009 22:21:39 -0700, Aitezaz wrote:

> Hi,
> I am facing problem in understanding the following. I have a wideband
> signal 30 MHz bandwidth with 30 kHz channels on 21.4 MHz center
> frequency. I have sampled it on 90 MHz, so the band is centered around
> 20 MHz due to sampling. I want to extract a single channel from this
> signal. For that I need to mix this signal from the carrier frequency of
> that particular channel. In that case, what I see is that the positive
> and negative parts of the spectrum overlaps. I am using gc5016 chip. Is
> there any way I can avoid this overlapping and downconvert the signal.
> I have a possible solution in mind. If i take hilbert transform of the
> signal and remove the negative spectrum, would i be able to perform
> this.
> 
> Aitezaz

Either filter ahead of your mixer, or use an IQ downconverter.

Trying to do a Hilbert transform after the fact won't help.

-- 
www.wescottdesign.com

Thanks Tim and Alexander for your replies.
@Alexander
I cant understand how i can avoid this overlap with this
specification. Whatever I multiply with the signal, the whole signal
will be down converted to baseband and overlap will still occur.

@Tim
Filtering ahead the mixer is quite an issue as with the change in
channel I would need a new filter. Plus sampling frequency is so high
I wont be able to get the desired filter response with the chip
specifications.

How can an IQ downconverter avoid this? What I know in IQ
downconverter you multiply cosine and sine with the same signal and
then low pass filter it. When on any channel lets say I, i will mix
the signal with a cosine, this overlap will still occur in individual
channels? isnt it? I would appreciate your patience in writing a few
lines for its explanation. Thanks

Aitezaz