> On Jun 11, 8:20 pm, Jerry Avins <j...@ieee.org> wrote:
>> I think he wants to know the difference between a complex filter (that
>> obviously has real and imaginary parts) and two filters, one for the
>> real part of the signal and another for the imaginary part. Do do I.
>
> cross products
Aha!
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by Vladimir Vassilevsky●June 12, 20072007-06-12
Jerry Avins wrote:
> I think he wants to know the difference between a complex filter (that
> obviously has real and imaginary parts) and two filters, one for the
> real part of the signal and another for the imaginary part. Do do I.
The equalizer must be the analytical function, i.e. it has to satisfy
the Cauchy-Riemann conditions. I.e. the complex combination of the real
filters is analytical only for the trivial cases.
Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
Reply by ●June 12, 20072007-06-12
On Jun 12, 12:22 am, "Ron N." <rhnlo...@yahoo.com> wrote:
> On Jun 11, 8:20 pm, Jerry Avins <j...@ieee.org> wrote:
>
> > I think he wants to know the difference between a complex filter (that
> > obviously has real and imaginary parts) and two filters, one for the
> > real part of the signal and another for the imaginary part. Do do I.
>
> cross products
Exactly. In the output of a complex filter, there will be some terms
that are proportional to the product of the imaginary part of the
coefficients and the real part of the input, and vice versa. This is a
product (no pun intended) of the complex multiplication involved in
calculating the output. If you do two separate real filters and sum
their outputs, you don't get these cross terms, so the overall
response is not the same as the complex filter.
Jason
Reply by Ron N.●June 12, 20072007-06-12
On Jun 11, 8:20 pm, Jerry Avins <j...@ieee.org> wrote:
> I think he wants to know the difference between a complex filter (that
> obviously has real and imaginary parts) and two filters, one for the
> real part of the signal and another for the imaginary part. Do do I.
cross products
Reply by Jerry Avins●June 12, 20072007-06-12
cincydsp@gmail.com wrote:
> On Jun 11, 6:58 pm, "John" <j...@excite.com> wrote:
>> My signal does not have a carrier, it is a baseband. do I still have to use
>> a complex equalizer ?"cpope" <cep...@nc.rr.com> wrote in message
>>
>
> Even though it's at baseband, in any practical system, your signal is
> probably going to be transmitted on a carrier at some point. Your
> receiver's LO is going to have some frequency error with respect to
> your transmitter's oscillator, which will cause a low-frequency
> complex sinusoidal modulation on your received symbols. Your equalizer
> can correct this if the error is small enough, but it needs both the
> real and imaginary components to do so (since the modulation caused by
> the frequency error is complex).
>
> Why are you so averse to implementing a complex equalizer anyway? The
> underlying structure is the same; you just need to allow all
> quantities to be complex and remember to put in complex conjugates at
> the right places, depending on the equalizer's type.
I think he wants to know the difference between a complex filter (that
obviously has real and imaginary parts) and two filters, one for the
real part of the signal and another for the imaginary part. Do do I.
Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
Reply by ●June 11, 20072007-06-11
On Jun 11, 6:58 pm, "John" <j...@excite.com> wrote:
> My signal does not have a carrier, it is a baseband. do I still have to use
> a complex equalizer ?"cpope" <cep...@nc.rr.com> wrote in message
>
Even though it's at baseband, in any practical system, your signal is
probably going to be transmitted on a carrier at some point. Your
receiver's LO is going to have some frequency error with respect to
your transmitter's oscillator, which will cause a low-frequency
complex sinusoidal modulation on your received symbols. Your equalizer
can correct this if the error is small enough, but it needs both the
real and imaginary components to do so (since the modulation caused by
the frequency error is complex).
Why are you so averse to implementing a complex equalizer anyway? The
underlying structure is the same; you just need to allow all
quantities to be complex and remember to put in complex conjugates at
the right places, depending on the equalizer's type.
Jason
Reply by John●June 11, 20072007-06-11
My signal does not have a carrier, it is a baseband. do I still have to use
a complex equalizer ?
"cpope" <cepope@nc.rr.com> wrote in message
news:466dcb50$0$17118$4c368faf@roadrunner.com...
>
> "Tom" <tomdarel@yahoo.com> wrote in message
> news:466d725c$0$12085$c3e8da3@news.astraweb.com...
>> I have two complex signals (Input and output) of a channel I need to
>> equalize. I can have 2 linear equalizers one for the real part and one
>> for
>> the imaginary and train them independently or I can use one complex
>> equalizer and train it using the complex signal. My questions are which
> one
>> is better? Is there any advantage of one on the other or they are
> equivalent
>> ?
>>
>> Regards
>>
>> Tom
>>
>>
>
> Unless you have absolutlely zero carrier offset (which is impossible if
> either receiver or transmitter is moving) you can not adapt real and
> imaginary independently. I think you have to use a complexe equalizer.
> That
> said, there are ways to implement the complex multiply with only three
> real
> multiplies, but I don't think that is where you were heading.
>
> -Clark
>
>
Reply by cpope●June 11, 20072007-06-11
Once you convert to magnitude and phase the operations are no longer linear,
right? The phase at least is modulo 2 pi.
The equalizer is just a filter. The filter can be applied in time or
frequency domain. In the frequency domain an FFT would need to convert the
time waveform to frequency domain, then multiply by the FFT of the equalizer
filter, and do an inverse FFT to get back to time.
But I don't see any way to convolve the data and the filter in
magnitude/phase terms?
-Clark
"Tom" <tomdarel@yahoo.com> wrote in message
news:466db4d4$0$10511$c3e8da3@news.astraweb.com...
> What if I transform the real and imaginary to Amplitude and phase signals
> and use them to train 2 equalizers separately and then transform back the
> output of the equalizers to real and Imaginary signals. Will this work in
> theory ?
>
> regards
>
> Tom
> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message
> news:mggbi.13524$2v1.1260@newssvr14.news.prodigy.net...
> > Becase the phase shift is different at the different frequencies.
> >
> > VLV
> >
> >
> > Tom wrote:
> >> Why the first option can not work ?
> >>
> >> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message
> >> news:8Aebi.25140$YL5.22280@newssvr29.news.prodigy.net...
> >>
> >>>
> >>>Tom wrote:
> >>>
> >>>
> >>>>I have two complex signals (Input and output) of a channel I need to
> >>>>equalize. I can have 2 linear equalizers one for the real part and one
> >>>>for the imaginary and train them independently
> >>>
> >>>^^^^^^^^^^^^^^^^^^^^^^
> >>>
> >>>This is completely wrong.
> >>>
> >>>
> >>>
> >>>>or I can use one complex equalizer and train it using the complex
> >>>>signal. My questions are which one
> >>>>is better? Is there any advantage of one on the other or they are
> >>>>equivalent
> >>>
> >>>If you are working with the real signal, you should use the real
> >>>equalizer. If you are working with the complex signal, you use the
> >>>complex equalizer. Which way is preferred depends on the application.
> >>>
> >>>
> >>>Vladimir Vassilevsky
> >>>
> >>>DSP and Mixed Signal Design Consultant
> >>>
> >>>http://www.abvolt.com
> >>
> >>
>
Reply by cpope●June 11, 20072007-06-11
"Tom" <tomdarel@yahoo.com> wrote in message
news:466d725c$0$12085$c3e8da3@news.astraweb.com...
> I have two complex signals (Input and output) of a channel I need to
> equalize. I can have 2 linear equalizers one for the real part and one for
> the imaginary and train them independently or I can use one complex
> equalizer and train it using the complex signal. My questions are which
one
> is better? Is there any advantage of one on the other or they are
equivalent
> ?
>
> Regards
>
> Tom
>
>
Unless you have absolutlely zero carrier offset (which is impossible if
either receiver or transmitter is moving) you can not adapt real and
imaginary independently. I think you have to use a complexe equalizer. That
said, there are ways to implement the complex multiply with only three real
multiplies, but I don't think that is where you were heading.
-Clark
Reply by Tom●June 11, 20072007-06-11
What if I transform the real and imaginary to Amplitude and phase signals
and use them to train 2 equalizers separately and then transform back the
output of the equalizers to real and Imaginary signals. Will this work in
theory ?
regards
Tom
"Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message
news:mggbi.13524$2v1.1260@newssvr14.news.prodigy.net...
> Becase the phase shift is different at the different frequencies.
>
> VLV
>
>
> Tom wrote:
>> Why the first option can not work ?
>>
>> "Vladimir Vassilevsky" <antispam_bogus@hotmail.com> wrote in message
>> news:8Aebi.25140$YL5.22280@newssvr29.news.prodigy.net...
>>
>>>
>>>Tom wrote:
>>>
>>>
>>>>I have two complex signals (Input and output) of a channel I need to
>>>>equalize. I can have 2 linear equalizers one for the real part and one
>>>>for the imaginary and train them independently
>>>
>>>^^^^^^^^^^^^^^^^^^^^^^
>>>
>>>This is completely wrong.
>>>
>>>
>>>
>>>>or I can use one complex equalizer and train it using the complex
>>>>signal. My questions are which one
>>>>is better? Is there any advantage of one on the other or they are
>>>>equivalent
>>>
>>>If you are working with the real signal, you should use the real
>>>equalizer. If you are working with the complex signal, you use the
>>>complex equalizer. Which way is preferred depends on the application.
>>>
>>>
>>>Vladimir Vassilevsky
>>>
>>>DSP and Mixed Signal Design Consultant
>>>
>>>http://www.abvolt.com
>>
>>