I have been working on digital QAM in FPGAs more than a decade and only now came to a basic fundamental self-questioning. We spend a lot of effort to process the I & Q channels in the digits until we finally hand them on to fellow RF engineer who will pass them through a mixer and into the real world. Nothing new here. The result of the mixer (though only one channel now) carries info on both original I/Q and mathematically, the mixer is a rectangular to polar converter. The question is: doesn't that also apply to any prior mixer? In many digital processing chains we do have prior stages of mixers, to separate channels and possibly to pre-assist the RF tuning then the same principle must apply i.e. the output of any mixer is the polar version of its rectangular input. So why do designers have to keep I/Q pair until RF stage, why not discard say the Q right from first mixer or so? I am assuming the signal phase stays linear and the signal is just delayed in time. Regards Kadhiem
QAM, back to basic question
Started by ●October 29, 2011
Reply by ●October 29, 20112011-10-29
On Oct 29, 6:52�am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> wrote:> I have been working on digital QAM in FPGAs more than a decade and only now > came to a basic fundamental self-questioning. > > We spend a lot of effort to process the I & Q channels in the digits until > we finally hand them on to fellow RF engineer who will pass them through a > mixer and into the real world. > > Nothing new here. The result of the mixer (though only one channel now) > carries info on both original I/Q and �mathematically, the mixer is a > rectangular to polar converter. > > The question is: doesn't that also apply to any prior mixer? > In many digital processing chains we do have prior stages of mixers, to > separate channels and possibly to pre-assist the RF tuning then the same > principle must apply i.e. the output of any mixer is the polar version of > its rectangular input. So why do designers have to keep I/Q pair until RF > stage, why not discard say the Q right from first mixer or so? I am > assuming the signal phase stays linear and the signal is just delayed in > time. > > Regards > > KadhiemI am confused about some of the stuff you are saying. I put together a flash tutorial on I Q modulation a few months ago. Perhaps this will help. http://www.fourier-series.com/IQMod/index.html
Reply by ●October 29, 20112011-10-29
>On Oct 29, 6:52=A0am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> >wrote: >> I have been working on digital QAM in FPGAs more than a decade and onlyn=>ow >> came to a basic fundamental self-questioning. >> >> We spend a lot of effort to process the I & Q channels in the digitsunti=>l >> we finally hand them on to fellow RF engineer who will pass them through=>a >> mixer and into the real world. >> >> Nothing new here. The result of the mixer (though only one channel now) >> carries info on both original I/Q and =A0mathematically, the mixer is a >> rectangular to polar converter. >> >> The question is: doesn't that also apply to any prior mixer? >> In many digital processing chains we do have prior stages of mixers, to >> separate channels and possibly to pre-assist the RF tuning then thesame>> principle must apply i.e. the output of any mixer is the polar versionof>> its rectangular input. So why do designers have to keep I/Q pair untilRF>> stage, why not discard say the Q right from first mixer or so? I am >> assuming the signal phase stays linear and the signal is just delayedin>> time. >> >> Regards >> >> Kadhiem > > I am confused about some of the stuff you are saying. I put together >a flash tutorial on I Q modulation a few months ago. Perhaps this >will help.Hi, Which stuff is confusing, that could help. I don't think it is about understanding IQ principles per se but I am asking specifically why don't we reduce the I/Q into one channel before RF stage. Kadhiem
Reply by ●October 29, 20112011-10-29
On Oct 29, 9:29�am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> wrote:> >On Oct 29, 6:52=A0am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> > >wrote: > >> I have been working on digital QAM in FPGAs more than a decade and only > n= > >ow > >> came to a basic fundamental self-questioning. > > >> We spend a lot of effort to process the I & Q channels in the digits > unti= > >l > >> we finally hand them on to fellow RF engineer who will pass them through > = > >a > >> mixer and into the real world. > > >> Nothing new here. The result of the mixer (though only one channel now) > >> carries info on both original I/Q and =A0mathematically, the mixer is a > >> rectangular to polar converter. > > >> The question is: doesn't that also apply to any prior mixer? > >> In many digital processing chains we do have prior stages of mixers, to > >> separate channels and possibly to pre-assist the RF tuning then the > same > >> principle must apply i.e. the output of any mixer is the polar version > of > >> its rectangular input. So why do designers have to keep I/Q pair until > RF > >> stage, why not discard say the Q right from first mixer or so? I am > >> assuming the signal phase stays linear and the signal is just delayed > in > >> time. > > >> Regards > > >> Kadhiem > > > I am confused about some of the stuff you are saying. I put together > >a flash tutorial on I Q modulation a few months ago. �Perhaps this > >will help. > > Hi, > > Which stuff is confusing, that could help. > > I don't think it is about understanding IQ principles per se but I am > asking specifically why don't we reduce the I/Q into one channel before RF > stage. > > KadhiemI think you are somewhat confused because I/Q modulation requires two mixers at the rf stage. If you are doing a strictly AM signal , you can do it with one mixer, but, I believe, that QAM effectively is a phase/amplitude modulation scheme. Wikipedia shows the QAM diagram as 16 states on the I/Q constellation plot. In order to turn the phase of the carrier you need two mixers with the appropriate I Q data. Bottom line - I think there are two mixers in the RF stage , whether you fully see it or not.
Reply by ●October 29, 20112011-10-29
>On Oct 29, 9:29=A0am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> >wrote: >> >On Oct 29, 6:52=3DA0am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> >> >wrote: >> >> I have been working on digital QAM in FPGAs more than a decade andonl=>y >> n=3D >> >ow >> >> came to a basic fundamental self-questioning. >> >> >> We spend a lot of effort to process the I & Q channels in the digits >> unti=3D >> >l >> >> we finally hand them on to fellow RF engineer who will pass themthrou=>gh >> =3D >> >a >> >> mixer and into the real world. >> >> >> Nothing new here. The result of the mixer (though only one channelnow=>) >> >> carries info on both original I/Q and =3DA0mathematically, the mixeri=>s a >> >> rectangular to polar converter. >> >> >> The question is: doesn't that also apply to any prior mixer? >> >> In many digital processing chains we do have prior stages of mixers,t=>o >> >> separate channels and possibly to pre-assist the RF tuning then the >> same >> >> principle must apply i.e. the output of any mixer is the polarversion>> of >> >> its rectangular input. So why do designers have to keep I/Q pairuntil>> RF >> >> stage, why not discard say the Q right from first mixer or so? I am >> >> assuming the signal phase stays linear and the signal is justdelayed>> in >> >> time. >> >> >> Regards >> >> >> Kadhiem >> >> > I am confused about some of the stuff you are saying. I put together >> >a flash tutorial on I Q modulation a few months ago. =A0Perhaps this >> >will help. >> >> Hi, >> >> Which stuff is confusing, that could help. >> >> I don't think it is about understanding IQ principles per se but I am >> asking specifically why don't we reduce the I/Q into one channel beforeR=>F >> stage. >> >> Kadhiem > >I think you are somewhat confused because I/Q modulation requires two >mixers at the rf stage. If you are doing a strictly AM signal , you >can do it with one mixer, but, I believe, that QAM effectively is a >phase/amplitude modulation scheme. Wikipedia shows the QAM diagram as >16 states on the I/Q constellation plot. In order to turn the phase >of the carrier you need two mixers with the appropriate I Q data. >Bottom line - I think there are two mixers in the RF stage , whether >you fully see it or not. >True the QAM RF mixer is complex and it receives I/Q inputs but eventually one output signal is passed to air (either output but surely not both). Am I right here? In this case we kept I/Q right to RF stage and processed two channels(I/Q) at double overheads. We could have as well applied a low frequency mixer and then extracted I output only well prior to that. The only issue I can think of in this scenario is that the RF designer has to deal with spectrum duplicate of single real mixer. Kadhiem
Reply by ●October 29, 20112011-10-29
kaz wrote:> So why do designers have to keep I/Q pair until RF > stage, why not discard say the Q right from first mixer or so? I am > assuming the signal phase stays linear and the signal is just delayed in > time.An analytical signal has one side spectrum. If you go to the real signal, you will have to suppress mirrored images by filters. This is a valid approach, too. Whatever is more appropriate method depends on application. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●October 29, 20112011-10-29
On Oct 29, 10:42�am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> wrote:> >On Oct 29, 9:29=A0am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> > >wrote: > >> >On Oct 29, 6:52=3DA0am, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> > >> >wrote: > >> >> I have been working on digital QAM in FPGAs more than a decade and > onl= > >y > >> n=3D > >> >ow > >> >> came to a basic fundamental self-questioning. > > >> >> We spend a lot of effort to process the I & Q channels in the digits > >> unti=3D > >> >l > >> >> we finally hand them on to fellow RF engineer who will pass them > throu= > >gh > >> =3D > >> >a > >> >> mixer and into the real world. > > >> >> Nothing new here. The result of the mixer (though only one channel > now= > >) > >> >> carries info on both original I/Q and =3DA0mathematically, the mixer > i= > >s a > >> >> rectangular to polar converter. > > >> >> The question is: doesn't that also apply to any prior mixer? > >> >> In many digital processing chains we do have prior stages of mixers, > t= > >o > >> >> separate channels and possibly to pre-assist the RF tuning then the > >> same > >> >> principle must apply i.e. the output of any mixer is the polar > version > >> of > >> >> its rectangular input. So why do designers have to keep I/Q pair > until > >> RF > >> >> stage, why not discard say the Q right from first mixer or so? I am > >> >> assuming the signal phase stays linear and the signal is just > delayed > >> in > >> >> time. > > >> >> Regards > > >> >> Kadhiem > > >> > I am confused about some of the stuff you are saying. I put together > >> >a flash tutorial on I Q modulation a few months ago. =A0Perhaps this > >> >will help. > > >> Hi, > > >> Which stuff is confusing, that could help. > > >> I don't think it is about understanding IQ principles per se but I am > >> asking specifically why don't we reduce the I/Q into one channel before > R= > >F > >> stage. > > >> Kadhiem > > >I think you are somewhat confused because I/Q modulation requires two > >mixers at the rf stage. �If you are doing a strictly AM signal , you > >can do it with one mixer, but, I believe, that QAM effectively is a > >phase/amplitude modulation scheme. Wikipedia shows the QAM diagram as > >16 states on the I/Q constellation plot. �In order to turn the phase > >of the carrier you need two mixers with the appropriate I Q data. > >Bottom line - I think there are two mixers in the RF stage , whether > >you fully see it or not. > > True the QAM RF mixer is complex and it receives I/Q inputs but eventually > one output signal is passed to air (either output but surely not both). > > Am I right here? In this case we kept I/Q right to RF stage and processed > two channels(I/Q) at double overheads. We could have as well applied a low > frequency mixer and then extracted I output only well prior to that. The > only issue I can think of in this scenario is that the RF designer has to > deal with spectrum duplicate of single real mixer. > > KadhiemIt is true that one output is passed to the air. But the carrier frequency must be manipulated in terms of phase and amplitude. The phase must be advanced sometimes and retarded sometimes and the amplitude must also be manipulated on the carrier. I/Q modulation breaks the carrier into two rf streams , then manipulates those two carrier streams with the I and Q digital data , then recombines the RF then sends out one data stream. If the carrier could be properly manipulated with one channel there would be no IQ modulation :-) Seriously, look at the flash programs I have. They might help.
Reply by ●October 29, 20112011-10-29
On Oct 29, 10:58�am, Vladimir Vassilevsky <nos...@nowhere.com> wrote:> kaz wrote: > > So why do designers have to keep I/Q pair until RF > > stage, why not discard say the Q right from first mixer or so? I am > > assuming the signal phase stays linear and the signal is just delayed in > > time. > > An analytical signal has one side spectrum. If you go to the real > signal, you will have to suppress mirrored images by filters. This is a > valid approach, too. Whatever is more appropriate method depends on > application. > > Vladimir Vassilevsky > DSP and Mixed Signal Design Consultanthttp://www.abvolt.comI doubt he knows what the hell an analytical signal is Vlad. You score a point for the peanut gallery.
Reply by ●October 29, 20112011-10-29
On 10/29/2011 10:42 AM, kaz wrote: ...> True the QAM RF mixer is complex and it receives I/Q inputs but eventually > one output signal is passed to air (either output but surely not both).True. There is no complex instantaneous value.> Am I right here? In this case we kept I/Q right to RF stage and processed > two channels(I/Q) at double overheads. We could have as well applied a low > frequency mixer and then extracted I output only well prior to that. The > only issue I can think of in this scenario is that the RF designer has to > deal with spectrum duplicate of single real mixer.The what-you-call duplicate spectrum is no small difficulty. The second mixer is cheaper than a filter that could replace it. Jerry -- Engineering is the art of making what you want from things you can get.
Reply by ●October 29, 20112011-10-29
On Sat, 29 Oct 2011 05:52:32 -0500, "kaz" <kadhiem_ayob@n_o_s_p_a_m.yahoo.co.uk> wrote:>I have been working on digital QAM in FPGAs more than a decade and only now >came to a basic fundamental self-questioning. > >We spend a lot of effort to process the I & Q channels in the digits until >we finally hand them on to fellow RF engineer who will pass them through a >mixer and into the real world. > >Nothing new here. The result of the mixer (though only one channel now) >carries info on both original I/Q and mathematically, the mixer is a >rectangular to polar converter. > >The question is: doesn't that also apply to any prior mixer? >In many digital processing chains we do have prior stages of mixers, to >separate channels and possibly to pre-assist the RF tuning then the same >principle must apply i.e. the output of any mixer is the polar version of >its rectangular input. So why do designers have to keep I/Q pair until RF >stage, why not discard say the Q right from first mixer or so? I am >assuming the signal phase stays linear and the signal is just delayed in >time. > >Regards > >KadhiemI don't think a mixer converts rectangular to polar, it just translates a signal in frequency. The signal can be represented in either rectangular or polar coordinates both before and after the mixer. Typically on a block diagram there are signal components labelled I and Q, both before and after some mixers, and if there is only one component coming out of a mixer (usually taken as the I component of the output), that means that only the real component of the output is needed to fully represent the signal at that point. Whenever the spectrum isn't symmetric about DC, which is true for a baseband signal or a signal that straddles DC in some fashion, then a full complex representation of the signal is required, in either rectangular or polar representation. If the spectrum is symmetric about DC, which it will be for real-valued signals, then only the real part is needed and the Q (or imaginary) component can be discarded. Many systems mix I and Q up to some IF frequency, and then only the real component is taken from there in the RF chain. Mixing the real-valued IF up to a higher RF necessarily involves filtering out the image frequencies created by the real-valued mix. An image-reject mixer uses both I and Q LO components to translate an analog signal in frequency, but they're not often used because it's difficult to keep the I and Q components exactly balanced in an analog circuit and the resulting distortion is often unacceptable. So it is possible to always use I and Q, even in the RF chain, and then just take the real part and send it out the antenna. It usually works out better to digitally translate the baseband I and Q signal away from DC so that only the real component is needed, so that only real-valued analog mixers are needed after that. This eliminates any distortion due to misbalances between the I and Q components of the analog mixing signal. Eric Jacobsen Anchor Hill Communications www.anchorhill.com
