There is a large literature on estimating and equalizing comms channels with everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result. That would give you the coefficients of a ma model whether it is min or nm phase. I assume the problem is noise but that too should average out.
Equalization
Started by ●February 27, 2014
Reply by ●February 27, 20142014-02-27
>There is a large literature on estimating and equalizing comms channelswith everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result.>That would give you the coefficients of a ma model whether it is min or nmphase. I assume the problem is noise but that too should average out.>Sounds good in theory, but it wouldn't make a very practical training signal. -Doug _____________________________ Posted through www.DSPRelated.com
Reply by ●February 27, 20142014-02-27
On Thursday, February 27, 2014 11:56:32 AM UTC-5, gyans...@gmail.com wrote:> There is a large literature on estimating and equalizing comms channels with everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result. > > That would give you the coefficients of a ma model whether it is min or nm phase. I assume the problem is noise but that too should average out.The issue with using an impulse is that all of your energy is concentrated into one instant of time. If your system doesn't possess a high dynamic range, then to avoid nonlinear effects you have to severely limit the strength of the impulse. So to get around this, use a signal that spreads the energy over time. For example analog tv used a frequency chirp on one of the video lines near the top of the screen to train the echo canceller (ghost removal). Other systems use pseudo random sequences for training. Clay
Reply by ●February 27, 20142014-02-27
On Friday, February 28, 2014 8:57:19 AM UTC+13, clays...@gmail.com wrote:> On Thursday, February 27, 2014 11:56:32 AM UTC-5, gyans...@gmail.com wrote: > > > There is a large literature on estimating and equalizing comms channels with everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result. > > > > > > That would give you the coefficients of a ma model whether it is min or nm phase. I assume the problem is noise but that too should average out. > > > > The issue with using an impulse is that all of your energy is concentrated into one instant of time. If your system doesn't possess a high dynamic range, then to avoid nonlinear effects you have to severely limit the strength of the impulse. > > > > So to get around this, use a signal that spreads the energy over time. For example analog tv used a frequency chirp on one of the video lines near the top of the screen to train the echo canceller (ghost removal). > > > > Other systems use pseudo random sequences for training. > > > > Clayah yes, makes sense, thank you
Reply by ●February 27, 20142014-02-27
On Thu, 27 Feb 2014 14:51:49 -0800 (PST), gyansorova@gmail.com wrote:>On Friday, February 28, 2014 8:57:19 AM UTC+13, clays...@gmail.com wrote: >> On Thursday, February 27, 2014 11:56:32 AM UTC-5, gyans...@gmail.com wrote: >> >> > There is a large literature on estimating and equalizing comms channels with everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result. >> >> > >> >> > That would give you the coefficients of a ma model whether it is min or nm phase. I assume the problem is noise but that too should average out. >> >> >> >> The issue with using an impulse is that all of your energy is concentrated into one instant of time. If your system doesn't possess a high dynamic range, then to avoid nonlinear effects you have to severely limit the strength of the impulse. >> >> >> >> So to get around this, use a signal that spreads the energy over time. For example analog tv used a frequency chirp on one of the video lines near the top of the screen to train the echo canceller (ghost removal). >> >> >> >> Other systems use pseudo random sequences for training. >> >> >> >> Clay > >ah yes, makes sense, thank youIt's essentially the same reasons that radar systems don't use impulses. It's much harder to get a decent amount of power in an impulse than in a longer signal. Spreading the power out over time with a signal that has good autocorrelation properties allows the channel (or targets) to be estimated with good resolution. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●March 3, 20142014-03-03
>There is a large literature on estimating and equalizing comms channelswith everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result.>That would give you the coefficients of a ma model whether it is min or nmphase. I assume the problem is noise but that too should average out.>Also remember that wireless communication systems are bandlimited. Firing a very short duration impulse implies utilizing a very large bandwidth, which is not allowed, even for channel estimation purposes only. Then, there come other issues with the analog and digital front-ends, originally designed to operate for the actual bandwidth used by information signal only. _____________________________ Posted through www.DSPRelated.com
Reply by ●March 4, 20142014-03-04
On Thursday, February 27, 2014 11:56:32 AM UTC-5, gyans...@gmail.com wrote:> There is a large literature on estimating and equalizing comms channels with everything from LMS to higher order stats. It occurred to me, why not just fire an impulse or string of them into the channel instead and then average the result. > > That would give you the coefficients of a ma model whether it is min or nm phase. I assume the problem is noise but that too should average out.Most systems are limited by peak power. Most systems are limited by dynamic range. Most algorithms suffer from poor time-bandwidth product.
Reply by ●March 4, 20142014-03-04
it works for the geologists... engineers whine about peak-to-average ratio... real men use dynamite... _____________________________ Posted through www.DSPRelated.com
Reply by ●March 4, 20142014-03-04
On Tue, 04 Mar 2014 14:42:18 -0600, "mnentwig" <24789@dsprelated> wrote:>it works for the geologists... >engineers whine about peak-to-average ratio... real men use dynamite...That's a very special class of Power Amplifier, though. ;) Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●March 4, 20142014-03-04
On Tuesday, March 4, 2014 7:46:35 PM UTC-5, Eric Jacobsen wrote:> On Tue, 04 Mar 2014 14:42:18 -0600, "mnentwig" <24789@dsprelated> > > wrote: > > > > >it works for the geologists... > > >engineers whine about peak-to-average ratio... real men use dynamite... > > > > That's a very special class of Power Amplifier, though. ;) > > > > > > Eric Jacobsen > > Anchor Hill Communications > > http://www.anchorhill.comReal men use this kind of Power Amplifier. http://www.slb.com/services/drilling/mwd_lwd/mwd/powerpulse.aspx