Hi, We have been using the algorithm for symbol timing error detection as proposed by gardner in our system, now for some reason, we have to reduce the bandwidth by decreasing the roll off factor of the RRC filter to 0.3 As suggested by gardner in his paper, that algorithm should not be used when roll off is less than 0.4 we have to see some other methods now. Is there any modification of gardner that anyone has seen to be working for lower roll off values? This is for a qpsk system. --------------------------------------- Posted through http://www.DSPRelated.com
gardner symbol timing recovery algorithm
Started by ●February 16, 2016
Reply by ●February 16, 20162016-02-16
On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> wrote:>Hi, > >We have been using the algorithm for symbol timing error detection as >proposed by gardner in our system, now for some reason, we have to reduce >the bandwidth by decreasing the roll off factor of the RRC filter to 0.3 >As suggested by gardner in his paper, that algorithm should not be used >when roll off is less than 0.4 we have to see some other methods now. >Is there any modification of gardner that anyone has seen to be working >for lower roll off values? >This is for a qpsk system. >--------------------------------------- >Posted through http://www.DSPRelated.comWhat was the original motivation for using Gardner? Is the system not coherent? Mueller-Muller is also very commonly used and works well. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●February 16, 20162016-02-16
The system is non coherent. I am not sure why gardner was used to be honest. I have associated with the project recently. Although I am exploring possibilty to use Mueller-Muller also, it would be preferable to make minimum changes to the system at this stage for us. If it can be made to work with some modifications to the gardner itself, it would be great. I am looking at a few IEEE papers also for modifications on gardner, but just wanted to ask if anyone has worked with any such modification already.>On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> >wrote: > >>Hi, >> >>We have been using the algorithm for symbol timing error detection as >>proposed by gardner in our system, now for some reason, we have toreduce>>the bandwidth by decreasing the roll off factor of the RRC filter to0.3>>As suggested by gardner in his paper, that algorithm should not beused>>when roll off is less than 0.4 we have to see some other methods now. >>Is there any modification of gardner that anyone has seen to be working >>for lower roll off values? >>This is for a qpsk system. >>--------------------------------------- >>Posted through http://www.DSPRelated.com > >What was the original motivation for using Gardner? Is the system >not coherent? Mueller-Muller is also very commonly used and works >well. > > >Eric Jacobsen >Anchor Hill Communications >http://www.anchorhill.com--------------------------------------- Posted through http://www.DSPRelated.com
Reply by ●February 16, 20162016-02-16
On 16.02.2016 21:57, Eric Jacobsen wrote:> On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> > wrote: > >> Hi, >> >> We have been using the algorithm for symbol timing error detection as >> proposed by gardner in our system, now for some reason, we have to reduce >> the bandwidth by decreasing the roll off factor of the RRC filter to 0.3 >> As suggested by gardner in his paper, that algorithm should not be used >> when roll off is less than 0.4 we have to see some other methods now. >> Is there any modification of gardner that anyone has seen to be working >> for lower roll off values? >> This is for a qpsk system. >> --------------------------------------- >> Posted through http://www.DSPRelated.com > > What was the original motivation for using Gardner? Is the system > not coherent? Mueller-Muller is also very commonly used and works > well. > > > Eric Jacobsen > Anchor Hill Communications > http://www.anchorhill.com >(1) It's possible to reduce self-noise by using pre-filtering. See 1991 "Zero-crossing DPLL bit synchronizer with pattern jitter compensation", 1993 "Design and analysis of a jitter-free clock recovery scheme for QAM systems", 1993 "Jitter-reduced Digital Timing Recovery for Multilevel PAM and QAM Systems", 1996 "Optimization of symbol timing recovery for QAM data demodulators". (2) There are two known modifications of the Gardner TED algorithm with reduced self-noice for low roll-off factors. See 2004 "A modified Gardner detector for symbol timing recovery of M-PSK signals" and 2006 "Extended Gardner Detector for Improved Symbol-Timing Recovery of M -PSK Signals". Hope this helps (but in case you've been actually exploring the literature I guess you know the answer already). Regards, Evgeny.
Reply by ●February 16, 20162016-02-16
On 16.02.2016 22:17, mohitdaksh wrote:> I am looking at a few IEEE papers also for > modifications on gardner, but just wanted to ask if anyone has worked > with any such modification already.Perhaps. ;-) Regards, Evgeny.
Reply by ●February 16, 20162016-02-16
On Tue, 16 Feb 2016 18:57:29 +0000, Eric Jacobsen wrote:> On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> > wrote: > >>Hi, >> >>We have been using the algorithm for symbol timing error detection as >>proposed by gardner in our system, now for some reason, we have to >>reduce the bandwidth by decreasing the roll off factor of the RRC filter >>to 0.3 As suggested by gardner in his paper, that algorithm should not >>be used when roll off is less than 0.4 we have to see some other methods >>now. >>Is there any modification of gardner that anyone has seen to be working >>for lower roll off values? >>This is for a qpsk system. >>--------------------------------------- >>Posted through http://www.DSPRelated.com > > What was the original motivation for using Gardner? Is the system not > coherent? Mueller-Muller is also very commonly used and works well.OK, dumb question here (dumb as in -- I should know the answer) What's the meaning of "coherent" in this context? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●February 16, 20162016-02-16
On Tue, 16 Feb 2016 16:09:22 -0600, Tim Wescott <seemywebsite@myfooter.really> wrote:>On Tue, 16 Feb 2016 18:57:29 +0000, Eric Jacobsen wrote: > >> On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> >> wrote: >> >>>Hi, >>> >>>We have been using the algorithm for symbol timing error detection as >>>proposed by gardner in our system, now for some reason, we have to >>>reduce the bandwidth by decreasing the roll off factor of the RRC filter >>>to 0.3 As suggested by gardner in his paper, that algorithm should not >>>be used when roll off is less than 0.4 we have to see some other methods >>>now. >>>Is there any modification of gardner that anyone has seen to be working >>>for lower roll off values? >>>This is for a qpsk system. >>>--------------------------------------- >>>Posted through http://www.DSPRelated.com >> >> What was the original motivation for using Gardner? Is the system not >> coherent? Mueller-Muller is also very commonly used and works well. > >OK, dumb question here (dumb as in -- I should know the answer) > >What's the meaning of "coherent" in this context?Whether the constellation is phase-locked or not. Since the OP indicated that the system is not coherent, a "rotationally invariant" TED is needed, which Gardner is usually the first choice, even though it's not quite actually "rotationally invariant". Mueller-Muller is often not the best choice for a non-coherent system. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●February 16, 20162016-02-16
On Tue, 16 Feb 2016 13:17:44 -0600, "mohitdaksh" <111830@DSPRelated> wrote:> >The system is non coherent. > >I am not sure why gardner was used to be honest. I have associated with >the project recently. >Although I am exploring possibilty to use Mueller-Muller also, it would >be preferable to make minimum changes to the system at this stage for us. >If it can be made to work with some modifications to the gardner itself, >it would be great. I am looking at a few IEEE papers also for >modifications on gardner, but just wanted to ask if anyone has worked >with any such modification already.Gardner was probably used because the system is non-coherent. Mueller-Muller may not be a good choice for such a system without modification. As mentioned by others, there are a lot of well-documented modifications to the Gardner TED that may suit your needs. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●February 16, 20162016-02-16
On 17.02.2016 1:23, Eric Jacobsen wrote:> On Tue, 16 Feb 2016 16:09:22 -0600, Tim Wescott > <seemywebsite@myfooter.really> wrote: > >> On Tue, 16 Feb 2016 18:57:29 +0000, Eric Jacobsen wrote: >> >>> On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> >>> wrote: >>> >>>> Hi, >>>> >>>> We have been using the algorithm for symbol timing error detection as >>>> proposed by gardner in our system, now for some reason, we have to >>>> reduce the bandwidth by decreasing the roll off factor of the RRC filter >>>> to 0.3 As suggested by gardner in his paper, that algorithm should not >>>> be used when roll off is less than 0.4 we have to see some other methods >>>> now. >>>> Is there any modification of gardner that anyone has seen to be working >>>> for lower roll off values? >>>> This is for a qpsk system. >>>> --------------------------------------- >>>> Posted through http://www.DSPRelated.com >>> >>> What was the original motivation for using Gardner? Is the system not >>> coherent? Mueller-Muller is also very commonly used and works well. >> >> OK, dumb question here (dumb as in -- I should know the answer) >> >> What's the meaning of "coherent" in this context? > > Whether the constellation is phase-locked or not. Since the OP > indicated that the system is not coherent, a "rotationally invariant" > TED is needed, which Gardner is usually the first choice, even though > it's not quite actually "rotationally invariant". Mueller-Muller is > often not the best choice for a non-coherent system. > > > Eric Jacobsen > Anchor Hill Communications > http://www.anchorhill.com >Eric, what do you mean by saying that Gardner TED is not "quite actually 'rotationally invariant'"? It's true that its output does not depend on carrier phase in case of zero carrier frequency error. It's also true that it's robust against moderate carrier frequency errors. Regards, Evgeny.
Reply by ●February 16, 20162016-02-16
On Tue, 16 Feb 2016 22:23:09 +0000, Eric Jacobsen wrote:> On Tue, 16 Feb 2016 16:09:22 -0600, Tim Wescott > <seemywebsite@myfooter.really> wrote: > >>On Tue, 16 Feb 2016 18:57:29 +0000, Eric Jacobsen wrote: >> >>> On Tue, 16 Feb 2016 09:54:00 -0600, "mohitdaksh" <111830@DSPRelated> >>> wrote: >>> >>>>Hi, >>>> >>>>We have been using the algorithm for symbol timing error detection as >>>>proposed by gardner in our system, now for some reason, we have to >>>>reduce the bandwidth by decreasing the roll off factor of the RRC >>>>filter to 0.3 As suggested by gardner in his paper, that algorithm >>>>should not be used when roll off is less than 0.4 we have to see some >>>>other methods now. >>>>Is there any modification of gardner that anyone has seen to be >>>>working for lower roll off values? >>>>This is for a qpsk system. >>>>--------------------------------------- >>>>Posted through http://www.DSPRelated.com >>> >>> What was the original motivation for using Gardner? Is the system >>> not coherent? Mueller-Muller is also very commonly used and works >>> well. >> >>OK, dumb question here (dumb as in -- I should know the answer) >> >>What's the meaning of "coherent" in this context? > > Whether the constellation is phase-locked or not. Since the OP > indicated that the system is not coherent, a "rotationally invariant" > TED is needed, which Gardner is usually the first choice, even though > it's not quite actually "rotationally invariant". Mueller-Muller is > often not the best choice for a non-coherent system.OK. I'm particularly dense today. What I think of as QPSK is that one sends A * cos(omega * t + phi(t)) where omega is the "carrier" frequency and phi(t) takes on four values depending on the bit values, and only changes on T_sym. At the receiver, omega is unknown, there's a random (and possibly slowly varying) phase shift added in, one doesn't exactly know T_sym, and there's an offset added to T_sym. So one must recover carrier and timing independently. So in this context, what makes a system coherent or not? TIA for your patience. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
