I have a question relating to injection-locked digital oscillators. Assume I have a DCO whose frequency control port is biased to produce some frequency F0. Also assume that I have a sine-wave signal whose frequency F1 is very close to F0. I wish to lock the DCO to F1 without using a pll by using injection locking. For injection locking to work, do I need to 1) add a small amount of F1 to the control port 2) add a small amount of both F0 and F1 to the control port 3) add the product of F1 and F2 to the control port Thanks for any pointers! Bob
Injection locked DCO
Started by ●June 25, 2013
Reply by ●June 25, 20132013-06-25
On Tue, 25 Jun 2013 10:12:38 -0700 (PDT), radams2000@gmail.com wrote:>I have a question relating to injection-locked digital oscillators. Assume = >I have a DCO whose frequency control port is biased to produce some frequen= >cy F0. Also assume that I have a sine-wave signal whose frequency F1 is ver= >y close to F0. I wish to lock the DCO to F1 without using a pll by using in= >jection locking. For injection locking to work, do I need to > >1) add a small amount of F1 to the control port >2) add a small amount of both F0 and F1 to the control port >3) add the product of F1 and F2 to the control port > >Thanks for any pointers! > >BobMany oscillators are designed to be resistant to injection locking, since it is often an undesirable effect, so it'll probably depend on the implementation of your DCO. Are you using a commercial part or something you built? And what do you mean by DCO? Is it a DDS-based oscillator or a digitally controlled synthesizer or what? Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●June 25, 20132013-06-25
On Tue, 25 Jun 2013 10:12:38 -0700, radams2000 wrote:> I have a question relating to injection-locked digital oscillators. > Assume I have a DCO whose frequency control port is biased to produce > some frequency F0. Also assume that I have a sine-wave signal whose > frequency F1 is very close to F0. I wish to lock the DCO to F1 without > using a pll by using injection locking. For injection locking to work, > do I need to > > 1) add a small amount of F1 to the control port 2) add a small amount of > both F0 and F1 to the control port 3) add the product of F1 and F2 to > the control port > > Thanks for any pointers!That depends heavily on exactly how your DCO actually works. If it consists of a phase accumulator followed by a sine generator, then your proposed scheme won't make a damn bit of difference to the frequency, although the output would be distorted. I think you'd need a DCO consisting of an actual IIR filter with adjustable poles and some sort of AGC to maintain amplitude, and I think that you'd need to inject your signal into one of the states, not the frequency input. I think that by the time you were done making your "simple" injection- locked digital oscillator you'd have a much more complex bit of algorithm than you'd have with a simple PLL and a DCO based on a phase accumulator. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●June 25, 20132013-06-25
On 6/25/2013 12:12 PM, radams2000@gmail.com wrote:> I have a question relating to injection-locked digital oscillators.Why would anyone need this?> Assume I have a DCO whose frequency control port is biased to produce > some frequency F0. Also assume that I have a sine-wave signal whose > frequency F1 is very close to F0. I wish to lock the DCO to F1 > without using a pll by using injection locking. For injection locking > to work, do I need toAny DCO could be seen as state machine which cycles at every T. If you alter the state at every T1, you can make machine cycle at T1.> 1) add a small amount of F1 to the control portYes. 2) add a small amount of both F0 and F1 to the control port Maybe. 3) add the product of F1 and F2 to the control port Maybe. Depends on the structure. Vladimir Vassilevsky DSP and Mixed Signal Designs www.abvolt.com
Reply by ●June 25, 20132013-06-25
Doesn't the injection signal couple directly to the DCO/VCO at the oscillation frequency (as opposed to the control input)? See slide 7, bottom right corner. http://rfic.eecs.berkeley.edu/~niknejad/ee242/pdf/eecs242_lect26_injectionlocking.pdf Intuition tells me that this is not the way to design a robust circuit, but maybe I'm too narrow-minded... _____________________________ Posted through www.DSPRelated.com
Reply by ●June 25, 20132013-06-25
On Tue, 25 Jun 2013 14:02:45 -0500, "mnentwig" <24789@dsprelated> wrote:>Doesn't the injection signal couple directly to the DCO/VCO at the >oscillation frequency (as opposed to the control input)?That's the idea. There's not usually an "injection lock input", since the idea of injection locking is that the systems influence each other in sympathetic ways and lock incidentally. The example with the metronomes is a pretty good one: http://www.youtube.com/watch?v=W1TMZASCR-I>See slide 7, bottom right corner. >http://rfic.eecs.berkeley.edu/~niknejad/ee242/pdf/eecs242_lect26_injectionlocking.pdf > >Intuition tells me that this is not the way to design a robust circuit, but >maybe I'm too narrow-minded...Usually injection locking is weak in that it can be broken more easily than a lock made via a PLL. But, since the OP hasn't given much info on the DCO being used, it's hard to say what's up with it. Injection locking is often something one wishes to avoid, so many systems are designed to be resistant to it. This may make it difficult for the OP to intentionally injection-lock a component designed to resist it.> >_____________________________ >Posted through www.DSPRelated.comEric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by ●June 25, 20132013-06-25
Reply by ●June 25, 20132013-06-25
More details .... I have a seperate frequency estimation loop that gets the DCO frequency within .2%. The reason that I can't use a pll is that the signal I want to lock to is very rich in harmonics, so a simple slicer will not extract a square wave at the fundamental. So I was hoping that injecting it into the control port in some way, either linear or non-linear, would pull it the last .2% Bob
Reply by ●June 25, 20132013-06-25
On Tue, 25 Jun 2013 17:55:44 -0700, radams2000 wrote:> More details .... > I have a seperate frequency estimation loop that gets the DCO frequency > within .2%. The reason that I can't use a pll is that the signal I > want to lock to is very rich in harmonics, so a simple slicer will not > extract a square wave at the fundamental. So I was hoping that injecting > it into the control port in some way, either linear or non-linear, would > pull it the last .2%Eh? You seem to be suffering from a fundamental misunderstanding about phase locked loops. No way, no how, do you need to derive a square wave from your signal before phase-locking to it. Presumably you are assuming that you have to use a phase detector that depends on edges. This just isn't so, particularly if your frequency is fairly close. If the fundamental is present, and if the ratio of sampling rate to fundamental is high enough, and if you can stand a loop bandwidth that's not too much smaller than your largest possible frequency error, then multiply the signal by a sine wave from your DCO, and use that. Life should be wonderful. -- Tim Wescott Control system and signal processing consulting www.wescottdesign.com
Reply by ●June 25, 20132013-06-25
