WHY LOOP FILTER IN PLL FREQUENCY SYNTHESIZER

pll site:www.edaboard.com

HELLO,ALL

WHY LOOP FILTER USED IN PLL FREQUENCY SYNTHESIZER,HOW CHOOSE THE BANDWIDTH,ORDER.

PLEASE GIVE APPLICATION NOTES,DESIGN COCEPTS.


THANK YOU TO ALL.

Loop filter is used to filter out the unwanted spur and also suppress noise of the control line for VCO. This helps the overall phase noise. Typically the trade off is settling time, stabilitiy and noise requirement

You may go to the following website and look at PLL documents
http://www.circuitsage.com/pll.html

the loop filter , is used to filter the output of the phase detector and remove the high frequency components on it
the BW of the filter determine many things , like lock time , settling time , loop stability , phase nose performance

u need to check dean book , from national it has a very good treatment

kouly

THANK YOU FOR YOUR HELP,
IF YOU HAVE THE SOFT COPY OF THE BOOK,PLEASE SEND MAIL.


THANK YOU,khouly

u can find it here , on the edaboard


khouly

The point is VCO works with a DC control signal, the output of the PD/PFD is not DC. So you use a loop filter to pass the DC only -> Low pass

You should either eliminate f or 2f depending on your phase detector -> design the filter accordingly.

yeah , this is another purpose of the filter is attenuate reference spurs as Sadegh said

khouly

thank you

Because we should get a clear Vtune, the order and size of the loopfilter are up to the Kvco, Icp, Fvco, and bandwidth of loop filter etc.

as for as loopbandwidth is concerned it must be less than 1/10 of reference frequency.

but practically loopbandwith is around 1/40 0f reference fre.

The main reason you need a loop filter is to keep the PLL stable. In terms of control theory, an oscillator has a transfer function of K/S, which has 90 degrees of phase shift. The return path has an integrator, or a quasi-integrator if passive, that also has a 90 degree phase shift as frequency goes higher. Add them both up and your open loop transfer function GH(s) has a 180 degree phase shift in it at some (or many) frequencies. Since the closed loop transfer function is
G(S)/(1+GH(S)), where the angle of GH(S) is now 180 degrees, you can see that the closed loop transfer function will have infinite gain at the frequency where the magnitude of GH is unity. That is, the PLL is guaranteed to oscillate/be unstable.

That is the only significant reason there is a loop filter. Other things like spur reduction, etc, are a secondary benefit only.

The loop filter is designed so that the angle of GH(S) goes from near 180 degrees back up to something closer to 90 degrees right around the frequency range where GH(S) magnitude is going to unity. This is a requirement for the PLL to meet if it is going to be a stable control loop.

biff44 said:

The main reason you need a loop filter is to keep the PLL stable. In terms of control theory, an oscillator has a transfer function of K/S, which has 90 degrees of phase shift. The return path has an integrator, or a quasi-integrator if passive, that also has a 90 degree phase shift as frequency goes higher. Add them both up and your open loop transfer function GH(s) has a 180 degree phase shift in it at some (or many) frequencies. Since the closed loop transfer function is
G(S)/(1+GH(S)), where the angle of GH(S) is now 180 degrees, you can see that the closed loop transfer function will have infinite gain at the frequency where the magnitude of GH is unity. That is, the PLL is guaranteed to oscillate/be unstable.

That is the only significant reason there is a loop filter. Other things like spur reduction, etc, are a secondary benefit only.

The loop filter is designed so that the angle of GH(S) goes from near 180 degrees back up to something closer to 90 degrees right around the frequency range where GH(S) magnitude is going to unity. This is a requirement for the PLL to meet if it is going to be a stable control loop.

Click to expand...

Sorry, I don´t agree with these explanations, because they can in no way justify the necessity of a loop filter.
The return path has an integrator, or a quasi-integrator if passive
Why is there an integrator ? If there is something like an "quasi-integrator" then this is the loop filter !
That is, the PLL is guaranteed to oscillate/be unstable.
That is the only significant reason there is a loop filter.
No, if the PLL is already unstable, an additional lowpass loop filter never can improve the situation. In the contrary, in makes the phase shift larger !

My explanation in short:
1.) Based on principles of control theory a kind of controller is necessary in each controlled loop to set the dynamic properties of the system.
2.) A lowpass filter is in most cases necessary for the PLL system to suppress unwanted outputs from the PD (sum of both frequencies)
3.) Thus, the controller consists of a low pass which must (a) ensure stabilty and good dynamic behaviour and (b) perform filtering.
4.) Because of the VCO phase shift of 90 deg in most cases a first order lowpass is used which - for stability reasons - exhibits a additional zero (lag filter).
5.) It is very rare, that in special cases a loop filter of second order (with a zero) can be used.
6.) Even if no filtering is required, a controller with a lowpass behaviour makes sense because it makes a second order loop which always has better dynamic properties (settling time,...) tha a first order loop.

sure sport!

biff44 said:

sure sport!

Click to expand...

I´m very sorry - but I don´t understand this comment (English is not my mother language).

It is a expression of incedulity.

Make yourself a PLL without a loop filter--For instance, a charge pump with a capacitor but no accompanying resistors. Turn it on, make some perturbation, like a phase step in the reference frequency, and tell me what the response is.

biff44 said:

It is a expression of incedulity.
Make yourself a PLL without a loop filter--For instance, a charge pump with a capacitor but no accompanying resistors. Turn it on, make some perturbation, like a phase step in the reference frequency, and tell me what the response is.

Click to expand...

Well, here is my answer: There will be a very small stability margin (under worst case conditions: zero or even less), because in the system as described by you the capacitor alone is NOT an appropriate loop filter.
Remember the original question of the topic: Why loop filter in the PLL ?. And, of course, the cap behind the charge pump is part of this filter. Therefore, I consider your former answer (25th of July) “The main reason you need a loop filter is to keep the PLL stable“ still as not correct.

As you have put a question on me - may I ask you also something:
1.) Do you know any written contribution or book on PLL theory in which the capacitor behind the CP is NOT regarded as part of the loop filter ?
2.) Quote (your reply 25.07): The return path has an integrator, or a quasi-integrator if passive....Which part or unit of the return path do you mean ?

My summary: I think the following answer to the original question of "ravisankar" is still valid: Without any loop filter the PLL is of first order and will in most - if not in all - cases not function properly. However, as a first order system of course it is always stable !
Then, the loop filter (mostly of first or second order with a properly placed zero) is necessary (1) to smooth the VCO control signal and to supress unwanted frequencies created during the PD process, and (2) to set the desired dynamic properties of the whole loop.

Last question: What is the reason for your "incredulity" ?

I always had trouble with the text books explanation of the Loop Filter. It was not
too practical , I understood it but it did not have "REAL" meaning to me.

I understand the loop filter as: A method to take a square wave pulse train (from
the Charge pump) and psuedo-rectifiy it to a DC(RMS) value, that will be used to
Tune the VCO (Varactor Diode) to a wanted Freq. Because the output of the Charge
Pump (Phase Detector) is a Square wave/Pulse that means that (Fourier analysis)
there are high Freq components and the Loop Filter (low pass) will also filter that
out. Thus we get a ~DC voltage with very suppressed High Freq Spurs (That will
wreak havoc in the system).

Now this answer I gave makes practical sense (to me), but as I have to work
on debugging VCO circuits I think my definition does not give me the
insight/understanding I need to debug. Can anyone expand upon my
explanation and fill in all the blanks I'm missing?
Thanks

In principle, you are right, however, some important comments are to be made.

A method to take a square wave pulse train (from
the Charge pump) and psuedo-rectifiy it to a DC(RMS) value, that will be used to
Tune the VCO (Varactor Diode) to a wanted Freq. Because the output of the Charge
Pump (Phase Detector) is a Square wave/Pulse that means that (Fourier analysis)
there are high Freq components and the Loop Filter (low pass) will also filter that
out. Thus we get a ~DC voltage with very suppressed High Freq Spurs (That will
wreak havoc in the system).

1.) Normally, the lowpass is only of first order because otherwise the loop cannot be stable (90 deg VCO phase shift and maximal additional 90 deg from the lowpass). More than that, because of these stability problems, the low pass must exhibit a zero in order to enhance the phase at the critical frequency region.
2.) Therefore, the filtering of higher frequency components is very bad and an additioinal filter circuitry is necessary with a rather high pole (corner frequency) which has only minor influence on the stability properties.
3.) These explanations apply to all kind of phase detectors - not only to the charge pump detector. However, the filter circuitry may look different for the various PDs or PFDs which exist.
4.) In summary, the loop filter exhibiting a special kind of low pass response has two tasks: (1) to attenuate unwanted frequencies and (2) to establish the dynamic behaviour of the control loop according to the rules of the control theory.

Well, if you wish, try to make a pll with a charge pump but without a capacitor. Let me know how it goes!

A charge pump plus capacitor forms a phase detector. A charge pump without a capacitor is useless, as it does not detect phase.

Let me know what other fairytale circuitry you wish to consider.

biff44 said:

Well, if you wish, try to make a pll with a charge pump but without a capacitor. Let me know how it goes!
A charge pump plus capacitor forms a phase detector. A charge pump without a capacitor is useless, as it does not detect phase.
Let me know what other fairytale circuitry you wish to consider.

Click to expand...

Yes, I totally agree as each control loop without a suitable controller is useless. But that´s common knowledge.
However, I doubt if this "answer" is helpful for the questioner element_115

Added after 3 hours 13 minutes:

biff44 said:

Well, if you wish, try to make a pll with a charge pump but without a capacitor. Let me know how it goes!
A charge pump plus capacitor forms a phase detector. A charge pump without a capacitor is useless, as it does not detect phase.
Let me know what other fairytale circuitry you wish to consider.

Click to expand...

To bif44:
Why are you so polemic ? For my opinion, engineers should be able to answer questions and to discuss different viewpoints in a more objective way.

As far as fairytales are concerned: There is a paper from a man named F.M. Gardner (known to people involved in PLL techniques as a "PLL-guru") which tells us something regarding "fairytales":

Excerpt from IEEE Trans Comm, COM-28, Nov. 1980
F.M. Gardner: Charge-Pump Phase-Lock Loops

Using Fig. 2(b), a pump current IP is delivered to the filter impedance ZF.....
.....
A large preponderance of applications utilize second-order PLL's. To obtain a zero-stabilized, second-order loop, consider a loop filter function
ZF2(s)=R2+1/sC
which is produced by a series connection of a resistor and a capacitor.
............
The frequency jumps inherent to the second order loop usually cannot be accepted and additional filtering is often included within the PLL in order to mitigate the ripple. The simplest ripple filter is an additional capacitor C3 in parallel with the earlier RC impedance
(end of quote).

Any further comments ?