Reply-To: "Kevin Aylward"
From: "Kevin Aylward"
Subject: Re: How to increase PLL order?
X-Newsreader: Microsoft Outlook Express 6.00.2800.1106
X-Inktomi-Trace: public1-pete2-5-cust33.pete.broadband.ntl.com 1039603616 29719 126.96.36.199 (11 Dec 2002 10:46:56 GMT)
Date: Wed, 11 Dec 2002 10:32:56 -0000
NNTP-Posting-Date: Wed, 11 Dec 2002 10:46:57 GMT
Christopher R. Carlen wrote:
> Kevin Aylward wrote:
>>> I think that depends on what the transfer function of the VCO is.
>> No it does not.
>>> course, we are used to Kvco/s, but for motors things go haywire.
>>> I am using the Z-state detector from a 4046, and it doesn't produce
>>> zero phase error with a motor, though it most certainly does with a
>>> normal VCO.
>> Then something is amiss, however, what do you mean by the Z detector?
> The Z-state detector is what Dan Wolaver calls the phase comparator II
> in the 4046, or other implementations.
>>> The reasons are in the control theory, at which I am not yet good
>>> enough to explain why this happens. But I have observed it
>>> nonetheless. A crude attempt at my explaining it would go something
>>> The motor/VCO transfer function is wierd, so that the loop filter
>>> transfer function is not that of a simple low pass filter with a
>>> finite high frequency gain. Instead it is a "zero-pole" as I have
>>> OPed. This filter doesn't integrate. Phase error is thus some
>>> non-zero constant.
>> Have you actually read the data sheet?
> No, I just plugged in the chip and hooked up signals randomly. :-D
Actually, it looks like it.
> Seriously, the phase detector II in the 4046 forces one to drive a
> passive filter, or I suppose an OP-amp integrator would work as well.
> According to Dan Wolaver, "Phase Locked Loop Circuit Design,"
> regarding the 4046 type II detector, "This circuit is intended to be
> used with a passive filter..."
> Now the reason is the high impedance state, which has to be averaged
> or low-passed before doing much of anything with it. Otherwise,
> you'd be PWMing the VCO at the phase detector output frequency.
This type of detector is a *sampled* data system. If the frequency/phase
is high it sends a correction signal to back off the VCO/motor, if too
low it corrects upwards. However, in lock the control voltage simple
stays constant on the capacitor, in which case the motor is run at a
constant voltage for most of the time. Sure, if it drifts out, then it
will have to get a correction voltage. That's what the loop is supposed
> With a motor as a VCO however, the VCO transfer function forces a
> radically different loop filter design, in order to stabilize the
> loop. This filter response is *high-pass*. You cannot feed a 4046
> phase detector II (z-state PD) into such a loop filter, or you would
> end up PWMing the motor as I said.
I don't know the details of your system, but you are suggesting that in
addition to the basic integrator response of the motor, there is another
pole(s) due to some yet unknown reason, but probably the transfer
function of your PWM drive. This may well be, but that does not mean
that you cannot use detector II as it is intended to be.
>In fact, that isn't even what
> happens with the z-state PD if you feed it straight into a high pass
> (or actually the zero-pole filter, as I have shown in the OP). It
> just doesn't work at all
>because the high impedance states demand a
> charge storage capacitor at the output of the PD. Hence, the
> statement that the detector is designed to be used with a passive
This all seems like darkness to me. No wonder your having problems. You
can't just piddle about like this without having any idea of how things
are actually working.
PD 11 is not just an integrator, it is a *sampled* non-linear system.
You can't treat it directly as an integrator, so don't mess with it.
> What I have actually done to make the 4046 phase det. II work with my
> high-pass zero-pole loop filter, is to pull up the output of the
> detector. This converts the detector into something quite different.
Indeed it does. What do you think it does? How is this new connection
actually acting like a phase detector? No wonder your having problems.
> I feed that into a low-pass filter to smooth the pulses into something
> reasonably averaged, with the pole of that filter far enough above the
> loop unity gain frequency so as not to disturb the effort of the loop
> filter in stabilizing the loop.
> This works,
Only goes to show that things can work for the daftest of reasons.
>but does not by any means produce zero phase error,
> again, the reason is in the control theory, not the phase detector.
The reason is that one has no idea, without a bit of analysis, what
tiger you have in your tank.
> The problem is that the phase detector cannot work as advertized in a
> motor PLL, for reasons having to do with the control theory, which
> force a particular design of loop filter, which force a different use
> of the 4046 PD II than what the datasheet says it should do.
Nonsense. Jesus wept dude. With all due respect here, you need to learn
what your doing before you make assertions of why you want to do
> The problem with using the 4046 in this manner, is that the detector
> can't reset in both directions, when using the pullup. Thus, the
> right thing to do is to move to a 3-state detector.
The right thing to do is understand the problem, not going off on a
tangent trying to solve a problem that isnt.
> Now it is possible I suppose to incorporate the integration of the
> 4046 PD II into my system,
>ie., let it drive a passive integrator as
> it is intended, then factor that 1/s into my loop stability
> calculations, and see what that requires from the main active loop
> filter in order to get it to work.
>But it cannot work with a motor,
> as the datasheet intends for one to use it, because that assumes a
> Kvco/s VCO, which is not what I have.
Ho humm. So what. Leave the phase detector alone as intended, then
design a loop filter on top of that that keeps the loop stable. Since
your driving a motor, your loop response can be very slow, therefore you
can simply add an additional opamp lead (gain increase with frequency)
stage with not too much trouble.
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.