Need to reduce the Phase by 45 degrees.

[Ted Wilson]

[email protected] (Ted Wilson) wrote in message

Hi

Sorry for the lag in this response, but I totally lost sight of this
thread for several days.

[snip]

If you imagine a sudden alteration in the waveform travelling through
your phase lead filter, the filter would have to produce a new output
waveform before it had received the necessary new input waveform that
gives it the necessary information on which to form its new output
waveform. Yeah? In other words it is a time machine. Hence it is
unrealisable.

Now you could in principle do it if the filter has access to a
prediction of what input is going to come along, and that is not hard
to obtain in many cases. Sometimes we know what lies ahead for the
input signal. And sometimes we dont know but can arrange to find out.

For some purposes you could use a relative future rather than real
future by delaying the whole signal to the input of the filter, but
letting the filter see the undelayed signal as well. This way the
filter sees the relative future of the delayed signal, and could
indeed produce a phase lead version of it... though with a fixed time
delay. An obvious way to implement this would be in software.



Well, I'd like to hear what you have in mind there.

What I have in mind is the basic differentiator circuit, and
variations of that which are used, for example, in control loops to
restore a sagging phase margin, albeit at the cost of elevating
closed-loop gain at a time when we would invariably like to get it
down to 0db.

Now it is a moot point whether what we actually see is a phase-lag of
270 deg, (that's not what they told us college, but then they did tell
us several porkies as I recall), but there is no disputing that the
net effect is to restore the phase margin to some extent and it
certainly behaves like the introduction of a phase lead inside the
loop.

I recall a discussion with a collegue some years ago on this very
topic and I planned to do an experiment with AM on successive cycles
of a sinewave, so that I could see which cycle of the two waveforms
belonged to which - never did get round to it.

I come back to my question: if we can create what certainly looks and
behaves like phase advance, why can't we do it without the attendant
+20db/decade, equivalent to an all-pass phase lag where we get the lag
without the attenuation?

Of course, if we could create such a circuit, we would have a panacea
for stabilising closed loop systems and the fact that nobody has done
it is a potent arguement for the fact that it can't be done. Does
nonetheless make for an amusing intellectual distraction, when you've
got half an hour on your hands.

At this time I could only offer a very basic and incomplete answer:
the fact that we can see the past doesnt mean we can see the future.
However I do think we can move beyond that.



Just as an aside, I dont know the circuits you have in mind, but is
the problem due to positive feedback? I have used stable positive
feedback on amplifiers, there are ways to do it... but the real points
on this one are further down the post.

Yes, the problem comes down to exactly that. You can create very real
negative impedances using variants of the basic negative impedance
convertor, but they all rely on the use of positive feedback around an
amplifier and as soon as you try and use the negative component in
conjunction with its positive counterpart you run into problems, due
to the frequency dependance of one of the components - at some
frequency you end up with more positive feedback than negative in the
composite circuit and the output will either latch in one of two
stable states, or oscillate.

If you know of a way round this, I'm fascinated to hear of it.

Regards, NT

Regards

Ted Wilson

 

[N. Thornton]

On 9 Apr 2004 04:12:56 -0700, [email protected] (N. Thornton) wrote:

Yes, the problem comes down to exactly that. You can create very real
negative impedances using variants of the basic negative impedance
convertor, but they all rely on the use of positive feedback around an
amplifier and as soon as you try and use the negative component in
conjunction with its positive counterpart you run into problems, due
to the frequency dependance of one of the components - at some
frequency you end up with more positive feedback than negative in the
composite circuit and the output will either latch in one of two
stable states, or oscillate.

If you know of a way round this, I'm fascinated to hear of it.

Hi Ted. Nice to read this, it looked to me like a particularly
interesting issue.

I'm going to digest most of your points more before responding, but I
can tell you some ways pfb instability can be tackled in rf
amplifiers.


1. Quenching. I hope I've remembered the term right This was the
approach used in the superregnerative receiver. Basically the circuit
was set into strong pfb, which rapidly produced uncontrolled
oscillation. However the osc took several cycles to build up, so the
circuit was quenched at the rate of 10kHz (then regarded as above
audible). So it gave high rf amplification for several cycles, then
the mounting oscillation was stopped, then it amplified the next
several cycles etc. There was no dc feedback path, so latch up could
not occur. Several cycles of ac feedback were needed for loss of
control. Removal of dc path strikes me as possibly being key with your
opamp circuit, perhaps.

2. Pfb is actually a perfectly stable configuration... upto a point.
Its only when pfb goes beyond a certain threshold that things get out
of hand. One way to control it is to make your amplifier non linear,
such that for low level signals it is unstable, but as signal level
gets large its reducing gain keeps it below the point of out of
control-ness. Now at low sig levels the amplification will be very
high, during silent input controlled sine oscillation will occur, and
for large signals oscillation will cease. This is how the classic
reaction receiver works. Again there is no dc path so latch up can not
occur.

3. PFB can also be combined with agc when instability is not met
within one cycle, so that a cycle of signal can be amplified greatly
before agc reins it back.

4. There are also assorted schemes involving pfb that were tried,
mainly in the 50s, to seek higher gain amplifiers. Google may turn
something up.

Its something I've done some work on, with 2 limited successes. One
was a success from a technical point of view, but the extra
transistors required to support the circuits building blocks meant
that one might just as well have used them conventionally. The other
was a more practical success, one others have also worked on, and is
used in some rf rx chips. It is basically a variation on the reaction
receiver, and is fairly straightforward once you understand all the
principles and elements of a pfb amplifier. Hence it is a nice
solution, if by no means unique.


Regards, NT

 

[Vlad]

System chosen depends on the phase stability required

Long time ago I work on a unit used to correct the differential phase
on transmitters that use a small piece of coaxial.
Later I design a unit used to convert black and white text to color. I
recall the units that provide the adjustable phase of the 3.58 M
consisted of a simple amp with differential output a couple of
components and a potentiometer. If that is what you are looking for I
can look for more information.

Vlad