VCO in a PLL

On Sun, 12 Jan 2014 12:05:04 -0800, gyansorova wrote:

> An old engineer told me that although in theory you get improvements in
> demodulating FM via a PLL, in practice you cannot make a VCO which has
> low enough phase noise to compete with say a Quadrature Detector. This
> is also the feeling in industry. I would have thought that a VCO could
> be made near perfect with an FPGA nowadays - or am I missing something?

If someone came to me and asked me to make a near-perfect VCO, then after 
I finished laughing the last thing I'd consider using would be an FPGA.

Or do you mean that a NCO with a digital PLL could be made better?

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

On Sun, 12 Jan 2014 13:44:06 -0800, radams2000 wrote:

> In a pll you typically apply the input to a comparator which feeds the
> phase detector, and whenever you have a thresholding operation you will
> sample the noise and this causes aliasing.

I think you're both misusing terminology and underestimating technology.

First, because when noise is critical and you want to square up an RF 
signal you generally use a limiter rather than a comparator.  Second, 
because limiting doesn't really "alias" noise.  If there's no clock, 
there's no sampling.  Given that the edges can happen at any time, that 
pretty much rules out "sampling".  Finally, back in my amateur radio days 
I used to keep track of some of the movers and shakers in the RF design 
world: it was recognized that taking a not-too-noisy signal and limiting 
it would actually _remove_ noise, because it would take out the component 
that was in phase with the carrier and showed up as amplitude noise, and 
leave only the phase noise behind.

> By contrast , a detector
> based on tuned LC circuits does not suffer from aliasing, but it
> typically relies on the open- loop phase or amplitude response as a
> function of frequency, which is not really a straight line. So the
> distortion tends to be high but the noise is quite good.

The detector circuits that I know of do, or can be made to, rely only on 
the phase vs. frequency relationship of an LC filter.  You've still got 
nonlinearities, but not the ones you're citing.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

On 12.01.14 21.27, robert bristow-johnson wrote:
> On 1/12/14 3:05 PM, gyansorova@gmail.com wrote:
>> An old engineer told me that although in theory you get improvements
>> in demodulating FM via a PLL,
>
> dunno what that theory is. but i *do* remember that "PLL front end" or
> something like that was a selling point for stereo receivers back in the
> 70s. i think the PLL was also part of the AFC function.

The PLL is disabled as soon as the channel is selected in this tuners.


Marcel

gyansorova@gmail.com wrote:
> An old engineer told me that although in theory you get 
> improvements in demodulating FM via a PLL, in practice you cannot 
> make a VCO which has low enough phase noise to compete with say 
> a Quadrature Detector. This is also the feeling in industry. 

This is reminding me of a few discussions from about 30 years ago.

One had to do with building PLL based FM modulators, but I
don't remember much of the discussion about noise. As I remember,
you add the modulating signal to the output of the phase detector,
and input to the VCO, and you have an FM modulator.

The digital FM demodulator I remember being advertized by
FM tuner companies was called pulse count, though I don't think
it actually counted them.

I also remember discussion about the 74S124 and 74LS124 (which
aren't so similar to each other) dual VCO, that it was not possible
to use both VCO on the package, as they would phase lock to
each other. 

> I would have thought that a VCO could be made near perfect 
> with an FPGA nowadays - or am I missing something?

But also about 30 years ago, for a similar but different problem,
the data separator for floppy disk controllers. The preference
of analog PLL based data separators vs. digital ones related to
the noise. But I believe the change in digital technology moved
along, and so digital data separators are now better. It might
be, though, that they are just cheaper.

-- glen

Tim

RF is not really my domain, so I'll take your word about the noise aliasing. What I was imagining was an unfiltered sinusoid with wideband noise, applied to a comparator feeding a type 2 phase detector, but in reality the signal is very narrowband by the time it arrives (and also heavily limited as you mention) so maybe this prevents the noise from turning into jitter at the pll input. 
Regarding the LC detectors I am moderately familiar with the S-curve method that dates back to the 50's I would guess. I recall it was in the Heathkit receiver I built in the 60's, and you were lucky if you got 0.5% THD. I found a good reference on FM detectors, below. 

http://www.radio-electronics.com/info/rf-technology-design/fm-reception/fm-demodulation-detection-overview.php

Bob

On Tuesday, January 14, 2014 1:40:31 PM UTC+13, radam...@gmail.com wrote:
> Tim
> 
> 
> 
> RF is not really my domain, so I'll take your word about the noise aliasing. What I was imagining was an unfiltered sinusoid with wideband noise, applied to a comparator feeding a type 2 phase detector, but in reality the signal is very narrowband by the time it arrives (and also heavily limited as you mention) so maybe this prevents the noise from turning into jitter at the pll input. 
> 
> Regarding the LC detectors I am moderately familiar with the S-curve method that dates back to the 50's I would guess. I recall it was in the Heathkit receiver I built in the 60's, and you were lucky if you got 0.5% THD. I found a good reference on FM detectors, below. 
> 
> 
> 
> http://www.radio-electronics.com/info/rf-technology-design/fm-reception/fm-demodulation-detection-overview.php
> 
> 
> 
> Bob

Limiters are not filters. They only convert amplitude fluctuations into phase noise. In the case of FM this is not a good thing since the derivative of phase is needed to demodulate it! Hence it is known that for poor carrier to noise ratios it is often better to get rid of the limiter altogether since it only amplifies the noise.

Again I'm not really an RF guy but my understanding is that your "classic" IF strip (circa 1970) in an FM radio has multiple stages of LC band pass filters with distributed gain and near the end of the chain the signal is limited pretty severely to keep AM noise from turning into FM noise and also to enhance the capture ratio. 
I don't remember if the AGC is also done in the IF strip, or in the front- end. 

Bob

On Tue, 14 Jan 2014 04:45:08 -0800, radams2000 wrote:

> Again I'm not really an RF guy but my understanding is that your
> "classic" IF strip (circa 1970) in an FM radio has multiple stages of LC
> band pass filters with distributed gain and near the end of the chain
> the signal is limited pretty severely to keep AM noise from turning into
> FM noise and also to enhance the capture ratio.
> I don't remember if the AGC is also done in the IF strip, or in the
> front- end.

I posit that you don't remember where the AGC was done because a lot of 
them didn't have AGC - it isn't needed with the typical limiting IF chain.

The NE604 datasheet shows a typical circuit with five cascaded emitter 
coupled pairs (with an external IF filter between gain stages 2 and 3).
An emitter coupled pair has (approximately) a tanh transfer function.  
Cascade a bunch and you have a pretty good limiter.

Regards,
Allan
I'm not really an RF guy either.

On Wednesday, January 15, 2014 1:45:08 AM UTC+13, radam...@gmail.com wrote:
> Again I'm not really an RF guy but my understanding is that your "classic" IF strip (circa 1970) in an FM radio has multiple stages of LC band pass filters with distributed gain and near the end of the chain the signal is limited pretty severely to keep AM noise from turning into FM noise and also to enhance the capture ratio. 
> 
> I don't remember if the AGC is also done in the IF strip, or in the front- end. 
> 
> 
> 
> Bob

That is pretty much right. The limiter attempts to stop AM getting into FM. This works ok for high SNRs but not for low however.

HowtodemodulatedtheFMsignalwithhighfrequencysuchas1GHZ,