PLL tricks

[Kevin Aylward]

"Bill Sloman" wrote in message news:m0csrk$2ci$1@dont-email.me...

How did you get the thermistor to accurately track the xtal temperature
to 1mK? Holding the xtal to 1mK is actually quite hard to do package
wise.

Put the thermistor as close as possible to the crystal and the Peltier
junction - in the aluminium plate/heat-spreader that couples the two
together. Double ovens help there - they minimise the temperature gradients
in the inner oven.

I am rather sceptical on a single oven doing it though. Temperature gradient
form xtal to sensor is a big problem. Both sensor and xtal are 3D objects.
Even the sensor is not at the same temperature throughout its body. Very
hard to actually measure the result at ppb frequency shift. Cycling over
temperature and setting to minimum frequency hysteresis is possibly the best
method. No frequency change hysteresis, and you know it's constant
temperature.

Kevin Aylward
www.kevinaylward.co.uk
www.anasoft.co.uk - SuperSpice

 

[Kevin Aylward]

wrote in message
news:adad796b-d1eb-4f04-850d-f999a6deefdd@googlegroups.com...

"Ringing the tank" doesn't produce anything that wasn't in the
drive waveform.

Absolutely FALSE.
James Arthur

I sense misunderstanding on terms here.

In the frequency domain, there can be no frequencies present in the output
that are not present at the input, if the system is linear. Not debatable.

However, different amplitudes and phases of the individual spectrum
components can result in time domain variations. For example, the duty cycle
of a square wave means that edges are not necessarily equal. If the exact
times between each x-ing edge matter, then there may well be issues that
don't occur for say, a radio system sensitive only to frequency content.


Kevin Aylward
www.kevinaylward.co.uk
www.anasoft.co.uk - SuperSpice

 

[DecadentLinuxUserNumeroUn]

On Tue, 30 Sep 2014 00:41:11 -0700 (PDT), Bill Sloman
<bill.sloman@gmail.com> Gave us:

The basic array of trignometrical identities got drummed into my
brain at secondary school in Tasmania in the late 1950's and I
have had occasion to recall them since,but not all that often.

The last state of the art gateway I built, tested, tore-down, and
shipped off to NBN Co was for Geeveston, Tasmania.

 

[Kevin Aylward]

"John Larkin" wrote in message
news:v3bj2a50dujvccu1vt05232agvkkpg0c28@4ax.com...


into an L-C tank. The tank then rings off, until the next pulse arrives.

Yes, but the output can only have a spectrum the same as its input,
although
with different amplitudes. There will be zero component due to the tuned
circuit frequency, even if if grossly off tune. The waveform is repetitive,
therefore can only have a discrete spectrum.

Yes, but if it's not a single spectral line, it will become jitter
once it's squared up. Predictable jitter is as bad as random jitter.

That does not really make sense to me. A repetitive waveform, whether its a
sine, square, or any arbitrary shape, is er... repetitive. The Ideal
squaring up an arbitrary repetitive waveform can not therefore produce
"jitter", predictable or random. Only noise in the system produces "jitter".

However...

I suspect that the real confusion here is one of definitions. Your jitter,
is not my jitter. Word misunderstandings have led to wars!

If a tank is off tune, it will produce effective zero x-ings that are not
equally spaced, despite still having a strict harmonic of the fundamental,
spectrum. I would guess that you call this jitter. If the application cares
that the time between every edge must be exact, then there is a potential
issue.

The pulse to pulse variations in the time domain will be effectively,
characterised by the level of sub-harmonics. In general, data channels can
tolerate a certain amount of sub-harmonic, say -30dBc. So long as the
resulting time error is not to large of a UI, say 0.1 UI.

Its quite feasible to get -50dBc for a double tuned circuit design, I will
have to go away and figure out what that means in ps for your application.

Interestingly, although a direct off the shelf 155 MHz vcxo would on the
surface, be a solution, one would have to check though that it did it as a
xtal at 155Mhz. Some oscillators at that frequency will use a multiplier.

The 155MHz is a data channel frequency, so one would assume that it is being
used for data applications,and therefore subject to those specs?

Kevin Aylward
www.kevinaylward.co.uk
www.anasoft.co.uk - SuperSpice

 

[Gerhard Hoffmann]

Am 30.09.2014 um 19:41 schrieb dagmargoodboat@yahoo.com:

On Tuesday, September 30, 2014 1:22:16 PM UTC-4, John Larkin wrote:

x2 is a nice special case, because pumping a sine wave through an
analog squaring circuit does, in theory, make a clean 2F sine wave.

There's a trig identity about that.

You can also do a pretty clean x2 from a sine with a bridge rectifier,
and a decent odd-order exciter with a schottky diode-quad clamp.

Both are lossy, so it surprised me (without calculating) that the several
gain stages needed don't add significantly to the oscillator's phase
noise, but both Kevin and Wenzel--and several other mfrs--are pretty
clear testimonials for that practice.

Yes, transformer 1:4 CT from Pulse, inc / digikey, 2 Schottkies (sp?)
and choke work really good upto 400 MHz in. But 9 to 12 dB loss.

Recently I tried 800 to 1600, but could not find a working balun for
the input side, not even 2 cascaded Pulse baluns worked, although being
spec'd to 1GHz. :-(

The -9 to -12 dB hurt.

A pair of BF862 in CG driven push-pull from a Pulse 1:1 center-tapped
make a doubler without loss.
49R9 || 10n in the sources, drains connected, 9:1 transformer to 50
Ohms. 13 dBm in, 13 dBm out at twice the frequency.

The output transformer is a 1:4CT wired as a "Spar-Transformator",
don't know the English word. Jörg?
That works to 40 MHz in. 5V for Vdd is enough.

For 100 MHz in, the output side must be tuned. That still gives
100->200 MHz at no loss. Above that, the BF862 is simply too slow.
I have tried NEC NE35???? but could not get them quiet. Somewhere
in the cycle they would oscillate at 10 GHz or so.

Gerhard

Any ideas for a balun that works at 800 MHz, maybe 1600 MHz???
But small! SemiRigid in double hole cores works.

 

[John Larkin]

On Tue, 30 Sep 2014 19:48:00 +0100, "Kevin Aylward"
<ExtractkevinRemove@kevinaylward.co.uk> wrote:

"John Larkin" wrote in message
news:v3bj2a50dujvccu1vt05232agvkkpg0c28@4ax.com...


into an L-C tank. The tank then rings off, until the next pulse arrives.

Yes, but the output can only have a spectrum the same as its input,
although
with different amplitudes. There will be zero component due to the tuned
circuit frequency, even if if grossly off tune. The waveform is repetitive,
therefore can only have a discrete spectrum.

Yes, but if it's not a single spectral line, it will become jitter
once it's squared up. Predictable jitter is as bad as random jitter.

That does not really make sense to me. A repetitive waveform, whether its a
sine, square, or any arbitrary shape, is er... repetitive. The Ideal
squaring up an arbitrary repetitive waveform can not therefore produce
"jitter", predictable or random. Only noise in the system produces "jitter".

However...

I suspect that the real confusion here is one of definitions. Your jitter,
is not my jitter. Word misunderstandings have led to wars!

If a tank is off tune, it will produce effective zero x-ings that are not
equally spaced, despite still having a strict harmonic of the fundamental,
spectrum. I would guess that you call this jitter. If the application cares
that the time between every edge must be exact, then there is a potential
issue.

In my application, that is jitter. It's what folks call
"deterministic" jitter, in that the edges jump around in a repetitive
pattern. But they still jump around.

The pulse to pulse variations in the time domain will be effectively,
characterised by the level of sub-harmonics. In general, data channels can
tolerate a certain amount of sub-harmonic, say -30dBc. So long as the
resulting time error is not to large of a UI, say 0.1 UI.

I'm doing precision timing, so any jitter, deterministic or random,
and any time drift vs the 10 MHz reference, are all system errors.

RF is easy. Pump something in here and something else comes out over
there. Timing to picoseconds is harder.

Its quite feasible to get -50dBc for a double tuned circuit design, I will
have to go away and figure out what that means in ps for your application.

Two cases we have cited here (Efratom and HP) both had 5x multiplier
chains with six LC resonators to clean things up.

Interestingly, although a direct off the shelf 155 MHz vcxo would on the
surface, be a solution, one would have to check though that it did it as a
xtal at 155Mhz. Some oscillators at that frequency will use a multiplier.

The nice little ones have too much close-in phase noise for me to
close my bang-bang loop around. Looks like I may have to spend some
healthy fraction of a kilobuck on a 155 MHz SC-cut OCXO.

The 155MHz is a data channel frequency, so one would assume that it is being
used for data applications,and therefore subject to those specs?

The system uses OC3 fiberoptic parts to distribute timing signals
around a biggish facility. We have to trigger something like 2000
"clients" at programmed times, to picosecond accuracy.

http://www.slac.stanford.edu/econf/C011127/TUAP069.pdf

Unfortunately, I have to make a really good 155.52 MHz clock from a 10
MHz sinewave reference. And I hate sine waves!

Kevin Aylward
www.kevinaylward.co.uk
www.anasoft.co.uk - SuperSpice

--

John Larkin Highland Technology, Inc

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

 

[rickman]

On 9/30/2014 9:31 AM, dagmargoodboat@yahoo.com wrote:

Rick wrote:
On 9/29/2014 11:48 PM, dagmarg...@yahoo.com wrote:
On Monday, September 29, 2014 10:48:38 PM UTC-4, Bill Sloman wrote:
On 30/09/2014 11:51 AM, dagmargoo...@yahoo.com wrote:

Just look at YOUR OWN CIRCUIT's zero-crossing spacings. They're AWFUL.

This may strike you as "spectrally dirty" and the zero crossings will
presumably be all over the shop, but an intelligent designer would have
tuned his tank circuit rather closer to a specific harmonic than you have.

Well that's been my whole bleeping point. The thing you said didn't
matter--tank-tuning--does matter, doesn't it?

Which now you admit, but suggest it's my fault for not tuning the tank
better? THAT WAS THE POINT. IF THE TANK IS OFF-TUNE, THE ZERO CROSSINGS
AREN'T *EXACTLY* WHERE YOU WANT THEM.

One thing I'm not clear about it that this tank circuit is a filter
right? I believe it could be called a linear filter, no? I believe by
definition a linear filter does not create new frequencies other than
what is in the input.

Do I misunderstand this?

The distortion in the zero crossings is because of the presence of the
harmonics. Your and John's comments seem to be saying the tuned filter
is not capable of passing a given frequency input without disrupting its
frequency. I'm pretty sure that is just plain wrong. It has nothing to
do with the tuning of the filter.

In particular, John seems to be unaware of the spectral content of a
square wave when he says,

If you ping a tank, it rings at its natural frequency. Period.
(So to speak.)

An L-C tank doesn't know or care when the next pulse is coming;
you're
implicitly arguing that it does.

I wrote that, not John.

You are correct, John wrote:

Imagine pinging an LC at, say, F, with its resonant frequency close
to, say 5F. In the time domain, after each ping it rings at its LC
resonant frequency, which is not precisely 5F. And it loses amplitude
between pings. Both make the ringing wobble and add jitter.

Which is essentially the same thing, talking about pinging and ringing
at its resonant frequency. This is a gross oversimplification of what
is happening and distorts any attempt to understand.

The L-C tank does "know" the frequency of each harmonic within the
square wave and it responds to each of them independently, without
altering their frequencies. "Pinging" the tank is a bit of a shooting
from the hip comment and is not accurate for this discussion where the
tank is a filter with an input of various frequency sine waves.

Let's dump the intellectual baggage.

John wants to produce *exact* zero-crossings. A frequency
multiplier has been suggested in the signal chain.

Simulate this better-than-life quasi-ideal frequency multiplier:

(10mA pulsed current source into 1nF + 22uH LC tank.
Pulse width=500nS, period=5.2uS.)

WIRE 96 16 0 16
WIRE 160 16 96 16
WIRE 96 32 96 16
WIRE 160 32 160 16
WIRE 96 128 96 96
WIRE 160 128 160 112
WIRE 160 128 96 128
WIRE 288 128 160 128
WIRE 304 128 288 128
WIRE 0 176 0 16
WIRE 160 176 160 128
WIRE 160 288 160 256
FLAG 160 288 0
FLAG 0 176 0
FLAG 288 128 out
SYMBOL current 160 176 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName I1
SYMATTR Value PULSE(10mA 0 0 20n 20n 500n 5.2u)
SYMBOL ind 144 16 R0
SYMATTR InstName L1
SYMATTR Value 22ľH
SYMATTR SpiceLine Rser=.2
SYMBOL cap 80 32 R0
SYMATTR InstName C1
SYMATTR Value 1n
TEXT -34 312 Left 2 !.tran 2mS

Look at the voltage waveform--are the zero-crossings uniform?
If the LC drifted or were mistuned, would the zero-crossings
stay fixed in time, or would they move?

That's the point.

Cheers,
James Arthur

That may be your point, but it helps if you actually can explain what is
happening. You don't seem to be able to do that. This is my point.

--

Rick

 

[Bill Sloman]

On Wednesday, 1 October 2014 00:49:58 UTC+10, dagmarg...@yahoo.com wrote:

On Tuesday, September 30, 2014 3:01:43 AM UTC-4, rickman wrote:
On 9/30/2014 1:15 AM, Bill Sloman wrote:

Kevin Alyward - who really does know about this stuff - talked about
double-tuned tank circuits, and Wenzel's gear fudges the multiplication
process to get reasonably pure harmonics. The crude example would be
exploiting the trignometric identity

cosx^2=cos2x/2-1

as a frequency doubler which you can realise with an AD834 multiplier..
More complicated schemes are possible.

I am not clear on your formula as the square notation is not clear what
you intend it to be applied to. Which of these did you mean?

cos^2(x) = cos(2x)/2 - 1 or
cos(x^2) = cos(2x)/2 - 1

I checked a few web sites and didn't find a formula for cos(x^2). In
context cos^2(x) makes more sense but I didn't find the identity you
provide.

Perhaps you meant
cos^2(x) = (1 + cos(2x))/2

Bill's just throwing out buzzwords.

On the contrary, my words are meaningful, but you lack the wit to understand them.

Of course you can use a more complicated filter and get a cleaner
waveform, that's obvious.

But that's changing the question from whether a pulsed LC tank has a
certain behavior, into whether a sufficiently elaborate filter can
select a particular component.

I've personally designed and used double- and triple-tuned multiplier
tanks. The Efratom rubidium standard John posted uses a multiplier
quite close to my sim (that Bill carped about), followed by ~6(*) tuned
stages for filtering.

Oddly enough the circuit diagrams all seem to use collector driven tank circuits. I'd corrected James Arthur's incompetent original sim by using exactly that technique to isolate the the tank circuit from the current source driving it.

The transistors in question are mostly 2N2639, with a 2N3553 driving the tripler (from 20MHz to 60MHz)

http://www.farnell.com/datasheets/1662493.pdf

http://static.elitesecurity.org/uploads/2/9/2993216/2N3553.pdf

The circuit shown are no closer James Arthur's sim than was my corrected sim.

See pg. 69.

https://dl.dropboxusercontent.com/u/53724080/Gear/Efratom.pdf



(*) Can't tell exactly how many tuned stages count without knowing
the transformer coupling factors, but T1, at a minimum, appears
to be double-tuned.

Why do you think that? If a transformer is going to be double-tuned, you do have to have weak coupling between the two coupled inductors.

Less obvious is that the result has ppm phase stability w.r.t. to
the excitation when the filter is detuned or drifts.

It probably doesn't. That's why you've got feedback loops, to correct for the drifts.

The KV638 varactor at CR1 is presumably part of that system. The whole point of the Rubidium clock is to get a 6.9GHz microwave output that exactly corresponds to the relevant hyperfine electron transition in the Rubidium optical absorbtion spectrum. The multiplication chain includes a small-scale slow modulation element that lets the hardware see the dip in the optical absorbtion when the microwave frequency is exactly on the transition, and keep the freuency excursions centred around it.

But I've already surmised, days ago, that Wenzel and others can use
low-order multipliers with special topologies to excite their filters
with unusually pure excitations, minimizing and avoiding most of those
considerations. Oh, and put 'em in ovens.

Ovens always help.

--
Bill Sloman, Sydney

 

[Gerhard Hoffmann]

Am 30.09.2014 um 22:23 schrieb John Larkin:

The nice little ones have too much close-in phase noise for me to
close my bang-bang loop around. Looks like I may have to spend some
healthy fraction of a kilobuck on a 155 MHz SC-cut OCXO.

You can look at Pascall, also, but the healthy fraction will probably
remain.

Gerhard

 

On Tuesday, September 30, 2014 2:14:09 PM UTC-4, Kevin Aylward wrote:

wrote in message
dagmargoo...@aol.com wrote:
Bill Sloman wrote:
"Ringing the tank" doesn't produce anything that wasn't in the
drive waveform.

Absolutely FALSE.

I sense misunderstanding on terms here.

In the frequency domain, there can be no frequencies present in the output
that are not present at the input, if the system is linear. Not debatable.

Understood. It's not linear though--the transistor collector is a varactor,
for one, and a parametric multiplier.

Further, the excitation is never pure, the duty-cycle and waveform are never
ideal. Even with an ideal switched current source into the tank, the time-
domain effects are manifest.

I designed a novel multiplier with a nearly ideal waveform years ago.
It made all sorts of things possible.

However, different amplitudes and phases of the individual spectrum
components can result in time domain variations. For example, the duty cycle
of a square wave means that edges are not necessarily equal.

If the exact
times between each x-ing edge matter, then there may well be issues that
don't occur for say, a radio system sensitive only to frequency content.

That was the original question, which Bill conflated with the other
issues.

Specifically, it all started with my statement that a ringing tank's
zero-crossings would not be accurately placed if the tank drifted or if
the tank were off-tune, at which point Bill jumped in to "correct" me.

The zero-crossings are of course critical to John's application, so that's
no small matter. Bill simply didn't appreciate either the need, or the
deficiency of his proposal.

Misaligned zero-crossings are also of little use as timing references to
better discipline a VCXO.

It wasn't entirely a waste--I did read in several places that 20-ish MHz
3rd overtone crystals have lower phase noise than 5th-overtone 100-ish MHz
crystals, even after multiplication. IOW, exactly as you said.

There's still the question of absolute phase drift in the multiplier's tuned
circuits.

If John had a fortune cookie for this project, I suspect his fortune would
read "I see an oven in your future."

Cheers,
James Arthur

 

On Tuesday, September 30, 2014 4:24:29 PM UTC-4, rickman wrote:

On 9/30/2014 9:31 AM, dagmargoo...@yahoo.com wrote:

Look at the voltage waveform--are the zero-crossings uniform?
If the LC drifted or were mistuned, would the zero-crossings
stay fixed in time, or would they move?

That's the point.

That may be your point, but it helps if you actually can explain what is
happening. You don't seem to be able to do that. This is my point.

No, your point was that the tank didn't add any harmonics that weren't
in the drive waveform, which was both irrelevant to the zero-crossings'
placement, and not strictly true (due to interaction with the driver's
collector).

I explained the issue at the outset: the tank rings at its natural
resonant frequency, irrespective of the excitation, which causes the
intermediate zero crossings to be misplaced.

For that reason I was cautious about the use of frequency multipliers,
lest drift or mistuning in their tuned circuits cause drift in absolute
phase.

At this point, that concern of mine has been vetted and confirmed.

All of this was quite clearly expressed at the outset.

Cheers,
James Arthur

 

[Dimitrij Klingbeil]

On 30.09.2014 01:48, John Larkin wrote:

I can talk to the customer about this. Losing the 10 MHz and
especially the 1 PPS would be a big deal. I think they would suspend
operations until everything was fixed.

The easiest thing to do would be to go back into acquire mode if we
lose 10 MHz. I suppose we could freeze the VCXO DAC value while the
10M is down... why not? But when it comes back up, we'll have to do
the cold-start acquire-and-lock sequence.

John

The SyncServer enters an internal holdover mode when it loses a time
source, in fact doing that is one of its main tasks. According to the
datasheet, it is supposed to provide holdover with a smooth transition
when the available time sources change. That looks like it's unlikely to
generate instantaneous time "jumps" when the GPS signal is acquired again.

How "smooth" the transition really is - and over what timespan the
resynchronization is accomplished - that unfortunately they don't
specify. Also, they don't tell if it continues to output 1PPS pulses
when in holdover or not.

It does however offer 3 options for holdover timing: TCXO (default,
unless ordered with a different option), OCXO, and Rubidium. They have
vastly different holdover performances. You may want to ask the NIF
guys, which version they bought (or will be buying).

If they've got the TCXO only one, you can probably expect to see a more
or less significant frequency shift for a while, when the S250
reacquires GPS after an outage and starts pulling its oscillator back to
where it belongs.

If they've got the one of the higher precision oscillator options, it
won't drift so far during an outage and so should need less of a
correction to get back on time (possibly resulting in less wander when a
truck with a GPS jammer passes by).

In any case, since you're doing picosecond timing and the S250 is not
supposed to provide that (its receiver and PPS out is basically in the
50-150 nanosecond class), you may have to deal with an input signal that
jitters considerably more than your output signal allows, so that your
PLL may have to perform some heavy jitter cleaning duty.

Please note also that a rubidium reference does not necessarily have low
jitter. It's accurate in the long term, but on a short timescale, it's
often much more jittery than a quartz (Time-Nuts has some details).

Please note also that even if the S250 has the rubidium installed, it
does not necessarily mean that the rubidium is driving the outputs -
there's likely a VC(X)O based PLL somewhere and that may be responsible
for synthesizing the signals for the sinewave outputs.

Anyway, it does not hurt to ask Symmetricom about the actual performance
of these output signals and the way they are driven.

Regards
Dimitrij

 

[Bill Sloman]

On Wednesday, 1 October 2014 11:22:19 UTC+10, dagmarg...@yahoo.com wrote:

On Tuesday, September 30, 2014 4:24:29 PM UTC-4, rickman wrote:
On 9/30/2014 9:31 AM, dagmargoo...@yahoo.com wrote:

Look at the voltage waveform--are the zero-crossings uniform?
If the LC drifted or were mistuned, would the zero-crossings
stay fixed in time, or would they move?

That's the point.

That may be your point, but it helps if you actually can explain what is
happening. You don't seem to be able to do that. This is my point.

No, your point was that the tank didn't add any harmonics that weren't
in the drive waveform, which was both irrelevant to the zero-crossings'
placement, and not strictly true (due to interaction with the driver's
collector).

The extra harmonic content from the collector imperfections isn't going to be big, and it's highly unlikely to be significant.

The fact is that it's the externally generated harmonics that mess up the zero-crossings, and you took a long time to articulate this insight in way that was comprehensible to onlookers who didn't have your - no doubt - deep but poorly articulated insight into how frequency multipliers work in practice.

I explained the issue at the outset: the tank rings at its natural
resonant frequency, irrespective of the excitation, which causes the
intermediate zero crossings to be misplaced.

That's one way of articulating the problem. Since there's very little - essentially none - of the tank's natural resonant frequency in the frequency content of the of the signal coming out of the tank circuit, it probably isn't a helpful formulation.

For that reason I was cautious about the use of frequency multipliers,
lest drift or mistuningin their tuned circuits cause drift in absolute
phase.

At this point, that concern of mine has been vetted and confirmed.

All of this was quite clearly expressed at the outset.

If it had been clearly expressed, several people might have understood it earlier. What you actually said was something very different.

Engineering insight doesn't lend itself to expression in natural language, which evolved to help people collaborate in quite different fields.

When I had to write up my first pulse-width-modulation D/A converter I spent a about ten days finding a form of words that worked. What is really embarrassing about the whole episode is that I'd pinched most of the design from the Motorala CMOS applications databook, and forgotten where I'd got it. It wasn't until I re-read the relevant bit of the databook a few months later that the penny dropped.

--
Bill Sloman, Sydney

 

[Bill Sloman]

On Wednesday, 1 October 2014 11:54:14 UTC+10, John Larkin wrote:

On Tue, 30 Sep 2014 18:22:19 -0700 (PDT), dagmargoodboat@yahoo.com
wrote:
On Tuesday, September 30, 2014 4:24:29 PM UTC-4, rickman wrote:
On 9/30/2014 9:31 AM, dagmargoo...@yahoo.com wrote:

Look at the voltage waveform--are the zero-crossings uniform
If the LC drifted or were mistuned, would the zero-crossings
stay fixed in time, or would they move?

That's the point.

That may be your point, but it helps if you actually can explain what is
happening. You don't seem to be able to do that. This is my point.

No, your point was that the tank didn't add any harmonics that weren't
in the drive waveform, which was both irrelevant to the zero-crossings'
placement, and not strictly true (due to interaction with the driver's
collector).

I explained the issue at the outset: the tank rings at its natural
resonant frequency, irrespective of the excitation, which causes the
intermediate zero crossings to be misplaced.

Actually, it doesn't. The zero-crossing more around because the tank is responding to the sum of the harmonics being fed in from the exciter.

For that reason I was cautious about the use of frequency multipliers,
lest drift or mistuning in their tuned circuits cause drift in absolute
phase.

At this point, that concern of mine has been vetted and confirmed.

" The last temptation is the greatest treason:
To do the right deed for the wrong reason."

http://en.wikipedia.org/wiki/Murder_in_the_Cathedral

T.S.Elliot does seem to write stuff that resonates with engineering.

All of this was quite clearly expressed at the outset.

If it had been, we might have got here rather quicker.

> See, it's all your fault that Ricky doesn't understand.

Rickman and I both understood what James Arthur was saying, and that he was saying something wrong. Now that James Arthur has explained what he really meant, or at least what he would now like us to think that he really meant, the world is full of sweetness and light.

Although I imagine you are still intending to use that inappropriate bang-bang phase detector.

--
Bill Sloman, Sydney

 

[Maynard A. Philbrook Jr.]

In article <m0d62n$k84$1@dont-email.me>, bill.sloman@ieee.org says...

It didn't address your crucial misconception. Kevin probably knows
too much about the subject to be able appreciate how you'd got it wrong.
I had my nose rubbed in the subject recently, when I was fooling with my
low distortion sine-wave oscillator simulations.

Low distortion? you'd be lucky if it even oscillated.

Your intuition is probably leading you astray. The simulation oscillates
happily enough. And my clever friend in London does seem to share my
opinion that the real thing will oscillate if I ever get around to
soldering the bits together.

--
Bill Sloman, Sydney

Did it ever occur to you that he's just telling you
that?

Everyone here knows how much of a wart you are if
they disagree with you.

"Proof is in the pudding"


Jamie

 

[Bill Sloman]

On 30/09/2014 11:31 PM, dagmargoodboat@yahoo.com wrote:

Rick wrote:
On 9/29/2014 11:48 PM, dagmarg...@yahoo.com wrote:
On Monday, September 29, 2014 10:48:38 PM UTC-4, Bill Sloman wrote:
On 30/09/2014 11:51 AM, dagmargoo...@yahoo.com wrote:

Just look at YOUR OWN CIRCUIT's zero-crossing spacings. They're AWFUL.

This may strike you as "spectrally dirty" and the zero crossings will
presumably be all over the shop, but an intelligent designer would have
tuned his tank circuit rather closer to a specific harmonic than you have.

Well that's been my whole bleeping point. The thing you said didn't
matter--tank-tuning--does matter, doesn't it?

Which now you admit, but suggest it's my fault for not tuning the tank
better? THAT WAS THE POINT. IF THE TANK IS OFF-TUNE, THE ZERO CROSSINGS
AREN'T *EXACTLY* WHERE YOU WANT THEM.

One thing I'm not clear about it that this tank circuit is a filter
right? I believe it could be called a linear filter, no? I believe by
definition a linear filter does not create new frequencies other than
what is in the input.

Do I misunderstand this?

The distortion in the zero crossings is because of the presence of the
harmonics. Your and John's comments seem to be saying the tuned filter
is not capable of passing a given frequency input without disrupting its
frequency. I'm pretty sure that is just plain wrong. It has nothing to
do with the tuning of the filter.

In particular, John seems to be unaware of the spectral content of a
square wave when he says,

If you ping a tank, it rings at its natural frequency. Period.
(So to speak.)

An L-C tank doesn't know or care when the next pulse is coming;
you're
implicitly arguing that it does.

I wrote that, not John.

The L-C tank does "know" the frequency of each harmonic within the
square wave and it responds to each of them independently, without
altering their frequencies. "Pinging" the tank is a bit of a shooting
from the hip comment and is not accurate for this discussion where the
tank is a filter with an input of various frequency sine waves.

Let's dump the intellectual baggage.

John wants to produce *exact* zero-crossings. A frequency
multiplier has been suggested in the signal chain.

Simulate this better-than-life quasi-ideal frequency multiplier:

(10mA pulsed current source into 1nF + 22uH LC tank.
Pulse width=500nS, period=5.2uS.)

WIRE 96 16 0 16
WIRE 160 16 96 16
WIRE 96 32 96 16
WIRE 160 32 160 16
WIRE 96 128 96 96
WIRE 160 128 160 112
WIRE 160 128 96 128
WIRE 288 128 160 128
WIRE 304 128 288 128
WIRE 0 176 0 16
WIRE 160 176 160 128
WIRE 160 288 160 256
FLAG 160 288 0
FLAG 0 176 0
FLAG 288 128 out
SYMBOL current 160 176 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName I1
SYMATTR Value PULSE(10mA 0 0 20n 20n 500n 5.2u)
SYMBOL ind 144 16 R0
SYMATTR InstName L1
SYMATTR Value 22ľH
SYMATTR SpiceLine Rser=.2
SYMBOL cap 80 32 R0
SYMATTR InstName C1
SYMATTR Value 1n
TEXT -34 312 Left 2 !.tran 2mS

Look at the voltage waveform--are the zero-crossings uniform?
If the LC drifted or were mistuned, would the zero-crossings
stay fixed in time, or would they move?

That's the point.

The tank circuit is resonant at 1.073MHz. The period of the excitation
is 5.2usec, or 192.3KHz, a multiple of 5.578. Moving the multiple to 5
or 6 produces a better looking waveform - with a much higher amplitude,
but still with an amplitude spike at the excitation points.

Two pole tuning clearly isn't good enough, and broad band excitation is
clearly putting a whole lot of energy into the tank circuit at at all
the other harmonics of the excitation frequency.

That James Arthur doesn't understand how to get a high quality waveform
out of a frequency multiplication system isn't all that interesting.
People with rather more expertise in the area do seem to be able to
manage it.

--
Bill Sloman, Sydney

 

[Bill Sloman]

On 1/10/2014 9:58 AM, Maynard A. Philbrook Jr. wrote:

In article <m0d62n$k84$1@dont-email.me>, bill.sloman@ieee.org says...
It didn't address your crucial misconception. Kevin probably knows
too much about the subject to be able appreciate how you'd got it wrong.
I had my nose rubbed in the subject recently, when I was fooling with my
low distortion sine-wave oscillator simulations.

Low distortion? you'd be lucky if it even oscillated.

Your intuition is probably leading you astray. The simulation oscillates
happily enough. And my clever friend in London does seem to share my
opinion that the real thing will oscillate if I ever get around to
soldering the bits together.

Did it ever occur to you that he's just telling you
that?

It's a theoretical possibility. He's never hesitated to tell me when he
thinks I've got something wrong in other cases. Mostly he's been right -
which is why I think that he is clever.

Everyone here knows how much of a wart you are if
they disagree with you.

Some people don't get stuff wrong as often as you do, or persist in
their errors with quite such enthusiasm. I'm happy to see my errors
corrected - it doesn't happen often enough to damage my self-esteem.
Other people seem to feel that they have more to lose.

> "Proof is in the pudding"

Intellectually speaking, you'd be more of a fruit-cake.

--
Bill Sloman, Sydney

 

[Bill Sloman]

On 1/10/2014 10:46 AM, dagmargoodboat@yahoo.com wrote:

On Tuesday, September 30, 2014 2:14:09 PM UTC-4, Kevin Aylward wrote:
wrote in message
dagmargoo...@aol.com wrote:
Bill Sloman wrote:
"Ringing the tank" doesn't produce anything that wasn't in the
drive waveform.

Absolutely FALSE.

I sense misunderstanding on terms here.

In the frequency domain, there can be no frequencies present in the output
that are not present at the input, if the system is linear. Not debatable.

Understood. It's not linear though--the transistor collector is a varactor,
for one, and a parametric multiplier.

Further, the excitation is never pure, the duty-cycle and waveform are never
ideal. Even with an ideal switched current source into the tank, the time-
domain effects are manifest.

I designed a novel multiplier with a nearly ideal waveform years ago.
It made all sorts of things possible.

However, different amplitudes and phases of the individual spectrum
components can result in time domain variations. For example, the duty cycle
of a square wave means that edges are not necessarily equal.

If the exact
times between each x-ing edge matter, then there may well be issues that
don't occur for say, a radio system sensitive only to frequency content.

That was the original question, which Bill conflated with the other
issues.

If that was the original question, it wasn't exactly clearly articulated.

Specifically, it all started with my statement that a ringing tank's
zero-crossings would not be accurately placed if the tank drifted or if
the tank were off-tune, at which point Bill jumped in to "correct" me.

What you actually seem to have said on Monday, 29 September 2014
10:57:24 UTC+10, was

">>> You bang the tank and it rings--so far so good. But if the tank
isn't *perfectly* tuned, it rings off frequency, and the 'ring' cycles
wander off phase, right?"

To which I responded,

"Wrong. It isn't the tank that generates the harmonics, but the
non-linear response of the multiplying diode. The tank can only respond
to the frequencies present in the output of the multiplier. It's
essentially a linear part, so can't do any kind of frequency
multiplication or inter-modulation on its own."

The fact that you were referring to the way the tank circuit reacts to
the simultaneous presence of several harmonics - all injected by the
same multiplier stage - wasn't entirely obvious.

The zero-crossings are of course critical to John's application, so that's
no small matter. Bill simply didn't appreciate either the need, or the
deficiency of his proposal.

I wasn't making any kind of proposal.

Misaligned zero-crossings are also of little use as timing references to
better discipline a VCXO.

If used with a linear phase detector, which can average out the
edge-to-edge fluctuations individual misaligned zero-crossings don't
actually matter. If the VXCO is pushing out well-aligned zero-crossings,
it doing it's job. The frequency stabilising circuit only has to push
out the right number of edges per unit time - here the time unit seems
to be about 2msec - to keep the VXCO frequency disciplined.

It wasn't entirely a waste--I did read in several places that 20-ish MHz
3rd overtone crystals have lower phase noise than 5th-overtone 100-ish MHz
crystals, even after multiplication. IOW, exactly as you said.

There's still the question of absolute phase drift in the multiplier's tuned
circuits.

If John had a fortune cookie for this project, I suspect his fortune would
read "I see an oven in your future."

One hopes he gets another one discouraging him from using non-linear
phase detectors.

--
Bill Sloman, Sydney

 

[rickman]

On 9/30/2014 9:22 PM, dagmargoodboat@yahoo.com wrote:

I explained the issue at the outset: the tank rings at its natural
resonant frequency, irrespective of the excitation, which causes the
intermediate zero crossings to be misplaced.

That is *exactly* the part that is not correct. When a tank is pinged
with an impulse there is energy at all frequencies. The tank will
resonate for some time. But the tank in this design is being stimulated
with specific frequencies which according to you and others, is *not* at
the resonance peak. So the tank can not "resonate" at that frequency.
It will resonate at the frequency it was stimulated with and those
frequencies will be attenuated much more than one at the peak of
resonance.

The tank circuit is a linear circuit. You seem to be saying that the
transistor turns the stimulating frequencies into *all* frequencies. I
don't think it works that way. The inter-modulation products don't
cover the entire electromagnetic spectrum, just the few frequencies that
can be obtained by the various sums and differences of the initial
frequencies and with reducing amplitude as you compound these sums and
differences. Eventually a wide range of individual frequencies are
obtained, but at enormously reduced amplitudes.

In the stimulating waveform, the phases of all the various frequency
components are such that the zero crossings are equally spaced. In
passing through the filter the different frequencies have different
delays so that the phases no longer align resulting in changes to the
timing of the zero crossings.

--

Rick

 

[John Larkin]

On Tue, 30 Sep 2014 18:22:19 -0700 (PDT), dagmargoodboat@yahoo.com
wrote:

On Tuesday, September 30, 2014 4:24:29 PM UTC-4, rickman wrote:
On 9/30/2014 9:31 AM, dagmargoo...@yahoo.com wrote:

Look at the voltage waveform--are the zero-crossings uniform?
If the LC drifted or were mistuned, would the zero-crossings
stay fixed in time, or would they move?

That's the point.

That may be your point, but it helps if you actually can explain what is
happening. You don't seem to be able to do that. This is my point.

No, your point was that the tank didn't add any harmonics that weren't
in the drive waveform, which was both irrelevant to the zero-crossings'
placement, and not strictly true (due to interaction with the driver's
collector).

I explained the issue at the outset: the tank rings at its natural
resonant frequency, irrespective of the excitation, which causes the
intermediate zero crossings to be misplaced.

For that reason I was cautious about the use of frequency multipliers,
lest drift or mistuning in their tuned circuits cause drift in absolute
phase.

At this point, that concern of mine has been vetted and confirmed.

All of this was quite clearly expressed at the outset.

Cheers,
James Arthur

James,

See, it's all your fault that Ricky doesn't understand.


--

John Larkin Highland Technology, Inc

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com