Searching \ for '[PIC] Fine-tuning oscillator' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/microchip/time.htm?key=oscillator
Search entire site for: 'Fine-tuning oscillator'.

Exact match. Not showing close matches.
PICList Thread
'[PIC] Fine-tuning oscillator'
2006\01\04@033342 by Jinx

face picon face
Does it matter which capacitor (the one on OSC1 or OSC2)
is variable to trim the crystal frequency ? I've looked in all the
MChip oscillator pdfs I can find but they don't say. Maybe it
doesn't matter or maybe they left it out

TIA

2006\01\04@063252 by Jose Da Silva

flavicon
face
On January 4, 2006 12:33 am, Jinx wrote:
> Does it matter which capacitor (the one on OSC1 or OSC2)
> is variable to trim the crystal frequency?

The input is going to be fairly sensitive to change.
A variable trimmer cap is also sensitive to change, example temp.
Therefore, you should modify the output side.

The ucontroller oscillator is not that much different from an inverter
oscillator.
Here's an example, which appears to show output modified as well.
http://www.ee.washington.edu/circuit_archive/circuits/F_ASCII_Schem.html#ASCIISCHEM_008

> I've looked in all the
> MChip oscillator pdfs I can find but they don't say. Maybe it
> doesn't matter or maybe they left it out

They probably aren't expecting to see the frequency modified at the
ucontroller.... in other words, you were thinking "outside the box" ;-)

2006\01\04@074240 by olin piclist

face picon face
Jinx wrote:
> Does it matter which capacitor (the one on OSC1 or OSC2)
> is variable to trim the crystal frequency ? I've looked in all the
> MChip oscillator pdfs I can find but they don't say. Maybe it
> doesn't matter or maybe they left it out

I think the one on the oscillator input pin (can't remember off the top of
my head whether that's OSC1 or OSC2) would be the better choice for using as
a trimmer.

But why do you want to do this?  You'd be adjusting to within a few pF and
then hoping that nothing else changes, like the temperature or the age of
the crystal.  Why not just buy a more accurate crystal?  How accurate do you
need?  You can get crystals down to 20ppm or so.  After that it gets more
complicated with whole oscillator circuits in temperature controlled cans.
After that it gets way more complicated, large, heavy, and expensive.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\01\04@074954 by Jinx

face picon face
> Here's an example, which appears to show output modified as well.
>
www.ee.washington.edu/circuit_archive/circuits/F_ASCII_Schem.html#ASC
IISCHEM_008

After another look around (instead of going to bed, yaaaaawn)
this one also shows a trimmer on the drive (gate output) side

http://www.euroquartz.co.uk/pdf/application-notes.pdf

That's kind of what I expected, although the PIC oscillator
is more complex than a couple of simple gates, so I thought
it best to get some opinions. For example, I don't know
whether it best to load an output when I don't know its drive
capabilities. Although I might assume that being a driver it
should be able to cope with minor variances

I'm surprised (oh, really, why ? ;-)) ) that MChip didn't show
a recommended configuration, given that they detail every
other aspect of their oscillator sections

2006\01\04@080832 by Jinx

face picon face
> I think the one on the oscillator input pin (can't remember off the
> top of my head whether that's OSC1 or OSC2)

OSC1

> But why do you want to do this?
> Why not just buy a more accurate crystal

I bought a load of 19.660800MHz for a product in which a small
deviation wouldn't be noticeable. Basically it has several modes,
from minutes measured in 100ths/sec or 10ths/sec to a few hours
in Hh:Mm or Mm:Ss. The customer has asked for a reasonably
stable time-of-day function to be added, which is no great bother,
except the crystals are around 130ppm fast (~19.663400MHz)
which will make a TOD about 11s fast per day. If I can get that
down to under 20ppm (say 1s/day or better) at an average ambient
(in this country probably 15C-20C) that would make me happy.
Or at least I could say I tried

> then hoping that nothing else changes, like the temperature or the
> age of the crystal  ?

I think it's a pretty safe bet the crystal will age

> You can get crystals down to 20ppm or so

That's true. I didn't expect these to be quite so off. Others I've
had for timepieces (usually 3.2768MHz) have been pretty good,
but I need the 19.6608 for this. I can replace one of the SMT
caps with a trimmer quite painlessly. There's no need in this
product for a TXCO, and I do know what a trial they can be to
set up

Although, had a TOD been required at the start I would have
made other arrangements for that part of it to be accurate

2006\01\04@091704 by olin piclist

face picon face
Jinx wrote:
> I bought a load of 19.660800MHz for a product in which a small
> deviation wouldn't be noticeable. Basically it has several modes,
> from minutes measured in 100ths/sec or 10ths/sec to a few hours
> in Hh:Mm or Mm:Ss. The customer has asked for a reasonably
> stable time-of-day function to be added, which is no great bother,
> except the crystals are around 130ppm fast (~19.663400MHz)
> which will make a TOD about 11s fast per day.

So why not calibrate in software?  Measure the real frequency at production
time, burn some sort of calibration factor into EEPROM or program memory if
need be, then have the firware adjust accordingly.

> If I can get that down to under 20ppm (say 1s/day or better) at
> an average ambient (in this country probably 15C-20C) that would
> make me happy.

20ppm is asking quite a lot, especially for crystals that are off by 130ppm
in the first place.  Are you sure your crystals are specified for parallel
resonant application?  Actually none of that would matter if you calibrate
to whatever frequency you ended up with at production time.  That would be
more stable too as you don't have to worry about a trimmer changing over
time or via idle fingers that think they are "fixing" something.  Basically
analog trimming sucks.

> I think it's a pretty safe bet the crystal will age

That was rather the point.

> I didn't expect these to be quite so off. Others I've
> had for timepieces (usually 3.2768MHz) have been pretty good,

The 32768Hz watch crystals are usually the most accurate for the money since
they go into applications where a few seconds per day matters.  You could
run a timer 1 oscillator perhaps and do the long term timing from that.

> but I need the 19.6608 for this.

Do you really need exactly 19.6608MHz, or any value 20MHz or a bit below as
long as you know what it is?


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\01\04@121506 by David VanHorn

picon face
Most watch crystals have the luxury of lugging around a large temperature
stabilized oven. Far larger than the actual device in fact.

Just for laughs, what loading cap was specified, and what cap values did you
use?

2006\01\04@144054 by ike, K8LH (sent by Nabble.com)

flavicon
face

You can also program a "software trimmer capacitor" of sorts in your ISR...  Just add or subtract 1 from TMR2 (or whatever timer you're using) for N number of interrupt cycles each second...  You could potentially realize accuracy of within plus or minus one Tcyc per second or 86,400 Tcyc per day (0.01728 seconds when using a 20-MHz clock and 1:1 prescaler on TMR2), not including temperature drift and crystal aging...  Unfortunately, if your frequency correction factor N is too high, it would interfere with the accuracy of your 10th or 100th second 'ticks'...

Good luck with your project...  Regards, Mike
--
View this message in context: www.nabble.com/-PIC-Fine-tuning-oscillator-t848165.html#a2208333
Sent from the MicroControllers - PIC forum at Nabble.com.

2006\01\04@144758 by Jose Da Silva

flavicon
face
On January 4, 2006 04:49 am, Jinx wrote:
> shows a trimmer on the drive (gate output) side

Now that I read your reply, other readers might think output xtal.....
I should have written output of "driver" too, because it is the output
from the chip we should modify. the output side of the xtal is just too
sensitive to disturbances, so you want that as stable as possible.

> I'm surprised (oh, really, why ? ;-)) ) that MChip didn't show
> a recommended configuration, given that they detail every
> other aspect of their oscillator sections

Trimmers are more of a radio issue since you tend to need a multiple of
the frequency and changes are somewhat noticeable. It's why I tend to
take 20ppm and numbers like that with a small grain of salt. Yes
crystals are rather stable long term, but from one crystal to another,
they are not exact frequency, so from an RF point of view, you tend to
have to tune the circuit.
Microprocessors are (or were) generally expected to get their long-term
timing from something else such as 60-hz line or an external time chip
(sort of following the 1980 IBM PC type model.
Right now, ucontroller stuff is sort of trying to do the dual tasks of
microprocessor stuff and also some stuff that tended to be on the rf
side, so you are asking for something microcontroller manufacturers
haven't really recognized as normal possibiliies yet.

By the way, 19MHz seems an odd frequency to use.
I don't know what you need it for, but since you mention seconds, I'm
assuming it's time related, which is okay for the trimmer, especially
since you see capacitor specifications all over the map ranging from
10pf to 60pf, so a variable trimmer is not going to cause pain.
I would suggest looking at putting a small 10pf in parallel with your
trimmer, and see if you can get one that tends to go the opposite
direction with temperature to your trimmer/xtal/etc.
If you were using the ucontroller to generate the 19MHz for a radio
signal (doubler, tripler, etc), I would suggest using an external
oscillator instead since all the mathematical movements happening
within the ucontroller is going to translate as noise on your rf (using
an external oscillator circuit isolates the oscillator from your
ucontroller currents/voltages/noises/etc).

Good luck on your design

2006\01\04@163142 by Jinx

face picon face
> By the way, 19MHz seems an odd frequency to use

It's actually very even !! ;-)

19.660800MHz breaks down very conveniently for one
other function, which is high-speed comms. It is also close
to the 20MHz speed limit for the F88, so makes short work
of all that needs doing

2006\01\04@163145 by Jinx

face picon face

> Just for laughs, what loading cap was specified, and what cap values
> did you use ?

OK Chuckles, here's the datasheet (only 40kB)

http://www.hoorayusa.com/spec_sheets/hc49u.pdf

I use 22pF on each side. The waveform looks OK on the
scope - sinusoidal and symmetrical

You say

http://www.dvanhorn.org/Micros/All/Crystals.php

"If your crystal wants 22pF, then you would start out with an estimated
value of 44pF for each cap. However, you must subtract a bit for the
parasitic capacitance, and the chip's input capacitance. The final formula
comes out as C=2(CL)-(CP+CI)"

Was fixed 22pF reasonable to start with ?

2006\01\04@165555 by Bill & Pookie

picon face
This sounds way to simple to be of use, but....
Just have the first minute after every even hour (except midnight) consist
of 59 seconds.  After all, as we have just seen, even god has to have "leap
seconds" every so often.

Bill

{Original Message removed}

2006\01\04@165810 by Jinx

face picon face
> So why not calibrate in software?  Measure the real frequency
> at production time, burn some sort of calibration factor into
> EEPROM or program memory if need be, then have the firware
> adjust accordingly

That's one thing I've been considering. I have another product
in which a central xtal-controlled 18F452 puts out a 1s timing
period to calibrate and synchronise satellite 16F88s and 12F675s
that all run on IntRC

For this product I have a couple of spare pins that are terminated
with a pin (unused until now) that I could inject a long accurate
period into, perhaps a day. It would be a minor pain to set up but
you'd get a better product out at the end. s/w inside the F88 can
work out how to reproduce that day. For example, TMR1 re-load
value is presently $4000 ( = -$C000) to give a 0.01s IRQ period.
For the TOD function this might now become some combination of
X * $4000 and Y * 3FFB, or whatever it turns out to be per second


2006\01\04@172058 by Jinx

face picon face


> This sounds way to simple to be of use, but....
> Just have the first minute after every even hour (except midnight)
> consist of 59 seconds

(the xtal is running fast BTW, so make that 61s)

Bill, a bit kludgy (it has a display and imagine the puzzled looks when
a time holds for 2s), but would work. It does involve knowing what the
error for every unit is however. I'm kind of leaning towards "teaching
by example". Show the micro what a real hour or day looks like and
then getting it to work out how to do it itself. What that would do is
actually similar to what you're thinking of, except it does it at the IRQ
level, by figuring out a combination of 0.01s and (0.01+a tad)s to make
the TOD run a tiny bit slower on average than it otherwise would have

Ideally I'd want to use something like Maxim's new TXCO

http://www.maxim-ic.com/view_press_release.cfm?release_id=654

but <oooh, sucks air through teeth> NZ$30, BGA package and
even at 1ppm that's still 30s per year

2006\01\04@173615 by David VanHorn

picon face
>
> OK Chuckles, here's the datasheet (only 40kB)


That data sheet tells me that this crystal can be ordered for values of
loading capacitance from 10pF on up, or in series mode.  Dosen't tell me
what this exact part number is built for.


> I use 22pF on each side. The waveform looks OK on the
> scope - sinusoidal and symmetrical


Yeah, it would.  But you're probably only loading it with around 15pF or so.
The two caps you see are in series (C1*C2)/(C1+C2) and that is in parallel
with an unknown stray value that I estimated at about 4pF


> Was fixed 22pF reasonable to start with ?


If the crystal wants about 16pF, then yes, but the data sheet you sent
dosen't say.

2006\01\04@173807 by David VanHorn

picon face
>
>
> (the xtal is running fast BTW, so make that 61s)


Symptomatic of not quite enough C.

Do you have a shortwave receiver with BFO?
Very handy for checking crystals without loading the circuit at all.

2006\01\04@180411 by Jinx

face picon face
> Do you have a shortwave receiver with BFO?

No, I've heard of those non-contact methods

> Very handy for checking crystals without loading the circuit at all

In this case I used a counter with a x10 probe, and measured it
at 19.66350-ish. Aw, stink bro'. I wasn't sure the probe hadn't
loaded the pin (although loading would have made it run slow, not
fast) so I ran the unit for a day, and it did actually work out to
what I'd measured

So, even if I could trust the measurements I make when trimming,
I've gone off trimming these units anyway

> If the crystal wants about 16pF, then yes, but the data sheet you
> sent dosen't say

I can compensate for a frequency error in s/w, but 22pFs would
be close enough to not do any long-term physical damge ?


2006\01\04@182721 by David VanHorn

picon face
>
>
> So, even if I could trust the measurements I make when trimming,
> I've gone off trimming these units anyway


Yeah, trimming is a PITA.

> If the crystal wants about 16pF, then yes, but the data sheet you
> > sent dosen't say
>
> I can compensate for a frequency error in s/w, but 22pFs would
> be close enough to not do any long-term physical damge ?


I don't think it's possible to damage the crystal JUST by having too small a
cap.. What's the drive level vs the spec?   What's the oscillation margin
(negative resistance)?  Overdriving is a frequently neglected source of
problems.

I'd be very tempted though, to fix it right by getting the loading caps up
to the right value, at least so the crystal is singing within spec'd
tolerance.

2006\01\04@192048 by Jinx

face picon face
> I'd be very tempted though, to fix it right by getting the loading
> caps up to the right value, at least so the crystal is singing within
> spec'd tolerance

I'll sort something out between the manufacturer and AN558 (230kB)

http://ww1.microchip.com/downloads/en/AppNotes/00588b.pdf

2006\01\04@194303 by Peter

picon face


On Wed, 4 Jan 2006, Jinx wrote:

> Does it matter which capacitor (the one on OSC1 or OSC2)
> is variable to trim the crystal frequency ? I've looked in all the
> MChip oscillator pdfs I can find but they don't say. Maybe it
> doesn't matter or maybe they left it out

The one on OSCin does the trick. The other one also works but it is
different.

Peter

2006\01\04@195644 by Bob Axtell

face picon face
Peter wrote:

>On Wed, 4 Jan 2006, Jinx wrote:
>
>  
>
>>Does it matter which capacitor (the one on OSC1 or OSC2)
>>is variable to trim the crystal frequency ? I've looked in all the
>>MChip oscillator pdfs I can find but they don't say. Maybe it
>>doesn't matter or maybe they left it out
>>    
>>
>
>The one on OSCin does the trick. The other one also works but it is
>different.
>
>Peter
>  
>
A client of mine would adjust his Hitachi HD64180 oscillator with a
variable cap of 0-10pF in
parallel with a fixed value (probably another 10pF), but he used a
specially-cut crystal
(NOT AT cut). Alas, he had a heart attack. I never saw his crystal spec.
but I believe it
was a "parallel cut" of some sort. The freq was nominally 6.144Mhz.

It kept excellent time, about 1/2sec/day.

All I did on that project was write the firmware.

--Bob



--
Note: To protect our network,
attachments must be sent to
spam_OUTattachTakeThisOuTspamengineer.cotse.net .
1-520-777-7606 USA/Canada
http://beam.to/azengineer

2006\01\04@201534 by David VanHorn

picon face
I looked at the app note, and thought it was a little misleading IMHO.

It seemed to say that all the caps should be considered to be in series,
which is not correct.

Say that you have Ca and Cb as the two main load caps. Then we have Cp for
parasitics and Ct for trim, in parallel with Cb

So we would figure this as the parallel combination of Cb and Ct, in series
with Ca, and that value in parallel with Cp.

Cp could be in the 3-10pF range, and Ct might be about the same, but if you
consider them all in series as the app note seemed to suggest, then you'd
end up with absurdly large values required to get the right number.. Or did
I misread the AN?

Otherwise, everything seemed in order. Oscillator drive adjustments, etc.

2006\01\05@071006 by olin piclist

face picon face
David VanHorn wrote:
> I looked at the app note, and thought it was a little misleading IMHO.

I haven't looked at the app note, but most writeups about crystal caps I've
seen have a big flaw in that they ignore the output impedence of the driver.
>From the comments here, it seems this app note is no different.

For example, a 20pF cap at 8MHz has about 1Kohm impedence.  It seems
entirely reasonable that the PIC oscillator output has an impedence that is
significant when in parallel with 1Kohm.  I expect it's less than 1Kohms at
8MHz, which would actually make it the dominant part of the impedence
driving the crystal.

I don't understand why everyone seems to ignore the driver output impedence.
I don't believe I've ever seen it mentioned by Microchip and I seem to be
the only one saying this every month or two when this topic comes up.  I
think my argument is sound, but if everyone else thinks I'm wrong lets hear
why.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\01\05@131751 by David VanHorn

picon face
> I haven't looked at the app note, but most writeups about crystal caps
> I've
> seen have a big flaw in that they ignore the output impedence of the
> driver.
> >From the comments here, it seems this app note is no different.


They do mention how to check the oscillation margin, and drive level, which
pretty much covers that.

2006\01\05@141934 by olin piclist

face picon face
David VanHorn wrote:
> They do mention how to check the oscillation margin, and drive level,
> which pretty much covers that.

That's not what I was talking about.  The impedence of the driver output
directly effects the choice of capacitor values, but I don't see that
discussed or included in the equations anywhere.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\01\05@143451 by David VanHorn

picon face
On 1/5/06, Olin Lathrop <.....olin_piclistKILLspamspam@spam@embedinc.com> wrote:
>
> David VanHorn wrote:
> > They do mention how to check the oscillation margin, and drive level,
> > which pretty much covers that.
>
> That's not what I was talking about.  The impedence of the driver output
> directly effects the choice of capacitor values, but I don't see that
> discussed or included in the equations anywhere.


I thought it would be netted into the results of what they show.
Am I missing something?  Do you have a link or doc on this?

2006\01\08@183327 by andrew kelley

picon face
> If I can get that down to under 20ppm (say 1s/day or better) at
> an average ambient (in this country probably 15C-20C) that would
> make me happy.

You could use in parallel with your trimmer (and/or regular cap) a NTC
(negative temperature coefficient) capacitor..  Would help with swings
in temp, but don't know offhand how to calculate what value you would
need or whatnot.

my $.002
--
andrew

2006\01\08@235916 by Jinx

face picon face
> You could use in parallel with your trimmer (and/or regular cap)
> a NTC (negative temperature coefficient) capacitor

This is what I'm going to try - (and I'll look into your suggestion of
an NTC cap as a further experiment)

According to Transpower, NZ's power generator, measuring a
24hr period from 3:30am to 3:30am should be accurate to one
or two cycles, +/- 1/25th second. This is an average, (which varies
because the nominal 50Hz goes up and down during the day, and
that pattern in turn varies with seasons and special events), so the
longer you sample (a year ? "Oi, where's that clock I ordered ?"),
the closer you'd get to 50Hz. But he reckoned 3:30am is about
when they'd put out the 4,320,000th and final cycle for the day.
You could, I suppose, measure a few days and average that

So, I'll use one PIC to count these cycles, putting out a pulse at
the start and end of counting. The start/end pulses of this 24 hour
period are measured by the target PIC with a fine IRQ, maybe
0.001s. It then works out how best to reproduce a second using
a combination of IRQ periods. Any fractional difference in that
reproduction can be sloughed, hopefully unnoticeably, at the minute
or hour increments, spread throughout the day. At 2 cycles original
error per day -> 91s / year, near enough. Present untrimmed
130ppm crystal error -> 4015s / year (20ppm -> 617s / year)

It's not a huge complication and many clocks can be done at
the same time

Besides this I'll be measuring the crystal frequency over a range,
eg 0C to 30C, to see if any temp co, however simple, will be
worth it. If I assume 15C as the average ambient, a clock based
on the above 24hr method will be more accurate than doing nothing

At the very least it's information gathering and playing constructively

2006\01\09@062744 by Howard Winter

face
flavicon
picon face
Jinx,

On Mon, 09 Jan 2006 17:58:53 +1300, Jinx wrote:

>...
> According to Transpower, NZ's power generator, measuring a
> 24hr period from 3:30am to 3:30am should be accurate to one
> or two cycles, +/- 1/25th second.

That sounds a lot more tolerance than I expected - having spoken to generator people, I understood that they
try to get the 24hr error as low as possible - zero in relation to their reference is their target, and
certainly less than one cycle, because of the number of systems that rely on it.

> This is an average, (which varies
> because the nominal 50Hz goes up and down during the day, and
> that pattern in turn varies with seasons and special events), so the
> longer you sample (a year ? "Oi, where's that clock I ordered ?"),
> the closer you'd get to 50Hz.

I'm not sure it would help - do you know that they carry errors over from day to day, rather than just working
to a 24hr timeframe?

> But he reckoned 3:30am is about
> when they'd put out the 4,320,000th and final cycle for the day.

Yes, I suppose they need to tweak it at a time when there is the smallest variation in load, so they probably
start working towards the correct cycle count at, say, midnight, with the appropriate delta to get it right by
03:30.  Since the Cook Strait link is DC, do the two islands have their own controls for their respective
grids, I wonder?

> You could, I suppose, measure a few days and average that

As I say, I'm not sure it would help - I don't think they, for example, let it slip a bit during the week and
catch up at weekends, I think they treat each day independantly.  I presume you don't have any sort of radio
time standard there, like MSF, DC77, and whatever the Colorado one is called, there?

Cheers,


Howard Winter
St.Albans, England


2006\01\09@073244 by Jinx

face picon face
> > 24hr period from 3:30am to 3:30am should be accurate to one
> > or two cycles, +/- 1/25th second.
>
> That sounds a lot more tolerance than I expected - having spoken
> to generator people, I understood that they try to get the 24hr error
> as low as possible - zero in relation to their reference is their target,

It's not that a day is any shorter or longer according to Transpower.
What he's saying is that because of frequency variations during any
given day, you can't predict with absolute certainty when the last cycle
of the day will be. IOW it might not be exactly at the precise time it
was yesterday. So the inaccuracy can be on the part of the measurer

For example, you'll have heavy load during meal times, eg 6-8pm,
and the frequency will go down perhaps 1Hz. In a city like Auckland,
say there's a huge jam on the motorway and many people get home
and start cooking (if at all) later than they would have done. This results
in load pattern different from the day before, which might have been a
public holiday, for argument's sake, and evening power demand was
down

Same thing when industry starts in the morning. Any load fluctuation
(perhaps the S. Auckland steel mill has a line down for maintainence)
will cause a change in the frequency pattern

> > But he reckoned 3:30am is about when they'd put out the
> > 4,320,000th and final cycle for the day.
>
> Yes, I suppose they need to tweak it at a time when there is
> the smallest variation in load, so they probably start working
> towards the correct cycle count at, say, midnight, with the
> appropriate delta to get it right by 03:30

Exactly. Masuring from 8am to 8am would probably result in a
lot more daily variation between those two times for the reason
above

> Since the Cook Strait link is DC, do the two islands have their
> own controls for their respective grids, I wonder ?

Wouldn't be surprised if they both had access to something
the same atomic clock

I asked at Meridian (a power retailer) and he said he wasn't
aware of any synchronisation. The regulations allow for some
quite wide tolerances

www.electricitycommission.govt.nz/pdfs/rulesandregs/rules/rulespdf/Pa
rt-C-1-Sep-05.pdf

Generally electricty flows from South to North, although it has
gone the other way once or twice

> I presume you don't have any sort of radio time standard there,
> like MSF, DC77, and whatever the Colorado one is called, there?

That's certainly something to look into. Cycle counting is easy
enough to do, but there's no actual guarantee you picked the
right time to start. There are times when it's more likely you'd
get a good result though

2006\01\09@081550 by Alan B. Pearce

face picon face
>> I presume you don't have any sort of radio time standard there,
>> like MSF, DC77, and whatever the Colorado one is called, there?
>
>That's certainly something to look into. Cycle counting is easy
>enough to do, but there's no actual guarantee you picked the
>right time to start. There are times when it's more likely you'd
>get a good result though

NZ does have a time standard source - it is a Caesium Standard that drove a
2.5MHz transmitter that used to be just down the corridor when I worked at
the DSIR in Gracefield. I believe it is used to derive the time pips used on
the radio. I know that at least once the Americans went out and calibrated
it, I presume using the GPS satellites, as they used a satellite receiver,
and established an "exact location" for it. The GPS network must have been
"quite new" at the time as this would have been the mid-late 70's.

As to maintaining the power cycles on the mains distribution, I was on one
of my apprenticeship courses with a guy from the NZED (as it was then) and
he related having a mains driven clock in a power station along side a
pendulum driven grandfather clock, which was used as the "time standard"
(presumably checked against the radio beeps each day) and during the night
they aimed to get the mains driven clock to "catch up time" to the
grandfather clock.

2006\01\09@161715 by Jinx

face picon face
> NZ does have a time standard source - it is a Caesium Standard
> that drove a 2.5MHz transmitter that used to be just down the
> corridor when I worked at the DSIR in Gracefield. I believe it is
> used to derive the time pips used on the radio

Ah, that's quite true

http://msl.irl.cri.nz/services/time/

"Colour Burst" looks interesting. A little less "fuzzy" than the others,
although I guess the pips could be detected with one of the under-
used LM567s lazing about here

I can feel an impending exercise comparing MSL to Transpower.....


2006\01\09@222251 by Jinx

face picon face
> > According to Transpower, NZ's power generator, measuring a
> > 24hr period from 3:30am to 3:30am should be accurate to one
> > or two cycles, +/- 1/25th second.
>
> That sounds a lot more tolerance than I expected

The above was what I got verbally. This is a written answer I
received a short while ago. The phrase "even a stopped clock
is right twice a day" comes to mind ;-)))

Maybe I'll take a closer look at the "pips" or colour burst to
fine-tune a PIC

======================

The short answer to your question is "no", there is not a set time of day
that you could start a counter on and be guaranteed 4,320,000 cycles
between then and a point 24 hrs thereafter. A slightly longer, and
hopefully more helpful answer, is as follows:

The terminology used for what you are referring to is the 'frequency time
error' or 'time error' for short, which is a measure of the average
frequency over a 24 hr period. Under the Electricity Governance Rules
(EGR), we are obliged to try and keep the 'time error' within 5 seconds
of NZST (which means the number of cycles per day should be 4,320,000
+/- 250). The EGR also obliges us to attempt to ensure that the 'time
error' crosses the zero point at least once per 24 hr period, but this does
not occur at any fixed time of day

While we will normally manage to meet these obligations in a given 24 hr
period, this is not something we can guarantee as there are factors outside
of our control that will influence this (for example the tripping of a large
block of load, or of a major generator can significantly distort the 'time
error' for a much longer period than would normally be expected). We would
therefore urge a degree of caution to anyone intending to use the system
frequency to implement any time-critical process as the accuracy is measured
in terms of seconds rather than milliseconds. Depending on what you are
doing however, this level of accuracy may suffice.

If you have not already considered this, another factor that can influence
the 'time error' as seen at any given point is the manner in which you
measure the system frequency. One of the more common methods with
micro-processors is by 'zero-crossing' detection (where the frequency is
calculated by detecting each time the AC waveform crosses the voltage
zero). While simple to implement, this method is prone to errors caused by
distortion in the voltage waveform, which can be caused by such things as
voltage phase shifts or harmonics. The quality of the voltage signal you are
receiving can therefore also be something you need to take into account.
Being in the vicinity of industrial loads with variable speed motor drives
and inductive or capacitive loads, or being on a connection prone to
phase-phase or phase-earth faults are typically factors that can distort the
voltage waveform. Again the degree of accuracy you require would be the
major factor.

-----

Hope this helps Joe.

Regards, Team Online


2006\01\09@234627 by Jose Da Silva

flavicon
face
On January 9, 2006 07:22 pm, Jinx wrote:
> The short answer to your question is "no", there is not a set time of
> day that you could start a counter on and be guaranteed 4,320,000
> cycles between then and a point 24 hrs thereafter. A slightly longer,
> and hopefully more helpful answer, is as follows:

If you are going to get this up and running what's stopping you from
averaging out 10 days?
This way, you measure about 43,200,000 cycles and divide by 10 to give
you 4,320,000. +/- a bit.

2006\01\10@181924 by Peter

picon face

On Tue, 10 Jan 2006, Jinx wrote:

>> NZ does have a time standard source - it is a Caesium Standard
>> that drove a 2.5MHz transmitter that used to be just down the
>> corridor when I worked at the DSIR in Gracefield. I believe it is
>> used to derive the time pips used on the radio
>
> Ah, that's quite true
>
> http://msl.irl.cri.nz/services/time/
>
> "Colour Burst" looks interesting. A little less "fuzzy" than the others,
> although I guess the pips could be detected with one of the under-
> used LM567s lazing about here
>
> I can feel an impending exercise comparing MSL to Transpower.....

With PAL B as you (and we) have you have to account for the phase jumps
in the carrier burst (every other line). Most PLLs for color TV use
follow the phase jump (the oscillator jumps phase every other line).
This reduces the available accuracy somewhat. A relatively easy way to
fix it is to slave a second PLL on the same frequency to the first, with
a long (100 seconds ?) time constant in the loop. That should give a
pretty stable signal for most normal purposes. PAL is specified as
4.43361875 MHz which implies ~2.2x10^-9. This is about 6 msec per month
or about 70 msec per year. IF the TX is coupled to a freqency standard
(such precision is good even for metrology - and I don't think that the
TV stations are trying very hard to apply it - cable even less, with the
color carrier being prepared in the set top box from a low precision
quartz). Anyway with 70msec per year error one can do some interesting
astronomy (e.g. earth rotation period disturbances and such).

Peter

2006\01\10@183725 by Jinx

face picon face
> > http://msl.irl.cri.nz/services/time/

> With PAL B as you (and we) have you have to account for the phase

> Peter

Thanks for that insight Peter. I've got a query into MSL for some
further information and practical examples

I think the mains cycle thing is a no-goer now after learning that
they "try" to keep within +/- 250 cycles per day. To answer Jose -

> If you are going to get this up and running what's stopping you from
> averaging out 10 days?
> This way, you measure about 43,200,000 cycles and divide by 10 to
> give you 4,320,000. +/- a bit

- I worry about what that "bit" could be. To find out what that
"bit" is, you'd need to compare it to a know standard. So why
not just cut out the middle man

2006\01\10@192035 by Jose Da Silva

flavicon
face
On January 10, 2006 03:37 pm, Jinx wrote:
{Quote hidden}

Here's something else to ponder in terms of colourburst....
Studio cameras are indeed accurate for the colour burst, if they get the
colourburst off of some atomic clock, but what would you make of
handycams and other portable cameras? They are portable, and therefore
not as accurate as you would like.

Now imagine you are watching the news and they show an on-the-spot
segment recorded from a handy-cam type camera.  I would think that you
are getting colourburst according to the portable camera and not
according to studio-quality colourburst.
I think the colourburst would be accurate for studio stuff (generated on
sight) or even film type shows and movies, but now you get a lot of
media recorded digitally, VCR tapes, etc, and pumped out via plain
VCRs, DVD machines, etc, ...all of which produce their own colourburst.
We may possibly be seeing the sunset on colourburst as an accurate
reference sent out by the TV services.

...just food for thought.

2006\01\10@194251 by Howard Winter

face
flavicon
picon face
Jose,

On Tue, 10 Jan 2006 16:25:06 -0800, Jose Da Silva wrote:

> Here's something else to ponder in terms of colourburst....
> Studio cameras are indeed accurate for the colour burst, if they get the
> colourburst off of some atomic clock, but what would you make of
> handycams and other portable cameras? They are portable, and therefore
> not as accurate as you would like.

Yes, but I'm pretty sure they regenerate it for transmission, *because* they have so many incoming sources -
studio cameras, handheld news cameras, telecine, handheld analogue and digital cameras and even phones from
members of the public.  Otherwise how would they be able to cut-in video form other countries, which use
completely different colour systems?  (BBC news often has US newsclips, for example).

Cheers,


Howard Winter
St.Albans, England


2006\01\10@212528 by Dwayne Reid

flavicon
face
At 05:25 PM 1/10/2006, Jose Da Silva wrote:

>Now imagine you are watching the news and they show an on-the-spot
>segment recorded from a handy-cam type camera.  I would think that you
>are getting colourburst according to the portable camera and not
>according to studio-quality colourburst.
>I think the colourburst would be accurate for studio stuff (generated on
>sight) or even film type shows and movies, but now you get a lot of
>media recorded digitally, VCR tapes, etc, and pumped out via plain
>VCRs, DVD machines, etc, ...all of which produce their own colourburst.
>We may possibly be seeing the sunset on colourburst as an accurate
>reference sent out by the TV services.

I think that you will find that all such external sources are fed
into a Time Base Corrector.  This synchronizes the external source to
the station master clock.

Harold H may be able to shed some insight on this.  I'll also ask a
couple of my TV Broadcast tech buddies about this later this week as well.

dwayne

--
Dwayne Reid   <dwaynerspamKILLspamplanet.eon.net>
Trinity Electronics Systems Ltd    Edmonton, AB, CANADA
(780) 489-3199 voice          (780) 487-6397 fax

Celebrating 21 years of Engineering Innovation (1984 - 2005)
 .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-.   .-
    `-'   `-'   `-'   `-'   `-'   `-'   `-'   `-'   `-'
Do NOT send unsolicited commercial email to this email address.
This message neither grants consent to receive unsolicited
commercial email nor is intended to solicit commercial email.

2006\01\10@222738 by Robert Rolf

picon face


Dwayne Reid wrote:
> At 05:25 PM 1/10/2006, Jose Da Silva wrote:
>
>
>>Now imagine you are watching the news and they show an on-the-spot
>>segment recorded from a handy-cam type camera.  I would think that you
>>are getting colourburst according to the portable camera and not
>>according to studio-quality colourburst.
>>I think the colourburst would be accurate for studio stuff (generated on
>>sight) or even film type shows and movies, but now you get a lot of
>>media recorded digitally, VCR tapes, etc, and pumped out via plain
>>VCRs, DVD machines, etc, ...all of which produce their own colourburst.
>>We may possibly be seeing the sunset on colourburst as an accurate
>>reference sent out by the TV services.
>
>
> I think that you will find that all such external sources are fed
> into a Time Base Corrector.  This synchronizes the external source to
> the station master clock.

Not any more. Most switcher are now 'all digital' which means
they effectively have a frame synchronizer on every input.
This is why you now commonly see 'lip sync' problems.
The video got retimed a few too many times without a
corresponding delay being added to the audio stream.

As we move to fully digital delivery of video streams,
this lip sync problem should go away.

> Harold H may be able to shed some insight on this.  I'll also ask a
> couple of my TV Broadcast tech buddies about this later this week as well.

Would be interesting to see the diversity of solutions out there from
old analog plant to modern completely digital ones.

Robert

2006\01\11@040748 by Alan B. Pearce

face picon face
>PAL is specified as 4.43361875 MHz which implies ~2.2x10^-9.
>This is about 6 msec per month or about 70 msec per year.
>IF the TX is coupled to a freqency standard

I don't know if they still do, but I believe when the PAL colour stations
were first set up they used a rubidium standard for generating this, as a
secondary standard (using a national Caesium standard as the primary), so it
was possible to use the colour burst as a highly accurate standard. However
with modern digital signal transmission I believe that the requirement for
tight accuracy of the colour burst has been relaxed somewhat as at the
digital-analogue interface all signals are resynced to the "local" (at that
point) reference. This means that the high accuracy is not needed right
across the studio network, so a lower accuracy standard (perhaps an ovened
crystal) can be used as a local standard.

2006\01\11@041924 by Jinx

face picon face
> I don't know if they still do, but I believe when the PAL colour
> stations were first set up they used a rubidium standard for
> generating this

They still do - if I can get access to it then that would be sweet
for calibration

http://msl.irl.cri.nz/services/time/

=====================

Colour Burst

Colour television in New Zealand uses the PAL system which requires
a stable sub carrier of 4.433 618 75 MHz ± 1 Hz. The colour burst signal
for TV1 and TV2 are locked to a rubidium vapour atomic standard at the
studio. The rubidium oscillator has a long-term drift of better than 1 part
in 1011 per month. The colour burst signal is transmitted in the form of
a burst of approximately 10 cycles of 4.433 618 75 MHz at the start of
each line of the colour picture. A crystal oscillator in the TV set is
locked
to this burst and thus to the rubidium atomic frequency standard. The
signal is easily extracted from a modern TV set using only three components

2006\01\11@072426 by olin piclist

face picon face
Jose Da Silva wrote:
> Now imagine you are watching the news and they show an on-the-spot
> segment recorded from a handy-cam type camera.  I would think that you
> are getting colourburst according to the portable camera and not
> according to studio-quality colourburst.

It doesn't work that way.  If you think about it, it can't because if each
camera had independently running sync and color burst, the picture would
jump every time it was switched.  A television station will have a master
sync and color burst generator.  Anything that can't sync to that directly,
and perhaps everything anyway goes thru a timebase corrector.

It used to be that TV stations would get a very accurate sync and color
burst feed from whatever network they were affiliated with, and these
networks had expensive sync generators that were regularly compared to the
NBS atomic clock.  The color burst deviations from the official standard
were even published.  This was some years ago before things were all
digital, and NBS is now NIST.  I don't know how/if things are done
differently now.

But aren't we losing sight of what Jinx wants to do?  I thought he wanted to
plunk some existing crystal on a PIC, then calibrate to the specific crystal
during manufacturing.  No matter how accurate the calibration is when it's
done, the crystal will age and there will be deviations due to temperature
changes.  I doubt he could specify the units to be better than 20ppm even if
they were calibrated to 0ppm error at manufacturing time.  This means that
about 5ppm for the reference is good enough, and going beyond that won't
allow a tighter spec on the product anyway.  So, the question is how to come
up with an accurate (5ppm or so) reference to calibrate the individual
crystals to.

I think he's already seen that the power line isn't good enough at any one
point in time.  I can think of several choices:

1 - Buy a frequency reference.  These are commercially available to better
than 5ppm using calibrated crystals that have been aged and are actively
temperature controlled.  HP used to make a portable cesium standard for only
a few 1000 $, but that sounds like rather overkill.

2 - Pick up your local radio time standard transmission.  New Zealand isn't
that big, so I'm assuming a suitably low frequency signal from a local
transmitter can be picked up reliably anywhere in the country.  This should
work fine as long as you can get the ground wave, which you probably can
since you'd be close enough.  I'm 2000 miles from WWVB in Boulder Colorado,
and couldn't use it's signal for that purpose here but Jinx probably has
much better reception from the NZ transmitter.

3 - Use a GPS that puts out a time signal.  GPS is based on very accurate
time measurments and is driven by atomic clocks way more accurate than
needed.  There are GPS units that tap into this and export highly accurate
and traceable time signals.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\01\11@075614 by Jinx

face picon face
> But aren't we losing sight of what Jinx wants to do?

The conversation has wandered a little but I'm still in the
market for a solution. Presently I'm waiting for answers
to queries to govt agencies. I will post what I find out and
decide to do as I'm sure it will be useful to others

> I think he's already seen that the power line isn't good enough
> at any one point in time

That was a disappointment - surprisingly rough

> 1 - Buy a frequency reference
> for a few 1000 $, but that sounds like rather overkill

For the value of this product and expected profit, yes

> 2 - Pick up your local radio time standard transmission

I'm waiting for an answer from RadioNZ on that one

http://www.gpsclock.com/specs.html

"It is not like the radio 'atomic' clocks that are accurate to
half a second or so" - true or ad spin ?

> 3 - Use a GPS that puts out a time signal

I've an on-going project that may involve GPS. If so, that
would be a convenient solution at that time. If not, I'm not
sure I'd want to buy anything that will get used just once.
Although I could tlook around for one to borrow. But I've
been doing a little surfing

http://tufi.alphalink.com.au/time/time_gps.html

One of the links includes the following warning -

"It has been reported by users in the USA, New Zealand and
Australia that the LCD time display on most, if not all, handheld
GPS receivers can be late by as much as 1 or 2 seconds.  This
can be readily shown by comparing the LCD display on a
handheld GPS receiver with a source of known accurate time,
such as WWV, telephone or true GPS time"

For the moment I'll just see what's available free. As MRL
promote their standards as being for exactly what I'm trying
to do, I've asked them for more information

2006\01\11@080515 by Shawn Wilton

picon face
Yeah, but if you read the statement on that page they say not to use the LCD
display.  Not that you should not use the time off the unit.  You could use
a serial link to pull the information.

On 1/11/06, Jinx <.....joecolquittKILLspamspam.....clear.net.nz> wrote:
{Quote hidden}

Shawn Wilton (b9 Systems)
http://black9.com

2006\01\11@084207 by Alan B. Pearce

face picon face
>> But aren't we losing sight of what Jinx wants to do?
>
>The conversation has wandered a little but I'm still
>in the market for a solution.


>> 2 - Pick up your local radio time standard transmission
>
>I'm waiting for an answer from RadioNZ on that one

I would have thought that you could have a "crystal set" (well, I appreciate
it really would need to be more than that) tuned to 2YA (511kHz ??) and use
an NE567 or similar to pick out the time codes. My understanding is that
they are 4 cycles on 1kHz each second for the 5 seconds before the hour, and
6 cycles on the hour, but I could well be wrong on the number of cycles.


>> 3 - Use a GPS that puts out a time signal

(changed the order of Jinx's mail here)
> http://www.gpsclock.com/specs.html
>
>"It is not like the radio 'atomic' clocks that are
>accurate to half a second or so" - true or ad spin ?


This sounds like a commercial version of Brooks Shera's time standard at
http://www.rt66.com/~shera/ which claims to lock a crystal to very high
accuracy.

>I've an on-going project that may involve GPS. If so,
>that would be a convenient solution at that time. If
>not, I'm not sure I'd want to buy anything that will
>get used just once. Although I could tlook around for
>one to borrow. But I've been doing a little surfing

I am looking to build Brooks project (one day) and have an ordinary 10MHz AT
cut crystal that I intend to oven in a thermos flask. It has one of the
turnover points pretty well smack on 40C.

{Quote hidden}

I believe the serial data time transmission can be quite inaccurate, which
is one of the reasons why a 1PPS pulse is available on OEM receivers. Maybe
you could try and hit up Rakon for one of those thumbnail sized GPS units I
saw reported recently
http://www.nzherald.co.nz/section/story.cfm?c_id=5&ObjectID=10363046

I guess what you want is an NZ wide version of the Rugby transmitter at
60kHz ;))

2006\01\11@101543 by Paul Hutchinson

picon face
>-----Original Message-----
>From: EraseMEpiclist-bouncesspam_OUTspamTakeThisOuTmit.edu On Behalf Of Olin Lathrop
>Sent: Wednesday, January 11, 2006 7:26 AM
>
<snip>
>
>1 - Buy a frequency reference.  These are commercially available to better
>than 5ppm using calibrated crystals that have been aged and are actively
>temperature controlled.  HP used to make a portable cesium
>standard for only a few 1000 $, but that sounds like rather overkill.

The Dallas/Maxim DS32KHZ 32.768kHz TCXO has a worst case accuracy of 2ppm
over 0 to 40 degC and 1ppm/year. They are in stock in DIP packaging for
$10.00 at DigiKey. Jinx might even be able to get one as a free sample from
Dallas/Maxim.

Maybe pair one up with a PIC to use for calibrating the other units.

Paul Hutch



2006\01\11@112406 by M. Adam Davis

face picon face
Using an OEM GPS receiver with a PPS output should be as good as and
in some cases better than other sources discussed.  For example, in
the US one can purchase the Lassen iQ for $50, which has a PPS output
rated to +/-50 nS of UTC according to the atomic clocks on the
satellites, which are referenced to the atomic clocks at the GPS nav
stations.  I don't know what their accuracy is - but since it's a PPS
that is within 50nS of the atomic clocks you can easily rely on it for
this type of application.

The fact that some GPS receivers report time incorrectly is not
related to the PPS output, and more likely to the way a specific
receiver converts GPS reported time to local time.  Since the position
fix is based on the atomic time, the PPS output has to be accurate -
else there is no position fix.

-Adam

On 1/11/06, Jinx <joecolquittspamspam_OUTclear.net.nz> wrote:
{Quote hidden}

2006\01\11@171302 by Peter

picon face

On Tue, 10 Jan 2006, Jose Da Silva wrote:

> Here's something else to ponder in terms of colourburst....
> Studio cameras are indeed accurate for the colour burst, if they get the
> colourburst off of some atomic clock, but what would you make of
> handycams and other portable cameras? They are portable, and therefore
> not as accurate as you would like.

Studio cameras are not very accurate wrt color burst. They become
accurate because they are either genlocked (to a time standard derived
signal distributed throughout the studio) or passed through a TBC which
genlocks them. If the processing is all-digital then they do not *have*
a color carrier signal, sending D2MAC or similar digital data through a
high speed switched dedicated network.

> Now imagine you are watching the news and they show an on-the-spot
> segment recorded from a handy-cam type camera.  I would think that you
> are getting colourburst according to the portable camera and not
> according to studio-quality colourburst.

Any magnetic recording will have the color carrier generated (not
regenerated!) at the last player. So its quality determines the
stability. Since this will often be a studio Betacam or DVPro machine
likely genlocked to the studio timebase it will be good no matter what.
Unfortunately this cannot fix wannabe movies shot with mini DV cams or
such but even then the image quality will be bad but the timing will be
perfect (well, almost).

> I think the colourburst would be accurate for studio stuff (generated on
> sight) or even film type shows and movies, but now you get a lot of
> media recorded digitally, VCR tapes, etc, and pumped out via plain
> VCRs, DVD machines, etc, ...all of which produce their own colourburst.
> We may possibly be seeing the sunset on colourburst as an accurate
> reference sent out by the TV services.

Probably true. The odd part is that cable and sattelite digital framed
signals are very accurately timed, probably traceable to a rubidium
standard somewhere, and the cable or sattelite box loses that and uses a
lousy 50ppm crystal to generate the output subcarrier. The reason the
cable and sattelite signals are very accurate is, afaik, the fact that
they have to keep a large switched network operating. For sattelite they
also have sideband and intermodulation concerns. E.g. I do not know what
stability the *carrier* of a sattelite transmission has but it should be
very very good imho. Maybe that is locked to a rubidium standard
somewhere. The same should be true for the bit rate in the digital cable
signal.

Perhaps the future of precision timing for the masses will require a
PLLd LNB on a sattelite dish and little else ? (An ordinary LNB can be
PLLd for sure, some already do this and the 2nd IF can be divided down
directly for frequency measurement). Now that would be nice. 10GHz with
+/-0.1Hz over 24 hours would be really nice even if weather doppler
would cause some problems.

Peter

2006\01\11@171655 by Peter

picon face

On Wed, 11 Jan 2006, Howard Winter wrote:

> Jose,
>
> On Tue, 10 Jan 2006 16:25:06 -0800, Jose Da Silva wrote:
>
>> Here's something else to ponder in terms of colourburst....
>> Studio cameras are indeed accurate for the colour burst, if they get the
>> colourburst off of some atomic clock, but what would you make of
>> handycams and other portable cameras? They are portable, and therefore
>> not as accurate as you would like.
>
> Yes, but I'm pretty sure they regenerate it for transmission,
> *because* they have so many incoming sources - studio cameras,
> handheld news cameras, telecine, handheld analogue and digital cameras
> and even phones from members of the public.  Otherwise how would they
> be able to cut-in video form other countries, which use completely
> different colour systems?  (BBC news often has US newsclips, for
> example).

Actually most studio mixing has been digital for about 10 years now
afaik. Foreign feeds often come as D2 streams or Betacam Digital tapes
or DV files. Standards are not so important (excepting for the infamous
frame drop/insert required to go to and from the 60 Hz standard).

Peter

2006\01\11@173446 by Peter

picon face

On Wed, 11 Jan 2006, Alan B. Pearce wrote:

>> PAL is specified as 4.43361875 MHz which implies ~2.2x10^-9.
>> This is about 6 msec per month or about 70 msec per year.
>> IF the TX is coupled to a freqency standard
>
> I don't know if they still do, but I believe when the PAL colour stations
> were first set up they used a rubidium standard for generating this, as a
> secondary standard (using a national Caesium standard as the primary), so it
> was possible to use the colour burst as a highly accurate standard. However
> with modern digital signal transmission I believe that the requirement for
> tight accuracy of the colour burst has been relaxed somewhat as at the
> digital-analogue interface all signals are resynced to the "local" (at that
> point) reference. This means that the high accuracy is not needed right
> across the studio network, so a lower accuracy standard (perhaps an ovened
> crystal) can be used as a local standard.

In studio mode digital actually *very* high accuracy is needed because
these people use the network at full bandwidth with switching and so on.
The digital studio network is completely synchronous afaik, and has very
accurate timing (unlike Ethernet). Also if they connect to foreign
networks then they almost certainly use at least a GPS derived clock (at
both ends). There is nothing as annoying as a TBC that drops frames once
every 10 seconds when the accumulated errors sum up to a full frame or
two and the TBC resyncs to the top of the next quad frame, when
something important is being passed over the link. The digital link is
at least as fussy although it hides the mistakes better.

Peter

2006\01\11@191839 by Peter

picon face

On Wed, 11 Jan 2006, Olin Lathrop wrote:

> I think he's already seen that the power line isn't good enough at any one
> point in time.  I can think of several choices:

I have been thinking some more so here is another choice:

TV color burst would be very nice, but it is locally derived. The
horizontal (and vertical), however, aren't. They should be related to
the signal rate from the digital feed, and that is supposed to be
accurate. So one could say that the TV's H and V rate should be
traceable to a good frequency standard. They should be significantly
more accurate then the mains in any case, probably a few ppm or less.

So why not use a TV's V rate (equal to mains frequency but more
accurate - beware of scan converting TVs) to drive a normal mains
powered clock and see what happens to the accuracy ?

Similarly, one could tune to 15625 Hz horizontal line freq. anywhere
near a TV (certainly in a house and withing at least 10 meters of any
working normal size crt TV), using a ferrite antenna not unlike tuning
into a LW atomic clock frequency. In fact I have built such receivers
before (receiving TV line freq. radiated as H field by the deflection
coil assembly), but for a different purpose. A clock referenced to this
frequency should be as good as the (digital) clock of the transmitter
chain, and better than the color subcarrier probably. Of course the TV
will have to be on air (cable box or sattelite). Watching DVDs or
Nintendo won't work. I don't think that any digital sattelite or cable
box can get away with an unsynched data stream. It would frame drop like
crazy (this is very annoying to spectators).

Peter

2006\01\11@192639 by Peter

picon face

How about asking whether the *carrier* of your local radio station (AM
or LW) is not locked to a standard ? Because that would be real easy to
use and very likely much more accurate than any off the shelf crystal
anyway.

Peter

2006\01\11@194051 by Peter

picon face

Reading some more on AM broadcast transmitter frequency stability:
Modern transmitters >25kW seem to come with +/-2ppm over temperature
range and can be 'improved by connecting to a GPS'. That would be at
least 10 times better than an unadjusted watch crystal even without GPS
and likely much better than 5 seconds/month.

This is very tempting to try out imho. Just take a cheap AM radio, PLL
it with its own IF and connect to a PIC ? Of course some work is
required.

I suppose that in more advanced countries the powers that be monitor the
big transmitters and send them love letters if they wander off their
allocated frequencies so they should be spot on, if for no other reason
then to be able to talk themselves out of remote (dx) interference
complaints.

Peter

2006\01\12@043532 by Enrico Schuerrer

picon face
----- Original Message -----
From: "Peter" <@spam@plpKILLspamspamactcom.co.il>
To: "Microcontroller discussion list - Public." <KILLspampiclistKILLspamspammit.edu>
Sent: Thursday, January 12, 2006 1:18 AM
Subject: Re: [PIC] Fine-tuning oscillator



{Quote hidden}

Due to the synchronizing technology of a TV the V + H rates are not exact
what you can name a "standard".

A much easier way is to use an ISDN telefon line or if  a SDH/SONET
transmission line with operators tact is in use to apply the tact frequency
to a quartz oscillator for synchronizing . The ISDN line tact or the
transmission line tact is in minimum bound to a time signal broadcaster
(better than 1 ^-12), mostly to a caesium or rubidium atomic clock (better
than 1 ^-16).

Regards

Enrico


2006\01\12@094123 by Vasile Surducan

face picon face
Yes, the output capacitor.

Vasile

On 1/4/06, Jinx <RemoveMEjoecolquittTakeThisOuTspamclear.net.nz> wrote:
> Does it matter which capacitor (the one on OSC1 or OSC2)
> is variable to trim the crystal frequency ? I've looked in all the
> MChip oscillator pdfs I can find but they don't say. Maybe it
> doesn't matter or maybe they left it out
>
> TIA
>
> -

2006\01\12@170006 by Peter

picon face

On Thu, 12 Jan 2006, Enrico Schuerrer wrote:

> Due to the synchronizing technology of a TV the V + H rates are not exact
> what you can name a "standard".

I agree, but integrating pulses over 24 hours will be *much* more
accurate than mains, especially in third world countries (when receiving
programs from sattelites or transmitters in first world countries).

> A much easier way is to use an ISDN telefon line or if  a SDH/SONET
> transmission line with operators tact is in use to apply the tact frequency
> to a quartz oscillator for synchronizing . The ISDN line tact or the
> transmission line tact is in minimum bound to a time signal broadcaster
> (better than 1 ^-12), mostly to a caesium or rubidium atomic clock (better
> than 1 ^-16).

I agree, but the digital streams that come down from a sattelite or from
cable have the same potential accuracy, when integrated over a long time
(24 hours or more).

Even with short integration times (say 1 hour) the accuracy can be
increased to better than 2ppm. F.ex. an AM transmitter sending at
nominally 1MHz and stabilized to 0.2ppm using GPS, and used over an hour
can sync a PIC that runs on a 16MHz crystal to the nearest 0.5ppm at
least. That's about one second per month for a clock.

Peter

2006\01\18@041714 by Jinx

face picon face
Just a quick update -

All the people that really should know about MSL timing
standards are still on holiday (lucky lucky #&$*%&) and
one in particular until the 30th

Radio stations' engineers have told me, don't trust our pips,
they're all over the place

Similarly TV techs (as some of you pointed out, thanks)
with their signals

In both cases above, they all pretty much said, yes, they
used to be good, but they ain't so no more

GPS is starting to look that little bit more attactive, and is
still in the running but we'll see what happens at the end of
the month with the standards lab

2006\01\18@050209 by Jose Da Silva

flavicon
face
On January 18, 2006 01:17 am, Jinx wrote:
> Just a quick update -
>
> Radio stations' engineers have told me, don't trust our pips,
> they're all over the place

I've noticed that with "some' radio stations here in Vancouver, Canada
during particularly wet or cold weather by having to retune the radio a
couple of times. Since my radio is indoors, and at somewhat constant
temp, I'm guessing it's problems at the other end.
The radio stations send their signals up a local mountain where they get
re-broadcast at the "correct" AM/FM frequencies at high power.
Those repeaters probably don't have temperature controlled huts at the
base of the TX antennas.

> Similarly TV techs (as some of you pointed out, thanks)
> with their signals
>
> In both cases above, they all pretty much said, yes, they
> used to be good, but they ain't so no more
>
> GPS is starting to look that little bit more attactive, and is
> still in the running but we'll see what happens at the end of
> the month with the standards lab

GPS should be particularly good and was used for sync.ing of radio cells
used for paging and possibly cell phone radio towers too, so you
should have a good chance that it works as advertised.
If you still need to use power-lines as a backup, incase the GPS doesn't
work, I think you might have some luck running, say, a couple of
counters in staggered formation counting maybe 10 days and staggered
maybe 1 day apart after each other, then add all 10 of them together
and take the average.
Extending the count to 20 or 30 days probably wouldn't be a significant
deal.
...just more food fer thought.

2006\01\18@054101 by Alan B. Pearce

face picon face
>I've noticed that with "some' radio stations here in
>Vancouver, Canada during particularly wet or cold weather
>by having to retune the radio a couple of times. Since
>my radio is indoors, and at somewhat constant temp,
>I'm guessing it's problems at the other end. The radio
>stations send their signals up a local mountain where
>they get re-broadcast at the "correct" AM/FM frequencies
>at high power. Those repeaters probably don't have
>temperature controlled huts at the base of the TX antennas.

Hmm, two things.

1. I would have thought they would get into trouble with the regulatory
authorities if the signal frequency drifted what would be a noticeable
amount on a tuneable receiver.

2. I would expect the "hut" to have reasonable thermal insulation if the
place gets cold, as items such as battery life would suffer otherwise. Also
I would have thought the operating equipment would keep the hut at a
reasonable temperature. Even with 20-30 degrees temp variation I would be
surprised if a crystal oscillator would drift enough to be noticeable on a
receiver - haven't worked it out in ppm, but did used to deal with VHF land
mobile R/T mounted in motor vehicles and we used un-ovened crystals to meet
12.5 kHz channel space specs.

2006\01\18@152454 by Jose Da Silva

flavicon
face
On January 18, 2006 02:40 am, Alan B. Pearce wrote:
> >I've noticed that with "some' radio stations here in
> >Vancouver, Canada during particularly wet or cold weather
> >by having to retune the radio a couple of times. Since
> >my radio is indoors, and at somewhat constant temp,
> >I'm guessing it's problems at the other end. The radio
> >stations send their signals up a local mountain where
> >they get re-broadcast at the "correct" AM/FM frequencies
> >at high power. Those repeaters probably don't have
> >temperature controlled huts at the base of the TX antennas.
>
> Hmm, two things.
>
> 1. I would have thought they would get into trouble with the
> regulatory authorities if the signal frequency drifted what would be
> a noticeable amount on a tuneable receiver.

I don't know if they use crystals for AM band.... however, it was on FM
that I noticed the problem, so crystal circuits would be expected.

> 2. I would expect the "hut" to have reasonable thermal insulation if
> the place gets cold, as items such as battery life would suffer
> otherwise. Also I would have thought the operating equipment would
> keep the hut at a reasonable temperature. Even with 20-30 degrees
> temp variation I would be surprised if a crystal oscillator would
> drift enough to be noticeable on a receiver - haven't worked it out
> in ppm, but did used to deal with VHF land mobile R/T mounted in
> motor vehicles and we used un-ovened crystals to meet 12.5 kHz
> channel space specs.

Probably about x9 the xtal ppm for FM assuming x9 multiplication which
isn't really that much drift for a radio receiver.
We used to build some mountain top receivers which had to be handled a
little extra carefully since you didn't have that much available power
you wanted to waste, so you would have to have compensation on the
crystal oscillator circuits to avoid xtal ovens, but those ran at a
higher frequency range.
Broadcast AM and FM have to be connected to the power grid if they send
huge amounts of power and therefore can't be that far from the grid.
Some stations have gone "off air" during power outages as a group since
they apparently share the same tower, so I don't think some towers
use backup, which may mean no need to control the hut's temperature for
battery backup.
...but like you say, xtals don't drift that much.
Maybe some other interference from Seattle since they are close enough
that we can still pick up some of their radio and TV as well, plus I've
noticed when they boost the power between day/night broadcasts.
Could be interference from other items, for example, one of the
neighbours has an electric car that gets charged at night.
Right now, don't have access to a spectrum analyzer or frequency
counter, so unfortunately will just have to drop it as an unknown since
I still put this under the category of "noticed" but not "annoyance".

2006\01\18@203351 by andrew kelley

picon face
> ...but like you say, xtals don't drift that much.

I did an experiment awhile ago using two PIC's.. One run off a crystal
and another off the RC osc..  RC varied widely.. xtal varied still..
at least +-30 Hz @ 4MHz from -28deg F to about 100 deg F..  Ive lost
the excel sheet by now..

andrew

2006\01\18@205632 by Jose Da Silva

flavicon
face
On January 18, 2006 05:33 pm, andrew kelley wrote:
> > ...but like you say, xtals don't drift that much.
>
> I did an experiment awhile ago using two PIC's.. One run off a
> crystal and another off the RC osc..  RC varied widely.. xtal varied
> still.. at least +-30 Hz @ 4MHz from -28deg F to about 100 deg F..
> Ive lost the excel sheet by now..

Thanks, looks like outside interference is my problem.

2006\01\19@070152 by olin piclist

face picon face
andrew kelley wrote:
> xtal varied still..
> at least +-30 Hz @ 4MHz from -28deg F to about 100 deg F..

That is +-7.5ppm, which is likely well within the crystal specs.  So what's
your point?


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\01\19@113901 by andrew kelley
picon face
> > xtal varied still..
> > at least +-30 Hz @ 4MHz from -28deg F to about 100 deg F..
>
> That is +-7.5ppm, which is likely well within the crystal specs.  So what's
> your point?

I figured that some people might like to know that it has been tested
outside of the factory.. or just FYI

--
andrew

More... (looser matching)
- Last day of these posts
- In 2006 , 2007 only
- Today
- New search...