DIY OCXO
Erv
My apologize if this subject was discussed before, I just recently found
this group.
I would like to experiment with DIY OCXO but I lack theory and
information. Any link will be appreciated.
I know there are some good texts at www.ieee-uffc.org but I don't have
password and joining don't support faculties outside US.
Thanks.
Best,
Erv
--
* remove NOSPAM to email me! *
Lynn Richardson
greatly affects the design. The simplest/cheapest OCXO I know of uses a 45ohm,
50C power posistor connected to 12vdc to heat the crystal case and keep the
critical, and hopefully nearby, oscillator parts warm. +/-2.5ppm over a -30C to
+60C range is possible with this. Higher stability requirements need more
advanced circuitry and enclosure design. SC cut crystal reference oscillators
need less temperature regulation than AT, but the crystals are expensive.
Perhaps a TCXO would be better? It certainly uses a lot less power. +/-1ppm over
the -30C to +60C range is possible with a 2 thermistor, 3 resistor compensation
netork.
Erv
I still haven't define design limits since I still don't know what
problems I can get. Results XO shuld be the best possible limited by hard
to found and expencive exotic parts and advanced tools.
TCXO look interesting too but I would like to try OCXO first.
I would like to use it within +10 to about +50 or +60 deg C, frequency
accuracy better then +/-0.5ppm. I will not complain if I can build
stratum 1 spec XO but I believe stratum 3 spec XO would be excelent DIY
XO. Maybe I'm too ambitious. :-)
I don't know prices and avaelability of AT and SC cut crystals in my
country so I still don't know what I will (could) use.
In short, everything is still open until I learn more about OCXOs.
Best,
Erv
In article <v16lut8splin2fvcq...@4ax.com>, lrich@TAKE-THIS-
OUTkc.rr.com says...
> What sort of stability over what temperature range are you hoping for, Erv? That
> greatly affects the design. The simplest/cheapest OCXO I know of uses a 45ohm,
> 50C power posistor connected to 12vdc to heat the crystal case and keep the
> critical, and hopefully nearby, oscillator parts warm. +/-2.5ppm over a -30C to
> +60C range is possible with this. Higher stability requirements need more
> advanced circuitry and enclosure design. SC cut crystal reference oscillators
> need less temperature regulation than AT, but the crystals are expensive.
>
> Perhaps a TCXO would be better? It certainly uses a lot less power. +/-1ppm over
> the -30C to +60C range is possible with a 2 thermistor, 3 resistor compensation
> netork.
Tom Bruhns
I just did a Google web search for "OCXO design theory" and got lots
of hits. I know that many of them will not be much use, but there are
some which look good. Suggest you try that. For example, one of the
hits was
http://www.corningfrequency.com/piezo/papers/itra.html, which looks
interesting.
I've always been impressed by the high percentage of practical content
on the Wenzel web site. You may find some interesting things relative
to OCXOs at this page and ones you can link to from it:
http://www.wenzel.com/oscillators.html
There have been some issues of "RF Design" magazine dedicated to
crystal oscillators...but these may not be terribly accessible to you.
You'll find some links to pdf files of a very few RF Design articles
on the Wenzel site.
General comments: crystal resonances have a temperature dependent
characteristic. AT-cut crystals have an "S" shaped curve starting low
at low temp, rising to a peak, falling to a valley, and then rising
again as temperature is increased. Adjusting the exact angle at which
the crystal is cut adjusts the shape of that curve. Crystals cut for
minimum variation over moderate temperature variation (for example, 0C
to 40C) would have a relative low peak and high valley, for low
variation in freq over the limited temp range, but they have a rather
high tempco outside that range. Crystals cut for operation in an oven
have higher peak and lower valley, but the zero-slope part of the
valley is at the desired oven temperature, perhaps 65C. (S-cut have a
single maximum and lower tempco in a range near that max, but are
rather harder to find and more expensive: cut on compound angle
relative to axes of the quartz crystal.) Anyway, you may well have
the option to hand-optimize and get to a very temperature-stable
oscillator: you could characterize the crystal and its oscillator
circuit over a range of temperatures, and pick the oven temp to be
just the zero-slope temperature. In your design, you may want to
allow for changing the oven temp, if you re-characterize the osc after
say a year of continuous operation.
Just a few things to pay attention to: crystal with high Q; crystal
in a holder which will be stable over time...hermetic sealing, good
long-term mechanical stabliity, ...; oscillator circuit which is as
independent of component variation as reasonably possible; load
isolated from oscillator circuit (buffer amp with very low reverse
feedthrough); stable, clean power supply voltages (see Wenzel site for
some power-cleaning ideas!); oven with minimum temp variation (double
oven, perhaps...or at least as isothermal as possible)--design a good
proportional controller; look up ap notes and "ideas for design" on
ovens. Beware of variations as you change the orientation of the
oscillator (because of gravity, most likely, but possibly because of
magnetic field variations).
I would expect you could get the thermal ambient dependence down to
where it's not a problem compared with aging, at least until it's been
operating a long time and aging has settled out. (Another Google
search for "crystal resonator aging" turned up lots more interesting
hits, for example http://dbserv.maxim-ic.com/appnotes.cfm?appnote_number=220.)
It's good to keep in mind that there are various types of "stability."
We use OCXOs here and worry about phase noise a lot: we select ones
based on low phase noise rather than on the best possible aging and
temp stability. But that's because this particular application is
more sensitive to phase noise. If you want a clock that will keep
time for several years, unattended, then it's crystal aging that
you'll be most worried about. If you want a frequency standard for
operation on a flight line in the arctic, you might care most about
immunity to vibration and to temperature variations.
Hope these thoughts help some!
Cheers,
Tom
zerv...@yahoo.comNOSPAM (Erv) wrote in message news:<MPG.16548e691...@news.hinet.hr>...
Lynn Richardson
a time reference. What I was talking about was for radios, not time servers.
Tom's post is very good, so I can't add much to it, just that if you have access
to measuring equipment and can do some computer programming, you can digitally
compensate the oscillator. This involves measuring the tuning voltage your
uncompensated voltage tuned crystal/oscillator needs over the temperature range
to be retuned to frequency, setting up a table of those voltages, then having
the dedicated microcomputer program constantly measure the temperature, look up
the proper voltage then drive a digital to analog converter to retune the
oscillator. If the ageing rate is known, the same microcomputer program can keep
track of the total time the oscillator has been running and generate an offset
to the tuning voltage to compensate.
Several companies offer such oscillators at reasonable prices, so you may be
better off just buying one:
http://www.temex-components.com/temex/product/crystal/index.html
http://www.ctscorp.com/reeves/fdefault.htm
http://www.vectron.com/products/tcxo/tcxo_index.htm
http://www.piezotech.com/General/mainproducts/productindex.htm
One note about aging, it's determined by how long the oscillator has been
running. Turn the oscillator off for a while and it's frequency, when turned
back on, will have 'relapsed' back to near it's initial frequency.
Erv
OUTkc.rr.com says...
> Whoops! Sorry, Erv. I'm an old radio man and didn't even think about you wanting
> a time reference. What I was talking about was for radios, not time servers.
Actualy I start to think about XOs when I needed one for my digital audio
project (DAC). When I saw I can't found TCXOs specific frequency in my
country I decide to DIY. One step to another and I came to OCXOs as
probably the best that could DIY so I decide to try it.
I know I don't need OCXO precision and accuracy, project shuld show me my
DIY XO limits.
> Tom's post is very good, so I can't add much to it, just that if you have access
Agree.
> to measuring equipment and can do some computer programming, you can digitally
> compensate the oscillator. This involves measuring the tuning voltage your
Thanks for suggestion. It is MCXO if I'm understand correct?
MC programing is out of my DIY experience but I hope I will found time
for that too.
> Several companies offer such oscillators at reasonable prices, so you may be
> better off just buying one:
Thanks for links, I found a lot of supplyers and manufacturers but would
like to see DIY progres first.
Also, I don't know prices for OCXOs today. I found just few offers for
about $600 and over $1000. This is toooo much for my student budget.
In that case TCXOs will be better invest for me.
Best,
Erv
Erv
Thanks for your long reply.
> http://www.corningfrequency.com/piezo/papers/itra.html, which looks
I found that site too. I was ordered free CD and waiting to recive it.
Look like a very nice site.
> I've always been impressed by the high percentage of practical content
> on the Wenzel web site. You may find some interesting things relative
Thanks for link. I didn't become aware of great texts I founded there.
Newer before look for sublinks.
> There have been some issues of "RF Design" magazine dedicated to
> crystal oscillators...but these may not be terribly accessible to you.
Unfortunately you are correct.
> General comments: crystal resonances have a temperature dependent
Here is my basic idea.
I shuld found dip or valley (max or min value) where crystal have min ppm
over few deg C. This will allowe few deg C change without too much ppm
change.
Crystal shuld be placed inside small Al block attached to some transistor
of FET taht will heat Al block to desiered temperature. Oscillator
circuit and Al block with crystal could be in some teperature isolated
chamber.
Outside isolated chamber could be some transistor/FET regulator and rest
of circuits.
With SMD components all design could be relativly small.
How about this idea?
I know my English is not very well but I hope you understand my idea.
> some power-cleaning ideas!); oven with minimum temp variation (double
> oven, perhaps...or at least as isothermal as possible)--design a good
Could you please explain double oven?
> It's good to keep in mind that there are various types of "stability."
> We use OCXOs here and worry about phase noise a lot: we select ones
> based on low phase noise rather than on the best possible aging and
> temp stability. But that's because this particular application is
> more sensitive to phase noise. If you want a clock that will keep
> time for several years, unattended, then it's crystal aging that
> you'll be most worried about. If you want a frequency standard for
> operation on a flight line in the arctic, you might care most about
> immunity to vibration and to temperature variations.
As I reply to Mr. Lynn Richardson, I will use it in audio DAC but that is
not important. I just want to see my DIY XO limits.
Tom Bruhns
zerv...@yahoo.comNOSPAM (Erv) wrote in message news:<MPG.165601cc...@news.hinet.hr>...
...
> Here is my basic idea.
> I shuld found dip or valley (max or min value) where crystal have min ppm
> over few deg C. This will allowe few deg C change without too much ppm
> change.
Yes, but do try for temperature control better than 1 degree C. Your
basic idea, below, is sound, and should get you to a fraction of a
degree, if the thermal regulator is well designed: high open loop
feedback control and good loop compensation. The loop compensation is
complicated by the fact that the thermal time constant of the block of
aluminum is relatively long. I would suggest you put at least two
heaters on the block, if you use transistors (FETs). Just mount one
on one side, and another on the opposite side, or mount four of them,
one on each of four sides. Use emitter (source) resistors to keep the
currents in all of them close to the same. Another way to heat the
block is to put several resistors on it, all in series, to heat evenly
all the way around. Put some insulation around the whole thing:
fiberglass or some sort of foam that can stand the warm temperature.
The idea is to make the whole block the same temperature, when it is
in steady state...think about things that would cause thermal
gradients across the block.
For best temperature control, I would put a thermistor on the crystal
itself, and use that for the ultimate control, but also put a
thermistor right on one of the resistors, to sense something like
"d(temp)/dt"...the more it differs from the crystal temp sensor, the
higher the rate of change of temp at the crystal will be. That can be
used in the controller to get better response. If you're not
comfortable with optimizing control loops, I'd say it is worth some of
your time to study this area, because it's important to getting the
best stability in the oscillator.
Perhaps you have found the AT-cut temperature curve already, but there
is a family of them shown at http://icmfg.com/gloss.html. If you use
a crystal that follows the "+8" curve there, you may stay within 1ppm
over almost 10 degrees C range, maybe +62C to +72C. But if you can
operate within under a 1 degree range, you might very well get to
0.01ppm, a big improvement. (To get all that improvement, you must be
certain you are operating at the zero-slope temperature, of course.)
> Crystal shuld be placed inside small Al block attached to some transistor
> of FET taht will heat Al block to desiered temperature. Oscillator
> circuit and Al block with crystal could be in some teperature isolated
> chamber.
> Outside isolated chamber could be some transistor/FET regulator and rest
> of circuits.
> With SMD components all design could be relativly small.
>
> How about this idea?
> I know my English is not very well but I hope you understand my idea.
>
> > some power-cleaning ideas!); oven with minimum temp variation (double
> > oven, perhaps...or at least as isothermal as possible)--design a good
>
> Could you please explain double oven?
An oven inside of an oven. The inner one would run at, say, 75C, and
the outer one at, say, 65C. The outer one then shields the inner one
from ambient temperature changes, and if you keep the thermal
resistance between the two the same all the way around, there will be
lower thermal gradients on the inner oven. But that is all probably
"overkill" for what you want to do!
Another thing to search for on the web: I recall that W. Stephen
Woodward submitted an "idea for design" some time ago about a very
accurate temperature controller. I think it was to control biological
processes to a small fraction of a degree. At least some of the
"Ideas for Design" are archived at
http://www.planetee.com/ed/ifd.html, and in particular there's one
from Woodward titled "Take Back Half: A Novel Integrating
Temperature-Control Algorithm" but I'm pretty sure there was a more
interesting one earlier than that.
Hope that helps some more... of course, as you know, you can make the
project as simple or as complicated as you like. Even if fairly
simple, if you pay attention to details, performance should be quite
good.
Cheers,
Tom
Doug Dwyer
For simplicity of thermal design create a single cavity for xtal,
maintaining circuit, and voltage regulators.
A dc proportional temperature controller will require temperature sensor
(thermistor or semiconductor) to be in close thermal contact with the
heater (semiconductor or winding) otherwise it will hunt upand down, a
two term controller can ease this situation. A simple circuit wont give
you a true thermal improvement of better than 100:1
Chose a crystal frequency in the range 5 to 10MHz and divide down to
your audio frequency.
Chose a crystal in a resistance weld package for low cost and 1ppm/year
aging.
Arrange frequency adjustment components varicaps and or cap trimmer to
be inside thermal environment.
Arrange final adjustment to be external by pot driven be temperature
controlled voltage.
Beware heat conducted down leads into cavity.
Make cavity of copper or aluminum.
Plot frequency/control temperature from 30 to 80C you should see a
"turnover" temperature ie one where the coefficient goes through zero
ppm/deg.
Normally cheap crystals are designed for minimum frequency excursion
over a temperature range. Either specify a crystal with a frequency or
zero coefficient at say 70C or (expensive) a minimum excursion over -
55+105C. The latter crystal will have a turnover at 70C +-20C (my guess
I havnt calculated it) however this crystal will be low in freq at 70C
and will need pulling up to nominal frequency.
Doug
n article <MPG.1655f8f38...@news.hinet.hr>, Erv
<zerv...@yahoo.comNOSPAM> writes
--
Doug Dwyer
James Waldby
...
> Another thing to search for on the web: I recall that W. Stephen
> Woodward submitted an "idea for design" some time ago about a very
> accurate temperature controller. I think it was to control biological
> processes to a small fraction of a degree. At least some of the
> "Ideas for Design" are archived at
> http://www.planetee.com/ed/ifd.html, and in particular there's one
> from Woodward titled "Take Back Half: A Novel Integrating
> Temperature-Control Algorithm" but I'm pretty sure there was a more
> interesting one earlier than that.
http://www.sciam.com/2000/0100issue/0100amsci.html shows a
simple 0.01-degree-C control (circuit diagram 1/6 of way
through text, brief description 5/6 of the way along).
Unfortunately, links to site with more details are broken.
Erv
thanks for link.
Best,
Erv
In article <3BECB3F6...@pat7.com>, j-wa...@pat7.com says...
> http://www.sciam.com/2000/0100issue/0100amsci.html shows a
> simple 0.01-degree-C control (circuit diagram 1/6 of way
> through text, brief description 5/6 of the way along).
> Unfortunately, links to site with more details are broken.
Erv
says...
> Most points have been covered. A few further suggestions.
Hi Doug,
Thanks for suggestions.
> Chose a crystal frequency in the range 5 to 10MHz and divide down to
> your audio frequency.
actualy I need 16.9344Mhz for current project and 24.576MHz for future,
24-bit project.
This lead me to another question.
I briefly take a look at crystals offer in my countryand found offers for
16 and 18MHz crystals, Frequency accuracy 20ppm @ 25 deg-C, temp
stability 30ppm (-10 to +60 deg-C).
Is there any limits or design downgradeing in voltage tuning frequency to
16.9344MHz and is it better to tune 16MHz up or 18MHz down to 16.9344MHz?
I recently found Analog Devices IC, ultrafast comparator AD8561. One of
aplication suggested is clock recovery and clock distribution.
Anyone have experince with this IC?
I believe it could be succesfully used on the end of the chain.
Doug Dwyer
MHz you wont succeed!
A maximum pulling for a crystal intended to be stable would be 30 to 100
parts per million.
Most countries have a crystal manufacturer for the amateur radio
community who will make a crystal to your specified frequency.
Look in local amateur radio magazines.
Sadly it is increasing practise to only make standard frequencies!
You could "synthesise" your frequency from a standard frequency.
Gets easier every day.
In article <MPG.16571e418...@news.hinet.hr>, Erv
<zerv...@yahoo.comNOSPAM> writes
--
Doug Dwyer
Gerhard Hoffmann
> Actualy I start to think about XOs when I needed one for my digital audio
> project (DAC). When I saw I can't found TCXOs specific frequency in my
> country I decide to DIY. One step to another and I came to OCXOs as
> probably the best that could DIY so I decide to try it.
> I know I don't need OCXO precision and accuracy, project shuld show me my
> DIY XO limits.
http://www.karlquist.com/professional.html
Remember, you need crystals for the right temperature
if you put them into an oven.
Gerhard
Tom Bruhns
> http://www.karlquist.com/professional.html
Yes, indeed! Plus, Rick is a really nice fellow. Got to meet him a
year or so ago when we both worked briefly in the same group. (Rick
has done some _very_ nice crystal oscillators...)
Cheers,
Tom
Tom Bruhns
> In article <wFFnnCAe...@ddwyer.demon.co.uk>, d...@ddwyer.demon.co.uk
> says...
> > Most points have been covered. A few further suggestions.
>
> Hi Doug,
> Thanks for suggestions.
I certainly wouldn't argue with Doug's suggestions...just thought I'd
point you to some additional info in the form of an article that
arrived today in "Microwave Journal" magazine. November 2001 issue,
Page 170, Timo Reinhardt, "Application and Operation of a Double
OXCO." It has illustrations of a number of the points that Doug and I
have made in our text postings. It's not on the Microwave Journal web
site yet, but I suspect it will be in a week or two...check out
http://www.mwjournal.com/mwjournal/. The article gives some good
discussion about design of high quality OCXOs, including some of the
important tradeoffs. It occured to me as I read the article that one
tradeoff you have the advantage with is manufacturability: you can
use tricks in a one-off hand-crafted oscillator that would be
difficult in production. That's not all one-sided, though, since
there are also manufacturing tricks that would be very expensive to
you.