I don't know, I looked for a name and could not find one.
Sometimes there is no substitute for a big block of copper when soldering.
I pick them up at flea markets, no one wants them.  I have a few.

Pete.

On 10/28/2016 12:27 PM, Adrian Godwin wrote:

The data sheet for the device only mentions 12V. Other data sheets (eg the
131) mention 5V and 12V, but don't indicate how to specify one or the
other.

Gerry Sweeny's article at
http://gerrysweeney.com/diy-hpagilent-53131a-010-high-stability-timebase-option/
mentions a part number suffix that distinguishes two choices but this
doesn't appear in the main datasheet. Other writers here have in the past
mentioned that the suffix has no predefined meaning and is used only to
distinguish design variants internally.

It appears that Isotemp do offer a choice of voltage options but don't like
to document it publicly. Tom Miller's strategy seems the safest if using a
surplus device.

On Fri, Oct 28, 2016 at 8:10 PM, Peter Reilley preilley_454@comcast.net
wrote:

It looks like that is the only device that could be damaged by 12 volts. Can
you find a replacement and try running at 5 volts?

----- Original Message -----
From: "Peter Reilley" preilley_454@comcast.net
To: "Discussion of precise time and frequency measurement"
time-nuts@febo.com
Sent: Friday, October 28, 2016 3:10 PM
Subject: Re: [time-nuts] Opening an Isotemp OCXO

Hi

Roughly 99.9% of all OCXO’s made go to large OEM customers. The percentage may
actually be a bit higher than that. There are relatively few markets that “catalog” OCXO’s
sell into.

Inevitably the first thing that an OEM wants is some form of customization. A specific
supply voltage, a certain output format, a wider (or narrower) EFC range … there
are lots of possibilities. For every OCXO that goes into production for these guys,
five or ten other designs are done (all equally custom) that never see the light of day
past the samples.

The spec’s that these parts are built to are negotiated between the supplier and the OEM.
In some cases they are the property of the OEM and the spec is their control drawing on
the part. In other cases the drawing is done by the supplier for that OEM and is property
of the supplier.

The OEM often has competitors. They would love to get access to the OEM’s control
drawings to see how the systems are designed. The supplier has competitors. They
would love to get access to the suppliers drawings so they can make cheap knock off
parts to those drawings. In both cases, the drawings (in general) have very real value.

The net result of this is that both suppliers and OEM’s put fairly fancy rules in place about
passing out drawings. More or less anything up to and including being boiled in oil is
(if legal in the jurisdiction) fair recourse under most of these rules. Needless to say
people learn pretty quick that you get fired for this sort of thing.

The net result is that the drawings for most OCXO’s simply do not exist in the public
domain. They do (or did) exist in some form somewhere. Getting at public copies of
them is highly unlikely. Going by “similar looking” drawings is not a real good idea ….

Bob

Bob
 
From: Peter Reilley preilley_454@comcast.net
To: Discussion of precise time and frequency measurement time-nuts@febo.com
Sent: Friday, October 28, 2016 2:10 PM
Subject: Re: [time-nuts] Opening an Isotemp OCXO

The reason that there was 12 volts on the unit was because I put it
there.  I should
have tried 5 volts first but the only datasheet that I could find said
12 volts.
All the eBay units that look the same say 12 volts.

Pete.

I do not think the designer was considering noise at all because tying
the inputs together would not do anything useful.  Emitter resistance
is inversely proportional to emitter current (26/mA) but putting them
in parallel lowers the current through each emitter so the total
emitter resistance stays the same.

Supply current is separate for each TTL gate so by using a single
8-input part, total power it is about half that of a dual 4-input part
and a quarter of a quad 2-input part depending on the exact operating
conditions.  Unused outputs should be high for lowest power.

74S30 Single 8-Input 5mA 10mA
74S20 Dual 4-Input 8mA 18mA
74S00 Quad 2-Input 16mA 36mA

One clever design I ran across used the 7420 dual 4-input NAND pinout
but wired the inputs which are pin compatible so the 7400 quad 2-input
pinout would also work.  This allows using a 74x00, 74x20, or 74S120
dual 4-input NAND buffer but it draws even more power.

74S120 Dual 4-Input 18mA 44mA

On Fri, 28 Oct 2016 13:31:38 -0400, you wrote:

Hi

Well, the first issue would be the one that started this thread…. what’s the supply and pinout.
Yes, you can tear a couple open and work that out. Next step would be to verify that the other
couple hundred in the batch all are the same pinout and supply.

Assuming that you have all the same parts, next up would be making the test fixtures for the
units. It does not need to be super complex, but you need a way to solidly connect to the part.
If the leads have been cut off, this is a bit more complex than if you have full length leads.

Most processes age the units first. That would involve putting all hundred pieces on power and
looking at each of them with a counter every couple of minutes for a month or two. Based on that
data you could get a pretty good idea of what the aging will run. You also will weed out some
percentage of the units that didn’t survive whatever process got them to you.

After stabilization, They would go into temperature test. Likely something like a dozen or two
per run. Since you don’t know the top end, I’d do a search for that first. I’d then do a search for
the low end. Based on data from a few runs (several dozen parts) you should have some idea
of the upper and lower temp limits.

A formal temp run over that range would be next. I’d probably do 5 degree steps. That way if
you are off on the endpoint guess you might be able to see the correct end points. You would
test the whole batch of units and then look at the data. My guess is that you would re-guess the
end points and re-test the batch at that point.

Assuming you know the correct load and EFC from the original tear down, they would move on
to some sort of bench check. If not, you would need to work that out.

On the bench check, I’d run each one over the (assumed) EFC range at something like 0.1V steps.
Here if your range is off, the data probably is still ok. That should give you a proper EFC setting
for each one. Since they have been aged and TC’d at this point, the EFC center point should
be pretty good.

After that, things like phase noise and ADEV would be on the list. Same thing, run them all
and see what they do. Make some decisions and toss out the outliers.

So, yes it can be done. Because of the tear down process early on and the data redundancy
needed, you have to get a pretty good sized group of units. The risk is that you get a group
of parts with a common (to that batch) defect. They were headed to the garbage dump and
went to eBay instead ….

Lots of Fun

Bob

On 10/28/16 1:09 PM, Bob Stewart wrote:

The real thing is "which requirement do I really, really care about"
beyond the straightforward frequency stability/phase noise specs.

You might have a specific "I care about 1 Hz to 100 Hz, but don't much
are about >100 Hz or <1Hz" and you're willing to let the mfr do pretty
much anything else for the rest.

I recently had a requirement where I don't much care about absolute
frequency accuracy, but I do care about phase noise and short term (<3
seconds) stability.  We got some quotes for OCXOs with the oven
disabled, figuring that we don't need to burn the power for the oven,
since that's more about frequency accuracy, and our environment (in
space) doesn't vary more than a couple degrees over hours.  That's
something you'll never see in a catalog.

We buy oscillators at JPL for things like landing radars - the operating
life is minutes, but it had better work for those minutes - we care not
what the aging or frequency accuracy is in this situation.

I know people that have bought parts for telemetry transmitters on
devices with life measured in seconds - there, the critical requirement
was "must start from -80C and run for 10 seconds"

A coworker was telling me about requirements for oscillators in fire
control radars attached to gatling guns - Now there's a vibe sensitivity
requirement.

So, as Bob pointed out, most oscillators go into applications that have
idiosyncratic requirements and a cost or delivery schedule requirement
pushing against the overall performance requirement.  Not everyone wants
a 1kg ultra stable oscillator from APL with a 3 year delivery time.

More information;

I added a picture to the dropbox from my Flir IR camera.  The picture
shows the copper block
that the crystal is attached to running at about 200 F.  In the IR shot
the copper block is to the
right.    In most of the regular pictures it is toward the bottom of the
picture.    This is with the
unit (minus the S30 chip) running on 5 volts for more than 10 hours.
Is that too hot?

While running at 5 volts the current is constant at about 4. amps, no
cycling.  At 12 volts
it cycled between .9 to .1 amps.  I would not expect cycling for the
temperature control
of an OCXO.  I would expect a linear temperature control circuit.

I looked at the tantalum capacitor on the bottom of the board.  The
marking is 39-10.
Does that mean 39 uF and 10 volts?  If so then it must be a 5 volt
unit.  The capacitor
did not explode at 12 volts.

Dropbox link:
https://www.dropbox.com/sh/52e9d1rva9kpb3w/AABmbIj1aK7Zk2J9SNMmu-JAa?dl=0

Pete.

On 10/18/2016 9:11 AM, Peter Reilley wrote: