[time-nuts] AN/URQ13 reference AT cut crystal? ==> Crystal Robot

[Bob kb8tq]

Hi

> On Feb 18, 2021, at 6:53 AM, Poul-Henning Kamp <phk at phk.freebsd.dk> wrote:
> 
> --------
> Bob kb8tq writes:
> 
>> Turning an “idea” into a production capable part involves making many 
>> batches of test samples. Think in the thousands of batches and hundreds
>> of parts in each batch. You have a “search” process at the blank chopping
>> level. You also have a search at the resonator fabrication level. Getting the 
>> chopping part right is only a small part of the whole process….
> 
> I realize this used to be a manual process, but today I would expect
> that you could automate a lot, of not most of it, if you wanted to ?

A normal crystal production line is *very* narrowly focused. Much of this 
narrow focus is related to the precision involved. It also is related to the
investment involved. As you move from production to design optimization
to research, the “range” of the equipment expands a lot. 

One random note: This whole “automation” process would have to go 
back to the quartz growing facility, since the bars are grown with a specific 
cut in mind.

Consider that a “normal” process involves cutting that whole bar all 
at the same angle. For your robotic process, you likely would want to
do something a bit different ….

> 
> It would still be a lot of work, and very expensive, but like
> biochemist trying out hundred of thousand compounds from their
> "libraries", robots really lower the cost.

The key point there indeed being cost to build all this gear. 

Let’s *guess* a bit here (and that’s all this is, a guess ….):

All of your experiments would need to be turned into resonators
and packaged in order to be characterized. Essentially you are 
making a crystal just like any other crystal. Making up a test of
50 in a group x 10 x 10 x 10 ( which is a pretty modest search)
still is building 50,000 crystals. 

Is this some sort of pipe dream? Not really. The quantities per 
batch are a bit small compared to what is normally run. The search
grid is a bit odd. However the total batch size is not at all unusual. 

In the open market, fully tested tested / characterized crystals 
aren’t going to sell for < $5. They are made and tested on the most
automated gear anybody has …. (this does not include whatever
your custom cutting process would cost). 

At this point, we have not discovered anything new. We simply have
done one step in examining a *very* small part of “the space”. 

For a production part, this exploration is specific to the frequency 
/ overtone / blank size you are digging into. The result might be:
“We now know more about a 16.384 MHz third overtone that will
fit in a T0-8 package.  We can now hit a resistance spec of 45 ohms
compared to the old spec of 60 ohms."

> 
> The real question must therefore be, if anybody reasonably expects
> there to be any superior "new" cuts to find in the first place ?

You really don’t know what you might find … There are already 100’s
of known cuts that are “better” in some way for some application. 
They aren’t in production anywhere, but they are “known”. 

> 
> What properties would you program a quartz-crystal-prototyping robot to search for ?

You very much want to (eventually) know about perturbations over 
temperature for a given resonator design at a specific frequency. That 
tends to be the “gotcha” on a new cut. Running 100% of the production
through a 150C sweep at 1C per minute is *not* going to make the
production people (or the customer) happy. 

> 
> Which parameter(s) of current crystal-cuts are "their weak point" ?

More Q is always nice.
Better ADEV never hurts
Lower aging is popular
Flat(er) temperature curves are attractive to an OCXO or TCXO designer

That *assumes* you are after an oscillator crystal. If you are after a thermometer,
accelerometer, pressure transducer, weigh scale, ……. you would probably
be looking for other parameters. 

Bob

> 
> -- 
> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
> phk at FreeBSD.ORG         | TCP/IP since RFC 956
> FreeBSD committer       | BSD since 4.3-tahoe    
> Never attribute to malice what can adequately be explained by incompetence.