[time-nuts] What sort of oscillator is this?

Jim Lux jimlux at earthlink.net
Sun Dec 14 16:14:11 EST 2014

On 9/28/14, 7:55 AM, Richard Karlquist wrote:
>> I find it odd that an instrument that probably cost $50,000 when new did
>> not have a TCXO as standard,  and perhaps an oven as an option.
>> But I think HP did this sort of thing a lot. Something that would have
>> cost
>> very little to add, became an expensive option. In some cases these
>> expensive options are nothing more than enabling a bit of software,
>> although the R&D cost of the software is probably a lot more than the
>> hardware cost of adding a better oscillator.

There's also a difference between the "kind of oscillator" in the 
instrument.. Rick can probably tell us for sure, but I've heard it 
rumored that counters typically got an oscillator optimized for accuracy 
and low aging, but not necessarily so hot for phase noise, while 
synthesizers and spectrum analyzers would get a good phase noise 
oscillator, but maybe with more aging, figuring that the "cal lab" at 
the customer's facility would reset the frequency every year anyway.

I'm sure there's also some aspects of whether customers were more likely 
to have a house standard or leave the equipment powered on vs connected 
to power (so that a "standby mode" could keep the oscillator powered on).

And that in turn was somewhat determined by whether the equipment was 
"portable" (has a handle, like a 8563 spectrum analyzer) or "rack/bench" 
(like a 8663 signal generator).  The portable units aren't going to be 
powered on all the time, so you want something that is "ready to go" 
within a short time after plugging it in.

> I worked for the HP Santa Clara Division for 19 years.  The reason
> why a customer would NOT want a precision oscillator in a high end
> instrument would be that he was going to use a "house standard".

> We of course made OCXO's at SCD and "sold" them to other HP divisions.
> It would not be impossible for a division to use a TCXO, but it would
> be out of character given that we transferred 10811's at "cost", which
> was then about $400.

More information about the time-nuts mailing list