[time-nuts] About HP10544A

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sun Mar 8 08:52:13 UTC 2009

Esa Heikkinen wrote:
>> 5370A: 1K + 10pF in series connected from 10544A output to input of an
>> ECL line receiver cascade.
>> 5345A: 1K + 910pF in series connected from 10544A output to input of CC
>> CB long tailed pair isolation amplifier.
>> 1K from CC base to ground.
> What do you think: if I would be able to create perfect buffer for it 
> somehow will it still have worse spectrum than tbolt ocxo for example? 
> With capasitive coupler test the 2nd anr 3rd harmonic was still high. 
> Maybe it's the best what it's possible to get with 10544 anyhow?
> So is it even reasonable to try or just get some other ocxo? I would 
> like to have as clean 10 MHz as possible with reasonable price. Let's 
> say that the spectrum what tbolt ocxo gives could be some kind of 
> minimum. But I'll need that ocxo inside thunderbolt in it's original place.
>> Since the 10544A phase noise floor is relatively high (~ -145dBc for the
>> later variants earlier versions had a significantly higher noise floor
>> spec) using 1K in series adds little additional noise.
> Is it so that 10544A isn't good for RF lab reference at all? Maybe it's 
> designed just for counters etc. where the spurious or harmonics are not 
> so big deal?
> Unfortunately I'm so rookie with these that clearly I'm seeking wrong 
> stuff. At very first I expected to have a good 10 MHz reference with 
> just using thundrbolt but then noticed that it's for time, not frequency 
> reference. Maybe 10544A is intended for timing applications also, not as 
> frequency standard?

A more efficient buffer like a white emitter follower using 2 x npn + 1
x pnp would be best.
If you want I can post such a circuit using BC546 + BC556 or similar.
This buffer needs to be located with 5cm or less of the OCXO connector
unless you use capacitance cancellation techniques.
If you elect to use an inductor to cancel most of the capacitance then
the total capacitance should be more than about 50pF or so if you use a
10% tolerance inductor.

Aside from the oven switching related spurs the 10544A isnt too bad.
Although the phase noise floor is a little high by modern standards.
Filtering out the second and 3rd harmonics is relatively easy.

A 10811A is somewhat cleaner at least it doesn't have any oven related
spurs and its intended to drive a 50 ohm load directly (+7dBm).

First you need to decide what output level you want:
+7dBm? +13dBm?

The how far do you need to suppress the 2nd and 3rd harmonics?

Do you need them to be 50dB below the carrier or 60dB below the carrier
or even more?
You need to set some actual specs.
Attenuating them by 20dB is easy if you need to suppress them much
further it gets more difficult.

Don't use a high Q bandpass filter as its phase shift tempco will be
relatively high.

The other issue is that you will probably need a high reverse isolation
frequency distribution amplifier if you intend that attaching or
removing various loads not affect the OCXO frequency by less than 1E-12
or less.


More information about the time-nuts mailing list