[time-nuts] Re: Low 1/f transistors for oscillator

[Richard Karlquist]

Generally speaking, the 1/f noise at, say 10 Hz, for a crystal
oscillator is determined 

entirely by the intrinsic noise of the crystal.  It's really hard to
screw up an oscillator 

circuit so badly that it actually contributes anything to the oscillator
1/f noise. 

For instance, I'm sure we will all agree that the HP 10811 is a pretty
decent oscillator. 

FYI, essentially the same crystal is used in the E1938A that I designed.
  Those two 

oscillators have ENTIRELY different circuits.  Yet both achieve the same
close in 

phase noise as what I measured for the intrinsic noise of the crystal. 
By the way, 

the HP 10811 uses a selected 2N5179 transistor.  But is NOT SELECTED FOR
 

1/f NOISE.  Instead it is selected for good gain at high DC collector
current, because 

the HP engineer who designed it had trouble with the oscillator starting
up correctly, 

in conjunction with the ALC circuit. 

If you use the same output circuit for your oscillator as the 10811, you
will also 

have no oscillator transistor contribution to far out phase noise.  
This is all 

very well documented if you want to read the details. 

In a previous life, I designed many many 5th overtone oscillators.  Your
crystal 

like the Croven ones I used in those days is not of the same caliber as
the 

10811 crystal, so therefore it is even more unlikely that the transistor


you choose is going to make any difference.  I always locked my 5th OT  

crystals to something like a 10544 oscillator to clean them up close in 

using PLL synthesizers.  The 10811 hadn't been invented yet. 

I hope this gets you back on the right track.

---
Rick Karlquist
N6RK 

On 2023-01-16 14:26, glen english LIST via time-nuts wrote:

> Hi
> 
> I am embarking on building up a block - a common collector Butler oscillator with an overtone  crystal at 98.304 MHz
> 
> on on FR4 , I've used the Philips NPN type  BFS17.
> 
> The results <=10 Hz phase noise were quite disappointing. (-69 @ 98 MHz)
> 
> Does anyone have any good suggestions. Maybe something slower (larger) like a 2N5179 (or its SMT counterpart MMBT5179) will be better.
> 
> regards
> 
> glen
> 
> On 13/12/2022 11:14 am, Lux, Jim via time-nuts wrote: 
> 
>> On 12/12/22 2:43 PM, Alex Pumm
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe send an email to time-nuts-leave at lists.febo.com