[time-nuts] how to find low noise transistors

[Gerhard Hoffmann]

Am 20.07.2015 um 01:57 schrieb KA2WEU--- via time-nuts:
> Good evening, this turns out to be a good discussion...
>   
> Any more inputs ?  73 de Ulrich
>   
>

1. To get a gut feeling about the virtues of nonlinear noise simulation: 
how much phase noise will we typically lose if we stay with linear 
simulation? I mean, we have been told so often how important it is that 
the amplifier offers a constant (low) impedance to the crystal and that 
the smallest nonlinearity would be an invitation to noise up conversion. 
It does not take a lot of conversion gain when one looks at -150 dBc. 
So, even if we use a HB simulator, the DUT will have to be pretty linear.


2. What do you consider the optimum AGC for, say, a Driscoll or Butler 
at 100 MHz? In my current work, most of the logic is triple module 
redundant and the oscillator is a single point of failure. Stopping 
oscillation at an EFC extreme would be a nightmare, but phase noise 
performance still cannot be ignored.


3. Is there any work on AGC vs. post tuning drift?


4. In [1] there are is some treatment about removing 1/f noise of a RF 
transistor by active LF feedback. It is applied to a BFR93A and the 
effect can be seen clearly. There are other faster transistors that 
would need that much more urgently, and for > 40 dB of 1/f noise 
probably more loop gain would be required. I can see a place for an 
ADA4898 here… Also, there are 1K resistors in the bases of both the RF 
and the AF transistors while we are discussing here replicating the 
transistors to shrink the effective base spreading resistance. It seems 
that the improvement could be much larger.


BTW I got -145dBc @100MHz @100Hz with mass production BFR93 transistors 
in Butlers, and the limit seemed to be ONLY the crystal; most crystals 
were much worse, even when they had comparable parameters and were from 
the same batch.


5. One must always find a balance between optimum close-in or far-out 
noise. The emitter input impedance of a 2 stage Butler sustaining 
amplifier may serve as an example. Make it small and there will be only 
a slight operating Q degradation - but less power available to the input 
of the sustaining amp. with a given crystal current; needing more gain 
and raising the floor.
Make it larger, and you get less operating Q and better floor.

Only 10% of a crystal batch seem to provide excellent close-in noise, 
the others being easily 10 dB worse. These others are more or less free 
(at least already paid for). They still could be used as a post-filter 
to shrink the noise floor. It would be necessary to de-Q them with 
resistors so that they can withstand the power and that they do not 
spoil the close-in noise.

Or use a bridge xtal filter that has no crystal resonance on the center 
frequency. That would require some discipline when tuning the oscillator 
to avoid blowing the filter crystals. Far out the noise still would 
decrease by 6 dB/oct Fourier-frequency-wise. 20 dB better makes the 
difference between OK and excellent.

[1] Rohde/Newkirk: RF/Microwave Circuit Design for Wireless 
Applications, Wiley

very short excerpt for a few days on < 
https://picasaweb.google.com/lh/photo/XUfeAuD8TvNqBOMuJiPtltMTjNZETYmyPJy0liipFm0?feat=directlink>

73, Gerhard, DK4XP