[time-nuts] Fwd: Re: homebrew 13 dBm distribution amplifier based on NIST design 5 to 100 MHz

hasweb at has.org.nz hasweb at has.org.nz
Mon Oct 13 23:37:10 UTC 2008

----- Forwarded message from hasweb at has.org.nz -----
     Date: Tue, 14 Oct 2008 12:34:56 +1300
     From: hasweb at has.org.nz
Reply-To: hasweb at has.org.nz
  Subject: Re: [time-nuts] homebrew 13 dBm distribution amplifier  
based on NIST design 5 to 100 MHz
       To: Gerhard Hoffmann <dk4xp at hoffmann-hochfrequenz.de>

Quoting Gerhard Hoffmann <dk4xp at hoffmann-hochfrequenz.de>:

> On Tue, 14 Oct 2008 08:35:55 +1300, you wrote:
>>     One trick that has been used for fixed frequency isolation amplifiers
>>     is to use a low Q series tuned LC circuit to short out the resistor
>>     in series with the base at the frequency of interest.
> Yes, but when I burn close to 3 Watts / channel and accept two    
> transistor chains
> to make it possible to get rid of of the transformers that hurt    
> Tom's application,
> then I won't easily accept new ferrite parts that might spoil that    
> hard won advantage.
> And all the impedance curves for ferrite beads that I have end at    
> 500 MHz or so
> and I would need them at 2.5GHz.
>>     Using a suitable ferrite bead instead of the resistor may allow a
>>     lower base to ground impedance at the signal frequency whilst ensuring
>>     transistor stability.
>>     Such instabilities are perhaps one reason why Spectracom and NIST use
>>     much lower ft transistors in their low noise 1-20MHz    
>> distribution amplifiers.
> The BFG31s have the advantage that I have a reel of them in the    
> drawer and the alternatives
> are few. Who else but NXP has a 1W PNP SMD wideband transistor in    
> active production?
> OK. At least TO-5?
> Killing ft is probably the easiest part, but precise delay comes
> naturally with large bandwidth.
> regards, Gerhard dk4xp
> (who starts thinking about T-coils and ft-doublers :-)
> Would 1 GHz distribution really be a barrier? Probably not.
> I have some experience designing 10 GBit/s fiber optic XFP modules and that
> should work nicely at 5GHz. )


You don't have to use ferrite cored inductors for the series resonant
base bypass networks.

I can remember forming Tcoil terminations by forming an inductor from
the lead wire of the axial leaded termination resistor and adjusting
the turns spacing whilst using a classical sampling scope to observe
the reflections from the input.

In practice one often needs to limit the bandwidth of a frequency
distribution system to keep the integrated phase noise at a reasonable

A 1GHz bandwidth distribution amplifier is really only necessary when
one has to distribute a 1GHz signal.

Even with more modest bandwidth distribution amplifiers using lower ft
transistors, the phase shift tempco of a typical double balanced mixer
at 5 or 10MHz can be much larger than that of the distribution
amplifier and can dominate the phase shift tempco of a system.


----- End forwarded message -----

More information about the time-nuts mailing list