[time-nuts] high rev isolation amps

Charles Steinmetz csteinmetz at yandex.com
Wed Mar 30 00:51:25 EDT 2016


Bob wrote:

>There were (and maybe still are) SOT-89 versions of the 2N3804 and 
>3906. They will handle more power than most of the other versions. 
>That gives you better Vce on the string.

Bruce wrote:

>The PZT3904 and PZT3906 are still available.

>With most of these old circuits reducing the LF noise contribution 
>to the emitter/collector current by the biasing circuit by utilising 
>lower noise power supplies and/or using improved biasing methods can 
>improve the clse in PN significantly.

The lower transistors on the totem pole operate in current mode and 
typically have only 3-5v from C to E, so their dissipation is very 
low even if the standing current is high.  I used TO-92 and SOT-23 
transistors there.  The top transistor, which must develop the output 
voltage, dissipates more than a TO-92 or SOT-23 should be asked to 
handle, even with a good heatsink, so I typically used TO-5 (TO-39) 
or SOT-89 transistors (with heatsinks) there.

It is definitely true that the published NIST designs did not 
minimize circuit noise.  I used *much* larger base bypass capacitors, 
and improved capacitance multipliers.  I also used separate 
capacitance multipliers for the base divider string and the collector 
supply.  I played with ultra-low-noise DC regulators on each base and 
for the collector supply, which gave modest improvements in both 
noise and distortion -- but the added complexity was not justified by 
the gains, IMO.  I did find that doing without the interstage 
resistors (from the collector of one transistor to the emitter of the 
next higher transistor) reduced both noise (modestly) and distortion 
(slightly).  In the very rare cases when a circuit oscillated (I 
experimented with *many* combinations of transistors), I used ferrite 
beads in place of the resistors or on one or more of the base 
leads.  Choosing transistors with very low base spreading resistance 
was another key to lowering noise.

Best regards,

Charles




More information about the time-nuts mailing list