[time-nuts] 5>10 doubler

Charles Steinmetz csteinmetz at yandex.com
Mon Feb 2 16:00:51 UTC 2015


Andrea wrote:

>I see. This configuration is in effect a common gate B-class (or AB, or
>"barely A") amplifier and the rectification is a side effect.
>But, what is the advantage between it and a couple of diode-connected
>transistors with a full A-class (more linear, so less spurs) 
>amplifier in front
>of it?

It is the rectification that causes the gross nonlinearities, not the 
amplification.  So no matter how linear an amplifier you make, the 
diodes (or Class B or AB amplifier) will cause gross nonlinearities 
that we do not want.  Furthermore, transistors have both even- and 
odd-order distortion products, while JFETs have predominantly 
second-order products.  So JFETs naturally tend to produce the second 
harmonic, while transistors also produce the odd-order products we 
are trying to avoid (as well as higher even-order products).

>I know that the circuit originates at NIST and thus there surely IS an
>advantage. Are it trading more spurs (that you can cancel out with filtering)
>for less phase noise (that you cannot recover anymore)?

I do not know precisely how the NIST circuit is biased, and as far as 
I know it is not general knowledge among time nuts -- so any 
substantive response would be conjecture.  I don't even know if NIST 
still uses it.  There are a few things to know -- NIST historically 
settles on something that works well enough, then sticks with it for 
a long time (until the phenomena they are trying to measure get 
distinctly better than their instruments).  NIST has lots of 
considerations besides pure performance, such as power consumption 
and fitting into old form factors, so they do not necessarily have 
the best possible solutions, even when they have just designed the 
next generation.  So, what we know for sure is that the JFET 
push-push doubler worked well enough for NIST's purposes when it was 
designed.  That does not mean improvements weren't possible.

>Adding negative feedback linearize further the "barely Class A" amplifier; so,
>it's good to sacrifice part of the gain of the push-push stage to reduce
>flicker noise (and thus add less phase noise) and at the same time spurs.

But it is the natural second-order distortion of the JFETs that makes 
it a particularly good way to build a push-push doubler.  We don't 
*want* to linearize it!

>If it's so, why use a nonlinear (or barely linear) gain stage to rectify?
>Using just one stage means in general less phase noise output (but with
>probably more spurs that can be filtered out), versus a more stage linear
>amplifier (perhaps with strong negative feedblack) followed by a rectifier?

The "barely Class A" push-push doubler does not rectify the signal -- 
it creates the second harmonic largely because of the device 
characteristic.  The design goal is to map the bias and input to the 
portion of the FETs' characteristic curve that has the best fit to a 
second-order transfer function, while at the same time holding noise 
down below the noise budget.  That is why medium-cutoff FETs like the 
J111 and J310 are the best choices, not sharp-cutoff FETs like 2SK369 
and BF862.

Best regards,

Charles






More information about the Time-nuts_lists.febo.com mailing list