[time-nuts] sine to square wave converter

bruce at ko4bb.com bruce at ko4bb.com
Fri Jul 11 03:50:06 EDT 2014

The Achilles heel of that biasing technique is that the emitter currents of the
pair of pnp's is affected by the noise on the 20V supply.
The 20V supply noise is only attenuated by a factor of 5 or so when both
transistors have equal collector currents.

I have both an LTC evaluation board for the LTC6957 and a Holzworth sine to CMOS
converter as well as a Timepod so I could measure the phase noise of both of
these for various 10MHz input signal levels.


> On July 10, 2014 at 8:21 PM Charles Steinmetz <csteinmetz at yandex.com> wrote:
> Bruce wrote:
> >Currently Linear Technology's sine to square wave devices with selectable
> >filtering (LTC6957 series) are better in that they are a closer
> >approximation to
> >the ideal zero crossing detector.
> >Failing that the next best is perhaps an AC coupled (both at input
> >and between
> >emitters) differential pair of 2N3906's or similar.
> My initial results with the LTC6957 did not produce lower phase noise
> at 10MHz than an optimized Wenzel two-PNP circuit (it may be possible
> to do better than my initial experiments with the 6957).
> Here is the circuit I use:
> Emacs!
> Using a 20v supply reduces the input feedthrough due to Q1's B-E
> capacitance, which tends to give the output square wave a sloping top.
> Using MPSH81s rather than 2N3906s helps with feedthrough, also, as
> well as reducing the rise and fall times (both about 2-4 nS with this
> circuit, depending on how hard it is driven, if it is built with
> proper attention to layout and stray capacitance).
> Some will insist that the LM329 is overkill, but the base bias can be
> a significant source (even the dominant source) of phase
> noise/jitter. The stability and low noise of the 329 improve
> performance materially -- even a TL431 or 1N829 is measurably
> inferior. An LM399 is somewhat better than the 329, but I have not
> found it necessary in practice. (Note that the pullup resistor is
> not shown -- 1.5k to 10k metal film from the 329 to +20v, not critical.)
> Some additional improvement can be achieved by using the PNP devices
> in an HFA3096 or HFA3128 array, but I have generally not seen the
> need for this in practic. As drawn, this circuit has lower residual
> PN than any 10MHz oscillator I have measured.
> Works best with input levels from 1 to 10Vpp (350mV to 3.5Vrms sine
> wave). There is a small duty cycle asymmetry (high longer than low),
> which depends on drive level. Using faster devices (such as HFA3096
> or HFA3128) reduces the asymmetry. If this is a problem, a resistor
> can be added from the base of Q1 to ground to trim out the asymmetry
> if the input level is well controlled. Otherwise, the mean output
> voltage can be detected, compared to a reference, and used to adjust
> either base voltage with a servo loop.
> Best regards,
> Charles
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.

More information about the time-nuts mailing list