[time-nuts] 1 PPS 50-ohm driver

Magnus Danielson magnus at rubidium.dyndns.org
Mon Apr 18 02:55:11 EDT 2016


Hi,

On 04/18/2016 06:14 AM, Bill Byrom wrote:
> For most oscilloscopes with bandwidth no greater than around 1 GHz, the
> response is Gaussian with a BT (bandwidth * risetime) product of about
> 0.35. So a 1 GHz oscilloscope has a risetime of about 350 ps. The 10%-
> 90% risetime of a series connected system is given by the square root of
> the sum of the squares of the individual risetimes.
>
> So if you had a 2 ns risetime pulse measured by a 350 ps (1 GHz BW)
> scope, the expected displayed risetime would be:
> sqrt( 4 + 0.1225) =  2.03 ns. So the scope bandwidth is unimportant in
> this case.

Indeed.

> More important is the oscilloscope probe (if used) and/or coaxial cable,
> launch into the cable,  and connectors. The slow rise after the initial
> edge in your scope image might be due to skin effect losses in the PC
> board, coax cable, and/or scope probe (or errors in an uncalibrated
> scope). But the rising edge looks great to me.

The slowly rising slope after the transition is complete is a sure 
tell-sign of skin effect. If you have reflections, they would give 
sections of higher/lower voltage and this one looks relatively clean in 
that regard. In practice, as long as you have a clean swift transition 
and no reflections large enough to be interpreted as another transition, 
you are fine. It may look ugly but still work for many digital signals. 
I this regard your transition is very clean.

> The problem with fast edges and fast login on the receiving end is that
> reflections due to impedance mismatch cause great difficulties. For
> example, if the load was a 50 ohm resistor in parallel with 20 pF to
> ground, a negative-going reflection would travel back to the source. If
> the round-trip propagation delay is 2 ns or less, the reflection will
> end up causing a reduced risetime or ringing at the source. Your setup
> has a good source match which should keep re-reflection to a minimum, so
> there shouldn't be any significant re-reflection problems.

Indeed. Making sure that both source and destination have relatively 
good impedance match is a good way towards clean signals.

Cheers,
Magnus


More information about the time-nuts mailing list