Thanks for that information, Marcus.
In time-nuts applications the information is usually in the leading edge
of the pulse, so that's what we worry about. But if you're latching
with a 100 ns clock, the PPS characteristics are a lot less critical --
you mainly need the pulse to be long enough so it's still there when the
next clock arrives.
Long, long ago I designed a simple circuit (TAPR sells it as the
"FatPPS") to stretch very short pulses from HP clocks and some GPS
receivers into something long enough to "catch" a PC serial port DCD
line for NTP use -- similar to what you described, the pulse needed to
be long enough to catch the next interrupt.
PPS can be surprisingly complicated!
John
----
On 8/27/25 07:44, Marcus Müller wrote:
Let me give a voice from the off:
- slew rate isn't as relevant for the Octoclock as clock/time source for
USRPs – these latch the PPS state on the rising edge of the internal
clock, which might or might not be derived from the 10 MHz from the
Octoclock and cleaned up after. That's why you'll find the schematics
[1] feature simple cascaded 7400 series inverters
- you're right, stating explicitly that this is the unloaded output
might have been nice, but then again, since characteristic impedance
doesn't matter for a signal that's DC for >= 100 ms, and since it's
"explicitly" omitted, I think that's a fair assumption. You'll even find
in the USRP X300 schematics (sheet 12) that the port is about 50 Ω-
loaded for sufficiently high frequencies, but at DC is effectively 3 kΩ-
terminated. Flat frequency response was seemingly not a design goal, but
rapid dampening of overshoot at low DC power was.
Cheers,
Marcus
[1] https://files.ettus.com/schematics/octoclock/
On 8/26/25 7:01 PM, John Ackermann N8UR via USRP-users wrote:
Agreed that it would be good for the datasheet to have mentioned this,
as it's a frequent cause of confusion.
While SMA connectors are designed for 50 ohms, it's really not safe to
rely on that to identify signal characteristics. After all, most
oscilloscopes use 50 ohm BNC connectors while their input impedance is
1 Mohm or more.
As I mentioned, there is no standardization for PPS signals
characteristics. As an example, older HP atomic clocks have PPS
output on a BNC connector that is well over 10 volts into 50 ohms
(unloaded it's way higher). They did that to increase the slew rate
to get more precise triggering in an era before modern logic gates
were widely used. But the width is very narrow (~20us) so the total
power in the pulse isn't that high.
Most modern systems use logic gate outputs don't always consider what
the load will be and that can cause issues. For example, running a u-
blox GNSS timepulse output into coax cable does not work well at all;
the source impedance is quite high and you don't get reliable
triggering at the far end of even a fairly short cable regardless of
termination. You need to buffer the signal before feeding it into coax.
I've designed several products with PPS outputs and use three 74AC04
gates in parallel to provide 5 volt no-load outputs with enough drive
capability to deliver at least 3 volts into 50 ohms. That has worked
very well driving many types of counters and other devices.
John
----
On 8/26/25 11:49, [email protected] wrote:
Hi John,
Thanks for this interesting insight!
However, I would think it is measured at 50 ohms as the outputs of
the PPS on the OctoClock-G are SMA connectors. As far as I am aware,
SMA connectors do not have a 1 Mohm impedance.
I found that inverters are placed before the PPS outputs, which have
a minimum output of 2.4 V and a typical output of 3.4 V. The input of
the USRP B210 (which takes in the signal from the OctoClock-G) has an
input range of 1.8 V to 5 V. So, I am not ruling out that mismatching
or matching at 1 Mohm can work too.
So it is not really an error in the datasheet. But, it would have
been nice if they had added an extra row in the table, pointing out
at which impedance this was measured, similar to the 10 MHz signal.
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
