usrp-users@lists.ettus.com
Discussion and technical support related to USRP, UHD, RFNoC
View all threads10 MHz Reference Initial Phase in the X310
Hello,
I am attempting to release a transmission from an X310 every second. To
accomplish this, I must measure, and calibrate the delay in the RF front
end of the radio for my chosen sample rate. I'd like the transmission to be
released within 1 clock cycle of the rising edge of the PPS.
I am feeding the X310 an external 10 MHz reference and 1 PPS, which are
produced by the same source, and are being supplied to the radio with
matched cable lengths. The source is a GPS receiver and in my lab I have 2
different generations of the GPS receiver.
While calibrating the front end transmit delay I noticed a discrepancy in
the radio timing between the separate GPS receiver generations. The 1st
generation of GPS receiver is 50 ns different than the calibration for the
2nd generation. When I look at the 1 PPS and 10 MHz output on a scope, I
noticed that in the 1st generation the PPS occurs at the top of a 10 MHz
cycle, and in the 2nd generation it occurs at the bottom of a 10 MHz cycle.
Half a cycle at 10 MHz is 50 ns. I suspect this is not coincidence because
I have now tested 6 different GPS receivers, 3 of gen 1 and 3 of gen 2, and
all 3 gen 1 calibrations are the same and they are 50 ns different from the
gen 2 calibrations.
Is this the expected behavior? Or is there a bug in the X310 code that
handles timing? I have never worked on hardware, but I would not expect the
initial phase of a 10 MHz reference to impact absolute time.
Thanks for your help!
On 06/23/2020 02:45 PM, Aaron Smith via USRP-users wrote:
Hello,
I am attempting to release a transmission from an X310 every second.
To accomplish this, I must measure, and calibrate the delay in the RF
front end of the radio for my chosen sample rate. I'd like the
transmission to be released within 1 clock cycle of the rising edge of
the PPS.
I am feeding the X310 an external 10 MHz reference and 1 PPS, which
are produced by the same source, and are being supplied to the radio
with matched cable lengths. The source is a GPS receiver and in my lab
I have 2 different generations of the GPS receiver.
While calibrating the front end transmit delay I noticed a discrepancy
in the radio timing between the separate GPS receiver generations. The
1st generation of GPS receiver is 50 ns different than the calibration
for the 2nd generation. When I look at the 1 PPS and 10 MHz output on
a scope, I noticed that in the 1st generation the PPS occurs at the
top of a 10 MHz cycle, and in the 2nd generation it occurs at the
bottom of a 10 MHz cycle. Half a cycle at 10 MHz is 50 ns. I suspect
this is not coincidence because I have now tested 6 different GPS
receivers, 3 of gen 1 and 3 of gen 2, and all 3 gen 1 calibrations are
the same and they are 50 ns different from the gen 2 calibrations.
Is this the expected behavior? Or is there a bug in the X310 code that
handles timing? I have never worked on hardware, but I would not
expect the initial phase of a 10 MHz reference to impact absolute time.
Thanks for your help!
These are external GPS receivers? What kind? Given your scope
measurements, how would this be related to a bug in X310? I'm confused
as to how you're linking the 10MHz/1PPS phasing on your external GPS
receivers to the X310 having bugs.
Marcus,
They are EndRun Meridian and Meridian II units.
I am very ignorant on this topic. Is it a standard that the 1 PPS should
coincide with the top of a 10 MHz cycle? I just wouldn't expect the front
end transmit delay, relative to the 1 PPS input, to depend on the 10 MHz
reference phase. I don't understand how the 10 MHz reference and 1 PPS
input are used to synthesize time. Is the 1 PPS detection done at the
master lock rate (200 MHz) or at 10 MHz?
On Tue, Jun 23, 2020 at 1:06 PM Marcus D. Leech via USRP-users <
usrp-users@lists.ettus.com> wrote:
On 06/23/2020 02:45 PM, Aaron Smith via USRP-users wrote:
Hello,
I am attempting to release a transmission from an X310 every second.
To accomplish this, I must measure, and calibrate the delay in the RF
front end of the radio for my chosen sample rate. I'd like the
transmission to be released within 1 clock cycle of the rising edge of
the PPS.
I am feeding the X310 an external 10 MHz reference and 1 PPS, which
are produced by the same source, and are being supplied to the radio
with matched cable lengths. The source is a GPS receiver and in my lab
I have 2 different generations of the GPS receiver.
While calibrating the front end transmit delay I noticed a discrepancy
in the radio timing between the separate GPS receiver generations. The
1st generation of GPS receiver is 50 ns different than the calibration
for the 2nd generation. When I look at the 1 PPS and 10 MHz output on
a scope, I noticed that in the 1st generation the PPS occurs at the
top of a 10 MHz cycle, and in the 2nd generation it occurs at the
bottom of a 10 MHz cycle. Half a cycle at 10 MHz is 50 ns. I suspect
this is not coincidence because I have now tested 6 different GPS
receivers, 3 of gen 1 and 3 of gen 2, and all 3 gen 1 calibrations are
the same and they are 50 ns different from the gen 2 calibrations.
Is this the expected behavior? Or is there a bug in the X310 code that
handles timing? I have never worked on hardware, but I would not
expect the initial phase of a 10 MHz reference to impact absolute time.
Thanks for your help!
These are external GPS receivers? What kind? Given your scope
measurements, how would this be related to a bug in X310? I'm confused
as to how you're linking the 10MHz/1PPS phasing on your external GPS
receivers to the X310 having bugs.
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
On 06/23/2020 03:18 PM, Aaron Smith wrote:
Marcus,
They are EndRun Meridian and Meridian II units.
I am very ignorant on this topic. Is it a standard that the 1 PPS
should coincide with the top of a 10 MHz cycle? I just wouldn't expect
the front end transmit delay, relative to the 1 PPS input, to depend
on the 10 MHz reference phase. I don't understand how the 10 MHz
reference and 1 PPS input are used to synthesize time. Is the 1 PPS
detection done at the master lock rate (200 MHz) or at 10 MHz?
The 1PPS is used exactly once, when you do a "set_time_next_pps",
after which the time-of-day clock on the board is driven by the master
clock which is phase-locked to the 10MHz external reference. So, the
time-of-day clock on the board runs at (in the case of the X310) 200MHz
by default, so each "tick" is 5nsec. The 1PPS signal is probably
"sensed" by logic that's running at the master clock rate. So two X310
units may
still have a small amount of residual ambiguity about when 1PPS
"happens", by perhaps as much as 5Nsec. But I'm not an FPGA designer, so
this is just an mildly-educated "guess".
There may be some lose convention about the phase of the 1PPS with
respect to the 10MHz generated reference, but would not expect it
to be an "enforced standard". Different manufacturers will have
different "servo" algorithms for steering the 10MHz clock output
(usually, it's a
voltage-tunable VCTCXO or VOCXO) with respect to the derived 1PPS
pulses. The error on the 1PPS signal is typically surprisingly large--it's
1PPS +/- a few 10s of nanoseconds, and the phasing of that 1PPS with
respect to the 10MHz signal isn't, I think, necessarily a "standard".
On Tue, Jun 23, 2020 at 1:06 PM Marcus D. Leech via USRP-users
<usrp-users@lists.ettus.com mailto:usrp-users@lists.ettus.com> wrote:
On 06/23/2020 02:45 PM, Aaron Smith via USRP-users wrote:
Hello,
I am attempting to release a transmission from an X310 every
second.
To accomplish this, I must measure, and calibrate the delay in
the RF
front end of the radio for my chosen sample rate. I'd like the
transmission to be released within 1 clock cycle of the rising
edge of
the PPS.
I am feeding the X310 an external 10 MHz reference and 1 PPS, which
are produced by the same source, and are being supplied to the
radio
with matched cable lengths. The source is a GPS receiver and in
my lab
I have 2 different generations of the GPS receiver.
While calibrating the front end transmit delay I noticed a
discrepancy
in the radio timing between the separate GPS receiver
generations. The
1st generation of GPS receiver is 50 ns different than the
calibration
for the 2nd generation. When I look at the 1 PPS and 10 MHz
output on
a scope, I noticed that in the 1st generation the PPS occurs at the
top of a 10 MHz cycle, and in the 2nd generation it occurs at the
bottom of a 10 MHz cycle. Half a cycle at 10 MHz is 50 ns. I
suspect
this is not coincidence because I have now tested 6 different GPS
receivers, 3 of gen 1 and 3 of gen 2, and all 3 gen 1
calibrations are
the same and they are 50 ns different from the gen 2 calibrations.
Is this the expected behavior? Or is there a bug in the X310
code that
handles timing? I have never worked on hardware, but I would not
expect the initial phase of a 10 MHz reference to impact
absolute time.
Thanks for your help!
These are external GPS receivers? What kind? Given your scope
measurements, how would this be related to a bug in X310? I'm
confused
as to how you're linking the 10MHz/1PPS phasing on your
external GPS
receivers to the X310 having bugs.
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com <mailto:USRP-users@lists.ettus.com>
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com
I guess I just need to think more about the problem. With your explanation
I still don't understand why the 1 PPS would be detected differently in the
two situations, and I still don't understand why the phase of the 10 MHz
reference with respect to the 1 PPS should matter. I can understand some
variation, but not 10 clock cycles worth. I know the PPS signal moves
around in time, but I know it doesn't move much in the GPS receivers I use,
especially over short time scales. As I said, perhaps I just need to think
more about what you said. Thanks for the insight.
On Tue, Jun 23, 2020 at 1:27 PM Marcus D. Leech patchvonbraun@gmail.com
wrote:
On 06/23/2020 03:18 PM, Aaron Smith wrote:
Marcus,
They are EndRun Meridian and Meridian II units.
I am very ignorant on this topic. Is it a standard that the 1 PPS should
coincide with the top of a 10 MHz cycle? I just wouldn't expect the front
end transmit delay, relative to the 1 PPS input, to depend on the 10 MHz
reference phase. I don't understand how the 10 MHz reference and 1 PPS
input are used to synthesize time. Is the 1 PPS detection done at the
master lock rate (200 MHz) or at 10 MHz?
The 1PPS is used exactly once, when you do a "set_time_next_pps", after
which the time-of-day clock on the board is driven by the master
clock which is phase-locked to the 10MHz external reference. So, the
time-of-day clock on the board runs at (in the case of the X310) 200MHz
by default, so each "tick" is 5nsec. The 1PPS signal is probably
"sensed" by logic that's running at the master clock rate. So two X310
units may
still have a small amount of residual ambiguity about when 1PPS
"happens", by perhaps as much as 5Nsec. But I'm not an FPGA designer, so
this is just an mildly-educated "guess".
There may be some lose convention about the phase of the 1PPS with respect
to the 10MHz generated reference, but would not expect it
to be an "enforced standard". Different manufacturers will have
different "servo" algorithms for steering the 10MHz clock output (usually,
it's a
voltage-tunable VCTCXO or VOCXO) with respect to the derived 1PPS
pulses. The error on the 1PPS signal is typically surprisingly large--it's
1PPS +/- a few 10s of nanoseconds, and the phasing of that 1PPS with
respect to the 10MHz signal isn't, I think, necessarily a "standard".
On Tue, Jun 23, 2020 at 1:06 PM Marcus D. Leech via USRP-users <
usrp-users@lists.ettus.com> wrote:
On 06/23/2020 02:45 PM, Aaron Smith via USRP-users wrote:
Hello,
I am attempting to release a transmission from an X310 every second.
To accomplish this, I must measure, and calibrate the delay in the RF
front end of the radio for my chosen sample rate. I'd like the
transmission to be released within 1 clock cycle of the rising edge of
the PPS.
I am feeding the X310 an external 10 MHz reference and 1 PPS, which
are produced by the same source, and are being supplied to the radio
with matched cable lengths. The source is a GPS receiver and in my lab
I have 2 different generations of the GPS receiver.
While calibrating the front end transmit delay I noticed a discrepancy
in the radio timing between the separate GPS receiver generations. The
1st generation of GPS receiver is 50 ns different than the calibration
for the 2nd generation. When I look at the 1 PPS and 10 MHz output on
a scope, I noticed that in the 1st generation the PPS occurs at the
top of a 10 MHz cycle, and in the 2nd generation it occurs at the
bottom of a 10 MHz cycle. Half a cycle at 10 MHz is 50 ns. I suspect
this is not coincidence because I have now tested 6 different GPS
receivers, 3 of gen 1 and 3 of gen 2, and all 3 gen 1 calibrations are
the same and they are 50 ns different from the gen 2 calibrations.
Is this the expected behavior? Or is there a bug in the X310 code that
handles timing? I have never worked on hardware, but I would not
expect the initial phase of a 10 MHz reference to impact absolute time.
Thanks for your help!
These are external GPS receivers? What kind? Given your scope
measurements, how would this be related to a bug in X310? I'm confused
as to how you're linking the 10MHz/1PPS phasing on your external GPS
receivers to the X310 having bugs.
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com