Ryzen mobos with serial port for Garmin GPS?
On 11/5/21 11:32 AM, Andy Talbot wrote:
Use FTDI USB serial ports - you can't go wrong with them.
https://ftdichip.com/
I have, at the last count, used something like 200 of their FT232 device in
one form or another on the shack PC. I know that, because device manager
has registered up to COM200. Every time a new one is plugged in, a new COM
port is set up.
Andy
www.g4jnt.com
On Fri, 5 Nov 2021 at 18:26, Alec Teal alec@unifiedmathematics.com wrote:
A friend of mine who lives and breaths this stuff (I wont tell you what
he does - but suffice to say he's authoritative) basically said to me on
something about serial ports that you can't go wrong with USB stuff,
even on Linux.
Would that work?
Serial ports certainly are getting scarce! You'd get 2 to a board an
embarrassingly long time ago!
A caution - FTDI has several series of chips (FT23x, FT245) that wind up
in commodity USB to Serial/RS-422/etc products. Some of them do not have
driver support for MacOS post Mojave (MacOS pulled the FTDI driver into
the kernel, but it only accepts some PID/VID values, etc).
I can't speak to Windows 10 etc since I've not tried it.
I've got a dozen or so RS232 dongles that are now useless in a general
sense - sure, they work with the old PC, but they lurk waiting to
frustrate you when you plug them into another computer that doesn't
support them.
And then, of course, there are the famous FTDI clones which "sometimes"
work.
For Linux, you may or may not care - depending on libusb, etc. But I've
had version compatibility issues there, too (Ubuntu, various LTS
versions we use at work).
Just budget for potential replacements.
And, because this is timenuts - any modem control signal (e.g. RTS)
going across USB is going to have a 125 microsecond uncertainty due to
the 8kHz frame rate of USB.
For what it may be worth, here are the current clock performance
measurements for the two stratum-1 time references on my home LAN. Each
computer has an AMD FX-8350 processor on an ASUS motherboard, with a DB9
serial header on the motherboard, and each system has a Garmin 18x
receiver wired to a DB9 serial connector.
The jitter (dispersion) in each case seems to be significantly less than
what I might expect from a USB-based connection. I don't have specifics
anymore, but I remember I tried connecting a Garmin 18x GPS to a PCIe
serial card once, and the result was way too unstable to provide any
sort of useful time reference. I assume this was because the PCIe
serial card was really just a serial interface to a USB chip, right?
Note that I'm using chrony here (not ntp or ntpsec).
(11/05 12:27:04 richw@equality ~ 2118) $ chronyc tracking
Reference ID : 47505053 (GPPS)
Stratum : 1
Ref time (UTC) : Fri Nov 05 19:26:59 2021
System time : 0.000001197 seconds slow of NTP time
Last offset : -0.000000044 seconds
RMS offset : 0.000001942 seconds
Frequency : 8.615 ppm slow
Residual freq : -0.000 ppm
Skew : 0.052 ppm
Root delay : 0.000000001 seconds
Root dispersion : 0.000030960 seconds
Update interval : 16.0 seconds
Leap status : Normal
(11/05 12:27:10 richw@equality ~ 2119) $
(11/05 12:27:39 richw@memoryalpha ~ 2000) $ chronyc tracking
Reference ID : 47505053 (GPPS)
Stratum : 1
Ref time (UTC) : Fri Nov 05 19:27:25 2021
System time : 0.000005428 seconds slow of NTP time
Last offset : -0.000024016 seconds
RMS offset : 0.000031762 seconds
Frequency : 30.816 ppm slow
Residual freq : -0.025 ppm
Skew : 1.607 ppm
Root delay : 0.000000001 seconds
Root dispersion : 0.000054530 seconds
Update interval : 16.0 seconds
Leap status : Normal
(11/05 12:27:43 richw@memoryalpha ~ 2001) $
Rich Wales
richw@richw.org
On Thu, 4 Nov 2021 22:36:29 -0700, you wrote:
(Yes, I know that PCI serial port cards exist -- however, these are not
suitable for connecting a GPS time reference because the PPS signal
timing is unpredictable -- so I've limited myself to using motherboards
with a builtin serial port.)
What timing issues did you have?
I've used a similar setup on two old windows machines (one XP and one
Win10) and I didn't see much difference between the built in serial
port and the ones on the PCI serial cards. They were not PCIe.
Angus wrote:
/What timing issues did you have? I've used a similar setup on two old
windows machines (one XP and one Win10) and I didn't see much
difference between the built in serial port and the ones on the PCI
serial cards. They were not PCIe./
The serial card I used (with unpredictable PPS signal timing) was a PCIe
card. But I think the more significant difference was that this PCIe
serial card was really a USB card with a serial-to-USB front end, and
this protocol conversion was the reason for the extra jitter.
Old PCI serial cards came from a time before USB was common (or even
existed), and they sent the serial data directly to the PCI bus without
going through any intermediate steps. If we could find PCI or PCIe
serial cards which worked in this same way, we would presumably have the
same high performance as in the old days. However, there is very little
incentive nowadays for hardware manufacturers to do this; the few serial
cards that are still made today (after all, who really needs, wants, or
even knows about RS232 anymore?) generally cut corners by slapping a
cheap serial-to-USB front end onto a cheap USB card, and except for
those of us who really need accurate timing, none of their customers are
likely to care or even notice.
Rich Wales
richw@richw.org
On Nov 5, 2021 9:14 PM, Rich Wales richw@richw.org wrote:
If we could find PCI or PCIe serial cards
The old Exar product line that MaxLinear acquired might still have PCIe
UART.
https://www.maxlinear.com/product/interface/uarts/pcie-uarts/xr17v352
If you could find those on a card they should be good performance. I
haven't measured the performance of the UARTs in super IO chips. Those
would share the upstream connection with a lot of other things, so
might vary in latency.
-- Chris Caudle
Hi
I would suggest that this is a very unusual way to do a serial card. There are
a lot of PCIe cards out there that go straight to the bus. Amazon is awash in them.
If you stick with the 1 or 2 port versions, you should not get anything to weird.
Bob
On Nov 5, 2021, at 10:14 PM, Rich Wales richw@richw.org wrote:
Angus wrote:
/What timing issues did you have? I've used a similar setup on two old windows machines (one XP and one Win10) and I didn't see much difference between the built in serial port and the ones on the PCI serial cards. They were not PCIe./
The serial card I used (with unpredictable PPS signal timing) was a PCIe card. But I think the more significant difference was that this PCIe serial card was really a USB card with a serial-to-USB front end, and this protocol conversion was the reason for the extra jitter.
Old PCI serial cards came from a time before USB was common (or even existed), and they sent the serial data directly to the PCI bus without going through any intermediate steps. If we could find PCI or PCIe serial cards which worked in this same way, we would presumably have the same high performance as in the old days. However, there is very little incentive nowadays for hardware manufacturers to do this; the few serial cards that are still made today (after all, who really needs, wants, or even knows about RS232 anymore?) generally cut corners by slapping a cheap serial-to-USB front end onto a cheap USB card, and except for those of us who really need accurate timing, none of their customers are likely to care or even notice.
Rich Wales
richw@richw.org
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.
Check Startech (https://www.startech.com). They have a range of RS-232
and similar I/O cards and at least the last time I bought one several
years ago, the chip emulated a 16550 UART and didn't seem to have more
jitter than any other serial board I'd tried.
John
On 11/6/21 12:32 PM, Chris Caudle wrote:
On Nov 5, 2021 9:14 PM, Rich Wales <richw@richw.org> wrote:
> If we could find PCI or PCIe serial cards
The old Exar product line that MaxLinear acquired might still have PCIe
UART.
https://www.maxlinear.com/product/interface/uarts/pcie-uarts/xr17v352
If you could find those on a card they should be good performance. I
haven't measured the performance of the UARTs in super IO chips. Those
would share the upstream connection with a lot of other things, so
might vary in latency.
-- Chris Caudle
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.
Although I didn't buy any of them for timing, I did make sure that
they were based on a 16550 UART since a lot of legacy hardware and
software expects that. Startech emphasizes the 16550 and no bridge,
but they are not cheap and I'm not convinced that they're worth it.
Mine were just cheaper 2 to 6 port ones, although it is possible to go
too cheap!
On Fri, 5 Nov 2021 19:14:55 -0700, you wrote:
Angus wrote:
/What timing issues did you have? I've used a similar setup on two old
windows machines (one XP and one Win10) and I didn't see much
difference between the built in serial port and the ones on the PCI
serial cards. They were not PCIe./
The serial card I used (with unpredictable PPS signal timing) was a PCIe
card. But I think the more significant difference was that this PCIe
serial card was really a USB card with a serial-to-USB front end, and
this protocol conversion was the reason for the extra jitter.
Old PCI serial cards came from a time before USB was common (or even
existed), and they sent the serial data directly to the PCI bus without
going through any intermediate steps. If we could find PCI or PCIe
serial cards which worked in this same way, we would presumably have the
same high performance as in the old days. However, there is very little
incentive nowadays for hardware manufacturers to do this; the few serial
cards that are still made today (after all, who really needs, wants, or
even knows about RS232 anymore?) generally cut corners by slapping a
cheap serial-to-USB front end onto a cheap USB card, and except for
those of us who really need accurate timing, none of their customers are
likely to care or even notice.
Rich Wales
richw@richw.org
time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.