Hi

I 100% agree that the TICC is a great choice. One very minor point:

Like anything else, you will need an accurate / stable “standard” to feed it with. This is also true of any reasonable alternative I can think of. There are a range GPSDO and telecom Rb options to sort through to meet that need.  Yes, there are a lot of possible choices, those are the ones that most folks turn to.

Bob

I've also been tinkering with building an STM32-based timestamper, though
my goal of making it more broadly available for sale keeps getting
postponed. It has 6ns resolution and ultimately will have 4 channels; I
project it'll cost about $50. However if you'd like to play with one I can
send you a 2-channel prototype.

As Bob mentioned, virtually every timestamper including professional lab
gear will require an external timebase to get the kind of stability you
want. The usual arrangement is for the timebase to feed a 10MHz reference
signal into the timestamper. The best option in my opinion is a GPS-steered
oscillator. Time-nut Nick Sayer has an excellent one (
https://www.tindie.com/products/nsayer/gps-disciplined-ocxo/). You can also
buy cheapies on eBay such as https://www.ebay.com/itm/166333117355; my
friend has that one and has been satisfied with it.

On Sun, Feb 11, 2024 at 9:41 AM Bob Camp via time-nuts <
time-nuts@lists.febo.com> wrote:

Still needs a stable 10 MHz reference clock.

Kind regards,

John

https://linkedin.com/in/john-h-6a131b12/

Juan,

If you can make it fit your budget, the TAPR TICC is a strongly
recommended solution. Feed it a 10 MHz reference, then you can configure
it to time-stamp two PPS pulses. I wrote code to integrate it to a
larger system, and that worked and was relatively easy to do. I find
that most issues I had was really due to me not having enough time to
clean the pipe, not the device itself. The back end is an USB that you
just hook up to your computer. On Linux is shows up at /dev/ttyACM0 and
from there it is fairly simple. It is recommended you learn how to
configure it on start-up. In a perfect world, I could drop a
reconfiguration anytime, and I've heard to test code exist to support
that, but I have not had the time to tinker around on it. As I said, I
did not have the time to clean the pipe, but my takeaways is never the
less that it is a great solution, compact and great performance.

Cheers,
Magnus

On 2024-02-11 13:37, juan--- via time-nuts wrote:

Juan, For a quick and cheap measurement I've used the TDC7200 alone with a
10 MHz reference.
It only measures from the PPS to the next clock edge (it can't resolve the
100 ns ambiguity) but it's ideal for comparing GPS PPS to Lab 10 MHz. GPS
doesn't move more than 50 ns between pulses, and the TDC's 55 ps resolution
is quite good enough for GPS.

Use the $15 Mikroe TDC Click board
https://www.digikey.com/en/products/detail/mikroelektronika/MIKROE-4770/15280756
with two SMA connectors.
GPS PPS connected to the Start input.
10 MHz goes to the Stop input.  (Must be a 3.3 v square wave, it doesn't
work with a sine clock).
The TDC specifies a minimum Start-Stop time of 12 ns. In practice I've
found that with a periodic Stop signal, it wraps very cleanly with no bad
behaviour. ie. Stop is properly gated, so it either captures the stop at ~6
ns, or it skips that one and captures the following stop at ~106 ns.
Nothing in between. This is fine for GPS PPS which you can easily unwrap.
Trigger the TDC as soon as you're finished processing one measurement,
it'll wait for the next PPS.
The Mikroe has an on-board TCXO, quite good enough over these short
intervals.

I've used it on breadboard with a Pi Pico and micropython. For a slightly
more solid solution there's also a "hat" for the full size raspberry pi,
that carries two or three click boards.
Everything after the 100 ns still depends on the clock of course.

Thomas

Hi Juan,

The STM32G4 is a great device for timestamping - it's what my timestamper
uses! It can run up to 170MHz which gives you a timing resolution of just
under 6ns.

As you noticed, some of the models (e.g. the STM32G474) do have the HRTIM
peripheral, and that was the source of my original excitement for
time-nuttery. It has a DLL of 32x on top of the 170MHz system clock, which
made me think I could use it to get 184ps resolution from a $10 chip.
Unfortunately when I sat down to finally write the code I realized that the
HRTIM only gives you that kind of resolution for output waveforms. Input
capture still runs at the system clock frequency.

In any case, I think it's still impressive to get 6ns resolution out of a
$5 chip (the STM32G431). My code and schematics are all open source; see:
https://github.com/jelson/rulos/blob/main/src/app/timestamper/timestamper.c
. My prototype board uses the STM32G431, with a 10MHz reference clock input
from the oscillator of your choice. I've used both GPSDOs and Rubidium
LPRO-101's. My board has a small analog circuit that changes the 0-centered
AC input of the reference clock to an above-zero waveform that can be used
as the stm32's system clock. The stm32 chip has an onboard PLL that I
configure to multiply the 10MHz ref clock up to a 170MHz system clock. Then
I use the input capture feature of the chip to acquire timestamps.

-Jeremy

On Thu, Feb 15, 2024 at 8:04 AM Juan Solano via time-nuts <
time-nuts@lists.febo.com> wrote: