PIO 1us Timestamp - solution with DMA, FIFO and TIMELR/TIMEHR registers

[mungewell]

About a year ago I was asking about time-stamping when a PIO goes high/low, and I've now got a solution using DMA to copy TIMELR/TIMEHR registers into a FIFO, and then another PIO copies TX FIFO directly back to RX FIFO.

App can then (leisurely) grab the precised timestamps from FIFO.

Example code:
dma_pio_timestamp.py.txt

So with a GPS providing a 1PPS pulse, I get a nice count of us values.

Starting test, waiting for trigger
Timestamp: 0x000000000069f625 (6944293 us)
Timestamp: 0x0000000000793865 (7944293 us)
Timestamp: 0x0000000000887aa5 (8944293 us)
Timestamp: 0x000000000097bce5 (9944293 us)
Timestamp: 0x0000000000a6ff25 (10944293 us)
Timestamp: 0x0000000000b64165 (11944293 us)
Timestamp: 0x0000000000c583a5 (12944293 us)
Timestamp: 0x0000000000d4c5e5 (13944293 us)
Timestamp: 0x0000000000e40825 (14944293 us)
Timestamp: 0x0000000000f34a65 (15944293 us)
Timestamp: 0x0000000001028ca5 (16944293 us)
Timestamp: 0x000000000111cee5 (17944293 us)
Timestamp: 0x0000000001211125 (18944293 us)
Timestamp: 0x0000000001305366 (19944294 us)

The eagle-eyed among us will have spotted that the last value is 1us 'different' from expected, this is down to my XTAL not being perfect on this board - but it's pretty damn good.

A longer (over night) run gives:

Timestamp: 0x0000000009888e6b (159944299 us)
Timestamp: 0x000000000997d0ab (160944299 us)
Timestamp: 0x0000000009a712eb (161944299 us)
Timestamp: 0x0000000009b6552b (162944299 us
...
Timestamp: 0x0000000938e08b34 (39608945460 us)
Timestamp: 0x0000000938efcd74 (39609945460 us)
Timestamp: 0x0000000938ff0fb4 (39610945460 us)
Timestamp: 0x00000009390e51f4 (39611945460 us)
Timestamp: 0x00000009391d9434 (39612945460 us)

So between second '160' and second '39609' (delta 39449s) the fractional portion went from '944299' to '945460' (delta 1161us), meaning my 12MHz XTAL is actually "(39500.00161 / 39500) * 12" = 12.000000353MHz (ie off by about 29ppb)... pretty sure I replace the XTAL with TCXO on that board ;-)
https://www.digikey.ca/en/products/detail/ecs-inc/ECS-TXO-3225-120-TR/5357946

Note: that's average... the log shows frequency fluctuations up to that point.

After that point in the log... something else happened.

Timestamp: 0x000000093a9b0c74 (39637 945460 us)
Timestamp: 0x000000093aaa4eb4 (39638 945460 us)

goes 'wonky' here....
Timestamp: 0x000000093aaa8d79 (39638 961529 us)
Timestamp: 0x000000093aaaabb6 (39638 969270 us)
Timestamp: 0x000000093aacf5c1 (39639 119297 us)
Timestamp: 0x000000093aad1467 (39639 127143 us)
...

Unfortunately this code doesn't work on the RP2350 at the moment, LED alternates with PIO input but no timestamps are printed.

[mungewell]

The 'wonky' bit is very close to the end of the log, perhaps I disturbed the cables when I came in and shut the log down.

BTW this is what the 1PPS signal looks like:

[mungewell]

rp2350 introduced some DMA changes...

[mungewell]

rp2350 uses a different timer base value....
dma_pio_timestamp.py.txt

MicroPython v1.27.0 on 2025-12-09; Raspberry Pi Pico2 with RP2350
Type "help()" for more information.
>>> 

MPY: soft reboot
Starting test, waiting for trigger
Timestamp: 0x000000000509ade7 (84520423 us)
Timestamp: 0x000000000518f02a (85520426 us)
Timestamp: 0x000000000528326e (86520430 us)
Timestamp: 0x00000000053774b1 (87520433 us)
Timestamp: 0x000000000546b6f5 (88520437 us)
Timestamp: 0x000000000555f939 (89520441 us)
Timestamp: 0x0000000005653b7c (90520444 us)
Timestamp: 0x0000000005747dc0 (91520448 us)
Timestamp: 0x000000000583c003 (92520451 us)

Yes, that's a stock XTAL...

[mungewell]

Wrapping this up with a script to configure GPS (Mosaic X in my case), trigger timestamps from the PPS signal, whilst monitoring the UART for GGA/ZDA messages.
gps_dma_timestamp.py.txt

Timestamp output tied to/references GPS message time.

MPY: soft reboot
Raspberry Pi Pico2 with RP2350
Starting test, waiting for trigger...
Timestamp for b'012251.00' is 0x0000000042c326a3 (1120085667 us)
Timestamp for b'012252.00' is 0x0000000042d268e7 (1121085671 us)
Timestamp for b'012253.00' is 0x0000000042e1ab2a (1122085674 us)
Timestamp for b'012254.00' is 0x0000000042f0ed6e (1123085678 us)
Timestamp for b'012255.00' is 0x0000000043002fb2 (1124085682 us)
Timestamp for b'012256.00' is 0x00000000430f71f6 (1125085686 us)
Timestamp for b'012257.00' is 0x00000000431eb439 (1126085689 us)
Timestamp for b'012258.00' is 0x00000000432df67d (1127085693 us)
Timestamp for b'012259.00' is 0x00000000433d38c1 (1128085697 us)
Timestamp for b'012300.00' is 0x00000000434c7b04 (1129085700 us)
...

[mungewell]

One more, that adds logging of CPU voltage and (averaged) temp.
gps_dma_timestamp.py.txt