@Bas, did you connect a USB-2 device to a USB-3 port or were both the device and the port USB-3?
usb-2 ttl-to-serial devices are quite common. usb-3 ttl-to-serial devices are not common or perhaps are nonexistent.
Non-FTDI cable (works great): https://www.adafruit.com/product/954
FTDI cable: https://www.adafruit.com/product/70
It seems that if a USB-3 port (PC-side) does interrupt time-stamping, then it shouldn’t matter if it’s a USB-2 or USB-3 device.
The Navisys GR701W uses a “Prolific Technology, Inc. PL2303 Serial Port” for the USB-serial conversion. While it works OK, accuracy isn’t ideal. This graph summarizes one experiment I ran.
Note that the Navisys PPS signal reaches the Rpi4 with a ~0.2 - 1.2 msec offset + high jitter.
hello @jos1264, thank you. that’s helpful. I’ve been emailing with sales@navisys.com.tw and they’ve been providing the same quote and payment mechanism. I’ve wondered why they don’t have their hardware online, but it seems they do. good find
Ok @stevesommars, the Raspberry Pi 4 has two usb2 ports and two usb3 ports. were you connected to usb3 ports when doing this test?
I understand latency in usb, but usb3 may have two mechanisms for accounting for these delays with (a) interrupt events rather than polling and (b) hardware timestamping.
So far nobody has been able to explain the favorable results from @Bas posted above on Feb 20 and Feb 22. It seems to be working with this particular device with PPS over usb3.
@stevesommars, I’ve been considering performing a similar test. the results you show are pretty interesting. they are more organized than expected. can you tell me more about this test? the test I’m considering will need to be in C (not python) and record delay from a known periodic source, like an arduino sending a small string at a fixed period, like 1000ms.
that results figure you’ve posted has an interesting pattern. it’s not totally random.