Actisense NGT-1-USB powers down after ~1 hr

[Martin Crick]

Hi,

This is not strictly a pure Signal-K issue, but the device is feeding a Signal-K server, so I'm hoping some experts may be able to help.

My NGT-1-USB works fine for a shortish period of time, then seems to power down. I can reactivate it by unplugging and replugging it into the Pi.

I have a Raspberry Pi 3B+ fed by an Actisense NGT-1-USB from my Raymarine NMEA2000 network.

I asked Actisense if there was a timeout parameter on the NGT-1 and they assure me there isn't. Incidentally, the support engineer had never heard of the device being used in this sort of situation - which surprised me as it's recommended all over the Signal-K documentation!

They advised me there must be something on the Pi that is causing the USB port to power off the NGT-1. Since a GPS antenna is running continuously on the PI via USB and showing no problem, as also is the file system for the Pi, I'm at a loss as to what that might be. 

The Pi is powered by a DC-DC voltage regulator giving 5V, 2.5A output from the boat's 12V supply.

https://www.amazon.co.uk/dp/B07P87KXNY/ref=cm_sw_em_r_mt_dp_NG7BGSKT86SJNZTV461X?_encoding=UTF8&psc=1

Any ideas? Suggestions for diagnostics to set - either on the Pi or in the Signal-K server?

Thanks in advance.

[Bradford Harley]

Bump the usb voltage to 5.2v.  You are probably getting voltage drop and the actisense may be more sensitive than the go’s puck.  

I run my actisense 7x24. 

Brad Harley. — Sent from my iPhone

On Apr 23, 2021, at 08:28, 'Martin Crick' via Signal K <sig...@googlegroups.com> wrote:

Hi,

--
You received this message because you are subscribed to the Google Groups "Signal K" group.
To unsubscribe from this group and stop receiving emails from it, send an email to signalk+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/signalk/d7fc56c1-ff4d-41fe-a5e1-469a2acda6a4n%40googlegroups.com.

[Martin Crick]

Brad,

Thanks for the confirmation it should work!

If you use a Pi on board, can you advise what power supply solution you use? The one I picked up does not have adjustable output voltage.

[Bradford Harley]

I have used the lm2596 successful and recently moved to mp1584en regulators with the theory that it would generate less noise.  Neither were noticeable with my boat.   I now run a mean well sd50 which is way overkill, but I never worry about power any more.  I would probably buy the sd-25 due to size. 

Brad Harley. — Sent from my iPhone

On Apr 23, 2021, at 10:07, 'Martin Crick' via Signal K <sig...@googlegroups.com> wrote:

Brad,

To view this discussion on the web visit https://groups.google.com/d/msgid/signalk/1878b00c-aac1-4318-97e2-7d350a29a593n%40googlegroups.com.

[beerca...@gmail.com]

I have been using the NGT-1 USB for years now without any problems. I do have a beefy power supply (fixed 5V, 5A):

https://www.amazon.com/EPBOWPT-Converter-Regulator-Supply-Transformer/dp/B01M03288J 

You also need to use nice fat copper power cabling to the RPi and USB hub in order not to lose too much voltage. 

My NGT-1 is connected through the powered USB hub.

The NGT-1 also worked fine earlier on when my power supply was not so good...

Regarding your debugging, do you see the NGT-1 fail when you hook it up to a windows machine running NMEA reader?

https://actisense.com/acti_software/nmea-reader/

[mats_a]

Hi !

Did you solve the problem ? 

The thing is that I have got the same problem. The difference seems to be that I have used the NGT-1 together with different RPi:s and SignalK for several years. Last summer it was with a RPi 4 and the hardware setup is identical to what I use today. The only thing that differs is that this year its a fresh install with latest OS and SK.

I have a backup, YDWG-02, which works OK so its not the N2K bus.

/Mats

[Brian Scally]

When I used an NGT-1-USB I used a 
https://www.amazon.com/dp/B01ENT1PPM/ref=cm_sw_r_cp_apa_glt_fabc_54E6GB638YT14XV39M4J
For the PSU. There are a couple of others similar but this seems good.
This was hard wired to a Pi3.

My ususal setup is Pi4 running on POE.  Data from an Yacht Device YDEN-2,

The data is logged to a time series database that is running on the vessel data store.

I wrote a simple macro to scan the data for gaps longer than 0.5s in one of the 10Hz GPS feeds.

My baseline shows no data gaps.

--

I ran several tests

Test 1
My usual Pi 4 with the latests installs, fully updated and an NGT-1-USB.  For power ( as this was a temporary test ) I just plugged into a Blue Sea Systems USB aux socket.

This is not necessarily a clean install and has plugins and some Node Red stuff running on it.

This showed the dropping behaviour as described. (Twice in 6h)

Test 2

I reserected a Pi 3 that I used several years ago and it did not reproduce the problem, with the same NGT-1-USB and 

I left the P1 3 as was and did not do any updates ( it has been off for a couple of years ) 

There was no problem

Test 3 

I took the NGT-1-USB and plugged it in to a PC and logged the output, and it ran for 6h.
Parsed the output for the 10Hz GPS

Test 4

I took my usual Pi 4 on the Blue Seas Systems USB socket and connected to my 

Yacht Device YDEN-2 over ethernet, and I saw no problems.

--

What this suggested to me is that may be ther NGT-1-USB is on the edge of the PSU or that there is a serial rate problem, and the Pi 4 driver has a problem.

More tests on Monday as I have gone sailing for the weekend so everything got put back.

But Monday tests are:

1- A Pi4 with a fresh install and  nothing else running on it ( I will need just the InfluxDB 2 plugin ) 

2- Changing the baud rate of the NGT-1-USB and seeing if this makes it better or worse.
Mine is set to the default 115200.

Actisence say this: 
https://actisense.com/knowledge-base/nmea-2000/emu-1-nmea-2000-engine-monitoring-unit/what-baud-rate-should-my-ngt-1-be-set-to/

3- Changing the PSU.

A question - what sort of NMEA 2K loads are you seeing?
I will switch stuff on or off to generate a similar load level.

My suspision is that it is a baud rate issue in the NGT-1-USB.

Brian

--
You received this message because you are subscribed to the Google Groups "Signal K" group.
To unsubscribe from this group and stop receiving emails from it, send an email to signalk+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/signalk/a3e3c1a6-da23-49e7-9ea6-cb41efb1c300n%40googlegroups.com.

[Martin Crick]

Thanks to everyone for suggestions.

Early advice was that power might be the problem. So I’ve tested with a powered USB hub between the Pi3 and the NGT-1. The hub is powered off the mains. This showed the same dropping out behaviour. (I’ve also bought some alternative power supplies to try, but suspect that isn’t the issue. Oh well!)

Like Brian, my setup used to work fine, til I upgraded to the latest Raspberry PI OS. 

I therefore suspect that Brian may be onto something with the baud rate tests - I look forward very much to hearing if that solves his problem. However, my current tests are running with pretty low NMEA2000 data sources - wind, depth, two tank gauges, autopilot computer (which of course includes heading etc. even while moored up). The chart plotter and AIS transponder are off. I’ve turned off the Signal K to NMEA conversion plugin, so the traffic should be only one way. The data connection report on the SignalK dashboard is showing 67.2 deltas per second, or 48%. 

[Bradford Harley]

What does the system and kernel log say?   Is the USB device getting errors?   Linux will normally tell you what's going wrong with a little digging.

To view this discussion on the web visit https://groups.google.com/d/msgid/signalk/2b1e02d6-ef59-449e-9002-2352f1071319n%40googlegroups.com.

--

Brad Harley

[Martin Crick]

Hi,

Latest status.

It stayed up all night (longest period by far) with:

- powered USB hub between Pi and NGT-1

- only wind, depth, auto-pilot sources on N2K

- no Signal-K to N2K transmission (only reception by Signal K)

Then turned on chart plotter. This adds GPS data to the N2K network. Stayed up for further 2 hours (again, longer than achieved previously). This gave loads on the Signal K dashboard for the NGT-1 of 87 deltas/s, 54%

Then turned on AIS. Load increased to 90 deltas/s, 55%. NGT-1-USB still running 3hrs later.

Now enabled Signal-K to N2K plugin. Atmospheric pressure and temperature enabled.
Within 5 minutes the Actisense has shutdown. Nothing obvious in /var/logs/syslog:

Apr 26 12:57:14 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[36m304#033[0m 24.326 ms - -#033[0m

Apr 26 12:57:34 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[36m304#033[0m 22.211 ms - -#033[0m

Apr 26 12:57:46 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[36m304#033[0m 25.416 ms - -#033[0m

Apr 26 12:57:55 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[36m304#033[0m 30.943 ms - -#033[0m

Apr 26 12:59:33 eowynpi signalk-server[310]: #033[0mPOST /skServer/plugins/sk-to-nmea2000/config #033[32m200#033[0m 10.910 ms - 45#033[0m

Apr 26 12:59:33 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[32m200#033[0m 20.890 ms - -#033[0m

Apr 26 12:59:37 eowynpi signalk-server[310]: #033[0mPOST /skServer/plugins/sk-to-nmea2000/config #033[32m200#033[0m 5.874 ms - 45#033[0m

Apr 26 12:59:37 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[32m200#033[0m 16.787 ms - -#033[0m

Apr 26 12:59:54 eowynpi signalk-server[310]: #033[0mPOST /skServer/plugins/sk-to-nmea2000/config #033[32m200#033[0m 13.977 ms - 45#033[0m

Apr 26 12:59:54 eowynpi signalk-server[310]: #033[0mGET /skServer/plugins #033[36m304#033[0m 32.664 ms - -#033[0m

Apr 26 13:01:57 eowynpi signalk-server[310]: DLE followed by unexpected char , ignore message

Apr 26 13:03:27 eowynpi signalk-server[310]: DLE followed by unexpected char , ignore message

Apr 26 13:04:29 eowynpi systemd[1]: Starting Cleanup of Temporary Directories...

Apr 26 13:04:29 eowynpi systemd[1]: systemd-tmpfiles-clean.service: Succeeded.

Apr 26 13:04:29 eowynpi systemd[1]: Started Cleanup of Temporary Directories.

Apr 26 13:05:01 eowynpi CRON[1578]: (root) CMD (command -v debian-sa1 > /dev/null && debian-sa1 1 1)

Not sure what the “DLE followed by unexpected char ,” means?

During the brief time this configuration was running, the load reported on the Signal-K dashboard was much lower, and fluctuated wildly between 0 and about 30 Deltas/s

Next test: leave the sk-to-nmea2000 enabled, but no PGNs within it enabled.

Again, this has stayed up for at least an hour with no problems.

My conclusions thus far:

- it isn’t the power supply

- If I disable sk-to-nmea2000 my problem seems solved for now

- the problem could be that sk-to-nmea2000 is sending a badly formed PGN that is crashing the NGT-1. That would require bugs in both the plugin (to send a badly formed PGN) and the NGT-1 (to not be able to handle it). This seems unlikely, though possible.

- the problem could be that there is insufficient bandwidth to both receive and transmit the required data volumes - the baud rate problem that Brian posited. 

I can’t easily change the baud rate as I don’t’ have a Windows machine on board - Mac and the Pi. For now, I’m going to disable the sk-to-nmea2000 plugin and use WilhelmSK to monitor all the data.

I look forward to hearing results of Brian’s tests!

Regards

Martin Crick

(Sent from my iPad)

On 24 Apr 2021, at 17:52, Brian Scally <scal...@gmail.com> wrote:



You received this message because you are subscribed to a topic in the Google Groups "Signal K" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/signalk/y5tqjgHPfC4/unsubscribe.
To unsubscribe from this group and all its topics, send an email to signalk+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/signalk/CACboiC_DDsHmHhB0aUjCF0fRLnd01sR%2BcfDcmy9kiVniwL-56A%40mail.gmail.com.

[Brian Scally]

Martin and all

My experiments on speed...
Combined with Martin's experiment...

I can confirm that:

-Increasing the baud rate reduces the crash interval, but does not eliminate it.
-Increasing the level of bi directional traffic makes more crashes.  I reduces the speed and it becomes much more prevalent.

- I split my nmea busses and used 2 signal k servers to act as a bridge between the 2.  I get <5min run time on the NGW-1.

I am wondering if there was a change to the serial port and how it handles full duplex operation with the later builds.

I can confirm that the pressure and temperature (BME680) does not cause the system to crash while testing with the YDEN-2 even at full my full N2K bus load.

Looking - briefly - I can't work out a way of changing the virtual port to half duplex.
But I can see how to do it with a real serial port.

So I don't think this is a server or plug in issue.
This is feeling strongly like a virtual com interaction with NGW-1 issue.

I have an NGW-1 wired to an actual serial port on a Pi-4 to see if this causes an issue.
So far an hour in and no problems.
The NGW-1 is running the latest firmware.
--

Last note.  I have noted that the NGW-1 I am using for testing is not at the latest firmware version.

I strongly suggest we are on the same (latest) version.

I am on 2.580.  Which appears to have all the serial port fixes.

But I suggest we update to the latest version and test on 2.690

Here are the release notes :
- https://actisense.com/wp-content/uploads/2020/01/NGT-1-NGW-1-ActiPatch-v2.690-Release-notes.txt
They seem to have been working with the baud rate algorithm.

Brian

[Brian Scally]

REPOST with some corrections 

Martin and all

My experiments on speed...
Combined with Martin's experiment...

I can confirm that:

-Increasing the baud rate reduces the crash interval, but does not eliminate it.
-Increasing the level of bi directional traffic makes more crashes.  I reduces the speed and it becomes much more prevalent.

- I split my nmea busses and used 2 signal k servers to act as a bridge between the 2.  I get <5min run time on the NGW-1.

I am wondering if there was a change to the serial port and how it handles full duplex operation with the later builds.

I can confirm that the pressure and temperature (BME680) does not cause the system to crash while testing with the YDEN-2 even at  full N2K bus load.

Looking - briefly - I can't work out a way of changing the virtual port to half duplex.
But I can see how to do it with a real serial port.

So I don't think this is a server or plug in issue.

This is feeling strongly like a virtual com interaction with NGT-1 issue.

I have the related NGW-1 wired to an actual serial port on a Pi-4 to see if this causes an issue.

So far an hour in and no problems.
The NGW-1 is running the latest firmware.
--

Last note.  I have noted that the NGW-1 I am using for testing is not at the latest firmware version.

I strongly suggest we are on the same (latest?) version.

I am on 2.580.  Which appears to not have all the serial port fixes.

But I suggest we update to the latest version and test which seems to be 2.690

Here are the release notes :
- https://actisense.com/wp-content/uploads/2020/01/NGT-1-NGW-1-ActiPatch-v2.690-Release-notes.txt
They seem to have been working with the baud rate algorithm.

Brian

[Brian Scally]

The NGW-1 ran for 7h.

So the Actisence core is not the problem.

I went to upgrade my NGT-1 to the latest version (2.690) and found that the installer will not run on my windows computer, even though it is a labview program and I can build labview executables. And the run the 2.210 NGT-1 firmware installer.

GRRR

The 2.210 Actisence firmware demonstrates the  the 'fault'.
I have logged a problem with Actisence over the firmware installer and will share any pertinent information.

The problem still appears to be a full duplex message conflict.
Since USB 2.x is inherently half duplex my supposition is that 2 messages are in transit at the same time causing a buffer issue in the NGT-1

One of the non fruitful debug paths I didn't mention was to make sure that the NGT-1 was not running in USB1.0 mode as this would mean 8byte packets and NMEA 2k packets spread across multiple USB packets nd we would likely run out of bandwidth.
Several of the early Pi keyboards were USB1.

Brian

[Teppo Kurki]

NGT-1 connection in SK server has the option to Collect Network Statistics. This enables conversion of NGT-1 diagnostic messages to Signal K, to get a little more insight into its internals.