Gigabit Ethernet Switch Issues with the T7

5 posts / 0 new

Wed, 02/19/2020 - 11:45

Chris Knudsen

We have a T7 with intermittent power-up connectivity loss.

We have tried mulitple T7 devices, with the same results.
The T7's as set with a static IP, on a local-to-the-system subnet.
The T7's are connected via Ethernet to a TrippLite NG16 unmanaged gigabit switch.
Powered via USB connector from a 5V system supply (a nice linear system supply).
The 5V supply, and switch are powered from the same AC source, and are physically in the same apparatus.
The AC source may be switched on/off with a main toggle switch.

There are many Ethernet devices on the connected switch (perhaps 10). The interfaces range from gigabit (control computer) to 10/100 (both commercial and in-house gadgets). None of the devices currently *need* to be able to achieve gigabit speeds.

*Sometimes* the LabJack cannot be found after a power-up. Connect LED's on the Ethernet jacks are illuminated (with some switches, see below).

So, we did some experimenting making a LabJack connections with different Ethernet switches after a power-up:

Allied Telesis AT-FS708 (10/100) Success
100% Success rate. This switch has a long working history with us.

TrippLite NG16 (Gigabit) Failure
Randomly fails initial connection for each power-up.

TrippLite NG24 (Gigabit) Success
100% Success. A dozen reboots or so. Many uneeded ports for our system.

Netgear GS724TP (Gigabit) Failure
A managed switch we had around. Failures for some of our slow devices, not so much with LabJack (inconclusive). However, it was possible to reliably get coms working by turning off auto-negotiating, and applying manual settings for individual ports.

Netgear JGS524 (Gigabit) Failure
100% Failure for LabJack! Worked with some other slow devices. Beware, Labjack users.

Netgear JFS524 (10/100) Success
100% Success rate. A dozen power cycles or so. Slower brother of the other Netgear JGS524 above.

HP 1420 (Gigabit) Success
100% Success rate. A dozen power cycles or so.

I'm wondering if LabJack or anyone in the user base has encountered issues with various switches, and perhaps has determined what to look for in order to avoid compatibility issues.

Suggestions?

Thanks, and best regards,

Chris

*Edits: formatted as a table, and added 2 more switch models, unformatted as a table because results were not as expected.

Wed, 02/19/2020 - 12:28

LabJack Support

I have not heard of any

I have not heard of any issues with specific switches. We use a TP-Link TL-SG1024 in our development and testing offices, so LabJack devices interact with a gigabit switch during most of our tests.

I am afraid I have no idea why the JGS524 would be so problematic.

The T7 has a watchdog feature that can be used to reset the device when it does not receive communication after a specificied amount of time. That will get the T7 to repeatedly try to link and communicate.

Another thought: Does the switch get powered up with the rest of the system? There may be some sort of startup issue which requires the switch to be powered / booted up before other devices try to connect.

Sorry I don't have a more direct answer. Please let me know if you figure anything out.

Thu, 02/20/2020 - 05:40

Chris Knudsen

The fine folks at LabJack

The fine folks at LabJack said: Does the switch get powered up with the rest of the system? There may be some sort of startup issue which requires the switch to be powered / booted up before other devices try to connect.

This is my concern. Maybe a power-up race condition. Perhaps glitchy, transient power during turn-on.

We have another round of switches coming in to try, and we will continue to exercise the (somewhat overkill) TrippLite NG24.

There are internet anecdotes about switches (even specfically the Netgear GS724TP) having issues with auto-negotiation and 10/100 devices. No data though.

https://stackoverflow.blog/2010/01/25/six-whys-or-never-trust-your-network-switch/

Yes, 2010.

Tue, 02/25/2020 - 08:37

Chris Knudsen

An update:

I've duplicated the above results on the bench with the various switches. I'm now using a new, duplicate HP1420 for torure testing, as it's our leading candidate going forward.

I'm programmically switching AC power, and waiting a random time from 10 to 60 seconds before attempting TCP connections to labjack and one of our custom gadgets, and then logging success/failure.

The HP1420 has survived 2000 power cycles so far, with 100% success rate.

I'm not really sure what to make of all of this. Observing some of the Gig switches during power-up, I've noticed that sometimes a link indication is present, and then go away (on the switch's RJ45 bulkhead connector). Sometimes it comes back, sometimes not. It feels like some Gig switches do not wait long enough for some sort of auto-negotiation response, and then assume no device is present.

I've also observed situations where a Gig switch was working fine, until I changed the order of the connections on the switch (while using the same switch ports). Good luck explaining that...

All 10/100 switches have worked reliably.

YMMV.

Tue, 02/25/2020 - 13:10

LabJack Support

It sounds like we have two

It sounds like we have two reasonable solutions:

* Start the switch up a minute before all the other hardware. Requires two power supplies and some sort of timing system.

* Carefully select the switch to use. A little more work when selecting components.