Low cost 10gbit ethernet testing

[kuon]

I am looking for a network tester for 10gbit RJ45 ethernet. I only need bandwidth measurement to ensure 10gbit will work. I also need a terminator, to measure wall sockets in different part of a building.

But I can only find cheap chinese testers which seems to simply test if there a connection or high end $5k testers.

Do you know any lower cost decent solution?

[TomWinTejas]

If this is a one time use case, just rent a cable certification tester.  If you (or your client) doesn't care about the cabling meeting the specifications and simply care about running at 10 gbps without errors, it all depends what confidence levels you are willing to accept.  Assuming you have a 10G switch in the IDF, you can simply run a ping flood at MTU from each port and look for interface errors at the NIC and switch port.  Better would be to get two 10G test sets and run them back to back, either running a smart loop on one end, or setting them up for a bidirectional RFC 2544 or BER test.  You might find used 10G test sets on eBay, you can look for brands like EXFO, Viavi, VeEX, Anritsu.  Another option would be to use a couple NIDs and run RFC 2544 or Y.1731 tests between them, but 10G NIDs are newer and likely more expensive... 1G would be fairly simple and inexpensive right now.  Last option would be a full DIY solution running a 10G NIC that supports DPDK and run a traffic generator like Cisco's TRex or supports XDP and run Ostinato.

All depends what you're actually looking to accomplish.  There's no such thing as a 10G terminator... if you mean loopback, for 10GbaseT, there's no passive loopback, you'd need a test set that actively loops traffic.  If you're testing fiber, it's rather simple so long as it's not BiDi, you just put in a hard loop so long as the power levels will be within spec.

[kuon]

By terminator I didn't mean it in the sense of terminating a line like Coax or CAN, but to be able to test a cable without having the two extremities at the same place. As you said, this would be a loopback.

I am rewiring my house and my parent's house, and I have the equipment for 1gbit but 10g is another story.

I guess as I have the NIC, going the DIY way would be simpler, but I have to find a way to make "portable" 10g computer, and unfortunately driving the NIC still require more than what low cost boards can drive.

[modoran]

Just buy 2 routers ( or only one ? ) which support 10 gbps link and return them afterwards in 14 days, works in EU very well.

[HighVoltage]

I bought a SignalTEK by Ideal (now TREND Networks) used on ebay for a great price.
It works like a charm.
If you are lucky, you can find a used one under Euro 1000

[kuon]

They look like nice devices, but the 10gbit version is not really available used, and is at about $3k-$4k.

[nctnico]

Maybe a cheap VNA could work but you won't be testing according to any standard. Just transmission losses and impedance mismatches.

[kuon]

I actually have a nanovna which I use for antenna design. But I wouldn't really know what would be the required values, also how to properly measure differential pairs.

But making a simple PCB for loopback and SMA adapter could be a fun little project, but I don't have enough ethernet knowledge to do it properly.

[nctnico]

For starters you'd need a balun to convert from single ended (50 Ohm) to differential (100 Ohm). These can be bought off-the-shelve. If you have two, you can put a 50 Ohm terminator at the other end or connect it to port 2 of the VNA. That way you can measure the loss (S12), impedance (S11) and VSWR (S11) but I have no idea what the fail limits would be.

[kuon]

So something like this ?

https://www.digikey.ch/en/products/detail/mini-circuits/TCM2-63WX/13927071

With primary connected to VNA, secondary to twisted pairs and on the other side the primary connected to 50 Ohm with ground?

[nctnico]

Yes. IIRC baluns can be bought off-the-shelve but I can't find them at Farnell right now.

[TomWinTejas]

I am rewiring my house and my parent's house, and I have the equipment for 1gbit but 10g is another story.

I guess as I have the NIC, going the DIY way would be simpler, but I have to find a way to make "portable" 10g computer, and unfortunately driving the NIC still require more than what low cost boards can drive.

If it's just a residential environment you can keep it simple.  You don't need to test full throughput, you just need to test for errors after link is established at 10G.  Testing at line rate simply means that to achieve the same confidence in a test you can run a BER for a shorter period of time to get the same amount of bits across the line.  They make Thunderbolt 3 10G NICs you can connect to a laptop and test back to a stationary server back in the wiring closet.  You can use iperf or even just a ping flood at MTU to get packets across the line and monitor for errors. 

I work on a backbone networking team for a large global network... while we will use expensive 100G test sets to validate BER on newly provisioned backbone circuits, when it comes to peering interconnectivity it's usually just simple ping tests and monitoring interface errors. 

[kuon]

I don't have a laptop with thunderbolt, but I have an HTPC with a free pci slot, I'll see if it can saturate a 10g NIC.

I like the VNA approach and it wouldn't cost much to make a simple PCB with SMA/balun/rj-45 port without magnet and a switch to select which pair to probe. But yeah, I don't know what value to look for and the actual throughput test will test the whole stack.

[cncjerry]

I wouldn't run 10gbitE over copper except maybe DAC patch cords.  Fiber is cheap as dirt with 30M LC-LC cables for less than $10 per in packs of 5.  And these were 3mm round cables that are very durable.  I put it in my house, have 5 devices with some bonded connected to a server using 10gigE.  Fiber and SFP+ as well as dual and quad channel cards are actually less than 2.5gigE.  I think the copper SFPs cost more. I use a Brocade / Rukus switch, only problem is power consumption on the switch is like 130w.   

[kuon]

I wouldn't run 10gbitE over copper except maybe DAC patch cords.  Fiber is cheap as dirt with 30M LC-LC cables for less than $10 per in packs of 5.  And these were 3mm round cables that are very durable.  I put it in my house, have 5 devices with some bonded connected to a server using 10gigE.  Fiber and SFP+ as well as dual and quad channel cards are actually less than 2.5gigE.  I think the copper SFPs cost more. I use a Brocade / Rukus switch, only problem is power consumption on the switch is like 130w.   

Doesn't it require expensive tools to terminate fibers onto patch panels? To be honest, I have not checked this for a long time.

[Kean]

Doesn't it require expensive tools to terminate fibers onto patch panels? To be honest, I have not checked this for a long time.

Yes, termination of the fibers requires the right tools and some experience, so instead you buy the fiber pre-terminated and then use matching couplers.

Something like this https://www.fs.com/au/products/35530.html

[nvmR]

If I were doing this all day, and I were more or less tech-savy, maybe I would brew up a custom setup such as this:
https://www.jeffgeerling.com/blog/2021/getting-faster-10-gbps-ethernet-on-raspberry-pi
(if you can get the parts).
An then print a custom box for it, a battery pack per piece, and ping away. It would be cheap, but takes some work.
For one-off solutions, renting something, or using a regular PC+nic would probably be more logical.

[nctnico]

Doesn't it require expensive tools to terminate fibers onto patch panels? To be honest, I have not checked this for a long time.

Yes, termination of the fibers requires the right tools and some experience, so instead you buy the fiber pre-terminated and then use matching couplers.

Something like this https://www.fs.com/au/products/35530.html

And be sure to get a fiber cleaning tool as well; you will need it.

[mikeselectricstuff]

Netgear 1g managed switches have cable test functions, maybe there is something similar in 10g?

[kuon]

Well, the thing is, I am a bit puzzled on what to do about copper or fiber. The thing is, 20 years ago, I was a contractor and I installed quite a bit of 100mb then 1gbit copper. I have a lot of experience with handling rj45 plugs and twisted pair cable. So I am confident with that, but I only have the real test (certification) equipment up to 1gbit (it was state of the art 20 years ago!) (I even installed some patch panel links with 1000CX with dsub cables).

Which is why I didn't really think of using fibers.

I am a bit "scared" of handling fibers, because bending radius in tubes, cleaving, splicing... Anyway, this is a bit off topic.

For copper 10gbit I will go the "iperf" route, I guess this is the most reasonable way of testing the whole "stack".

[TomWinTejas]

Pretetminated fiber is easy to deal with and pretty future proof if you go with single mode for residential purposes.  Bend insensitive fiber is cheap nowadays and you can get armored fiber if you're running thorough spaces like attics or crawlspaces where rodents could be a problem.  Duplex fiber with LC connectors is the most versatile for single runs and if you need to run >3 to a given location you can look at MPO with cassette breakout to LC on both ends. 

Fiber cleaners are nice to have, but honestly, for residential applications you are fine without them.  I've got a big supply of Cletop cassette cleaners, IBC single click, and wet cleaners for work on multi terabit DWDM line systems... but for 10G or 100G client optics it's rarely an issue.  Unless you are operating at the margins you will find that the nastiest fiber works just fine.  You do run the risk of pitting the fiber face over time, but that's usually with repeated insertions that you wouldn't see in a residential environment.

[kuon]

I must be missing something, but for pre-terminated fiber, you have to measure the length of every tube beforehand I guess?

I did some splicing a long time ago, but with equipment costing like $50k (what the guy installing the fiber in the town back then said, had no real idea, we had a chit chat and he let me splice a few chute), it's really easy with the right tools, but I have none

[justme1968]

just coil up the excess and hide it somewhere. much cheaper, easier and more reliable than terminating yourself.