Skip to content

Instantly share code, notes, and snippets.

@darconeous
Last active November 8, 2024 16:20
Show Gist options
  • Save darconeous/8c7899c4d2f849b881d6c43be55066ee to your computer and use it in GitHub Desktop.
Save darconeous/8c7899c4d2f849b881d6c43be55066ee to your computer and use it in GitHub Desktop.
Hacking the Rectangular Starlink Dishy Cable
@jbowler
Copy link

jbowler commented Feb 10, 2023

my Peplink router won't recognize that an ethernet cable is connected. It sometimes shows "Connecting..." and sometimes "No Cable Detected". (see attached screenshots)

The screen shots aren't there, so far as I can see. This is a cable continuity problem; "no cable connected" means one or other pair is intermittent or open circuit or maybe the channel pairs are swapped.

In this case test end-to-end, or rather end-to-middle. I.e. use an ethernet cable connector from the Neutrik to the RJ45 that goes into the McCown and make sure it's showing the right connection. In this case that is 1-2, 7-8 straight through and 3-6/4-5 swapped; the tester needs to be one of the ones that shows the actual connections, not just pass/fail. Then check from the McCown data in/out RJ45 back to the RJ45 that goes to the router; the connections should be identical to those on the other side. A simply way of doing this in one shot is to take the RJ45's that go into the McCown and plug them into an ethernet pass-through connector. Then the two ends (Neutrik-router) should show straight through. The shield should be continuous in all cases.

Most likely this will be fine; if it isn't fix the problem by testing segments of the link. My original design was like yours; I swapped blue/green pairs before and after the PoE. I came to the conclusion that I don't like this because the wiring of the PoE connections to the antenna is just plain deadly for any normally wired passive PD so I am moving to swap blue/green as soon as possible out of the antenna (that would be the cable to the Neutrik in your case) then back again immediately after the data in/out port of the injector. That's just so much safer.

You can also check the McCown jumper setup; for some reason mine was shipped with the StarLink power arrangement so it had to be changed for "normal". If you plug the McCown in with the PSU disconnected then use a tester on RJ45s at each end (e.g. plug the end RJ45 into the RJ45 port on the NF-488 and do a continuity test) you should see each pair shorted. If you pull all four jumpers off the McCown that should be the only shorts. With the jumpers in the correct positions IRC 1236 should be shorted together along with 4578 (so both pairs are connected in each channel). Once again don't quote me on this - check very carefully!

If you have a good straight-through connection all the way then you need to suspect that the SPX connector into the antenna is fried. This is what happened to my original 75ft cable; I've split and tested the thing all the way back to within 10cm of the connector and orange is shorted (30ohm) to ground, apparently somewhere inside the connector at the cable end. My test kit consists of a broken open StarLink ethernet dongle with the cable cut and an RJ45 plug crimped onto the thick wires. I can plug the SPX connector from the antenna end into this then take the RJ45s on the two ends and do a continuity/cable test. Once again the connection should be straight through in your setup; in mine it is too but that is because I have both RJ45s with blue pair/green pair swaps.

I had cabling problems all the time I was doing this; my error mostly. I found that I can verify the cable setup correctness by plugging the RJ45 that goes into the router into a switch. If everything is wired correctly the switch detects a 1000MBit connection to the antenna about 5s after the power is connected. In fact the router can then be plugged into the switch and it should all work; the switch gives you access to 192.168.0.0/16 so it's possible to see what is out there...

You can also check the resistance of each pair while connected to the antenna with a suitable breakout RJ45 and testing with the PSE disconnected. This helps if there is some poor connection in there. I can't find the core resistances of the Wurtz electromagnetics but you should see each pair with a resistance just a few ohms. When I measured I was seeing maybe three ohms across correctly connected pairs. You should also see a low resistance between the two pairs corresponding to + and the two corresponding to - and around 30kohm between + and -

  • I purchased the Starlink ethernet adapter, and inserted it into the stock setup. So I'm using the SL router for power. Peplink is still showed "Connecting...", forever.

That sounds exactly like my original problem! My 75ft cable antenna connector was fried and the router was fried. With a new cable, a new dongle and the original router I had an internet connection but no ethernet on the dongle. Somehow the short in the SPX connector had killed something in the router that makes the dongle work. Once again plug the RJ45 from the dongle into a switch and see if the lights come on. It took me about 2 weeks to persuade StarLink customer support to send me a new router. At the end I'd already bought a new cable and they didn't respond to my request to refund that cost.

The SPX connector is pretty much garbage. In my case I plugged it in in summer and it failed in winter. I can only assume that the very small amount of water vapour in the sealed chamber that is created when the plug is inserted condensed and shorted enough out to fry the plug. The jack in the mast can apparently be pried out; I'm very tempted to pull it out, cut it off and solder cable directly to it with a swapped RJ45 at the end.

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@jbowler
Copy link

jbowler commented Feb 10, 2023

the hAP is operating in Bridge mode, between WAN and WiFi. the WAN port status is "Link OK, running, not slave", but no significant traffic is flowing over this port.

Connect to the WiFi then connect to http://192.168.100.1 (HTTP, not HTTPS). This will give you the antenna web page. Go to Settings/Advanced/Debug Data, scroll down to "DISH" then scroll down even further until you find "EthSpeedMbps"; it should be 1000, but if there is a cable wiring problem it still works but at 100Mbps or maybe lower. I had orange-white disconnected on an RJ45 jack and I got 100... The debug data can be copied by clicking on the weird pages icon at the top right; this makes it easier to read.

You may need to add a static route to 192.168.100.x on your router to get to the page, though it will probably work just changing your machine IP to 192.168.100.42

EDIT: for that matter, given that there are four LAN ports, just plug both the StarLink and the Peplink router into two of the LAN ports (in bridge mode they should all be LAN ports, but to be safe use the ones that are marked as such.) The MikroTik:

https://mikrotik.com/product/RB962UiGS-5HacT2HnT

is just a switch in that mode and if the switch can get packets in it should have no problem sending them out to the router.

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@jbowler
Copy link

jbowler commented Feb 10, 2023

I measure any of (1,2,3,6) to any of (4,5,7,8) as 3.4 to 3.6 ohms, on cable connected only to dishy.

That's correct if the cable is not swapped, i.e. you are checking the surge-suppressor connection, not the jack to the McCown (since you put the swap between the two). On regular PoE wiring groups (1245) and (3678) are connected together (via the centre taps of the electromagnetics, and ideally some diodes). In either case the resistance between conductors from each of the two groups should be the PoE signature resistance, I measured it at around 30k.

I no longer have any switches :(

A bridge is a switch and the two LAN ports on the Peplink are too. I'm suspicious of ethernet channel 1 (1236, unswapped) because the Peplink ports are apparently only 10/100Mbps, so they only use channel 1. Hence my suggestion to check the debug data and to try dish->MikroTik->Peplink

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@jbowler
Copy link

jbowler commented Feb 10, 2023

I very briefly got access from PC<>SL ethernet adapter<>Dishy with 2nd SL router and original cable. 40Mbps Up/.04 Mbps down. I could see two devices (PC and cell phone) attached via my cellphone Starlink app. Then the app showed 0 devices attached and all data stopped! Now the app shows my cell phone attached, but not the PC.

There are three ways of getting information from the dish; the router if it is connected to the dish, 192.168.100.1 (dishy.starlink.com) which may not work if the StarLink router is connected (so far as I can determine) and, most useful in this case, the StarLink app on your cell connected via the internet; not via the StarLink router.

The latter connects to the antenna via the satellite. No working local network and no StarLink router required. It should display "ONLINE" and the debug data ("THIS DEVICE") should show "Non-StarLink IP address", though if your are connected to the internet via the dish it will show "Starlink IP address". At this point you can "COPY DEBUG DATA" and paste it into something useful, like an email. This is the only way other than 192.168.100.1 of getting the full debug data; the StartLink App displays a sanitised version and it has to be copied out of the app.

Even so the only information in the debug data about the local connection is the speed of the connection.

I can see how if Dishy won't provide any IP to my PC, it won't with a DHCP-configured WAN port on a router, either.

The dish does provide an IPv4 to the router; it doesn't need to be statically configured. I believe it is using DHCP but it's supplying a CGNAT address; the address it provides is the one on the internet side of the gateway. It may well be that there is only one of those per antenna and that it just gets bridged across the antenna, indeed that seems most likely.

@jbowler
Copy link

jbowler commented Feb 10, 2023

I used a GL.iNET Slate (AR750S) in Access Point network mode (basically a bridge mode).

The AR750S also has 10/100/1000 ports, like both Mikrotiks.

Dishy<SL original cable>SL ethernet adapter<>SL Router #2 
                              ^--<eth cable>GL.iNET router<eth cable>Peplink router

The images you posted aren't coming through but I assume this didn't work. It's triple NAT; SL Router #2 isn't in bridge mode so SL Router #2 is offering DHCP in the 192.168.1.0/24 network and it offers it on the ethernet adapter too. The AP is irrelevant; the Peplink gets some random 192.168.1.x IPv4 on its WAN port (this should show up in the Peplink management pages) and tries to add a third layer of NAT to that.

There's a point of failure with the Peplink but it's a failure in a far simpler system where the Peplink is connected directly to the dish. If you want the StarLink router in the picture for power and, more important, customer support put SL#2 into bypass mode and use the above system including the GL.iNET router, at least to start with, to make sure you get 1000Mbps connection lights on the GL.iNet. If you do you should get a Peplink connection that works or at least fails without complaining about broken cables.

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@torrmundi
Copy link

torrmundi commented Feb 10, 2023 via email

@jbowler
Copy link

jbowler commented Feb 10, 2023

ethSpeedMbps: 100, in three different status snapshots.

At least one of the eight conductors is broken or shorted, or, worse, the magnetics inside the antenna are damaged. The fault is intermittent. That's better; it suggests a fried connector which from what you have said would have to be the dish jack or the cable dish-end plug. I suggest connecting Dish->latest official StarLink cable->StarLink Router#2, copying the debug data and editing out everything except the ethernet speed, which will apparently be 100 but should be 1000, then opening a customer support ticket, or reopening the one that got you SL#2 and demanding a fix. If they haven't sent you a new cable yet they will, eventually.

The traditional cable-guy approach is to walk round the connectors and wiggle the wires going into and coming out of them. Sooner or later the problem becomes non-intermittent. If you can repro the 100 speed with the stock kit then my experience with the stock StarLink kit is that a connector is fried. I saw several (3 or 4) stress points in the 75ft cable and I stripped out each of those until I was left with the connector and 10cm of cable. It was also a double failure; when the connector on the end of the cable fried replacing the cable wasn't enough, the router had to go too.

If you don't want to deal with Customer Support and assuming the 100 problem exists in the stock kit you could try cleaning out the connectors. I did this but have so far been unable to remove the (30 ohm) short in the plug hence my previous comments about removing the connectors from the antenna end.

@bghira
Copy link

bghira commented Feb 15, 2023

well, i followed the instructions here with swapped connector from the dishy cable to the Tycon, then the swapped connector to a normal t568b to the router, and nothing happens when i plug it all in. i've got a power switch on a 24v->48v boost transformer which can push 250w. that unit does actually light up, but nothing happens to the dish when i plug it in. no ethernet light goes on on the router port. any ideas?

@morehardware
Copy link

I think I figured it out. These connectivity problems had to do with power and POE. In typical POE applications the switch injects DC 48V that after 100 meters transmitting, drops the voltage down to DC 39V which can still satisfied the requirement. The Starlink POE need way more juice than typical POE devices. I got my configurable high power 12v powersupply and learned some things after finally adjusting it to a point where I got a consistent successful Starlink ip connection with speed. The power supply has three adjustments, Input power under voltage protection, current and output voltage. When you attach a 12 volt power supply, there is a yellow LED which indicated that the unit incoming voltage is set too high (default 24 volts). The under current (UV) pot is turned counterclockwise till the yellow light goes out. Next I adjusted the output voltage to 48 volts. I left the current as is since I dont have a good current meter. This initial setup failed like all my other setups. the computer would see the ethernet cable, try to handshake, then disconnect in a cyclic manner. Next I turned up the voltage to 56 volts and the same thing happened but in a longer interval. I watched the powers supply while it was cycling and noticed that the low voltage protection light would flash ever so slightly followed by a loss of connection to DISHY. Basically a warm reset. I think what is happening is that the power demand when the DISHY boots drops the input voltage to a point that Dishy reboots. I then adjusted the current higher but it kept cycling at the same interval. Then I adjusted the low voltage protection lower to the point where I no longer got an led flash when DISHY booted and that has led to a stable 12 volt setup.

This power supply worked when the Cheapo Chineese Bricks, and the Mean Well did not. It has no Documentation in the box but if you read the QUESTIONS AND REVIEWS you can ascertain how to adjust it. Some comments also claim success with the Starlink as well.

DC-DC Converter Module Boost DC Step Up Voltage Regulator CV Stabilizer Power Supply Module 10-60V to 12-97V 1500W 30A
https://www.amazon.com/gp/product/B07NM52VV5/ref=ppx_yo_dt_b_asin_title_o00_s03?ie=UTF8&psc=1

@WIMMPYIII
Copy link

well, i followed the instructions here with swapped connector from the dishy cable to the Tycon, then the swapped connector to a normal t568b to the router, and nothing happens when i plug it all in. i've got a power switch on a 24v->48v boost transformer which can push 250w. that unit does actually light up, but nothing happens to the dish when i plug it in. no ethernet light goes on on the router port. any ideas?

Try with dishypowa or "800 gige poe apc"

@morehardware
Copy link

morehardware commented Feb 15, 2023 via email

@WIMMPYIII
Copy link

Uploading IMG_20230207_184808.jpg…

@WIMMPYIII
Copy link

Uploading PXL_20221221_045705930.jpg…

@WIMMPYIII
Copy link

Uploading Screenshot_20221107-143754.png…
This is the one that comes it a enclosure. Slightly different layout but pins are the same.

@morehardware
Copy link

morehardware commented Feb 15, 2023 via email

@bghira
Copy link

bghira commented Feb 15, 2023

i'm using two 400Ah 12v LiFePO4 batteries in 24v series connection, their overall storage is more than 9kWh and i'm capable of pulling more than 400 watts from them without substantial V-drop. the 48v step-up is capable of putting out more than 200 watts, as determined with a load tester. i actually under-sized the fuse at first, and it blew right through the 5A (135w). it doesn't burn up a 7.5A fuse, which is 202 watts.

the one aspect of your suggestion that i feel has merit is the precision required for terminating the connectors. it's entirely possible that my RJ45 crimping tool is inadequate for shielded through-connector termination. i'll have a go at making a "normal" patch cable out of the many dozen feet i've got remaining of the starlink cable, and see if it works that way. if it doesn't, that'll help me determine where to go from there.

this crimping tool does alright with some of the more "supple" CAT5e cable i've terminated, but its blades also seem to be wearing out.

@morehardware
Copy link

morehardware commented Feb 15, 2023 via email

@bghira
Copy link

bghira commented Feb 15, 2023

i plugged the Swapped -> T568B cable into something and the port didn't light up :( i wonder if my crimping tool isn't pushing hard enough

@jbowler
Copy link

jbowler commented Feb 15, 2023

i actually under-sized the fuse at first, and it blew right through the 5A (135w). it doesn't burn up a 7.5A fuse

I think putting a fuse in the PoE supply (i.e. between the PSU and the PoE) is an extremely good idea. I just put a 2A fuse into my system, so I have a 2A APM (the small automobile size) blade fuse between the PSU:

https://www.amazon.com/gp/product/B08889XW1F rated 48V, 3A

and the Tycon injector:

https://www.tyconsystems.com/poe-inj-1000-wt rated 2.25A at up to 80V.

I haven't had any problems so far, but it hasn't been on long yet. Certainly no problems during the boot. I've also turned on "preheat" as a test, I think that maybe adds 250mA to the load.

The fuses I'm using are underrated - they are limited to 32V and this is a 48V system. Technically I should be using the classic cylindrical glass tube, such as a 20mm, but at my own risk I chose to use the blade fuses because I have a whole load of them for various amperages. Full specifications of APM fuses are also readily available:

https://www.eaton.com/content/dam/eaton/products/electrical-circuit-protection/fuses/bussmann-series-supplemental-fuses/automotive-blade-type/bus-ele-ds-2048-atm-series.pdf

The important thing here is that I put a 2A fuse into a circuit that I know is limited to 2A continuous and which, by experiment (with preheat/snowmelt off) I also know doesn't take even 1.5A for more than 1 second continuous. You can see from the Eaton specification that with this fuse (well, at least if it was made by Eaton...) for the 2A fuse to fuse in 1 second the average current would have to be about 3.4A and I know that the 3.4A transient draw occurs for a lot less than 1s; more than 250uS but certainly less than 1s.

@WIMMPYIII
Copy link

i actually under-sized the fuse at first, and it blew right through the 5A (135w). it doesn't burn up a 7.5A fuse

I think putting a fuse in the PoE supply (i.e. between the PSU and the PoE) is an extremely good idea. I just put a 2A fuse into my system, so I have a 2A APM (the small automobile size) blade fuse between the PSU:

https://www.amazon.com/gp/product/B08889XW1F rated 48V, 3A

and the Tycon injector:

https://www.tyconsystems.com/poe-inj-1000-wt rated 2.25A at up to 80V.

I haven't had any problems so far, but it hasn't been on long yet. Certainly no problems during the boot. I've also turned on "preheat" as a test, I think that maybe adds 250mA to the load.

The fuses I'm using are underrated - they are limited to 32V and this is a 48V system. Technically I should be using the classic cylindrical glass tube, such as a 20mm, but at my own risk I chose to use the blade fuses because I have a whole load of them for various amperages. Full specifications of APM fuses are also readily available:

https://www.eaton.com/content/dam/eaton/products/electrical-circuit-protection/fuses/bussmann-series-supplemental-fuses/automotive-blade-type/bus-ele-ds-2048-atm-series.pdf

The important thing here is that I put a 2A fuse into a circuit that I know is limited to 2A continuous and which, by experiment (with preheat/snowmelt off) I also know doesn't take even 1.5A for more than 1 second continuous. You can see from the Eaton specification that with this fuse (well, at least if it was made by Eaton...) for the 2A fuse to fuse in 1 second the average current would have to be about 3.4A and I know that the 3.4A transient draw occurs for a lot less than 1s; more than 250uS but certainly less than 1s.

It's pretty impressive that it can cook 135w fuse.

@jbowler
Copy link

jbowler commented Feb 15, 2023

@WIMMPYIII : It's pretty impressive that it can cook 135w fuse.

What can cook a "135W" fuse and what is a "135W" fuse; fuses have amperages(A) and fuse times(s).

@bghira
Copy link

bghira commented Feb 16, 2023

i guess the voltage across pins 1 and 8 should be there even when nothing's plugged in, right? if so, i think this Tycon is DOA because it is not putting out any Voltage.

@WIMMPYIII
Copy link

i guess the voltage across pins 1 and 8 should be there even when nothing's plugged in, right? if so, i think this Tycon is DOA because it is not putting out any Voltage.

Or something got shorted.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment