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
@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 24, 2023 via email

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 24, 2023 via email

@jbowler
Copy link

jbowler commented Jul 24, 2023

I also found a Kill-A-Watt meter, and in summertime (right now), it's drawing 122-123V and 0.35-0.4A with Snow Melt off or on automatic. In pre-heat, it draws 0.9A.

It sounds like you are measuring the current into the router. The power consumption of the router itself should be fairly constant. There will be an extra power draw from the dish, but you can work out the raw power required by the router simply by powering it on with the dish not connected (do not connect or disconnect the dish to a powered on router; this causes a very significant power surge to the dish which is the most likely candidate for frying one or more of the four connectors). See my earlier comment:

https://gist.github.com/darconeous/8c7899c4d2f849b881d6c43be55066ee?permalink_comment_id=4479847#gistcomment-4479847

I got my figures using the Tycon power injector described in this gist (not the router) so I was able to measure the current being sent down the cable. I also put a 2A fuse inline with the power delivery so that limits the short-term average current. The fuse never blew even though there is good reason to believe that there is a startup current/power surge which might be as much as 7A for a (small) fraction of a second (hence my comment about about not connecting a "hot" router to the dish). My detailed measurements start here:

https://gist.github.com/darconeous/8c7899c4d2f849b881d6c43be55066ee?permalink_comment_id=4463407#gistcomment-4463407

If you look at one of the posts that follow:

https://gist.github.com/darconeous/8c7899c4d2f849b881d6c43be55066ee?permalink_comment_id=4466245#gistcomment-4466245

You will see the graph of what actually happens during boot and after. This is using a Fluke 189 logging multimeter, pretty much the industry standard when it was bought for me it but I admit I've owned it for coming on 30 years and never got it recalibrated. The most telling figures might be the ones in the first comment: the dish apparently does consume close to 90W for short periods of time. 90W is the capacity of the router, some injectors can handle more but certainly not all of them.

I believe I left "pre-heat" on after my first post and you can see that the actual dish consumption hits a max of 1.6A but averages somewhere in the range 0.6 to 0.9A. This corresponds to a maximum power of 77W and an average over 5s in the range 29W-43W, consistent with the stated router capabilities.

It's almost impossible to guesstimate the actual power going to the router without a calibration of the router efficiency; the router takes 110-250V input and converts it to its own internal power requirements (5V or 3.3V, probably 5) and the dish requirements (48V). Efficiencies will be in the range 80-90% but the design might be dumb.

Nevertheless whatever excess current is delivered to the router when the dish is connected it is certainly an underestimate of the current going to the dish, because that current is at ~48V.

All the evidence so far is that the design of the dish produces significant surge currents under some circumstances. In particular there is evidence that there might be a very high initial surge if a charged, hot, PoE injector is connected to the dish. This is what I was doing with my Tycon setup and the tester I was using (which I don't trust) reported a 7A surge.

The surge should be no big problem for the components involved because they are mostly passive, maybe all; it's not clear how many diodes StarLink put into the circuit but it sounds like "none", that's one way of getting more power. My own tests prove that in my system the thing that fried was the connection (the connector) at the dish. It fried in a way that is consistent with an instantaneous surge. What caused that surge is impossible to say.

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 25, 2023 via email

@jbowler
Copy link

jbowler commented Jul 25, 2023

This is consistent what what I've been reading (maybe here?) that the dish use 0.30A with Snow Melt completely off.

That's consistent with my numbers; it would correspond to about 0.6A going to the dish. A rule of thumb assuming 80-90% efficiency in the router is that the current flowing to the dish MIGHT be about twice the extra current to the router. Maybe "pre-heat" increases it slightly but I suspect that will be thermostatically controlled (this doesn't cost enough money for StarLink to cost-reduce it out of the dish). I measure in February, before the current heat but it was still well above 0C.

Our dish is always connected, so what does that mean? It's not as if I climb the ladder and monkey around with the connector.

Don't start up, power cycle, the router without the dish cable connected. I.e. don't plug the dish cable into the router when the router is connected to the power supply; this causes a surge because the router socket is already powered up and the dish is waiting for juice. It says this in the instructions; if you follow the instructions step by step the router is connected to the power after the router is connected to the dish. Of course you have to look at this like a lawyer or a computer programmer to see that as a clear instruction.

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 25, 2023 via email

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 25, 2023 via email

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 25, 2023 via email

@jbowler
Copy link

jbowler commented Jul 26, 2023

I did my best to read through the whole thread. It's frustrating because I have only a limited background, and among other things the acronyms have thrown me at times.

It's a long set of comments and there are more questions than answers (like why don't systems running of 12V batteries work as well as 110V systems?)

My answers are cryptic because I do assume a certain knowledge set but for your purposes there a really only two important things to understand. The first is the electrical power equation:

P = V × I

That matters because the dish requires a certain amount of power with any given configuration and the router (or PoE) has to delivery that power; voltage and current don't matter, only the product counts. The second equation is Ohms law:

V = I × R

This matters because the cable itself takes power; the power to drive the current (I) down the line, and that creates a voltage drop across the cable (V).

Maybe the important thing here is that what I was saying is that my numbers (as reported in the various links) are completely consistent with yours, except, maybe, for the pre-heat behavior. The difference is that you measured current at the router power inlet. That inlet takes a voltage between 110V and 250V whereas all my measurements are the current on the cable which has a power supply of 48V (I used a PoE but it is a 48V PoE like the router IRC.)

So the router has to convert 122V (AC) with 0.3A for the dish to 48V with whatever power the dish requires. The conversion isn't 100% efficient but it will be in the range 80%-95%; assume 90% (arbitrary). This would mean that the router/cable assembly are being supplied with 33W at 48V (DC) which is a current of 0.69A.

That's pretty much identical to what I measured; a 5s average of 0.6 to 0.8A, looks like 0.7A to me!

The pre-heat figure is, however, a bit of a problem; 0.9A less 0.07A for the router means 100W, at 90% efficiency that means the cable assembly is getting 90W and this is the quoted (faceplate) limit of the router. It's still well within the capabilities of the cable but it really is flat out for the system design. It's close to 2A up and down the cable (so the cable is carrying 4A total, still fine so far as I can see). Maybe that's the way StarLink designed it and maybe I didn't manage to switch preheat on properly (or maybe StarLink had it disabled in SW Oregon?)

Another question would be whether the cable problem might be not the cable wiring but the connectors. Note that I'm referring to a standard 50-foot cable connected to a Gen 2 dish.

I think it is the connectors but I don't see how anything the router delivers continuously is going to cause a problem.

Then you don't want me to connect the dish cable into the router when it's connected to the power supply, which I take to mean plugged into the wall.

Yes, because of the way the dish power supply inside the router seems to work. I believe it's a fairly basic passive PoE design (I haven't located a circuit diagram for the router so I can't be sure.) It is a passive design (or the hacks on this gist would not work) and I believe it directly converts the AC power at 60Hz/110V or 50Hz/220V to the 48V. This means that it has to retain sufficient energy between the two points where the AC hits 48V; it has to keep on delivering 48V even though the input voltage is less than 48V.

When you connect the dish to the router with the router already turned on my hypothesis is that this energy floods down the cable to charge the various power supplies inside the dish. This is an "inrush" current and it can be very large. Poorly designed (IMO) 19.2V laptop power supplies had a nasty effect of causing an arc - lightning, accompanied by a bang, thunder - when plugged into a power outlet, this happened for the same reason.

So my further hypothesis is that it's simply a bad idea to plug the dish into a powered on router because that inrush can cause an arc in the connector and, given the connector, easily damage it.

Guess what I did when I was first playing around with my new dish? I had it outside on the lawn and, using the 'phone app, I was testing locations. Because I didn't want to have to go back into the house to plug the router into the outlet I disconnected and reconnected the dish cable connection at the dish. As I reported above the connector between the cable and the dish is now fried; both sides. StarLink sent me a new cable but not a new dish... There's still a significant extra resistance in the connector within the dish mast. I've given up on StarLink but if I hadn't I would have got round to pulling out the dish connector (it's not impossible) and soldering my own CAT5e cable to the wires.

That's just my story of course, but I don't believe any of the reports which imply that the router can deliver way, way, more than 90W. (Of course I have a couple of routers now that I can destruction test, maybe I will one day...)

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 26, 2023 via email

@jbowler
Copy link

jbowler commented Jul 26, 2023

Is there enough resistance in four 24AWG 50-foot wire pairs to care about in the context we're discussing?

No.

The standard StarLink cable is 75ft, there are 8 24AWG wires in there, but they are twisted together so they are very slightly longer than 75ft. Not enough to matter I believe. They are also stranded in the StarLink cable; they are each 7 (or 11, I should count them :-) much thinner (31AWG?) wires but the "24AWG" should take all that into account. I normally find "engineeringtoolbox" to be a good resource:

https://www.engineeringtoolbox.com/awg-wire-gauge-d_731.html

So each wire is around 0.2mm² in diameter and 8 of them are 1.6mm² This is a "cable assembly" so the whole thing has to be considered as a unit. It is equivalent to a 15AWG wire (a single wire). See this table:

https://www.powerstream.com/Wire_Size.htm

The "maximum amps for power distribution" column is based on power loss; the resistance causing excessive loss of power. The value is 4.7A, so the CAT5E is approaching reasonable limits for power loss (and this is, of course, the point of this gist). The "maximum amps for chassis wiring" is the figure for temperature rise. The figure is 28A; this is the point at which the wire (cable in this case) starts to overheat. 28A is in'n'out; the router would be supplying 14A. That's 672W; well within the capability of a 15A US 110V power supply (1650W) but well beyond what the white tombstone can deliver.

For voltage drop it makes more sense to go back to the resistance of 24AWG wires because the fact that they are grouped together doesn't matter and it's just too easy to make mistakes through unnecessary complication. The standard StarLink cable is 75ft long, the longest they actually sell is 150ft long. The powerstream site gives 25.67Ω per 1000ft for 24AWG; that's actually solid and I don't know which number gets priority to determine AWG, resistance would make sense but I don't know. Using that number there are four conductors in and four out for a resistance of 6.4Ω per foot over twice the distance (there and back) times 150ft, 1.93Ω, pretty much 2Ω. I believe my measurements of the resistance of an actual cable were less than that so that seems a possible upper limit.

So that's a 4V loss for 150ft at the max of the StarLink router (2A). Hence the idea to go to a 52V supply for longer, e.g. 300ft; the standard maximum of the ethernet signal. 300ft at 24AWG (the other recommendation is to go to 23AWG CAT6) drops 8V which, IRC, is the lower limit of the PoE standards and StarLink isn't standard in this regard. For the standard cable the drop is 2V; irrelevant.

These figures are consistent with a reasonable assumption that the StarLink engineers checked what they were doing. I can criticize what they did in other ways, but not this one.

@crdiaz324
Copy link

Have you guys seen this? Here is a kit that plugs right in, avoiding the need to cut or invert your wires. https://techcharmer.com/products/custom-poe-injector

@torrmundi
Copy link

torrmundi commented Jul 30, 2023 via email

@jbowler
Copy link

jbowler commented Jul 30, 2023

The YAOSHENG adapter avoids the need to cut the cable; https://www.amazon.com/YAOSHENG-Rectangular-Adapter-Connect-Injector The YAOSHENG PoE injector is rated at 3A, the TechCharmer PoE looks like a Tycon custom build in which case it would be 2.25A. The StarLink ethernet adapter ($25) can be hacked to make an adapter for the end of the cable as well. Cutting the hole removes the need for the dubious SpaceX connectors. I suspect the white plug is a standard PCB connector like one of the JST connectors.

@JustOneGuyHere
Copy link

JustOneGuyHere commented Jul 30, 2023

The standard StarLink cable is 75ft, there are 8 24AWG wires in there, but they are twisted together so they are very slightly longer than 75ft.

Sorry, the standard Starlink cable is 50 feet. It's in their specs, and I confirmed it by measuring the old one that I pulled out.

https://www.starlink.com/specifications

Starlink doesn't say whether it's Cat 5e or Cat 6a. I've seen it described both ways, but as I write I'm inclined to say Cat 5e. That would be 24 AWG. Ampacity for 24 AWG is 3.5.

https://www.engineeringtoolbox.com/wire-gauges-d_419.html

Now for the fun stuff. When my router is plugged in without the cable plugged in, it registers 0.07A at the indoor outlet. When the cable is plugged in, it registers 0.37A. Both numbers vary a bit but not much; the dish is drawing about 0.3A minus whatever resistance is in 50 feet of 24AWG and two connectors. That 0.3A is measured at the outlet, but there's a step-down transformer in the router that converts 120 volts to 48 volts. This turns 0.3A into 0.75A. Doesn't change the wattage: still 36 watts

My next door neighbor, also a Starlink user whose cable crapped out, measured 5A at the wall outlet last winter with Snow Melt running. Subtracting the 0.07A that the router uses, you get 4.93A delivered to the 48 volt power supply. Voltage stepped down, amperage stepped up to 12.5 over wiring rated for 3.5A . Either way, about 600 watts. (Okay, 591.6 minus a smidgen of power loss in the transformer.) It's becoming clear to me just why so many Starlink cables are failing. Would you use Cat 5e (or Cat 6a) to run power to a space heater?

Now that Starlink is fully commercial, it's becoming popular in rural areas. In places like mine, where it's cold enough to trigger the Snow Melt thermostat off and on for maybe half the year if left on the default "automatic" setting, I think this wiring is getting damaged a little bit each time until it finally gives out. I also note that the Starlink router label shows that it's a 2A device, yet my neighbor observed that it was drawing 5A when Snow Melt was running. I wonder if Snow Melt is frying both the routers and the cables.

My big remaining question is: How long does Snow Melt go on when it goes on? My guess is "not for long enough to fry the cables right away." This is why, when I click around and see reports of bad cables, the story is almost always that it worked for X months then suddenly stopped. In any case I will check that this coming December. When there's a storm, I'll sit next to the meter, turn on Snow Melt and watch what happens, and for how long. I see that Starlink says the dish will melt 1.5" of snow per hour. Sorry, that tells me nothing. At what angle? And what triggers its activation?

I am not an electrician. Didn't take Electrician 101 at the community college. I've had to do a bunch of research to figure this out, but the closer I look the more I think this explains a widespread Starlink problem. Finally, we might ask how something this basic and frankly stupid could happen (if I'm correct, that is.) The answer would be rocket science. Musk & Co. put all their brain cells into the rockets, the satellites, and the dishes, and wound up "assuming" the routers and cables. Kind of like when the space shuttle geniuses assumed those O-rings. Smart people are no less inclined to be oblivious than anyone else.

FINALLY ... if anyone can tell me how I'm wrong about this, I am much, MUCH more receptive to correction than it might seem.

@bghira
Copy link

bghira commented Jul 30, 2023

Sorry, the standard Starlink cable is 50 feet. It's in their specs, and I confirmed it by measuring the old one that I pulled out.

used to be 75. but it's honestly too much distance. i was always cutting and shortening.

@TyraelTLK
Copy link

TyraelTLK commented Aug 11, 2023

I can't find passive poe injectors with 4 pairs. Is it possible to use 2 of these making 2 custom Y cable to use only the 2 powered pairs of each injector?

https://www.alfa.com.tw/products/apoe03g?variant=39871024889928

@jbowler
Copy link

jbowler commented Aug 11, 2023

Is it possible to use 2 of these making 2 custom Y cable to use only the 2 powered pairs of each injector?

Yes but you have to know exactly what you are doing and it's pointless since the Tycon, as documented above, works just fine so long as correct wire pairs are swapped. If you don't want to do that the one I posted links to on Amazon doesn't require the wire swap. Seriously though, the cable has to be cut with any of these solutions (except the two most recently posted) so attach the RJ45 modular plus and it really is no more difficult to crimp those on with the correct wire swap than it is to crimp them on without a wire swap. (Both are error prone of course.)

A single APOE03G costs the same as a single Tycon PoE (USD15 from Amazon, USD10+postage direct from Tycon).

@TyraelTLK
Copy link

From Germany I see the Tycoon no more available from your link. From another it's 30€ but long time delivery, I ordered 2 days ago and it should arrive 24th but it hasn't been shipped yet (but it's too late for me in anycase). The APOE03G is 9.85 + 5€, total 25€ and should be here in 2-3 days.
Thank you!
Let's see if I'll be able to put together something that'll work ;)

@sjkjs
Copy link

sjkjs commented Aug 12, 2023

The Reolink power supply is out of stock and I can't find an equivalent replacement where I live. DC power supplies don't seem to advertise whether or not the DC negative is connected to ground, it seems to be luck of the draw.

How can I make sure that the CAT5e shield is correctly grounded when using a POE-INJ-1000-WT PoE injector? It only has + and - inputs. There's no terminal for ground. Is it safe to take any random 48-52V power brick and ground its DC negative output?

@jbowler
Copy link

jbowler commented Aug 12, 2023

How can I make sure that the CAT5e shield is correctly grounded when using a POE-INJ-1000-WT PoE injector?

It's well documented in the Tycon specification; the PoE does a pass through on the shield. You need to use a PSU which is floating but general purpose PSUs should do that; the only one I've found that did actually ground the output was a LinkSys PSU designed for a particular piece of equipment. Test it first.

@sjkjs
Copy link

sjkjs commented Aug 12, 2023

It's well documented in the Tycon specification; the PoE does a pass through on the shield. You need to use a PSU which is floating but general purpose PSUs should do that; the only one I've found that did actually ground the output was a LinkSys PSU designed for a particular piece of equipment. Test it first.

The shield is meant to be grounded isn't it? I will be connecting the dish (via the PoE injector) to a network switch which is powered off a DC barrel jack and has a floating ground. So, if I used the Tycon injector and used its shield pass through, the shield would still be ungrounded at all 4 plugs (2 on the dish side, 2 on the switch side).

In this situation don't I need to explicitly provide it with a ground, or accept the risk of being ungrounded?

@jbowler
Copy link

jbowler commented Aug 12, 2023

In this situation don't I need to explicitly provide it with a ground, or accept the risk of being ungrounded?

Yes, the shield should be grounded somewhere. You have to make an explicit decision where to ground it unless the PoE injector does it. (The PoE injector for my other PoE aerial, from my original ISP, does ground the shield.) Normally nothing else in the cable run (including surge suppressors) will ground the shield.

In my StarLink arrangement I had the dish plugged in to a surge suppressor then from there to the Tycon WT then all the way to my router; this was in a box outside the house. I ran the drain wire out of one of the RJ45 plugs and just connected that to the same ground as the surge suppressor.

In the arrangement with my prior ISP I put a surge suppressor on the outside of the house, in a box, and had waterproof RJ45 sockets on that. IRC I had a shield pass-through to the PoE but I could have grounded the shield of the incoming RJ45 inside the box and arranged not to connect it on the outgoing connection to the PoE, or broken the connection to the PoE by using an unshielded cable in the last step (wall jack inside the house to PoE.)

NetGear PoE switches do ground the shield; for the ones with the wall-wart and the low voltage DC supplies there is a grounding screw on the case, for ones with built in AC power the shield is connected to the AC ground.

The rule, if there is one, seems to be to ground at the power supply; the PoE injector. All the switches I have provide shielded RJ45 sockets and all those shields are connected together. On one (a new TrendNET 10G switch) the shield is also connected to the power supply socket (12V 1A barrel connector). On another NetGear switch (non-PoE) the shield is isolated from the barrel connector. Since you are using a a switch I would expect every shielded cable plugged into it to be connected together so if any get grounded they all will be.

@sjkjs
Copy link

sjkjs commented Aug 13, 2023

I ran the drain wire out of one of the RJ45 plugs and just connected that to the same ground as the surge suppressor.

Thanks for the explanation. That all makes sense.

I'm using a Netgear GC110P PoE switch which doesn't have the shield grounded and also doesn't provide a grounding screw. The shield seems to be disconnected from everything.

I like your idea of using one of the spare RJ45 ports and running a drain wire to somewhere else that does have a ground. I might look at doing that too.

@darconeous
Copy link
Author

For what it's worth, I grounded my 200' setup at two places:

  1. Using a Cat-6 surge protector close to the antenna, ~120ft away from the house. This grounds the shielding and provides an arc path to ground for the twisted pairs.
  2. Grounding the shielding at the service entrance for the house, no surge.

The ground at the antenna is it's own ground spike. The ground at the house is the shared house ground (luckily I could run a wire to that on the exterior of the house). Unclear if I'm going to ultimately have galvanic problems. Hopefully not. Resistance between the two independent ground spikes is fairly low.

Originally both ground points were going to be full CAT-6 surge protectors, but I was having trouble with them increasing the resistance slightly so at some point I decided to just ground the shielding at the service entrance. That was before I switched to 52V, so perhaps I should revisit that... I'm hesitant to cut the wire though, since I can't un-cut it if it doesn't work out.

@jbowler
Copy link

jbowler commented Oct 6, 2023

Resistance between the two independent ground spikes is fairly low.

I'm very interested in knowing what that resistance is :-)

@darconeous
Copy link
Author

I'm very interested in knowing what that resistance is :-)

If I remember correctly, it was a few ohms. I don't remember what it was exactly, I just remember thinking "wow, that's lower than I expected". I have an outlet nearby with a ground that is connected to the house ground (not the ground that the antenna is connected to) so one day I got curious and pulled out the multimeter. But a low resistance makes sense, we have a lot of minerals in our clay-type soil that also has considerable moisture content. And I really buried that rod quite well.

I'm curious now, we just had a lot of rain. I can check again tomorrow.

@crdiaz324
Copy link

Is it possible to use 2 of these making 2 custom Y cable to use only the 2 powered pairs of each injector?

Yes but you have to know exactly what you are doing and it's pointless since the Tycon, as documented above, works just fine so long as correct wire pairs are swapped. If you don't want to do that the one I posted links to on Amazon doesn't require the wire swap. Seriously though, the cable has to be cut with any of these solutions (except the two most recently posted) so attach the RJ45 modular plus and it really is no more difficult to crimp those on with the correct wire swap than it is to crimp them on without a wire swap. (Both are error prone of course.)

A single APOE03G costs the same as a single Tycon PoE (USD15 from Amazon, USD10+postage direct from Tycon).

I've successfully used the techcharmer.com kit with my modified dishy without having to cut or swap any wires. The process involved connecting the provided cable from the kit to the plug in the back of the dishy, and subsequently to the POE's power output. Then I connected the POE to my router utilizing a standard ethernet cable I acquired from Amazon. This setup has proven to be incredibly reliable, running seamlessly for approximately 4 to 5 months. The kit was roughly $30. Hard to beat that IMO.

@JustOneGuyHere
Copy link

JustOneGuyHere commented Oct 20, 2023 via email

@jbowler
Copy link

jbowler commented Oct 20, 2023

@JustOneGuyHere; @crdiaz324 was commenting on his previous post from July:

https://gist.github.com/darconeous/8c7899c4d2f849b881d6c43be55066ee?permalink_comment_id=4645398#gistcomment-4645398

It's worth noting that TechCharmer now sell that Tycon lookalike PSU with what appears to be YAOSHENG adapter I posted a link to (two posts down from the above). Here: https://techcharmer.com/collections/our-products/products/starlink-dc-adapter-kit The Amazon link I posted no longer works but it's here:

https://www.amazon.com/YAOSHENG-Rectangular-Adapter-Connect-Injector/dp/B0BYJTHX4P

It's a nice move by TechCharmer; they've undercut the YAOSHENG very expensive PSU ($81) with an adequate and well proven power supply (assuming it is one of the Tycons), but they certainly seem to be selling the YAOSHENG adapter as part of it. Everyone benefits. The only downside to the TechCharmer solution is that there is no surge protection that I can see, but there is none on the official StarLink router (aka incredibly expensive PSU) so far as I know.

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