Consumer-grade routers on puny power supplies (2024)
blog.apnic.net
blog.apnic.net
1. Downstream of the mains power supply are DC-DC converters that run the router hardware. Those contain the filters and capacitance you think you're fixing. Nothing in that router actually cares about mains power quality. They absolutely do not care about perfect sinusoids.
2. If you were seeing insufficient power to the router, you would observe crashes and faults -- not slowdowns.
3. Two different routers showed the same behavior, which suggests that the fault lies outside the router+power supply and more to do with something common (e.g. network, laptop).
The dip shows a reduction in voltage, and a larger one than I would like, but without a scale on either time or voltage, it's difficult to guess if it actually matters. I would suspect not, since the device does boot successfully. Again, the voltage doesn't matter, since the router runs off its internal DC-DC supplies, not the external power supply.
I'm happy that the capacitor and new supply has fixed the issue, but I'm unconvinced by the explanation. Check grounding between inverter and laptop.
Possibly he's not describing the problems right. I can certainly believe a shitty enough power supply would cause problems.
But that was an all-or-nothing failure mode in which they would power up but never do anything else. Performance changes is a claim that requires much stronger evidence.
As for the article, I'm left with more questions. Surely the solar inverter wasn't running all the time in series, like a double conversion UPS, right? So how how would the mains waveform have been significantly affected in normal operation?
Also those scope traces, what's the scale? Are we talking 100mV dip, or a 1V dip? And is that a storage scope, or is that one cycle of the supply's ongoing ripple? And the complete lack of any dip on the "fixed" trace with extra capacitance makes me wonder if they even got the triggering right.
I have questions. What's the point of a dedicated solar circuit? They are leasing from their "local lines company". If the lines company is in the leasing business, this system must be grid-tied right? Why the dedicated solar circuit?
Also, is it common for full sine wave inverters to produce power less clean than the grid? Maybe when the batteries are low? Curious.
Could also be affecting the analog circuitry more if the droop is too bad or its browning out. That could be loss of gain, SNR, etc.. that could cause packet loss and retransmissions
So yeah, seems unlikely the only impact would be a sluggish dashboard. Maybe the device was churning on error re-transmissions from the brownout? Like the CPU itself was OK but the ethernet ports weren't?
I think this is more likely. Two different routers impacted. Crappy grounding or induced noise causing high BER on the links.
Granted, cheap consumer devices are much simpler than that, but it's still something that can be added to the SoC.
Ebon Upton started Raspberry Pi to help Qualcomm dump stock they couldn't get rid of otherwise.
Look at every single Pi to come out - it's been faster than what came before it, but in a matter of weeks half a dozen competitors have better boards with faster processors for cheaper - that don't have all the nonsense like RPi foundation repeatedly fucking up the power supply so vendors could milk people on "pi compatible" USB power bricks.
A Pi 5 16GB costs $120. Plus case ($10) plus power supply ($12) plus video adapter ($10)...$152. That is absurd.
I think RPI is starting to lose sight of that with the Pi 5 and especially the 16gb models. It's starting to just be expensive and the better support for an ARM SBC starts losing a lot of value when it's butting up against x86 mini-pcs like the large number of N100-based options at around the $150 price point[1]. They aren't really any less efficient, and x86 software support is still consistently better than ARM software support.
1: eg https://www.amazon.com/Beelink-Intel-N100-Computer-Desktop-D... - $170 for a complete unit w/ 16GB RAM & 500GB SSD.
About the best thing they did was adding a choke on it.
On top of that, it's a crappy 40MHz analogue scope. You're not going to see anything useful.
Finally a savvy helpdesk person had him move his ADSL router. I believe he moved it off a cheap power board and plugged it directly in to the wall.
All problems immediately solved.
>[...]
>Ideally, we’d want manufacturers to put capacitance to that effect into their router power supplies or at least routers. But we realise that capacitors cost a few cents each and aren’t really required on stable grid supplies. And manufacturing costs matter.
Weird this was being seemingly blamed on "puny power supplies", and that "ideally" manufacturers should be overspecing their power supplies to accommodate his use case. Does he also think manufacturers should "ideally" add a battery to routers on the off chance that there's a transient blackout?
More to the point, he supposedly has a "4500W peak solar sinewave inverter", and his router probably consumes no more than 50W. That 50W is unlikely to cause issues with such a system, and even if it was the straw that broke the camel's back, it's weird to blame it on the router, rather than the inverter or the entire system. If you had a take-home salary of $10k, blew it all on gambling, then your card got declined on a starbucks purchase, you wouldn't characterize this as "my coffee habit was using more money than my income can supply".
The inverter probably only does a few step approximation of a sin wave which is fine for most things, but clearly not for this power supply which browns out (I’m guessing during an extended period of 0v on the stair step sin approximation).
Whether your average power supply should be tolerant is unclear, but if it’s the only device playing up it’s clearly less tolerant of imperfect power than everything else.
Unless that little dip on the supply is constantly happening, I highly doubt that's causing any of the issues. That dip could be from in-rush current or from the router powering up, both of which would be happening before any data transfer. It means nothing without any sort of voltage or time scale.
It'll be a switch mode power supply, where the power is rectified, converted to high-frequency AC (50-200 kHz), then sent through a certain specialized transformer.
The other problem they create, then, is high frequency noise which must be filtered out. Some cheaper adapters just dump the noise back into the mains and cause problems for _other_ devices.
Come to think of it there's a decent chance it's a /different/ adapter somewhere on his system that's causing the problem and not the size or capacity of the products native adapter.
There absolutely is. The core material is not iron, it is a material specifically engineered for the purpose. There's a very specific size gap between the two halves of the core. There's an extra sense coil for the feedback circuit.
The gap is controlled by the cross section of the core but also contributes to EMI. The sizing concerns are typically related to the latter problem.
An SMPS can easily use an external feedback circuit and may do so for certain types of applications. For mass manufactured low current applications you can often buy them with one integrated.
Unfortunately I'm unable to find any examples of them used outside of voltage-multiplying circuits in obsolete electronics.
That's what the article is talking about - the output capacitors in their cheap power supply weren't big enough for the peaky load of the router. (Well, probably were assuming a sinusoidal input, but not big enough when the stairstep input has big gaps).
Or get the modern western ~300V DC input ones, there's no good (technical) reason to feed AC from a non-grid-tie inverter to a SMPS-only circuit.
The DC can be quite passive, just switched in via a relay and anti-reverse-current-diode together with bridge-rectified mains, she otherwise just bare battery bank voltage There's a reason modern data centers use that style of setup (except with a central power factor correcting rectifier, not a plain bridge rectifier, and often hanging the batteries on it directly using controlled rectifier output voltage to control battery charging).
"Rummaging through my old parts box unearthed a power supply that was a bit more powerful than the one that came with the router, and even fit the connector." Is he describing a "wall wart" type power supply? A picture would help. A teardown would help more. There are large numbers of really crappy wall-powered DC supplies available, mostly from China.[1]
Look for a UL approval marking and a UL approval number. That indicates at least some attention to quality. UL is mostly concerned about fire safety, but their basic test for power supplies includes connecting it to a load box at the power supply's maximum rating, then running it for a while to see if it overheats. This catches power supplies with exaggerated ratings.
Avoid no-name power supplies. This is a well-known headache.
Edit: don't repeat myself
Then it's just a matter of picking out one that's adequate for the selected device, it's a many-to-many mapping with lots of valid solutions, not one-to-one.
The correct and effective means to have safe and correct products is to monetarily and criminally punish corps that produce faulty goods.
It isn't a mark of quality, or of fitness for purpose.
It does mean that the device(s) submitted for testing are unlikely to burn your house down, though.
The AC -> DC power supply is tested for safety by UL, but the thing powered by that power supply is not tested at all.
Normal UL testing and listing deals with mains-powered devices, not low-voltage DC-powered devices.
I remember waking up early on Saturdays in 90s to load some game from tape on my C64 before the neighbor starts his sawmill. Couldn't load anything when it was running :)
The type that can do this also fail at a rate that depends on temperature. Which is why they shouldn't use them internally. If you truly need to condition your power than you need to do it separately as almost no one shares in this problem and would not be benefited by adding these degradable parts.
The PSU manufacturer has cheaped out on the capacitor in the input, in hope that the circuit which switches the primary coil would still be able to extract enough power even if the instantaneous mains voltage drops to something as low as 30V (which it does twice every period). So, on a pure sine wave input, they only need to cover this (relatively short, below 3%) below-30V gap twice every period.
But with the UPS feeding the circuit, the situation is different: most of the time, the rectified voltage is 340V, however, there is a significant period of exactly 0V that still needs to be covered by the capacitor.
I realized the inverter was using about 5 watts idle and that was about half of my average power budget for the camp (counting night time), so I cut the adapter's cable and re-wired it directly to the battery. Worked fine. Most components in a computer that nominally run at 12V are passing it through some sort of adaptive converter (either a buck converter or directly to a speed-controlled motor) so they have a fairly wide voltage tolerance. I obviously didn't find out if it reduced the lifespan at all.