Comment by gruez | Hacker News Reader

Comment by gruez | Hacker News Reader

>Clearly, the router is pulling more current during the dip than the power supply on my solar circuit can supply.

>[...]

>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 50W load (if that) isn’t the problem - a lot of cheap power supplies assume a perfect sinusoid which you get from a spinning mechanical source of electricity.

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.

The wallwart for the router is going to be a dead simple design that doesn't in the least bit care about how dirty the sine wave is. It goes into a step-down transformer (12VAC usually), into a bridge rectifier (now 12VDC), add some decoupling caps to smooth out the 12V. Inside the router, there will be LDOs that bring that voltage down to something usable along with more decoupling caps to filter the input and output of the LDO. LDOs have high ripple rejection especially at low frequencies so they can take on pretty dirty voltages.

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.

There's no way a router in 2024 is using a big iron transformer. They're too expensive.

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.

There's nothing special about the transformer in SMPS other than it's size which is due to the high frequency.

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.

> the transformer in SMPS other than it's size which is due to the high frequency

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.

It depends on the switching frequency, and not all SMPS run at a high enough frequency to justify ferrite. On top of that ferrite is effectively just iron in a ceramic.

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.

If you aren't fussed about galvanic isolation there are many SMPS topologies that don't use transformers. Cheapest step down converter would be a simple buck regulator - no transformer needed!

Piezo transformers are dirt cheap.

I've never heard of these before, so thank you for making me aware of them, they are very cool.

Unfortunately I'm unable to find any examples of them used outside of voltage-multiplying circuits in obsolete electronics.

> along with more decoupling caps to filter the input and output of the LDO

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).

There are server power supplies that won't work on non-sinusoidal power sources. (read: cheap UPSs) I learned that the hard way.

Use one that has the "240VDC (China only)" rating, and feed them proper DC.

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).

Reply on news.ycombinator.com