In-Wall Network Switch

[DolphLundgren]

I've been thinking for awhile that it would be nice if someone made a network switch that fit into a standard 1 or 2 gang box and could be used to take a 10G uplink and break it out into 1G ports.
I think this could work really well in homes that are heavily connected without having to do as many runs as it would take to feed all the devices in an room.
You could run a single fiber or Cat6a drop to a room, plug in the switch and have a clean way to connect up your devices. Would be especially cool to power it off PoE and/or pass-through some PoE to other devices.

I went as far as looking for parts to try and make one myself (This Broadcom chip looked promising) but that's about as far as I could go. I tried out KiCad but I've gotta admit that this isn't the easiest project to learn on and I'm pretty sure high speed data lines and BGA chips aren't the most fun at the best of times.

So I've come here to ask if anyone else is interested in this idea, if they could help out or point me in the right direction.

[rdl]

Looks pretty close to what you want at a glance, but this kind of device does seem pretty rare. Kind of expensive too, I think.

https://www.amazon.com/PoE-Texas-GBT-4-IW-Gigabit-Extender/dp/B07Z59SG17

[Ranayna]

Those exist, but are expensive, and are not flush with the wall:
https://www.cisco.com/c/en/us/products/switches/catalyst-micro-switches/index.html#~models

[DolphLundgren]

Looks pretty close to what you want at a glance, but this kind of device does seem pretty rare. Kind of expensive too, I think.

https://www.amazon.com/PoE-Texas-GBT-4-IW-Gigabit-Extender/dp/B07Z59SG17

That's quite close to what I was thinking however it's 1G in and out with a focus on PoE delivery. Honestly not that bad of a price; for a switch like I'm thinking of under $100 would be the goal. Running more drops can get expensive pretty quickly and if it could connect to fiber lines then getting faster speeds over time would just require replacing the switch and not rerunning the system.

Those exist, but are expensive, and are not flush with the wall:
https://www.cisco.com/c/en/us/products/switches/catalyst-micro-switches/index.html#~models

That seems vaguely similar to what I was thinking of. Do you think the size is a limitation due to heat or just because it's designed more for business/enterprise that is more likely to use large cable ducts? Also whatever the price is I wouldn't be surprised since it's Cisco and they are charging a lot on name value, software and support.


Combining the two ideas basically creates what I was thinking. From the Cisco switch, side mounted SFP+/SFP28/QSFP+/QSFP28 port that is comparable in size to a 1gang box with the faceplate, mounting and possibly (debatably useful until more home devices support it) PoE capabilities of the PoE Texas switch.

I'm going to look more into the PoE Texas company, see how large they are and what they have for products; I could see them making something like this.

Also I would take any advice that I can get on actually designing something like this. The chip I was looking at seems like it should almost be plug-n-play with the basic setup I saw in the documentation; bearing in mind that I'm very new to this it seemed like it would only take wiring to the proper connectors and a power supply to get it working in an unmanaged fashion but I, obviously, can't be sure of that. Also even if it is that easy, I still don't know how on earth to handle designing a circuit board for a BGA chip.

[Ranayna]

I have no experience with those micro switches, and only little with catalysts in general.
Also, i might have been premature mentioning them: They were announced in January, but apparently you still can't buy these, they are still not even in the pricelist yet.

I would assume that the size is mainly a thermal issue. The default DC power supply (external!) is 80 Watts, and it can provide 65 Watts PoE power to downstream devices. I think that can get quite toasy under load.
Additionally, these are fully managed as far as i know (of course only if you pay extra for the license), so that will also take additional space.

[madires]

There are tons of DIN rail ethernet switches, some with SPF. But I haven't seen any with 10G yet. My guess is that 10G would create to much heat for this application (closed in-wall box).

[DolphLundgren]

Yeah, taking DC input to pump out PoE+ instead of forwarding a bit of extra PoE power alongside being fully managed would definitely eat more of the thermal/space budget.

Fully managed and PoE output would be nice but I'd be fine throwing both of them away to get within size/thermal constraints. Too bad these guys aren't actually properly in-wall mount either.

I'm worried this means that the only option is a custom solution because unless someone knowledgeable in designing something similar to this can pass on some wisdom or take up the mantle, I'll probably be stuck for quite some time.

There are tons of DIN rail ethernet switches, some with SPF. But I haven't seen any with 10G yet. My guess is that 10G would create to much heat for this application (closed in-wall box).

I'm unfamiliar with DIN rail switches, I thought they wouldn't be in wall.

It's a very real possibility that it could be too hot for closed in-wall boxes, the only power usage/temp information I could pull together when considering designing this myself was that the chip itself would use under 5W max, 3.8W typical. I haven't figured out how much an SFP+ adapter would use or anything like power circuitry.

I was thinking that it might be possible to not "enclose" the switch entirely by putting venting in the front faceplate in the space around the jacks and either arranging the geometry to provide some thermal air flow or using a mini-fan.

[Fred27]

I've been thinking for awhile that it would be nice if someone made a network switch that fit into a standard 1 or 2 gang box

I think the problem may be that there's a "standard" per country, vastly reducing the market for the product or requiring quite a few variants.

It's far easier and more flexible to have a single RJ45 on the wall and a switch external to that. If you're going to have multiple connected devices in the room anyway, one network switch won't make much difference. Personally I have a small network switch in the cable tray under my desk. Unobtrusive and easy to swap a 5-port for an 8-port when I needed to recently.

[DolphLundgren]

I think the problem may be that there's a "standard" per country, vastly reducing the market for the product or requiring quite a few variants.

That's a possibility for why it's not a more common release. I do think it may be possible to come up with a design that is modular enough to fit different styles though, which would be an interesting idea to explore at some point.

It's far easier and more flexible to have a single RJ45 on the wall and a switch external to that. If you're going to have multiple connected devices in the room anyway, one network switch won't make much difference. Personally I have a small network switch in the cable tray under my desk. Unobtrusive and easy to swap a 5-port for an 8-port when I needed to recently.

I agree that can work perfectly fine in a lot of situations, I just like the sleekness that could be had with it in wall.

[ajb]

10G makes this a lot harder, both in terms of manufacturing and marketability.  There aren't that many 10G switch chipsets you can buy from the open market, and AFAIK those that are on the market are on the higher end of the capability spectrum, so in addition to needing more careful circuit design/layout with BGA packages they also require more software development to bring the switch up (it's common for higher end switches to have built-in processors for management and complex switching logic, and you'd have to dig into the docs to see how much you need to do to get even basic packet switching).  You'll also need to use external PHYs with many of them, which means routing MII signals which have more strict signal integrity requirements than the wire side MDI, and the whole board has to get bigger.  Then there just aren't that many applications that really demand 10G at the moment.  Even with gigabit being common now, how often do you think those links get anywhere close to saturation?  If you're doing high-bitrate video editing or something else that requires bouncing a ton of data across the network then you probably just want 10G all the way from the PC to the storage, which requires a different kind of switch.

If you're willing to drop the uplink to gigabit then it all gets a lot easier.  You can buy single quantities of ~8 port gigabit unmanaged/lightly managed switch ASICs in QFP that can be configured by pin strapping or an external EEPROM, so just assemble the board and you're done.  No external PHYs needed, so the only thing you're laying out is power and the copper MDI which is going to be much more forgiving than any MII interface (which you'd need to connect a switch to an external PHY).  The MDI bandwidth is also only ~100MHz compared to ~500MHz with 10G.  All of this can be done on a cheap-as-chips four layer board with a standard stackup from your favorite fabricator.

[DolphLundgren]

10G makes this a lot harder, both in terms of manufacturing and marketability.

I can understand that, it's just the point I think is interesting because it lets multiple devices run on a single connection at full 1G without blocking.

There aren't that many 10G switch chipsets you can buy from the open market

That's true but I did manage to find the Broadcom one I mentioned in my initial post and it seems almost perfect for this.

AFAIK those that are on the market are on the higher end of the capability spectrum, so in addition to needing more careful circuit design/layout with BGA packages they also require more software development to bring the switch up (it's common for higher end switches to have built-in processors for management and complex switching logic, and you'd have to dig into the docs to see how much you need to do to get even basic packet switching).

The particular chip that I linked to is the Unmanaged version and, while I can barely understand the doc, it seems like it shouldn't really need much software and is all built-in.

You'll also need to use external PHYs with many of them, which means routing MII signals which have more strict signal integrity requirements than the wire side MDI, and the whole board has to get bigger.

I know at least for this chip the 1G PHYs are internal and you wouldn't need one. The 10G I'm having more trouble with understanding the docs. They claim it's XFI which, if I understand correctly, means that it's signal is compatible with SFP cards directly. If that's true then the 10G PHY can be handled by the SFP+ card of the user's choice.

Then there just aren't that many applications that really demand 10G at the moment.

Fair point. Also while technically this would be an application demanding 10G since feeding multiple 1G devices without blocking needs it, it's also not super common to saturate multiple 1G connections at once especially in a home setting. I personally, just like the idea of being able to run a 10G connection to a room and then hook up multiple media devices and PCs and never worry about them interfering with each other.

If you're doing high-bitrate video editing or something else that requires bouncing a ton of data across the network then you probably just want 10G all the way from the PC to the storage, which requires a different kind of switch.

Oh, I know that. I already have my PC hooked up via 10G so that I can transfer data to/from my NAS faster.

If you're willing to drop the uplink to gigabit then it all gets a lot easier.  You can buy single quantities of ~8 port gigabit unmanaged/lightly managed switch ASICs in QFP that can be configured by pin strapping or an external EEPROM, so just assemble the board and you're done.  No external PHYs needed, so the only thing you're laying out is power and the copper MDI which is going to be much more forgiving than any MII interface (which you'd need to connect a switch to an external PHY).  The MDI bandwidth is also only ~100MHz compared to ~500MHz with 10G.  All of this can be done on a cheap-as-chips four layer board with a standard stackup from your favorite fabricator.

I'm certain of that too. I've seen a cool project that fits a 5 port 1G switch into a tiny 45x45mm footprint. I could basically shove one of these in a gang box with no issues or at least use the same parts try redesigning the board.

[ConKbot]

Broadcom device for a higher end market that actually has the datasheet   usually you have to contact them, and ytell them how many 100k per year you want to make to even be considered for the privilege of having a datasheet bestowed upon you.

[DolphLundgren]

Don't say it too loudly they might take it down. 

[DolphLundgren]

Bump: Still looking for anyone that could give me advice on how to design a circuitboard for this thing or any products similar to what I'm thinking of.