ไม่สามารถเล่นวิดีโอนี้
ขออภัยในความไม่สะดวก
We're building a high-end router. Here's the update for June 2024.
ฝัง
- เผยแพร่เมื่อ 12 ส.ค. 2024
- Thanks to PCBway for sponsoring this video! www.pcbway.com/
Thanks to Anni for the beast PC I use for development: anni.si/
➡️ If you'd like to participate in the pricing survey for this device, please fill out this form:
research.typeform.com/to/KegR...
🚀 Intro to the series, if you haven't seen it yet: • I'm making my own high...
Bootlin embedded linux materials: bootlin.com/doc/training/embe...
CHAPTERS:
0:00 Finally! Development!
0:24 I went to the Netherlands!
1:45 Power supplies
4:14 Where to put the boot drive?
7:12 Wi-Fi support update
8:43 Rant.
11:02 PCBway
11:24 u-booted!
16:22 recap for June 2024
17:29 Logistics
"If you're new to embedded, fuck you, figure it out" is just so true. There is so much institutional knowledge that you need to have to work in this space, it's kind of frustrating.
@@Dygear when RTFM gives you the middle finger... Fortunately, Tomaz figured it out and promised tutorials on how to do it! 😂
This was definitely my favorite part of the video haha. There was a lot of emotion there.
True! I had to unpack and disassemble (with Ghidra) the FW of my ISP supplied router. The Device Tree files in there are obscure, it took me ages to comprehend them.
I have been learning embedded since fifth grade and it never ceases to amaze me how closed and opaque everything is. For anyone starting, it's so much faster to go to an active community (local or online) and have someone point you in the right direction.
It’s much more rewarding when you have to figure everything out by yourself though. As frustrating as it can be.
"Let me explain it to you like I'm 5"
_Proceeds with a college freshman level of an explanation_
My Slovenian friend, please don't keep us in the dark. Those difficulties you find on your way are EXTREMELY relatable. Any tips/walkthroughs you have for us are extremely helpful. Love you, brat
Thank you! :)
An indepth look into the development process and iteration on the kernel loading from the dev PC and stuff would be super interesting and cool!
@@MaximinoReyes That is super interesting, unfortunately I am woefully unskilled, and definitely not the right recipient for your comment :D
Maybe repost it as a new comment under the video
I used to work for a switch/router company. It was pretty interesting. In the past, the company would buy a special switch (from broadcom, for example) and a license for the SDK. In general, the SDK would come with an old linux kernel, some command lines and a 90's web interface. Then, we would add features (almost from scratch) to the product. The hardware part of the company was pretty good for a reasonable small company (500 engineers). Later the company started building it own OS that could be used with almost any device that support linux. They started with their GPON (fiber) line of products. I think the company is not doing very well, many engineers moved to the Netherlands and US for better jobs, and the government stopped heavly invest on public fiber infrastructure. I think it is very hard to compete with chinese routers, but, at the same time, if you can provide a reliable and quality product, their is still a market for it.
Are you starting from an existing framework (like openwrt), SDK from the chip that you chose or are you building something from scratch?
Are you talking about MikroTik? 🤔
@@Daniel15auLithuania, Latvia, Malaysia, Vietnam, and China for Microtik, so a good guess. I’m not sure of any network hardware manufacturers that don’t have any operation in china.
@@Daniel15au No, the name of the company is datacom from Brazil
By the way, Mikrotik is pretty good for its price, but it feels like it is heavy CPU oriented, with almost no specific hardware processing
Which is good to make many products, but can lack performance under heavy load
I really hope he's not starting with the vendor's SDK as they are wildly out of date, and for a device with swapable parts it's not a good idea.
oooooh ya! tutorial videos about boot stuff would be awesome!
I've just made more work for myself haven't I? Good thing I enjoy it haha :D
Awesome!! I think I can speak for most everyone here - anytime you think should I or should I not make a video about this - haha make it. I'm not even into development or hardware like this, but super interesting to learn about. I know the videos really take time and effort to do and appreciate your willingness to continue to update us and make additional videos. Really looking forward to the Kernel compiling machine that you had built and especially what sort of hardware makes a machine better or worse at compiling kernels.
Thanks for the encouraging words - they mean the world to me!
10:33 that actually makes sense, uboot is opensource bootloader with big community behind and they should make good job of documenting their tool for different boards, not the other way around.
Your borderline breathless enthusiasm and urge to share as much information as possible as fast as possible is deeply endearing. Good luck with the project! 🌞😉
I feel you on the documentation issues. usually stuff written by a third party is wayyyyyy better written.
The notes from people who have had to actually use the stuff, sure. But it gets dated rather quickly, too.
fascinating video, I'm enjoying your journey of development for this router!
the "FU, figure it out" is very common in embedded development. I've been an embedded HW EE for decades, and the era when we could get good support from everyone is long gone. ST is pretty good. NXP is definitely not.
I always liked NXP microcontroller documentation, back when I last used it (maybe 10 years ago). If you opened the chapter on e.g. I²C, it started off by telling you what else you need to configure (bus clocks, GPIO alternate pins, etc.) to get it to work, and referred you to the right sections. ST on the other hand just assumed that everything was configured fine and described just the I²C peripheral. You had to read and remember the entire 1400 page manual to get anywhere at all. That said, screw whoever at NXP thought it was a good idea to put the contents page at the back and omit the index.
Great video thanks. I’m interested in a hands on tutorial video.
fast SSD/nvme storage is good if you want to offer packet capture in you feature set, one would love to see that feature in high-end router. eMMC has limited wear and tear (TBW), Tesla had a recall because of this, specially if you are going to use it for logs
Hey Tomaž,
I find it fascinating and commendable how detailed you are in sharing your development process on TH-cam - there's a lot I can learn from it. I do have one question: some of your steps are technically detailed to the point where I wonder (from an investor's or customer's perspective) if you're sometimes "reinventing the wheel" where it might not be necessary. Is this primarily to optimize costs, or is it driven by specific community/user needs?
Wishing you continued success and best regards.
No, this is how you build an embedded system from scratch.
Good point. The thing is, I could have bought an existing PCB, package it into our own custom enclosure and call it a day. But there are several problems with that. First the margins are insanely slim at that point, second, you can't open-source much (if anything at all) and third, the system can be close to your desired specifications, but never perfect.
And to not reinvent the wheel, we're using u-boot bootloader, Linux kernel and likely OpenWRT filesystem :)
some of these steps could be in theory skipped but if you are doing something like he is its probably good idea to do it to do it right and you need to do it right. he is showing us how that is done
@@bigpod Tomaz* might be reinventing the wheel, but boy is he* going to make it round and smooth. And blue, as far as the color goes :)
(* I mean the entire team)
@@cristinelcostachescu9585 i would more say he isn reinventing the wheel but making it fit his cart
Nice work man! I am so happy to see the progress I love this project!
amazing project! cant wait to see July! amazing work
Im working on a similar problem at work right now, im excited to see how you end up solving this!
ALL the BEST to you guys succeeding this project! great work, nice insight in the process and well produced videos!
cheers and good luck from your neighbour, carinthia/austria
this is awesome!! super interesting seeing the beginnings of an embedded device
I think the idea of including an m.2 sata ssd for boot is actually probably a good one, there are some quite cheap devices and it will let power users upgrade if they want.
Thanks for keeping us posted. Excited for project!
I'm enjoying this series a lot, I love the amount of information you share about the entire process. Looking as much forward to the journey as to be able to order my router 😀 By the way, I'm in favour of the tutorials. I will never develop something like this myself, but I agree working with people in the integrated systems sphere, that the institutional knowledge is a hurdle. If you could share some of yours, that would certainly make it easier for others to enter the space and we'd all end up with more variety of devices that suit our needs. And i would watch them too.
Gosh, this is so interesting and educative, I doesn't even know the price point yet, but these videos make me absolutely certain, I'll buy one when the time comes and promote it within my peers. Keep up the good work and let that Kickstarter roll!
Keep up the good work. Staying subscribed to see this to life
Very interesting. Good luck!
What networking stuff is going to be supported on the router? Any dynamic routing protocols? Any advanced switching protocols? I hope you guys also focus on ease of monitoring. Something ive missed from medium enterprise routers and network devices is Somethings more than snmp etc.
Užival sem v tem upjdetu. Hvala! :)
Hurry up with that router ;-). I'm starting to look for a new router now, mainly considering protectli or building something. I'm slow enough that maybe yours will hit the market by the time I actually do acquire something.
Can just pick up a used Thin Client, MiniPC, Open Router, 1U server or old hardware run OpnSense on it and there ya go, Enterprise level and done
Thanks for sharing. You can go ahead and make the detailed tutorials mate. That would be very cool. We'll done in unleashing in regards to the documentation. Consolidering The price they charge for the dev board , it should come with a perfect documentation.
this is really cool to see. makes me wonder if getting that fritzbox made you so angry that you had to make your own, better router haha :3
jokes aside, i really love to see this kind of insight and transparency in the making of devices millions rely on daily, especially all the embedded linux stuff :3
im personally pretty nerdy about that kinda stuff so its really cool to learn more about it (yes i would like the more hands on/in-depth video ^^)
(also woah, i didn't know u-boot lets u load stuff from the network into ram and boot that, thats really cool)
Interesting, I work on the OS for a datacenter switch / router vendor. When I saw "making a high end router" I didn't have a WiFi access AP in mind, but this is an incredibly fascinating video nonetheless. Thank you for doing this in the open!
I guess "high end router" really can mean vastly different things.
Edit to add: ohh, the wifi chip is just for dev iteration! Ok, I need to watch more of your stuff, this is fascinating!
Looking forward to the build video on the Anni system
by ditching the pmic wont you lose the ability to monitor the voltage that each psu is outputting?
Feels exactly like during my studies. We had also embedded systems as a lecture for two semesters. It was exactly like this. RTFM, Git gud. Reading through a 3000 pages manual of a developer board without any clue really is a bit..."demoralizing" :D
Yeah. If only we'd get some basic pointers - it would save sooooo much time.
Why not you use logical voltage converter between 1.8v and 1.2v
Too complicated. WAY too complicated. There's a lot of signals to convert, and I don't think it would make sense.
"Fuck you, figure it out." -- this was my exact experience getting into embedded "The Wrong Way", granted when I started we didn't have no Arduinos! haha
@tomaz regarding the FLASH interface:
Why not use a level snifter IC ? There are some good one i know that are bidirectional and when it come o unidirectional ones there are some that can do RMII they are like 0.5$
You can also play with Vio from the FLASH and MPU and make it close enough to not make problems
I know this comes a bit to late now, but would it make sense to use the MediaTek MT7988A (Filogic 880)? Because it seems to have much more connectivity than the NXP-LS1046A, for example the Banana Pi BPI-R4 has two mini pcie, and one m.2 slot.
Asked him this on twitter, he said in the next couple of videos it will become clear why the NXP route is their better choice
Yea, I was thinking the same. I think that the HW network stack on the NXP have more bandwidth. But who knows.
The cool think is that MTK have hardware blcoks to accellerate/offload WIFI packets when using their chips.
It is basically all mainline code so also future SW updates are easier, no stupid outdated/vulnerable BSPs etc.
Turris Omnia Enterprise is based on same family of NXP SOC. Could talk with then about their experience.
Super cool project! I hate bad documentation!
I like your „Rant Time“ ;)
What you should keep in mind before moving to the production version of the firmware is to remove the device tree from BL3 of the U-Boot for security reasons. Can be activated later to perform "over-the-air" online Firmware Upgrade but it would not be as secure against attacks from the network, there are workarounds but what I said first is essential 😅
I want to do the it the other way around. Open source everything then patch the holes with the help of this community. :)
Out of curiosity is there anything (other than cost) that really prevents you from using a bunch of high speed transistors to do the logic level translation so you can use high capacity nands? I don’t know the exact conditions you have for that but I don’t think there’s anything obvious preventing it.
i think it is not fast enough and the signals get distorted , unless u slow it down enough
We're using the same SoC for a Network product. I'm interested to know If you decided to go with Buildroot or Yocto for your own Linux distribution, also are you considering enabling OP-TEE?
Is SPI booting an option? You could get an SPI flash chip, which are very common but kind of expensive per byte, or even just use an sdcard in SPI mode. It might be kind of slow, depending on hardware support(maximum clocks, QSPI, Octal SPI, etc.).
Thanks for the update, love the amount of details you provide!
However I didn't understand exactly how you can remove the PMIC from the system. How does the P-GOOD pin replace the PMIC?
As far as I know, powering a CPU is mostly an issue of sequencing the different voltages in the correct order, and fast enough to be ready for each next stage. Can you please elaborate?
Oh and, one last thing: remote file systems FTW!!
Let me write this down and answer it the next month because Ken is working on these, so I want to make sure I understand correctly myself before giving an answer, if that's ok.
You chain the PGOOD output from the first regulator to the enable of the second one - then the PGOOD of the second regulator to the enable of the third one, and so on.
@@tomazzaman yes, please!
@@triffid0hunter yes, I thought of that as well. However, some CPU supply stages require very precise timings in order to initialize correctly. If P-GOOD does not come up in time, or comes early, a CPU supply stage might come too late, or it may initialize too early when previous stage(s) did not finish making their feature ready. In some cases the CPU might be partially usable, but spits out complete garbage, which makes it hard to debug where the problem comes from.
@@cristinelcostachescu9585 Yea, what is odd is that there are some PMIC chips from asin brtands that are way cheaper than 5 bucks. And sometimes for VCORE, and VMEM those are way cheaper than single chained DC DC converters.
Wasnt the the PMIC point to reduce cost and complexity? Even RPI decided to go with one of those nowdays.
PMIC are also useful if you want to add cpufreq with voltage scaling when supported. Repowrt power usage metrics etc.
On my letest bringup I've forgot the proper timing of one of the power bus,
I solved with a brutal mosfet on the backplane board that cut the power of the 24V rail until pgood for the 3.3v level is good.
In this case the 24V rail is the main one. And used to make the 3.3 DVVD, 5.0 AVDD and 3.3 AVDD so powering that up post the logical rails was not possible. I've forgot about htis issue, so I bodged the power sequence in the backplane between the PSU and the single boards.
I'm just a DIY guys on the EE side :D
I would like to see the/a video/series on how to navigate embedded system, I may never use it, but I do want the knowledge.
The rant just won you a like and subscription from me.
So... based on experience (ISP/Network provider/MSP)... do not do eMMC ....
Multiple reasons:
90% of our failed devices are using onboard storage - we toss them when they fail.
If you use an SD card then it can be upgraded/changed ... but higher failure rate.
Lower/worse read/write performance - although it may not be required to have a high performance storage for a router ... it actually does make a difference.
Bootup/startup time, navigating UI for management (eg loading files from disk to render/send), logging, general data storage.
Overall ... rather just go SATA and use a 2.5" SSD - it's cheap, fast enough - and upgradable if someone wants to use more storage.
Also, the components will be less and ofc, less failure conditions - not requiring a mux etc...
Interesting, I mean I've had a number of SD cards fail on me, but I wouldn't think eMMC does as well, given it is just an interface - and a controller - to NAND (at least in our case).
Thanks for the suggestion!
@@tomazzaman Yep.... thus why SATA :P
Not really sure why, maybe endurance quality difference ...
But SATA does give the best of both worlds and you have available IO so overall, easier to implement and lower TCO ...
BUT.... if you get a 2.5" SSD and open it's case, it's bloody small .... so you could use that :D
OR.... m.sata interface .... it's still sata, but NVME form factor
Why 2.5“ SATA rather than M.2 SATA?
@@Daniel15au see my add comment... i forgot about them lol
eMMC is ok for reads. It's write cycles that really kill them. For just holding kernel + ramdisk, it'd be fine, but if you're storing logs, or frequent configuration changes, use something with better resilience. And, of course, don't solder it directly to the board!
Oh man, needing to figure everything out for yourself because of terrible documentation is such a frustrating thing to experience regardless of what it is related to. But at least you seem to know what you're doing (or at least getting there) which is good :)
Absolutely more about u-boot and kernel how to!
12:32 What's about Secure Boot ? how ROM validates if BL2, BL31 firmware are legit?
Check the NAND's Vil vs. the CPU's Vol and the NAND's Voh vs. the CPU's Vih. You may just get away with 1.2v NAND w/ 1.8v IO.
Your rant is but one of the reasons why about any vendor of large scale networking gear has been shifting to x86_64 arch. The others being cost - yes, an embedded x86 is very competitive with ARM, thanks for the price gauge ARM vendors have been playing - and availability of both hardware - thank you Intel and AMD - and updated documentation. I work on a medical industry and our RnD group has gone thru a lot of what you’ve been and their workload took a new fresh direction when the base hardware changed.
Did you guys use a BSP kernel or something like upstream LTS 6.6?
I've tried several kernels actually. Regular one (Torvalds repo), both versions 6.9 and 6.6, then I tried BSP one (based on 6.6 and 5.5), and finally OpenWRT (5.5). All of them built manually, no Yocto or buildroot, and they all work fine and pretty much the same as long as I pass the right device tree.
Are you planning on launching a pre-order offer for the router?
Hi Tomaz. I understand you perfect when you had to build the U-boot. I'm playing right now with a Banana Pi BPI-R4 board and I was able to build U-boot + AT-F to be able to get NVME working. I worked about 1 week until I found a patch for U-boot that activates PCI-e 3.0 for the chip to be able to see NVME SSD and now I'm in the phase of building the Kernel and Root FS on NVME.
My ideea is to use U-boot entirely with bootmenu using an u-boot-env on NVME wich is generated by OS each time a new kernel is installed (similar with Grub for x86). I want to use SPI flash just for the firmware.
I'm wanting for you device to be available for sale to play with it and "reflash" it with my own U-boot. 😁
Is your power supply low noise for streaming audio?
If you stream audio, the power supply on a router doesn't need to be low noise, because it doesn't matter.
As a layperson , any reason why the router won’t have “4x4 MiMo” ? Or is that reserved generally for cellular modems and not really necessary in WiFi?
Great Video.
As a potential customer, I was wondering if you are planning to make your custom Linux kernel open-source or not?
Yes, we plan to open-source not just kernel, but also u-boot and any code on the software side.
Just take my money@@tomazzaman
I'd love to see a tutorial style video!
Isn't this something like the r86s but arm? Is there any advantages with the cusom hardware? You can always use custom OS/kernel with UEFI
There are linux based router OS such as VyOS with redundant images to mitigate failed upgrades. If you are building your own os all these needs to be considered. I don't think OpenWrt supports this natively unless you have u-boot/partition magic like WRT32.
PS: How VyOS is doing with open-source it another problem
I don't think you will get any hardware routing/switching accelerations but there are tools like DPDK or crypto offload that can make the software router perform better in some use cases.
Also making routing configuration easy and safe is a important consideration. If you're considering OpenWrt, it is good for basic DNS/DHCP configs. But it doesn't have good support for more advanced routing protocol even like bgp. Then there are BSD based os like OPNsense with way better support for more advanced routing.
Agreed on many things, but we have to start slow - we don't have the capacity to go that high at this point yet. Also, the CPU that we chose does have dedicated networking hardware (called DPAA), so we will get acceleration both for traffic and crypto.
@@tomazzamanYou mentioned it will be running openWRT initially.
In my experience one big weakness of OpenWRT is no implementation of hardware offload.
As far as I know, only one Mediatek SoC is supported for hardware network offload.
Crypto offload might be ok just to the extent of using neon.
Are you planning to put some work into OpenWRT to get it working, or will that be later in the Opnsense port?
Would love some tutorials!
Maybe put u-boot on NAND and then let it boot from nvme? Or put kernel into NAND. Or entire base system. Either way 8GB is huge space.
Wouldn’t level shifters to NAND flash be better than an SDII mux?
I was also thinking about it. And there most probably will be a speed penalty, but probably not as much as using a mux... Maybe even the "dirty" level-shift trick works where you only clip the voltage down and use the 1.2 output voltage directly on the CPU input pin that will probably go high at 0.8V or so.
It should be possible, but we'd have to level shift A LOT of pins so the design would become much more complex compared to a 6-channel 2-way mux.
I would like to build some simple x86 intel SBC but there is no documentation or schematics available unless you work for some big company and you signed NDA.
There is also no info from them on how do you even get started with things like BIOS, do I need to code bios myself? are there some examples ready to go I could use? Just another thing you're expected to know already.
In the PC world, the BIOS is something you go buy. (linuxBIOS/Coreboot is a pain in the ass, even on platforms other people have "figured out"... there's only so much one can know without NDA'd chip specs, and as we all know, docs are usually wrong.)
@@jfbeam That's disappointing
and you thought this was going to be easy as finding the right hardware and designing the right data paths.
I have serious concerns about if you're going to release your own router software. Software security in a router/firewall appliance is no joke and while you seem well intentioned, I do have concerns about the financial resources needed to continue to update the software to patch the different packages appropriately as vulnerabilities are discovered.
Free updates for life are already a thing with Pfsense, OPNsense, and Mikrotik.
It amazing that any router gets sold and supported more than a year after launch.
I know this is hard enough as it is but can you clarify if this product will have multi-WAN support and fast VPN support like WireGuard? Will you try to compete with products like the TP-Link ER605 or the TP-Link ER8411, or are you trying to make a "gamer router"?
Is really SDIO that common in m.2 WiFI cards? I only heard of a single M.2 SDIO WiFI module used in rather obscure thin client that in fact the opposite issue - no PCI Express at m.2 wifi port at all (Dell Wyse 3040). Every card I know uses PCIe for signalling. SDIO is common in embedded, soldered in modules used in cheap devices.
I don't think losing SDIO from WiFI port is really an issue.
A lot of tri-radio cards use SDIO for Wi-Fi, yes. Unfortunately. I'd prefer if they didn't :)
please keep that remote root as a recovery
Tutorials would be great
I will buy a Router from you
I'd be more than happy to get one
Thank you!
I WILL fund the operation if building a new router, as long as you can ensure to me that it WILL be open source for its entirety. even if the company grows, I want that everything is well documented, at least the hardware!
what specifically makes the router "high-end" ?
great to hear about the progress
also when it comes to m.2 mechanism What is wrong with standoff and screw
also good conversation on twitter/x about booting linux(the one with kernel panic and me saying i do things with rootfs)(would have put a link but rather not for visibility reasons)
also i hope this will have its last bootloader unlocked unlike phones
Also this inspired my to build a rootFS for this device when i get my hands on it im not a embedded guy but i do like RootFSs
also also also also ewwww device trees, ACPI for the win
The problem with standoffs is that they need to be soldered onto the substrate. But if you want to support multiple lengths of M.2 cards, then they can get in the way. But if you use the latch that we're considering, then you can simply move the latch from one position to another.
@@tomazzaman couldnt you just have threads in PCB and movable standoff(ok i would guess that gets expensive)
also on the topic of devicetrees as someone who want to make a distro for snapdragon X elite laptops i already feel their pain cause you know potentially each laptop may require a special kernel for that laptop if its stage 3 bootloader doesnt support passing the devicetree into the kernel, or compiling as many as possible into the kernel (i hope this gets done at mainline level PLEASE I WANT THOSE LAPTOPS TO SUCCEED ON LINUX)
Maybe, we didn't bother with that because these latches are cheap and easy to work with.
@@tomazzaman i may be a special case(because ima techie) but i hate them in same way i hate every PCIe locking mechanism
Going back to your storage issue, I'm not sure I understand what you are having trouble to source... 1.8v NAND chips are common in the 8GB range... 1.2v is usually the voltage for RAM chips, not NAND flash...
Not really. Try to find a 32GB NAND in 1.8V :)
@@tomazzaman Ah, yep, over 8GB is not anywhere to be found... I think you might be ok with 4x 8GB though, and a small address decoder circuit to drive the chip enable pins (assuming the controller has a large enough address space)...
How about using network storage to connect the flash to the IC?
Wee bit of info - a high end router is not made in software on a SoC style chip with network ports.
You are so cool. 😎
One thing i still dont get and returning to the video and relistening didnt help why would you want to in any case use this as wifi ap(because if you are thinking of buying such a router you likely are already owner of proper APs or are planning to buy proper APs) and imo storage is kinda importsnt for logging and similar
Giving people a choice.
Why not pfsense?
Because it's an ARM Architecture. Opnsense currently does only support x86
@@phiwatec2576 There is a fork of OPNsense for Arm, but it isn't official. He posts on the OPN forums every single update.
In previous update Tomasz mentioned the target os to be pfSense. I guess the linux kernel is used for initial fast prototyping until they get the basic system work.
Zakaj nemrete mesto Linux kernela modificirati illumos kernel???
How are you guys navigating patent issues around everything? Even something as simple as that M.2 latch can be covered under horrible US patent laws and result in a big electronics player like Asus coming around demanding payment.
We're not. Nothing we're doing violates any patents so far. At least not that we're aware of any.
I don't understand the need for two wireless network interfaces. I doubt the target audience for this router will put a Wi-Fi card and use it as an all in one solution ala cheap tp-link.
My point is one slow slot for "some radio" should be enough.
It's all about choice (and best practices). You can prefer tri-radio for home automation and the wifi is mostly for the iot. Or, you can prefer high-speed wi-fi 6 (and even 7). These two don't come in the same device, unfortunately.
@@tomazzaman The 3-radio slot makes sense. I don't understand the replacement of storage slot for fast WiFi. I can imagine more use cases for bigger/faster storage (pcap, logging, etc.) than for fast WiFi.
@@hipantcii both have their pros, cons and use cases, we just decided to focus on networking rather than storage. Storage can still be expanded via a USB 3.0 port on the back, if necessary.
Zakaj k vragu nisu mogli napraviti čip z više interne ROM i vse te uklopiti v jeden bootloader? Tri bootloadera - ampak je te fakat bedastoča...
Because each bootloader is responsible for their own thing. I guess.
@@tomazzaman ma zakomplicirali su celi proces podizanja bez veze!
This is a cool project but I don’t understand the use of the word “high end”, modern high performance routers are now commonly 800G per port, with tens of Tbps and billions of packets per second. 🤔
By the time you come out with a full on working prototype the wifi standards will have changed lol. Wifi 8 will be out probably in the next 2 years .Then you'll need something that can support speeds up to 200Gbs.
Fun project but i only see a future when there is a proven past. Any router marketed as a bit high end must have a track record of non stop software updates and bugfixes. Recompiles with all parts renewed on a regular base. And no just providing sources to do compile yourself is only for a very small public. Just making the product even if it is very good is just half of the marathon. The support in the years following has the same weight in importance. I don't say it is not possible. I have a functioning pfSense setup so i can wait and watch the progress. I am interested but not convinced yet. The reason i go for pfSense over OpnSense has to do that every minor fix can be installed and uninstalled at will without any knowledge of managing firmware or sources. Just as organised updates in the web interface. Mayor builds contain the feature enhancements and mandatory fixes over a period so a version update is more permanent and the chain of fixes strats to grow again from scratch. I am not bashing on OpnSense, pfSense just has my preference.
I understand completely. But, well, you can start a project with half of the things done. It always starts with 0 :)
I know your new to this, but everyone's documentation sucks. No one puts time into docs. Several times, I've had to go look at the verilogic code to know how my idiot coworkers changed things. (and yes, 90% of that code is the power sequencing to get everything to boot correctly. I can't see your plan of chaining one rail off another not being a nightmare.)
An american mind in the body of a slovenian.
Glede Linux jedra svetujem sledeče - ne izberi preempt-rt patcha, si boš prihranil ogromno časa in živcev. Imam slabe izkušnje z preempt-rt jedrom na x86 in še slabše na ARM arhitekturah
Regarding the Linux kernel configuration I suggest - avoid the preempt-rt patch, you will save yourself a lot of time and nerves. I have a bad experience with running a preempt-rt kernel on the x86, and even worse on the ARM architectures.
This is starting to sound like a lost cause. I supported this at first, but cost and time is starting to take it toll when such a board already exists. So I think I'm just going to go with the youyeetoo router board.
hands on video plz.
Maybe I'm missing something here, but why do you need to manufacture your own hardware, when you can turn literally any existing PC running Linux into a router?
By this logic, because there are plenty of old PCs out there, routers shouldn't exist? :)
They can exist - so much Cisco, Palo Alto and Juniper, to name a few, have shifted to x86 instead of their old PowerPC/ARM devices for anything more capable than a potato.
@@hquest the x86 processor isn't doing the actual routing and switching.
@@adampope5107 True, the ASICs do the R&S job, but the x86 is the one programming the ASICs to do their job.
@@hquest yeah, so it doesn't have to be a particularly robust processor. It's only handling control plane traffic. I'd imagine they're not going with an x86 due to power and/or cost reasons.
Zaman se trudiš 😢