I have considered it before, but I think that 12V PC fans are relatively large, so when the fan rotates, it may get tangled with the coaxial line of the antenna;
However, there should be a demand to make the case larger so that the fan will not interfere with the coaxial line. I will see if I can change one of the two fan connectors.
For me, it looks like Banana Pi is widening the range of router PCBs 
:
I.
Have you ever considered (in some way) decoupling router board development and CPU performance?
For example, by using something like a CM5 board?
II.
Orange Pi awakened my attention by using a replaceable eMMC:
Has this your attention too? 
III.
2x M.2 M-key for NVMe SSDs 
3x M.2 B-key for 5G Modules 
IV.
Introducing the USB-C debug port increased usability!
In the past years, we often had people who encountered problems with the use of DIP switches in general. I didn’t want to question the use of NAND and eMMC.
But I was thinking about how we can make this more usable:
I’m not sure that a layman in their first days would be interested in boot topics so much! They are glad that the system works! Maybe there will be people who never overcome this hurdle?
I’m not really sure how we can solve this without impeding advanced users. But solving the DIP switch topic will cause less frustrated people.
Because some people use these routers as their daily drivers. And changing the boot configuration mistakenly (by mechanical action) could cause time-consuming debugging. And this is not good on a day when you really have something else to do and need internet access.
so, the case will not compatible with BE19?
I. You already know what I’m planning, but heat dissipation will be a challenge
II. Many other users still want to directly onboard eMMC, so this design is not considered for the time being.
III. Also, many people will need 2~3 5G modules
IV. Because it is a development board, we need to consider the different requirements of NAND/SD/eMMC boot, so we set a DIP switch. We also noticed that BPI-R4 is prone to accidental touch, so the switch of R4Pro will be moved a little bit inside the board, and a component that requires more force to switch is used.
And we will suggest to MTK that the ROM Code of the next chip can be set to have a priority boot mode instead of bootStrap, such as SD>NOR>NAND>eMMC
There may be several different versions of the case 
Could you put capacitor and power filters on the 5v and 12v lines, we suspect that due to the lack of power filters and caps on the BPI-R4 it causes higher noise therefore reducing signal strength, if you could add that it would solve the problem
(1) Please remove the SFP+ power mosfets (leaving the power permanently on), or attach them to a separate GPIO pin so that it is not connected to SFP moddef. The current BPI-R4 design is causing issues with several GPON ONU modules that requires hardware changes, documented here:
(2) Please consider adding pads (and maybe accompanying silkscreen) to the SFP+ control signals to make troubleshooting easier. There should be some space under the SFP cage which can be used for that.
Example from mellanox:
The BPI-R4 currently exposes these signals through 0201 resistors which can be tricky to work with
Hi simon
Are there electrical issues (e.g. reduced speed due to weaker contact points) with the socket way? It would make upgrade more easy.
Really?..imho 1 would be enough 
This will be a great change! Bit how to determine boot from emmc if nand/nor is also flashed with valid bootrom? I would take emmc as second to have system there (because of space) and leaving nand/nor as recovery when emmc is broken.
Another great change would be having 2.5g ports on the internal switch to not leave them unused like here and keep the dual-sfp without external switch.
Also adding poe to internal phys would be better that change one sfp to phy only for adding poe.
SD > eMMC > NAND > NOR
Thank you for opening the floor for BPI members to share their comments …
Consider these comments based on my experience with BPI-R4
My point here (the aluminum case design is bad ) where its mentioned previously by BPI this case will absorb the heat but actually its not that much … you can find many case like this
To be honest I don’t see any benefit from adding another two slot for 5G modems ,
Can you combine the speeds of the first 5G modem with the second modem, each having a separate SIM card? If not, then what is the benefit of placing 3 or 5 modems on the board?
My point here : the design of sim slot should be changed (Putting the sim on tray its contain a high of risk to damage the board due to not focusing . it should be replaced with pushing type ( Like Huaweii , ZTE, Nokia ,and etc… all other companies) using this type
I hope these comments can help to improve current builds and new builds
very nice- it looks cool, i like it. my thoughts:
actually many manufactures are more and more interested in poe routers (or “AIO” device). TP_Link ER7212PC, Unifi UDM SE, Mikrotik RB5009Upr… If we can have an openwrt poe AIO device it would be a killer. it would be great for the small-medium business market.
OK，R4Pro will add testpoint to I2C BUS/TX_Disable/MOD_ABS/SFP_LOS/Power，
and of course leaving the power permanently on too.
HI Frank, Sorry for the late reply.
Vias, sockets, and socket instability, etc., each additional item will increase the overall risk. In general, there is no big problem with using sockets. The cost increase caused by adding sockets is not the key, but the defect rate will be slightly higher than that of direct onboard, which is not something that enterprise users who deploy in large quantities would like to see.
Generally, one 5G Module is enough for an individual, but some companies may use two. In China, many people/companies may use three.
These are just some of my personal thoughts. I will suggest MTK later, but it depends on whether they will accept it. 
The latest schematic diagram adds two 10g phys, which are combo with SFP+.
Both internal phys(2.5G WAN) and 10G SFP will be layout at the same time, and switching can be achieved by simply changing DTS in software, without changing hardware
Hi Thank you very much for your suggestion!
My current idea for R4Pro is to design two cases:
no wifi and 5G power consumption is not very high, so the goal is to make it as light and thin as possible, within 200x140mm;
The second one uses wifi/5G, which consumes a lot of power, so the size will be increased, and there may also be a place on the top of the case to install a 12CM standard computer fan.The noise is not too loud and easy to find a replacement.
As for the 5G module part, each module corresponds to a SIM card (it has been replaced with the R3mini’s SIM slot without card tray), which are independent of each other and can do Link Aggregation, But this is not a common scenario for individuals
We have already started to deal with the BE14 issue.
BE19000+ePA has always been a product I wanted to make, but there are too many things to do, so I can’t take care of it for the time being.
We will discussed with MxL to see if can manage Mxl86282 through SFP’s I2C.
If possible, we can make a POE switch for BPI R4Pro, connect it through DAC cable, and use MT7988 to manage it.
I think the r4Pro could be slightly larger to directly expose all Ethernet ports without relying on FPC cables for extension. By integrating all ports, the board would lengthen enough to accommodate the BE19 component securely, avoiding a floating/unstable design. For users preferring a compact size, the r3mini remains a sufficiently small option. The r4Pro’s positioning should focus on addressing the shortcomings of the r4 while enhancing expandability. As for PoE support, this should be handled by dedicated PoE switches—the RT5400PoE’s power output cannot reliably support 48W, making its inclusion here less practical.
I hope if there is better way to access tom EMMC directly And 1 sim card And debian11 image for emmc with all necessary patch to make sim data connection work
