(Solved) WR802N v1 HW upgrade + latest version (?)

I understand, that support for this device as it is ended a while back due to the amount of RAM and flash available (by default).

With the possibility of adding a 64Mbyte or potentially a 128Mbyte ram chip and a 16Mbyte flash chip, would it be possible to compile the firmware for this particular device?

At the end my requirements would be:

  • no IPv6
  • 1 Wireguard interface for very low traffic (less than 1Mb/s)
  • low power consumption
  • wifi AP

I already have the device, and I do understand that there are routers that would be fairly cheap and capable, but I already have the flash chips and sourcing RAM would be also fairly easy, and it is a really good feeling breathing new life into an existing, but somewhat abandoned device.

has anyone done this before? I could find topics where people upgraded RAM and also others where the flash has been upgraded, so I know it's not impossible, but here the challenge for me is that I would also need to compile the firmware.

Any help would be appreciated

There is little-to-no reason to attempt a hardware upgrade of this type given the age and limitations of that device.

Consider that you need to have the right tools (i.e. soldering equipment for SMT parts), a flash programmer device, and the replacement ICs... not sure if you already have all of those, but if you don't, you need to buy them first. Then you need the skills to replace surface mount ICs... done wrong, this will hardware brick the device and thus be a complete loss.

It is much wiser to get a more modern device which can be purchased new or used for fairly cheap. If it needs to be a travel router type form factor, there are several options in this category that are fully supported (although some of them may only be 8/64 devices and thus will not have future support beyond 23.05).

Thank you for the quick answer.
Sorry, your points are totally valid regarding the parts and tools and also the skills to be able to solder the components and if I did not have those I would close this out right now.

Apart from the RAM chip I have all the tools (IC programmer, soldering station), flash chip and I have already done smd soldering, so that part should be fine.

I am interested to see if this is doable from a SW/programming side or something that would be only doable for a seasoned sw dev/veteran. In which case I will try to stay clear of this :slight_smile:

It is certainly doable from a software perspective... but consider that you will be likely starting from a tiny image from 18.06 -- this is very very old, EOL and unsupported. There are many unpatched security vulnerabilities, so consider that this shouldn't actually be used anyway.... but should you decide to go forward, also understand that tou'll be almost entirely on your own in this effort (people may offer some best-effort support if they are so inclined, but don't expect much if anything). This goes for making the changes to the code, compiling, and even configuring the device (it's so old that the syntax has materially changed over the last 6 years).

So...

  • Possible? Yes, likely (assuming there are no hardware limitations about addressing the larger flash and ram space).
  • Good idea/worth it? Not really.
  • Effort? not insignificant, possibly significant.
  • Supported? No.
1 Like

Thank you for the thorough answer, it was the not insignificant effort that I wasn’t sure of.
I will use the device as is in it’s current form, in a fairly secure environment as a wireless ap in times I only need a low power wifi for IoT devices.
I was eager to do something/upgrade it, but at the same time I have plenty of other projects.

Thank you for the clear cut answer! :slight_smile:

No problem.

The truth is that editing the relevant files for the new flash/ram size isn't, in itself, terribly hard. But the reason it's a non-trivial amount of work is that there can be a reasonably steep learning curve for compiling the firmware as a whole (and you'll also have to compile all your desired packages, too... your customized firmware image won't be able to pull from the standard repo) -- and you'll likely be doing it on your own so if you run into problems, it may be extra effort to resolve them.

But -- it is a good learning opportunity, and you could always try your hand at getting the software side working and only then make the effort to upgrade the hardware. I still would recommend against it as I think you'd be better off with more capable hardware from the start... but then again, some people love a challenge (myself included :stuck_out_tongue_winking_eye: )

https://openwrt.org/docs/guide-developer/toolchain/start

1 Like

Flash upgrades are relatively easy, soic8/ soic16 is something you can deal with (parallel flash on older devices or NAND would not be).

RAM upgrades are a topic I would avoid, even if they'd be -desperately- necessary for anything with 32 MB RAM or less. Size- and number of pins is just excessive, finding the right -compatible- chips is non-trivial (and might need several attempts). Desoldering these is hard, so unless you get a new separate chip (and don't have to harvest it from a larger module) and basically cut away (destroy in the the process) the old chip, it's typically not something you can do at home (without a lot of practice, hot-air desoldering station and -again- even more practice).

Software-side, the changes are reasonably 'easy', but as flash size/ partitioning and RAM size are hard coded in the DTS, you will have to keep fixing the software for 'all eternity' into the future - as you won't be able to flash an unmodified firmware again. Care is needed to find a flash chip compatible with the existing bootloader, otherwise you'd have to fix that first (that might be hard), this limits the maximum size to 16 MB on most devices (32 MB flash chips need special support from the bootloader, as its accessing is different). Warning, keep the RAM size in mind for sizing the flash chip, this needs to fit into RAM during upgrades - so a 16 MB flash chip would be hard to fit into the RAM.

Yes, it's possible, if you choose your components wisely and manage the soldering. But keep in mind that it will need constant attention into the future (flash an unmodified firmware and the device bricks), that also implies rebasing your patches regularly (kernel upgrades may break your patches without you noticing, as your patch still applies to the old kernel, while the new -unpatched- kernel is installed).

…and at the end of the day, the device remains old, slow and with slow single-band 802.11n 2.4-GHz-only wireless. I would not touch draft-n/ ar9132 based hardware anymore (only 400 MHz, very fragile wireless, not worth the effort), yours is fortunately newer than that (qca9533) - but the single 100 MBit/s ethernet port is still 'not great'. While you can do it, I would not consider this to be a sensible approach for this kind of hardware (and even less for even older ones). This situation might be different for a tl-wdr3600/ tl-wdr4300 (8-128 --> 16/128, where the flash upgrade would be reasonably easy, RAM sufficient, and the general hardware more sensible for continued use).

Don't even look back at 18.06 for this, current OpenWrt still has all the necessary support - even if it would never fit into 4 MB flash anymore (and is therefore untested!), it has not been actively removed yet (although a clean-up may happen in the future, as we're past the point where simply dropping some packages would make a buildable 4 MB image - and 32 MB RAM does mean sacrificing even more (e.g. WLAN altogether)).

EDIT: given the RAM size (32 MB), I would not go beyond 8 MB flash for this device; if you do -successfully- upgrade the RAM, there wouldn't be any reason not to go 16 MB flash.

BUT, still keep an eye on the economics of all this. Even if you have the soldering equipment (>>100 EUR) and the practice, the spi-nor writer and clamps, with shipping and everything you already get close to 5 EUR (and more) for the components alone - but for that kind of money you can already find more suitable devices on the various used markets (e,g, the aforementioned tl-wdr3600/ tl-wdr4300) in working condition.

1 Like

@psherman: That last link... that dangling carrot :smiley: something I should not open... :smiley: I will go down that rabbit hole, but probably not today. I will try to keep my common sense as I do understand the hurdles (exactly what I was cautious about as that was my gutfeel)

@slh: Thank you for the lengthy explanation, and the insights into the SW part, which suggests to me , that the targets and device details are still in the source code, and given the flash and memory upgrade it should be doable through config updates and building the firmware? (trying to simplify it to the extreme)

As I mentioned, the flash (16Mb) is already given, (which provides plenty of space for the packages I need) ram chips wouldn't be hard to get especially if I do not want to go to the extreme 128Mb (DDR1) and settle for a 64Mb chip... which should be more than plenty for what I would need (no IPv6, LAN port acting as WAN + Wireguard tunnel with very low throughput requirements = IoT devices chirping back to HomeAssistant) 64Mb chips are fairly easy to get from a 512Mb sodimm stick which uses TSOP-66 chips... Or buy the single chip from a reputable site (NOTaliexpress in this case)...
Also soldering is not an issue, so "unfortunately' I am more inclined to set up an environment and check out the firmware building process.. :slight_smile: just when I was about to settle for using the device as a simple low power AP :smiley:

And I do get it... for not much more money (than the components) I could get a GLiNet MT300Nv2 or an AR300M16 and do all of this out of the box and would have a better throughput... and it would also be an overkill to serve as a drop in remote low throughput "jump box".

I think I will set up a VM linux with the firmware-building environment for now and see what I can achieve without falling into the aforementioned rabbit hole :slight_smile:

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.