This statement by daniel is what led me to believe that my module might work. Does anybody know of a GPON module that definitely works? I'm open to spending money if needed. I just want that thing running 
I have ten od ma5671a with diferent firmware, i have finaly time to play with my isp, i don't think that will be problems to make this work. I have also three of bpi r4 , one with snapshot and two with ofw. This weekend will be playfull.
But that's not a problem.
That's wrong information. BananaPi restricts their stock firmware and hard-codes USXGMII 10G mode there, for the sake of simplifying support and not having to deal with compatibility issues caused by having to auto-detect the module speed.
For the record, and again, from the author of the driver (me): The hardware supports 1000M, 2500M, 5000M and 10000M modules. Period.
In (real/vanilla) OpenWrt and Linux we do support also 1000M, 2500M and 5000M SFPs with the R4.
That's very true, but also this you can override in software by instead of driving Moddef0 as input driving it as output which will force power on (at the cost of no longer being able to detect module presence/absence in software, hence the module has to be plugged at boot to be detected then).
1 Like
@daniel is flow control supported on the sfp ports?
if supported how do i enable that?
I was working on a PR to add support for that in the Ethernet driver, it's not hard to do, but got distracted by other thing... I may take care of it in the next couple of days.
4 Likes
This makes me very happy
So, if the module is not recognized in latest (yesterday) snapshot, this means that the only thing I have to do is forcing the power to be on?
Does this need a patch and own compilation of the driver or does "software" mean that I can tell the driver to do this from the outside? This could be my first journey into Linux kernel/driver stuff. As a Java guy this might become interesting 
For the full picture: You also said that the stock FW also forces other things - this is not relevant here, then? I can confirm that the module works flawlessly in the hexS
sorry for the misinformation and thank you for the correction
1 Like
Started poking around power consumption a bit. All values measured at the MAIN (so it's a bit overestimate of actual usage due to AC/DC conversion loss):
R4 idle (only serial console attached and boot off the SD card): 4W
Each GbE connected to the RJ45 switch: +0.1W each
USB 3.0 thumb drive inserted: +2W**
USB 2.0 thumb drive inserted: +0.5W
Anyone tried to look at minimising idle power consumption?
MT7988A's Cortex-A73's seem doing a terrific job already managing its idle power. So undervolting the CPUs alone yielded nothing.
Tried to suspend USB controller, two ethernet controllers (for SPF+ slots) yielded nothing. Perhaps already automatically managed well.
Tried to turn off all LEDs and yielded nothing (measurable at the MAIN).
Tried to turn of 'PWM fan' (I don't have a fan actually connected) and yielded nothing.
Any thoughts or successful attempts?
** Haven't measured USB consumptions on embedded systems before and I'm kinda shocked to see USB 3.0 consumes so much power relatively speaking. No wonder hearing so many people complaining about hot steaks^H^H sticks.
Now thinking about it.. so many consumers plug in a USB stick and run some sort of a 'little linux distro'. Be it Entware for a lot of all-in-one consumer routers, or OpenWrt's own idea of 'extroot'. The additional 2W power draw doesn't make it a sound idea. I meant all such recommendation should have added that: stay on USB 2.0 port if no need for higher speed. 
mine chucks along under 5w. measured via usb c power meter.
both sfp+ connected via DAC. and two lan ports in use. boots and runs on sdcard. running on performance cpu governors which mean 1.8GHz fixed.
at this power draw. imho doesn't make sense to try to make it any lower.
cpu temp 45-47 deg in closed factory casing with passive heatsink.
I agree R4 is good on idle power. Nothing need to start worrying about.
Just that I've a retiring MediaTek MT7621 wired router, idling at 2.7W measured at USB meter. So perhaps near 3W at MAIN. Essentially doing the same job for me at the moment as R4 under test.
So I'm not greedy. Hope if we could get R4 down to 3W idle. LOL
for me it replaced a Edgerouter infinity sucking almost 40w. so what is 2w?
2 Likes
Just wondering what data you have got for heating here?
BE14 uses 12V so for sure the heat should be significant in comparison to regular wifi modules on pcie....
Could you please show us some data about it?
1 Like
With the latest snapshot, it drops 0.1W. So now I have, 3.9W
Figured out that I had to switch off devices in 'BIOS' (i.e. Device Tree) to take effect. I don't need the four PCIe slots and the two SPF+ cages for the near future. So I turn them off.
PCIe 0, 1, 2, 3 OFF: -0.6W
SFP 1, 2 OFF: -0.2W
Now, R4 idles 3.1W
Also figured that once I removed the SD card and switched to eMMC, I will get another reduction of 0.1W.
Bravo. Mission accomplished. This cures my OCD. I'm happy up to here @ 3.0W for next year or so (until I figure what expansion cards to add).
Bonus.
For more serious OCD than me, you could get further close to 0.1W drop by idle @800MHz and turn off the built-in USB 2.0 controller dynamically.
The VIA USB chip inside R4 is a smart hub. Premium stuff, LOL. Can use 'hub-ctrl' or 'uhubctl' to turn off PPPS in a smart hub.
So the ultimate lowest idle power at the MAIN shall be around 2.9W for now.
Until OpenWrt snapshot has new & surprise improvement...
1 Like
To cure your OCD, use USB-C PD power, you can see from my previous post, even with some traffic passing I don't even see it using more than 3W.
Where did you get it? I tried a generic northbridge HS from ali but 30mm height is too much for the official case. If anybody knows one that is 20mm high, I'd be interested to hear.
I don't need PCIe either but I really do not feel like building my own images, any chance we get that as a runtime or uboot switch?
Best would be a pure alloy block + heat dissipation pad. Then the complete enclosure would serve as heat pipe.
Agree in principle but it's overkill for the wattage to be dissipated.
2-3W @ 1Gbps should be credible. When I did a casual 1Gbps speedtest.net, only see 0.1-0.2W extra at the MAIN.
AC/DC conversion for the common 12V adaptors with 'Efficiency Level V' marker is only about 60-70% efficiency between 3 to 5 W range.
So for my R4 idle at 3W measured at the MAIN, I would estimate it's about 1.8 - 2.1W at the logic board level.
I have a good USB PD power supply and cable. Type A USB meter. A bit shameful that I don't have a Type C USB meter around.
Mine is 10mm and quite a misery story**. 25mm is about the right size. 30mm may touch the top cover (?).
A heroic member on Bananapi forum tried out Alpha Nova tech heatsink. You may want to look up their LP50 series, available at various height.
Do note that: according to his comment, he also had to perform some gymnastic on the push pins to go down. LP50 series has 60mm hole distance. R4's logic board has ~59.3mm. So it's not a butter smooth fit.
**The one I got is likely fake or rejected OEM batch. It doesn't fit at all ! The holes are 3mm too farther apart. I had to work really hard to trim the 3mm aluminium..