Upcoming Maintenance Alert:

The UBNT Community will be upgraded at 5pm MDT on April 25th. During this time the community forums will be set to read-only status.

Learn more

×
Reply
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

forum.mikrotik.com/viewtopic.php?p=323323#p323323
read the entire thread
this isn't a game about your manhood littlebill
-sent using tapatalk


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Established Member
Posts: 1,175
Registered: ‎06-18-2010
Kudos: 196
Solutions: 10

Re: Mikrotik RB750UP fatally flawed

no its apparently about your lack of something... thus making it up with your God complex.
i stand by what i said. i never said there weren't issue's
Regular Member
Posts: 360
Registered: ‎12-11-2009
Kudos: 16
Solutions: 2

Re: Mikrotik RB750UP fatally flawed

just to make it clear: MTK's power supply and POE injector are both at our control box, from there we have a 280' CAT6 cable run to the 750UP, the 750UP is fed via POE, which in turn feeds a Bullet 2.4 on a 110' CAT5(the Bullet is fed via POE by the the 750UP). So, the 24v 2.5a is feeding both the MTK and UBNT.
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

just to make it clear: MTK's power supply and POE injector are both at our control box, from there we have a 280' CAT6 cable run to the 750UP, the 750UP is fed via POE, which in turn feeds a Bullet 2.4 on a 110' CAT5(the Bullet is fed via POE by the the 750UP). So, the 24v 2.5a is feeding both the MTK and UBNT.


That is a REALLY long link for POE. Just throwing that out there.


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

no its apparently about your lack of something... thus making it up with your God complex.
i stand by what i said. i never said there weren't issue's

lol you are impossible. There's a couple of EEs talking about the flawed design of the RB750 circuitry on the poe input on that thread, if you would put your pride aside and actually read it.
5.18 addressed *some* issues, but didn't fix the underlying problem. That has been fixed, in many situations, by replacing the factory capacitors on the board.
These issues have also been talked about here on this forum.
What I'm getting at, and it may or may not apply to the OP in this situation due to the length of his run, is that there are documented problems in both the software and the hardware of the RB750 when it comes to powering the device off the POE port.
A firmware patch cannot fix the hardware part of the issue, and any software adjustments are band-aids at best.
If you'd spend more time researching some of the things you talk about, and were a bit less aggressive, many here would see you more as "helpful".
I don't "dislike" you, and often agree with you in some situations, but you need to be less "sand paper" and be more "cocoa butter" when it comes to some of your replies.
Note: This is coming from a combat vet who spent 2 1/2 years fighting the aggressiveness of being in a war zone for an extended period of time. If that doesn't tell you something, nothing will.


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Established Member
Posts: 1,175
Registered: ‎06-18-2010
Kudos: 196
Solutions: 10

Re: Mikrotik RB750UP fatally flawed

lol there is a poster above you that proves it works, with some pretty impressive lengths, yet you continue to try to say it doesn't work. the 750 is definitely a YMMV situation, it obviously works for some but not for others its not a all or nothing situation
at least your laughing, that's more then i can say for the other "guy" that has issues with me... I just call it like I see it. guess some people don't like that Willy NillyWilly NillyWilly NillyWilly Nilly
you started with "wrong" that bothers me, eh try a nicer a word lol w/e
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

lol there is a poster above you that proves it works, with some pretty impressive lengths, yet you continue to try to say it doesn't work. the 750 is definitely a YMMV situation, it obviously works for some but not for others its not a all or nothing situation

at least your laughing, that's more then i can say for the other "guy" that has issues with me


There is large number, I don't know the percentage, that use a certain type of crappy electrolytic capacitor that is overrated and is prone to failure. Mikrotik swears they are "okay" now, but people are still getting the "newest ones" from their distributors with them in it.

Mikrotik has this habit of completely ignoring issues like they don't exist. If you wanna see "rough" ie like sandpaper, try explaining to normis that you have .. say.. brown hair. He will explain to you, very caustically, that you're ****ing wrong, it's an optical illusion, and that when we die we turn into giant chickens who lay baby velociraptors and then eat them for sustenance.

Those guys are ****ing impossible to deal with, so I just don't even post.

I do probably read that forum as much, if not more, than this one though.


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Established Member
Posts: 1,175
Registered: ‎06-18-2010
Kudos: 196
Solutions: 10

Re: Mikrotik RB750UP fatally flawed

yes i have posted over there several times but it tends to be fewi i deal with. i agree difficult.
eh i just happy we don't get censored on this forum. i hate when i can't speak my mind
i did not know about the cap issue. i did see someone fix a cap a while back, but i also saw alot of sucess stories.
i read enough, alot of times unless i have a reason to learn something i don't learn it. i figure i will run out of space and forget where i live
Established Member
Posts: 1,175
Registered: ‎06-18-2010
Kudos: 196
Solutions: 10

Re: Mikrotik RB750UP fatally flawed

where is this post about defective caps? i went looking don't see it in the 5 minutes of looking
im aware of the cap coxwireless posted. i believe it was posted on ubnt first, but i don't see anything on mikrotik about defective from the factory and miko support actually said cox fix is bad(if it works who cares) so i don't really see the solution
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

do a search for "bad capacitor" on the mikrotik forum. the problem hits rb750up s especially hard because of their poe output capability.
-sent using tapatalk


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Established Member
Posts: 1,025
Registered: ‎02-23-2008
Kudos: 41

Re: Mikrotik RB750UP fatally flawed

wrong. read the mikrotik forum. the problem still exists after 5.18


Yes, it does.

I am waiting, and crossing my fingers, for a firmware fix.

Meanwhile, it looks like I may need to get myself some caps.
Established Member
Posts: 1,000
Registered: ‎07-13-2008
Kudos: 129

Re: Mikrotik RB750UP fatally flawed

hi,
I have now purchased a 25W 10 ohm and 100 ohm pot to stick inline with the bench PSU here. this will be my variable line length simulator.
I have some Ceramic 47UF caps and also some tad expensive 22000uf pulse caps as well as some standard long life 4700uf axial electrolytics.
this is just to resolve problems with live kit by finding out which solution works.
The RB750UP has a micro controller on board, this has an 11 bit ADC and the I/O pins are all muxed to it. They measure the input volts after the power cut switch and the volts after a current sense resistor bank on the input (5 0.1ohm resistors in parallel).
this is how they measure overall system current by measuring the difference.
on each output they have another small resistor on the input to each port switching transistor. they compare the voltage port side of this with one of the two voltages above to detect current.
Our initial solution was to remove these resistors so there was no measurable current flow per port and therefore, you would think, no false port shut offs from the micro.
This did not work. This means the software issue is a race hazard type issue when measuring the input volts and the port output volts with the spikey nature of a transmitting/receiving rockets power draw etc..
as i mentioned the micro has a mux, this means to measure voltages you need to select the ADC to the right pin, allow it to settle then take a measurement. Then you can move on to the next measurement on the next port etc..
My thinking, as a designer, is that they are not taking measurements of voltages the right way.
They should be doing :-
measure input volts after main current sense.
measure each port output volts then remeasure input volts before measuring the next one.
this way the differential will be correct if measured fast enough that the input voltage does not have time to change.
What i think the 8 year old programmer has done is this :-
Measure input volts once. store it..
measure port voltages in a loop, disable port if overload detected.

So on a POE fed input , the input volts will drop correspondingly with the load from the device attached, say a now transmitting rocket, due to the impedance of the cat5/6.
the stupid code sees this voltage drop and compares it with the stored value and says "volt drop excessive, must be an overload".
From this, the simple cure is to kick the 8 year old off the dev PC and let somebody who has a clue about impedance and power sort it. while they are at it, they need a power circuit designer too, under the scope the layout is god awful and the lack of understanding is in avundance.
Of course it has been months now since the problem was flagged up and Normis (is that Latvian for moron?) keeps on telling people "we will fix it soon".
for our solar type apps we also bypass ALL the diodes as the input power DC jack or POE has a diode in series . this is because all the people who use their kit are assumed to be morons and need this "protection".
each port has a diode too, short them out. this way you don't lose a volt or so of precious power for off grid setups.
If of course you wire em up wrong or plug in power back into the other ports then perhaps you are the moron they assume we all are Man Happy
Although strapping C across the input will probably fudge the problem, the cure should be the code. Another of course is possibly to lump some large C across the input ports to the micro so the spikes are not seen.
With the ubi poe switches, i do hope they can provide a user port, a single on/off bit of output would be useful (like the user led on mt stuff).
this would allow those with power budget constraints to use that (as we do the user led output) to feed an external dump load controller . this way it becomes a charge controller too Man Happy
Will, when i get time, do the tests on the bench to reproduce the power problem and report back on what works best as a fudge.
P.S. 47UF ceramic caps strapped across the existing input caps on standard RB750's work well to stop them blowing up due to inrush. Oh and a couple of 100uf ceramics (6.3V) replacing the electrolytic on the output of the switcher.
Bill Lewis
Kijoma Broadband
UK 5 GHz WISP - ISPA UK, RIPE Member
Senior Design Eng , RF / Network Eng.
Highlighted
Established Member
Posts: 1,025
Registered: ‎02-23-2008
Kudos: 41

Re: Mikrotik RB750UP fatally flawed



Will, when i get time, do the tests on the bench to reproduce the power problem and report back on what works best as a fudge.


I look forward to it.

For me, to goal is to have an easy to implement solution, even if it means spending a couple of dollars extra in parts.
Established Member
Posts: 1,025
Registered: ‎02-23-2008
Kudos: 41

Re: Mikrotik RB750UP fatally flawed

kijoma, anything new to report yet?
Established Member
Posts: 1,000
Registered: ‎07-13-2008
Kudos: 129

Re: Mikrotik RB750UP fatally flawed

hi,
if it would stop raining here then we might be able to put some time into this.
At the moment i have it all connected up to a bench PSU, 100 OHm wirewound pot in series and although no definitve tests (i.e. i haven't had time to hook up some rockets to it and set them so i can transfer stuff to/from em from the 750), i have hooked up two 20000 uf Pulse caps in series (they are 12V) and stuck them across the input of the 750.
interestingly i can cut the power for about 3-4 seconds and restore it without the router even blinking.
So this may be a bit OTT and they are expensive (in the UK at least as we have to import em from the US) but if it works, I may use em anyway as we bought 10 of em.
They are small and fit easily inside the 750 case.
Anyway, am wet and need a drink as just got in from a rather wet day up on the picker. ad its gone 9pm here Man Surprised
Bill Lewis
Kijoma Broadband
UK 5 GHz WISP - ISPA UK, RIPE Member
Senior Design Eng , RF / Network Eng.
Established Member
Posts: 1,000
Registered: ‎07-13-2008
Kudos: 129

Re: Mikrotik RB750UP fatally flawed

Hi All,
Right, i have spent some of my Sunday here in the UK analyzing the RB750UP problem.
The test setup.
RB750UP - Thurlby Bench power supply set to 24VDC - 10 Ohm Wirewound pot inline with + feed (to emulate different cat5/6 lengths). The power coming in via a cat5 plug on port 1, as if it was coming via a run of cable.
Two Rocket M5's plugged into port 3 and 5 respectively. laptop into port 2
Too simulate high traffic i used the internal bi directional speed test in one rocket and tested to the other over the air.
Capacitors purchased to test with, one 4700uf/25V axial electrolytic. Two AVX 22000 uf Pulse caps (12V type and in series). The latter easily fit inside the 750UP box, however are quite costly.
POE on each rocket port set to "on".
Here are the results.
First i tested the rig with the pot fully closed and no mods/caps, ran a load of tests and no issues occured. full bandwidth each way and nothing turned off etc..
So as already known, with a low impedance DC source the device is stable.
Next was to set the pot to a typical 50 Metre length of cat5 - 26AWG (worse case cable). this is typically 40 ohms per 1000ft , therefore about 13.3 ohms per 100M , so i went for about 6.6 ohms, measured on the pot with the power off on a DMM.
At this length the two rockets would not stay powered up even when idle. the PoE logic on the 750 kept turning them off and eventually left one of them off.
NOTE: i can validly use the resistance of a single core here as normal PoE usee the blue/brown pairs so you have half the resistance of a single core per pair, however you have two pairs in series so the equivalent loop resistance is the same as a single length of one core.
I didn't bother to test longer lengths (higher resistance) at this point as it was futile.
First i added the sledgehammer, the two 22000 uf AVX pulse caps in series. one end soldered to the ethernet socket array screen for ground and the other end to the 750 side of the input series diode.
This stabilized the unit at the 50 metre length. the rockets would stay on and even after numerous speed tests between them it all remained stable.
So they work! . next i increased the resistance bit by bit and re-run the tests. Once i went past 7.5 ohms the 750 started to drop one rocket under load.
I then swapped the cap with the 4700uf/25V electrolytic one.
This also continued to work at the same points and fall over past 7.5 ohms. This shows that 4700uF is ample to slow the rate of change sufficiently to keep things alive at 7.5 ohms or less. Any more C appears no benefit.
So the summary of those tests shows that the 750UP can be used if 4700uf is strapped across the input when run from 24V at the base end as long as the cat5 run is typically under 55 meters or a bit longer for 24awg cat6.
This of course assumes just two rocket M5's are attached and in use and with Full TX/RX traffic. More radios not being hammered throughput wise or lower power radios should mean you can use all 4 poe ports.
Other things i tried:-
Hanging some extra capacitance on the A2D inputs on the micro. that didn't help and it just cycled resetting the poe ports.
There is a zener diode which looks to be used as a reference for the micro, i lobbed a cap across it, it made no difference.
I shorted out the current sense resistors on the main input side (5 or so in parallel) , this made no difference either.
One last thing to try is to bypass the current sense resistors on the PoE ports and see if this changes the stability points. I doubt it tbh as this is the first thing we did to other 750UPs and they still have the issue, hence we are investigating a solution while the MT guys work out how to just switch on/off the ports and stop trying to be so damn clever while treating us as stupid.
The voltage monitor on board is very inaccurate , reads over 1.3volts low on this particular unit. each unit reads differently so the statement by one of the support people that "it is accurate" is the usual crap.
Another interesting thing to note is the main switcher that runs the 750's own electronics has stability issues. especially without the added cap on the input. this can be heard from the buck reg coil.
I will post some pics later and results of shorting out the current sense on the ports.
cheers
Bill Lewis
Kijoma Broadband
UK 5 GHz WISP - ISPA UK, RIPE Member
Senior Design Eng , RF / Network Eng.
Established Member
Posts: 1,000
Registered: ‎07-13-2008
Kudos: 129

Re: Mikrotik RB750UP fatally flawed

Hi,
Removed the current sense resistors from the two ports used for the rockets and can now operate to about 8.7 ohms. at this point the volts on the 750UP drops to ~13V when the rockets are hammering so there isn't much point going beyond this anyway.
linking out the diodes will improve the headroom though.
I suspect at this voltage drop the micro on the 750 is detecting an absolute under voltage and turning off the port(s) under load.
right some pics attached of the PSU, the cheap cap, the uber pulse caps and them in position. Also a close up of the linked out current sense resistors. these are the ones originally marked 220 as per the ones still in place on the other ports.
We will be using the pulse caps as they were expensive, fit neatly and we have them Man Happy
NOTE: NO permission is given to share this information in part or whole on the Mikrotik Forums or to Mikrotik themselves. They would only deny the problem anyway then ban you from their forums.
Anything you do to your router is at your risk blah blah etc..
Bill Lewis
Kijoma Broadband
UK 5 GHz WISP - ISPA UK, RIPE Member
Senior Design Eng , RF / Network Eng.
psu.jpg
cap.jpg
ubercap.jpg
ubercapsinpos.jpg
reslinks.jpg
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

Just a note, I have several 750UPs on 5.18 that are powering 2 Aircam's a piece at up to 150ft per aircam (each), on shireen cat6 FTP (23 AWG). *SO FAR* I have not seen any issues (4 days). I am powering the 750's via the power port with a 24V 1AMP AC/DC converter. I haven't tried pushing data across the lan ports until it croaks, but when I get a chance to I'll let you know.
www.shireeninc.com/wp-content/uploads/2011/08/Specs-DC-20211.pdf
(Note: the boxes suck, and the cable is so large I spent a week scouring the internet for connectors that would fit)
(Additional note: the new Ubiquiti Toughcable Pro has some pretty nice boxes, and is almost identical to Shireen Cat5e FTP in every way).


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Established Member
Posts: 1,000
Registered: ‎07-13-2008
Kudos: 129

Re: Mikrotik RB750UP fatally flawed

hi,
there is no problem if you are powering the 750up locally, it is only if you fit it remotely and feed it via POE.
Bill Lewis
Kijoma Broadband
UK 5 GHz WISP - ISPA UK, RIPE Member
Senior Design Eng , RF / Network Eng.
SuperUser
Posts: 21,696
Registered: ‎11-20-2011
Kudos: 7787
Solutions: 233

Re: Mikrotik RB750UP fatally flawed

i have 41 750ups at this point. it still has issues when being feed via the adapter with smaller awg cable, just not as often as via poe powered.
edit: my latest test run is several posts above.
-sent via tapatalk


isp builder | linux sorcerer | datacenter automation conjurer | blog: blog.engineered.online
link to our slack channel on the blog
Reply