Hi ...

Question:

Under Winbox (ROS 5.2) if we change some parameter of a R52Hn, at the bottom of the interface pane cames a short message: ... initializing.

Well, this means what? ROS is initializing (again) the driver under a new config? Or this series of RB cards (52n, 52Hn, 52nM ...) are empty and "programmed" on the fly by the driver?

The reason I'm asking this: a R52Hn that works ok most of the time but under certain circunstances shows high noise floor (alternating -115 <> -98), freezes if a speed test is realized from a client (Btest), sometimes "snooper" works, sometimes cause router to reboot.

BTW, it's in test on a bench X86 router.

Regards;

Follow up:

Investigating a bit more the R52Hn. This is a backup card and was sealed for almost 6 months since I brought. When I tried to put it in service last november I had problems because the noise floor locked at -97dBm and makes clients disconnect. I was at this time replacing a R52n in use at one RB433UAH. Always the latest ROS (5.2 now, 4.X then).

So I thought: a defective unit. But ... I brought 3, 2 was in service. A 33% failure rate?

I left this card aside. No repair, but which part of hw can be intermitent & influencing nf readings? LNA? Antenna switch? Noisy pll?

Following MT guys sugestions, the periodic calibration should take care of this nf figure after 1 minute (default cal interval = 1 minute). If not, manually I should set the noise floor threshold to a "standard" value (-110 ... -115) and repeat the tests.

No success. Low throughput, router rebooting, suddenly the bandscan stops showing APs, snooper sometimes works, sometimes not and router reboots after that. I gave up of this R52Hn so.

Then I ordered a pair of R52nM. Since I still using legacy modes on 2.4, I want to use a N card with two antennas & both chains enabled to have kind of spatial diversity. Works nice on R52n btw. And the nM was choosed because its low power sink (~1.8W).

On my X86 PC with ROS 5.2 I did a test with an old OEM AR5211 based card. 802.11B ... all functions ok, alignment, bandscan, snooper, as AP, client, etc, no locks, no reboot. Just fine, udp btest 5.8 ... 6.2Mbps single direction with ping to client pc rising from 5ms to 600ms under btest. 1% CPU.

Removed this card an decided to try one of the fresh R52nM.

As a new interface, it went up disabled. I set its basic parameters, folowing my APs standard:

- 2.4 only G ap-bridge
- Rates configured (6M basic, 6, 9, 12 ... 54 suported)
- Distance = 1Km
- tx/rx chain = 0
- hw retries = 7

... then enabled the interface. Test notebook delayed a bit to connect but succeded. Btest on it, router rebooted.

Again, using eth, tried to snoop, bandscan, freq usage etc ... and sometimes router reboot, sometimes interface simply does not received any surrounding AP signal, nf alternating from -98 to -115.

Ops! R52nM & R52Hn with the same behaviour?

The second R52nM was installed. This time I tried a different aproach. Selected "2.4 B" and configured as above, except the rates (now 1M basic, 1, 2, 5.5, 11 suported).

Bandscan ok, snooper ok, btest ok, fast connection from notebook, freq usage ok.

Well, if it works on B mode, why not on G mode? Switched to G and ... all work ok, 36M/36M, udp btest ~ 13Mbps, ping times under btest less than 5ms, cpu 2%.

Ok ... this card is healthy and I have one R52nM & one R52Hn damaged.

As I live in the "jungle" and we don't find R52 at the supermarket, I decided to give both cards a second "chance", starting with B modes.

Same thing, both caused the router to reboot etc. Then ... why not simply use winbox "reset configuration" and start everything from zero? So I did.

Surprize: the "defective" R52nM came to life, no problem at all.

Replaced it by the "defective" R52Hn. Well, this guy still exibiting -97 as noise floor, despite of the enabled cal routines. Crazy connect speeds from laptop, even side by side with the X86, -50dBm signal. Set NF manually to avereage noise floor figures I saw on R52nM (-112). Then laptop connects ok.

Tried bandscan ... ok, freq usage ... ok, snoop ... router rebooted (the X86 power supply fan even reduced speed a bit).

Again, set NF to "auto" and there was -97.

I used winbox "reset configuration" again. Set up G modes again, ap-bridge. Status shows the -97dBm noise floor. Wait 5 minutes to see if it will "calibrate it" to -11XdBm. No luck.

After any other try I run freq usage, because I know that noise floor readings appear there. Well, started with -97dBm noise flor but a pair of seconds later this figure dropped to -113.

When I stopped freq usage and watch status pane, -97 became -112 ... -113 and still like that.

After that, everything run ok. No more reboots, btest ok, etc etc. Overnight btest, no problem at all. Rebooted the router, no problem. Swapped all 3 cards, no problems.

*** This cards store any settings on it ??? Like some hardware (e.g. some nics, sound cards & etc) that on PNP OSes are configured when installed and keep some info like bus address & irqs on it ??? Others require bios setup to be configured as NO-PNP because before OS rise up bios must configure such hw ??? ***