I have an SR71-15 in a RB600A running ROS 4.6.

The Ubiquiti datasheet lists power specs for the various modes. (http://ubnt.com/downloads/sr71_15_ds.pdf)

For example, MCS7/MCS15 lists 19dBm. But in ROS the default power rates listed for the card put MCS7/MCS15 at 13dBm.

Can I safely raise this power to 19dBm or is there something going on with how ROS is reporting the power levels that I should be aware of?

I need to know this as well... possibly the reason we have major issues with N working reliably ? I have asked this in the UBNT forums as well and never got any straight answers other than use the default... but it doesnt look right to me.

I asked at the UBNT forum and they said to ask Mikrotik. I'm having a hard time understanding the 802.11n competely. Here's what I do know.

The same link using the same boards and same ROS version produces a stable -55dBm link with a pair of XR5s @ 48mbps. ROS says that speed is +24dBm on my XR5.

At MCS7 on a pair of SR71-15s, my power levels are -65dBm or 10dBm less.

The 24dBm of the XR5s minus the 10dBm difference between the two cards would put my output power of my SR71s at 14dBm which is close to the 13dBm the ROS interface says.

There are alot of assumptions in my reasoning, but perhaps if all this is true then I can do a manual overide to 18 or 19dBm without harming the radios?

edit: I'm only using one chain so perhaps this affects my power output?

Using only one chain should be a 3dB drop in tx and rx regardless of the brand card or power output.

oldman u're the king

i still havent gotten an answer on the SR71-15 offsets... some people say 10db, you ask the mfg and they are keeping it secret. what is the hardware offset on the sr71 cards? anyone?

its 6 dBs, on xr's card is 10db

can you point me to anything written by UBNT that shows this? how do you know it's 6db? If thats correct I would love it, but I've now heard 6db and 10db and 0db.

if you want to be sure u must ask in the ubiquiti forum to Mike specifically, but in XR's cards its showed the correct information,but in other brand doesnt,

if you want to be sure u must ask in the ubiquiti forum to Mike specifically, but in XR's cards its showed the correct information,but in other brand doesnt,

sorry, but mike is not helpful in this matter. i already started a thread over there and his answer was "ask mikrotik" which is why this thread exists.

no sooner do I say that than I find out he answers.

http://www.ubnt.com/forum/showthread.php?t=19121

u are confused changeip, this chip are controlled by the driver and if you want to power up, manufactors adds RF amplifier in the rfoutput of these chips. so in some card's eeeprom is corrected but in other doesnt, but the real rf output is that the RF amplifier give to the antenna

for examples in the xr's it has a 10dbs offset so will show you as a 18dbm but really is 18+10=28

im not as confused as you think : )

routeros understands the offset for certain cards. the sr71 is not one of them I believe. If it was the tx powers column would list it both numbers. this means if I use the default of 19dbm, add in the 7db offset, then I'm at 26dbm. Correct? If that is so, then the default settings aren't driving the card to its fullest.

for examples in the xr's it has a 10dbs offset so will show you as a 18dbm but really is 18+10=28

Here is a screenshot of the XR5 and how routeros knows it has an offset:



Funny thing though, ros is using 17db + 10db = 27db, when the specs on the card actually shows 28db. Does this mean the defaults with the xr5 aren't using the full capability of the card?

ive XR2 and another engenius 600mw in 3.30 ros



This is a XR2 tx power with the eeprom modified showin the correct TX power (but only showing)



this one is another brand 28 dbm (600mw) but in the tx power not showing as 28 but I know that it is works at 600mW 18+10 dbs offset =28dbm

OK... so to the point. How do we fix our SR71-15 cards to account for the offest issue?

Thats what Im trying to get to the bottom of : ) The UBNT data sheet shows 25dmb and their website states 27dbm... so whichever one it is, set default power to that minus 7. UBNT does a great job with their marketing, but damn their technical details suck.

Thats what Im trying to get to the bottom of : ) The UBNT data sheet shows 25dmb and their website states 27dbm... so whichever one it is, set default power to that minus 7. UBNT does a great job with their marketing, but damn their technical details suck.

changeip, you are not understood well what "offset" means, the output that shows you in "tx power" is not the real, the real one tx + 7dBs. In terms of electronics, the chip handles the part of RF output but it is connected to another amplifier and that is what generates the offset.

so this graphics means
Rate tx power realtx power
6Mbps 19 19+7=26dbm~398.1mW +-2dbm
9Mbps 19 19+7=26dmb
36Mbps 18 18+7=25dbm~316.2mw
54Mbps 15 15+7=22dbm~158.5mW
HT20-1 18 18+7=25dbm~316.2mw
HT20-2 18 18+7=25dbm~316.2mw
HT20-3 17 17+7=24dBm~251.1mW
HT20-4 16 16+7=23dbm~199.5mw
HT20-5 15 15+7=22dbm~158.5mW
HT20-6 15 15+7=22dbm~158.5mW
HT20-7 14 14+7=21dbm~125mW
HT20-8 13 13+7=20dbm=100mW
HT40-1 18 18+7=25dbm~316.2mw
HT40-2 18 18+7=25dbm~316.2mw
HT40-3 17 17+7=24dBm~251.1mW
HT40-4 16 16+7=23dbm~199.5mw
HT40-5 15 15+7=22dbm~158.5mW
HT40-6 15 15+7=22dbm~158.5mW
HT40-7 14 14+7=21dbm~125mW
HT40-8 13 13+7=20dbm=100mW

I hope u will undertaand with this table

some cards doesnt saved in the eeprom the correct value.

correct, i do understand that. routeros defaults to 19dbm, which puts the cards output at 19+7=26dnm. Correct? Depending on which spec sheet at UBNT you believe it's either under 1db or over 1db. Their main page shows its a 500mw card at 27dbm, but their spec sheet says its 26dbm, which is ~400mw. So which is it? Either way the default in RouterOS is wrong for this card, correct? It should either be using 18 or 20, not 19.

correct, i do understand that. routeros defaults to 19dbm, which puts the cards output at 19+7=26dnm. Correct? Depending on which spec sheet at UBNT you believe it's either under 1db or over 1db. Their main page shows its a 500mw card at 27dbm, but their spec sheet says its 26dbm, which is ~400mw. So which is it? Either way the default in RouterOS is wrong for this card, correct? It should either be using 18 or 20, not 19.

I understand u, but it says Avg. TX Power 27dBm, +/-2dB, and also when u comparing 100mW to 1000mW=1 the relation is 10 times, but as the RF poweroutput is logaritmict 100mW=20dBm and 1000mW=1W=30dBm so the relation is only 1.5 times so its no so much a few dbs +-

power=10Log(P/1mW)
P is express in mw

but it is 100mw that we're talking about ... enough to overdrive the card if its too much. 400mw vs 500mw.

From Mike @ ubnt (hope he doesnt mind I share this, its in their forums as well):

The card has a 7dB offset. You should set the TX power to 20dBm in the MikroTik as I do not think they account for the offset. however you CANNOT set all data rates to 27dB as there is required backoff for the higher MCS rates.

Thanks,

Mike

OK, so now how do we properly compensate. There doesn't seem to be a really good way to change the MCS power settings in winbox. Is this a command line fix only?

There is real TX-power measurement for XR5. Add 10dB attenuator value and ~2dB pigtail attenuation to measured power value.

For 54mbit rate

if you use 10Mhz, and 5 mhz channel width how much dbs do you gain?

I need to know this as well... possibly the reason we have major issues with N working reliably ? I have asked this in the UBNT forums as well and never got any straight answers other than use the default... but it doesnt look right to me.

I don't want this thread to die...!

Had a working 31km link (19.2miles) (two RB600s@533MHz, Ubiquity XR5 (600mW) at the "bridge" end, Ubiquity SR5 (400mW) at the "station-wds" end) -63/-61dBm, 100/100CCQ, 54/54 TX/RX rate... everything was working OK, except that I needed more bandwidth.
I replaced the XR5/SR5 pair with two SR71-15 cards.
I was expecting the same signal strengths, but they dropped to -68/-68 dBm, 91/92CCQ.
I left them in 1X1 configuration until the ordered dual pol dishes arrive.
At 20MHz channel width they connected at 58-HT/58-HT...(if I recall correctly)
It wasn't much of a progress...
Then I set the HT Extension Channel to "above control" and got 108-HT/108-HT... but the real TCP bandwidth was maxed out at 38Mbps (half-duplex) and the link wasn't stable at all.
I ended up putting these cards back to 802.11a standard. (5GHz, non-N)

I'm puzzled too with these offset values... Mine are different!
So, this is a SR71-15 miniPCI card with a newer chip (AR9220)... RB600, ROS 4.5, 5GHz-a/n



As you can see, there are NO differences between the columns!
Why am I getting different readings?

Regards

I was in your position with the RB600 and moving from XR5 to SR71-15. After much legwork I found that RB600 is the bottleneck on TCP speeds. When I moved to RB411AH platform I started getting 30mbps symmetrical TCP traffic on a 40MHz channel. This was with a 26mi PtP link (single chain.)

So I suggest that you may have two problems. First, the power offset issue is obviously impacting your signal strength. But, secondly, the RB600 is also limiting your throughput. The only solution is either to go to a x86 stand alone PC build (get Core2 system, SATA to CF adapter, PCI to MPCI adapter, etc) or get an RB800 or RB411AH/RB433AH solution if you want to stay with relatively low-power consumption. If you are doing dual-chain I suggest either RB800 or use a PC as RB4xxAH might not be able to keep up. (I have not tested dual chain)

There has been some discussion about the offset issue and over at ubnt.com, Mike has said you can safely raise your power levels on the SR17-15 cards over what ROS states as the card's defaults (i.e., 13db @ MCS7/HT8.) I have experimented with this and found that raising your MCS power levels in winbox is not very easy to accomplish. You basically need to set your values to manual and go into the terminal and hard code by hand if you want to affect the correct power levels for each modulation level. You cannot manually hard code your power values to be in line with UBNT specs through winbox at this time (plz fix this microtik, kthx.) If you just have to do it rhough winbox then you can manually override all power levels but you must be careful since higher MCS cannot operate at the approved power levels that lower MCS can so be mindful of that.

Also, I found that moving from an XR5 solution to an 802.11n (be it R5nH or SR71) resulted in some sacrifice of CCQ. XR5 solution was rock solid 100% CCQ but SR71 can fall into the high 80s and low 90s percentiles under a full load. Another thing I noticed going from XR5 to SR71 is that my power level reading in ROS fluctuate GREATLY which I feel is probably a bug or something since link always holds steady. When I say fluctuates, I mean bouncing around from -15 to like -98 in terms of readings. I just have to ignore it and pay attention to overall readings.

Last thing I will mention that I observed is that under full load bandwidth test I end up in the high 80s/low 90s CCQ but after a few minutes of stress testing the cards or ROS completely wig out and CCQ drops into the 40s/50s and the cards downshift and the whole thing becomes a mess. I have to cancel the test and everything goes back to normal. Not sure what that is all about but I see it happen frequently.

Oh, and I would move to the new 5.0 beta firmware. I got a performance boost when I did that. Also, play around with your nstreme framing modes and interference avoidance settings. I managed to squeeze out some extra speed doing so.

I cant resist getting into this fray. There is a fundamental issue that seems to be overlooked (forgive me if I missed it up this thread).

In what context could Real Power ever be in to make this seemingly obvious metric? More like Unreal Power

Tx Power --- Real Power --- Total Power SAY WHAT!!!

Power levels like 28(18)? If its 28dBm, why do I care if there is an offset? I must be missing something. Cant really do anything about it unless I really want to mess things up.

If you set the power on a very close device to 8dBm, guess what happens? I want <10dBm often in some closeby (<1mi) connections. Depending on the rev level of ROS, different things happen. The power amplifier aspect of this conceivably brought some light to this topic, but again, what can we do about it? Someone mentioned running one chain (!). Well thats special, even though I have had to do it for some estoric reasons, specifically using the Spectral Scan. This thread just proves that MIMO isnt ready for primetime.

Lest you think I havent used it......I have 1500 foot link from 700ft above a baseball field (from across the street from a skyscraper). Card is rated 300Mbps. I am so happy to achieve the 85Mbps downstream that is actually stable, and 100/100 CCQ to boot--it acutally streams video very successfully! Bitrates achieved by lowering configured rates and power made this work. However, it seems that the nature of the MAC/PHY mechanism is such that it leverages anomalies, deviations, and bouncing signals, at least for the spatial diversity of 80211.n. It just doesnt seem to work very well in perfect LOS environments. IMHO, there isnt anything to fix or leverage. That is far too simplistic, but it makes me wonder why it is so hard, not only to tell what power you are running (including the estoric knowledge that twice the power is 3dB), and the new alphabet soup we have been accosted with, that it takes about ten times the time to get only a marginal increase in speed. Those manufacturers who package radios dont seem to have done much better. I want ROS control of my devices, but I can see why they are attempting to package it--hide what either no-one understands or doesnt really work.

The only MIMO-like radio I have ever seen operate properly (forgive the slight off topic comment) is the 'Pre-Wimax', venerable, reliable, ridiculously expensive, and now almost ubiquitous (no pun) legacy Orthogon. I accidently broke the warranty seal on one of those gems. Wear your sunglasses if you ever dare to attempt such a thing. You might even hear chiors singing, and bells ringing. There is so much well thought out precision hardware in there, you finally realize why it costs so much, and why it can only stop working with an Act of God such as lightening. No wonder the twenty, at least 5 year old links, are still hummin along, and passing data exactly as provisioned and expected. Wonder why they got bought out.......