ACK Timings

[tony]

When adjusting the ACK values, use these guidelines:

Atheros ar5212 cards - default ACK values:
802.11b mode: 450.
802.11a mode: 180.
802.11g mode: 180.

Atheros ar5211 chards - default ACK values:
802.11a mode: 180
802.11b mode: 230

The values above represent the hardware defaults for the cards we have. When you wish to adjust for a longer link, increase the numbers beyond the values shown. Setting the value back to 0 will use the card's default.

For example, if you wish to extend a 5.8 link, try using a value of 400 or more as a starting point.

All values are 1/4 of a microsecond (usec). 100usec equals an ACK value of 400.

[lonnie]

Some numbers to report.

Our test data
At 0.7 km the minimum should be 125
At 1.7 km the minimum should be 150
at 19 km the minimum is 568, suggest 575 or 600

Customer supplied data
At 6.2 mile between 400 and 405.



Nothing precise for longer yet, but I have a report that at 27 km 5,000 did not work and 9,000 did. 9,000 is too high but the customer was not keen on trying every possible setting. When it worked they left it alone.

The symptoms of a value that is too low is the first signal reading is indicating a reasonable number and the second one (which is ACK RSSI) is showing -95. This indicates that the ACK is not getting through and a higher ACK time is required. The two numbers will typically be the same, although the ACK RSSI will sometimes be higher. We are still not sure why they differ, and we are looking into the reason.

[datawork]

played around with the 9000 ack setting last night ranging from 8500-9000.Speed really didn't change much on our 6.2 mile shot. Though, anything below 8600 would break the link.

However, I did play around with our default setting of 755 and found the sweet spot to be between 400 and 405. This change resulted in a 1500KBps increase, giving us a 4500 KBps link

[datawork]

Our 32 DBi Equinox grids just arrived and we'll be putting them up this week. Weather permitting, should have some ten mile numbers by the end of the week.

[CompuTron]

Just thinking about how Star-OS could have a link tuning feature.

I'm thinking, between 2 neighboring boxes, the highest MAC address is elected the master. The master would tell the other box to adjust it's timing and the master would then perform a short bandwidth test. Then, the master would adjust it's timing and do another test. This process would repeat. There would be a fail-safe value where both boxes would return to if they lost communication during the tests for a specified amount of time.

It sounds like incrementing some of these numbers & testing can really take some time when it's the perfect type of task for automation - repeditive.

Travis

[lonnie]

Once we get a bit more data we'll have a simple number in miles like we have for the Orinoco radios.

[bkehoe]

I've just set up a 44KM (27 miles) 5.8 link using the pacific wireless 29dB dishes and, at the remote end a mikrotik a/b minipci card in a mt routerboard, and at this end a 5054 in a standard pc (was supposed to put in a mt a/b card, but mixed them up and havent got a chance to change it yet). Anyway, I've managed to get a signal of around -88 according to staros, qual of around 7 or 8 (10 or 11 according to the ap side). Anyway, was wondering if anyone has any suggestions for the ack timeout setting? 9000 wouldnt work, though 8500 did, as did 8000, and I've put it down to 2000 (approaching MT recommended values for 45KM), but am hesitant to go any lower till i'm actually at the remote site in case i loose the connection to it! ;) However, I'm having a problem with speeds - 6KB/sec approx to be exact. The link is perfectly stable, no droppeed packets, low pings of around 1ms, etc. Because of this i think its an ack issue, though I'm also suspicious of the 5054 card (i've seen reports off people using the MT cards for over 55KM with plenty of bandwidth). So, I was wondering if anyone has any suggestions of things to try before replacing the 5054 with the mt a/b card it should have always been? I also have to fine-tune the antenna at the remote end. Approx 10ft of LMR400 at each end as well. Have shorter pigtails to use as well - just need some connectors for them - there's got to be a good bit of loss through the 18 or so inches of that ultra-thin cable used in the mt ones... Using staros 1.13.3 on the ap end, 1.13.4b2 at the client/remote end.

Thanks,
Brendan

[dkii]

I thought I saw somewhere that the star-os settings should be appx 4x the mikrotik settings, something about the difference in timing schemes. I have a 7 mile link set at 9000(although I know it doesn't haev to be that high, it should be around 700, but I don't want to change anything yey) and am having no problems with speed. I don't think over-shooting the ack setting will cause you speed problems. I picked up about 3db of signal on one end when I soldered a short lmr100 pigtail to the card, in place of the 18" tiny u.fl pigtail. I've seen 6" u.fl pigtails on fab-corp.com if you're not comfortable soldering on the cards.

[bkehoe]

Yep, I thought going over wouldnt affect it as much as going under. I've messed around with different channels, and am now getting a Qual reading of 12 and signal of -83, but with no difference in speeds, so imho it isnt the signal giving the low speeds, leaving either the acks or the 5054 card. I've been going on the 4x MT values as well (they recommend 405us for 45km - x4 = 1620 in staros - wont adjust any lower till I'm in a position to log into the remote end locally though in case it stops working;) ). I've got 2 of the fab 6 inch pigtails sitting here, but I ordered the wrong sma connectors to put on the lmr400, so have to wait on the right ones to come! :( I'll give the fab pigtails a go first before soldering. :)

[szern]

Is it very difficult to solder?
Can anyone please post some enlarged pictures of the soldered Mini-PCI card?

Thanks in advance!

[bkehoe]

Something just came to me there - polarisation. Around 35KM of this link is over water, so has anyone done testing to see what kind of polarisation is best? Fresnel is clear (sites are 1000ft asl and 400ft asl). I'm using vertical polarisation at present.

[georgew]

(they recommend 405us for 45km - x4 = 1620 in staros - wont adjust any lower till I'm in a position to log into the remote end locally though in case it stops working;)

So do what I do.... set the watchdog timer to reboot if the link is down too long, and when you are testing your settings, don't save them, only apply them. When you get one that doesn't work, the box reboots itself, and you are back in business... just don't save settings untill after they are proven to work.

[lonnie]

Please make sure you have the latest beta for testing. It has some tweaks we added for quicker linking and a guaranteed hunt on the client when you change the AP.

I calculate a setting of about 1500 should adequate for 27 miles. Something that helps is to set rate to 6 mbps until you have the antennas aligned and the ACK adjusted.

If you have the ACK too low you will not see them link. They will briefly flash a reading but will continue to hunt for the AP (as evidenced by the Client side showing a channel scan).

[mp3turbo]

yes, bkehoe, I'd definitely try horizontal polarisation. We have seen several problems arising from signal being reflected from water (well, and it was a small river!) - the polarisation changed all that.

bye, mp3turbo.

[bkehoe]

OK, changed the near end to a nice mikrotik a/b card like it should have been and put a 6inch pigtail on it. Quality now up to 21 - 23 and signal of around -73dB. Also changed the antennas to use horizontal polarisation. Upgraded the ap end to the newest beta as well.

Result; not a bit of difference. Still around 6KB/sec speeds. If its any help, they're quite slow to associate, and it sometimes takes a few changes of ack times on either end to get them associated. Once associated, the link is rock solid and doesnt loose any packets and stays up till i reset it.

So, any more suggestions to try before I see if mikrotik works on it? One other possible thing I can think of is that there's a stay wire close to/in the path of one of the dishes, around 10 - 15M away. I doubt this is causing what I'm seeing, but is there any chance it could be it? ;) I'll move the dish to test it if anyone thinks there's a possibility this could be the cause! ;)

Brendan

[bkehoe]

I'm guessing then that nobody has any ideas about what to try then? I'm curious - why does staros have an unlimited ack time setting when mikrotik say that the cards max ack timing is 409us?

[tony]

I'm guessing then that nobody has any ideas about what to try then? I'm curious - why does staros have an unlimited ack time setting when mikrotik say that the cards max ack timing is 409us?

It depends on how you represent as an ack. We allow a higher range as to allow fine tuning while we are still experimenting (values are scaled down support the hardware). Future releases will have a value in 1/2 usec intervals which will have a cap of 409 as a value. (hardware limitation).

Thanks!

[bkehoe]

So do you mean that in the latest beta, entering 409, which is the max that can be entered, the actual ack value will be 204.5usec, or did you change your mind and the values in the latest beta are the actual usec values?

[tony]

The max ack value that can be entered for Atheros hardware is 204.5ms, and the values as entered in StarOS are now in 1/2 usec intervals making a maximum value of 409.

Thanks!

[bkehoe]

Isnt this half of what mikrotik allows?

[bkehoe]

I have just confirmed this. On a link that needs approx 75usec to associate with mikrotik, staros wont associate untill in the region of 150. So, this obviously explains why my 43KM link wont work, as it needs an ack of around 400usec, but staros is only going up to 204.5! So, can you fix it? ;)

[tony]

I checked, and we are allowing a value of 409us in the latest v1.13.4b3 release (enter 818) though it has not been tested.

If it works as expected, please let us know.

Thanks!

[bkehoe]

It works it works!!! I'm just so delighted to see this thing finally working! I had been planning on replacing it with mikrotik and all, so I'm just delighted that its working with staros. Managed to upload the latest beta to the far end (took over an hour), upgraded set ack to 818 and rebooted it while upgrading the near end, and once it rebooted off it went! Getting 500KB/sec over the link now, and hope to do some more tweaking. Rate is showing as 36MBit at present. :D
Thanks a lot tony! :)

[tony]

I am delighted to hear it works!

[gueswhoiam]

This thread is very confusing. What are the acceptable values for ACK for release 1.13.3? In this thread I see numbers all the way up to 9000 posted here and then I see numbers that say the max is 818. In Star-Os it says "usec" next to the field, but then we're suppose to multiple the usec by 2 to get the value to enter. Huh? Can sombody clear this up in one simple message? Thank you.

[lonnie]

We will soon have a number in miles. We are getting enough confidence in the numbers to now make the conversion. For now simply start at about 800 and work down. Or check the numbers near the top of the topic and set near the ones we have have verified.

[bobbyc]

Are the 5.8ghz atheros timings alot different than 802.11b hermes? I was wondering about the possibility of offering PTMP 5.8ghz by mounting the AP directly behind the 16dB radiowaves 5.8ghz sector (sector-pigtail-5354 card/WRAP), and then using a 5354 card/WRAP/ and the newly planned pac wireless 5.8ghz rootenna as CPE.

Question is this: Will a atheros AP operate in a similar fashion as the hermes distance timing?

Assuming that the calculations work out for my atheros AP to be useful to 8 miles, will all clients be able to associate to the atheros AP if its ACK is set to 8 miles?
Thanks,
Bob C

[lonnie]

Yes that should work fine.

[bobbyc]

Cool that sounds good.
I see the hardware limit is 27 miles... would a 32 mile link result in the radios not locking? If they would lock, do you think it would have packetloss, or just be slow(er)? Assuming enough signal (33dB Maxrad dishes on each end with no cable loss except u.fl pigtail), do you think I could get 6 megabits/sec?
Just thought you might know without testing...
Bob C

[bobbyc]

Hi guys, I am going to be upgrading both ends of my 1.5 mile 5.3ghz link tonight to the latest beta. Right now it is on a older version that doesn't have the new ACK increments. What is a safe number for 1.5 miles?
Also I need a safe number for a 4.5 mile link.
Thanks!
Bob C

[lonnie]

For any unknown link use 400. Then do a binary sweep, by cutting the difference in half each time. If 400 locks, then go to 200. If it locks then go to to 100. If there is no lock then go to 150. If it locked at 100 go to 50. Hope you get the idea.

This will get you to the optimum value in 9 steps or less. Then just add 10 to 15 for good measure. Soon we will have this ACK think to a simple number in miles.

[bobbyc]

Well I didn't have much time to play with the ACK settings because I didn'
t want to piss off my customers too much by re-activating changes over and over again, but on my 4.5 mile link, I have the AP set to 400. It is 1.13.4b2 (old ack values). The client is 1.13.4b6 (new ack values and has a ACK of 100.) I think I tried 50 also on the client machine. What happens is that every once in a while the AP and client will associate for a few seconds and then go away for a while. The signal on the client screen shows -87, but the AP association list shows -70/-70
Does this look like a ACK problem at this point, or do you think my signal sucks that bad? I probably won't have a chance to play with the ACK settings till next sunday night/early morning, but if anyone could reccommend some ACK numbers for a 4.5 mile link that would be sweet!
Bob C

[lonnie]

I would suspect that you have a noise issue. -87 dB is real low. The 70/70 at the Ap assocaition is the ACK RSSI/ Packet RSSI and is not the local/remote as in the 2.4 GHz gear.

ACK will not link if it is too low. If it is too high you will get a slower retransmit on errors since it waits longer to time out. On a solid link you woul not notice much difference from proper ACK to the maximum.

[bobbyc]

Good call, I changed to channel 165 and it still had the same problem... so i tried 161 and that did the trick... signal is -66. AP end is AG520 with soldered on 12" pigtail and 10' LMR-400 to 26dB pac wireless grid.
Client end is AG520 with soldered 12" pigtail and 50' LMR-400 with same type antenna.
Power overide is set to 30, I'll put it on 20 now that the link is up and running.
When I made the lmr400 cables I soldered on the crimp center pins, but quickly crimped the center pins also while the solder was still hot.
I think these are the last grids I'll buy, since a panel would have better f/b ratio and it seems like there is a lot of junk on at least one of these towers in 5.x spectrum. I'll probably replace these 26dB grids with a couple 23dB flat panels.
Bob C

[bobbyc]

Damn I spoke too soon. When I started pinging the client from the AP, I watched the client signal level go from the 60's to 80. Now it flucuates from 77 to 81. So I'll probably down the road replace the AG520 with 5004 cards, and panel antennas. But for now I can light this link up and make good use of it.
Bob C

[lonnie]

When I made the lmr400 cables I soldered on the crimp center pins, but quickly crimped the center pins also while the solder was still hot.
Bob C

Bad move. The solder provides a physical as well as electrical connection and while it is cooling the rule is never ever move it. After it is cooled you also do want to stress it physically and crimping certainly rearranges things.

By soldering after you crimp you will improve the connection and since it was adequate as a crimp only, you are making things better. I would recommend that you use a solder with 5% silver for the very best electrical connection.

As long as you are in the bottom -80's you should have a solid connection.

[bobbyc]

OK I'll re-do the connectors. It actually looks like my link is gone this morning, so I'm gonna have to try some more channels, and put panels in place of the grids.

[bobbyc]

Just an update: I never did have the link on channel 161, it was always on 165 when it started to act weird. Towards the end of the week the link became stable; so I think perhaps I forced some outfit to change channels on their gear :)
Anyways I have it on channel 161 now and locked the rate to 9 (it still gets way faster than that when doing tests). What's nice though now is that my signal on the client screen stays at -68. Before it would hover around the high 70's low 80's.
If pings stay 0 loss into mid week I think I'll be switching the routes over to this 5.8ghz link, and then switch to panel antennas down the road... but I think I'll keep my soldered/crimped N connectors since they seem to be working great, and my signal problems probably were a noise issue, not bad connectors.
Thanks!
Bob C

[rasimoes]

From a support site... Maybe help! :)

range ack-timeout
5GHz 5GHz-turbo 2.4GHz-G
0km default default default
5km 52 30 62
10km 85 48 96
15km 121 67 133
20km 160 89 174
25km 203 111 219
30km 249 137 268
35km 298 168 320
40km 350 190 375
45km 405 - -

[georgew]

From a support site... Maybe help! :)

range ack-timeout
5GHz 5GHz-turbo 2.4GHz-G
0km default default default
5km 52 30 62
10km 85 48 96
15km 121 67 133
20km 160 89 174
25km 203 111 219
30km 249 137 268
35km 298 168 320
40km 350 190 375
45km 405 - -
Just for the fun of it, I tested a 8.59km link and found the ack tipping point to be at 89usec for a 2.4 g mode link.

88 was dead, 90 worked good, 89 was the edge.

Assuming the values are linear I used your chart to extrapolate a value of 86.4usec for the link using your chart. Since the 86.4usec varries substantially from the 90usec setting I found to work, I conclude that your chart is not accurate for our driver/card set. You can use it to guess your starting point, but you can expect the setting that works to be higher... at least on some cards. The numbers are close...

[robok]

Where can find ACK settings??

[lonnie]

It is now just set by the distance setting.