Page 1 of 1

DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Sat Oct 05, 2019 1:58 am
by SergeiF
Recently I have setup two Disc Lite5 ACs in PtP bridge configuration ~300m apart (clear sight over land).

I have tried with NV2 and noticed a series of hiccups where the throughput drops to 0 for a few seconds. These hiccups are quite regular (once a minute).
If 802.11 is used there are no hiccups.
The other issue is that at 300m I expected much better performance than ~150Mbit/s (both NV2 and 802.11 yield this throughput).
Both discs run 6.42.11 (long term).
Channel		5660/20-Ceee/ac/DP
Wireless Protocol		802.11
Tx Rate		866.6Mbps-80MHz/2S/SGI
Rx Rate		866.6Mbps-80MHz/2S/SGI

Tx/Rx Signal Strength		-34/-31 dBm
Tx/Rx Signal Strength Ch0		-36/-33 dBm
Tx/Rx Signal Strength Ch1		-38/-35 dBm
Noise Floor		-105 dBm
Signal To Noise		74 dB
Tx/Rx CCQ		97/97 %
Overall Tx CCQ		97 %
Distance		1 km
RouterOS Version		6.42.11
and
Channel		5660/20-Ceee/ac/DP
Registered Clients		1
Authenticated Clients		1
Overall Tx CCQ		97 %
Distance		35 km
Noise Floor		-106 dBm
Notice one of the dishes detects distance as 35km...

Can anyone spot anything wrong in these stats?

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Sat Oct 05, 2019 11:02 am
by mistry7
Welcome,

there is nothing wrong on your Setup, Limit transmit Power would possible gives a Bit more speed (you are about -34, this is to much, -50 ist ideal)

But for real ac Speed you Choose the wrong vendor

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Sat Oct 05, 2019 11:13 pm
by scampbell
I am not sure what is wrong in your config but I agree the signal strength is too high (devices yelling at each other) - please publish your wireless config so we can see what country, regulatory settings, tx-power and aerial gain you are using.

Below is a shot of a NetMetal running Ceee AC with connect speed (air rate) of 780Mbps/780Mbps showing actual data throughput of approx 390Mbps - data rate being typically half air rate. Tx/Rx signal on this link is -50dBm and protocol is NV2. I'd compare with 802.11 but it is a production link and I cannot make changes.

These NetMetal's are connected to mANT30's (with Sleeve) and running on a 5Km link - the mANT30 is being used as it is an urban to rural link and we want to reduce interference so to compensate for the extremely high gain aerials we reduce the TX power of the Netmetal to achieve -50dBM. The mANT30 & Sleeve reduce side lobes and interference from behind - this improves the SNR.

/interface wireless
set [ find default-name=wlan1 ] adaptive-noise-immunity=ap-and-client-mode \
antenna-gain=30 band=5ghz-a/n/ac channel-width=20/40/80mhz-Ceee country=\
"new zealand" disabled=no frequency=5xxx frequency-mode=regulatory-domain \
hw-protection-mode=rts-cts installation=outdoor mode=bridge \
multicast-helper=full nv2-cell-radius=15 nv2-preshared-key=xxxxxxx \
nv2-security=enabled radio-name=xxxxxxx rx-chains=0,1 ssid=xxxxxxx \
tx-chains=0,1 wireless-protocol=nv2-nstreme-802.11
netmetal.PNG

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Sat Oct 05, 2019 11:27 pm
by scampbell
Welcome,

there is nothing wrong on your Setup, Limit transmit Power would possible gives a Bit more speed (you are about -34, this is to much, -50 ist ideal)

But for real ac Speed you Choose the wrong vendor
I agree on the TX-Power.

Also what method of speed test is being used and are you running it from devices either side of the radios that can generate a minimum of 500Mbps themselves without maxing out the CPU's e.g test from A->D where B&C are your radio link. A-->B<-->C--->D

Try testing both UDP and TCP. Remember TCP is always slower than UDP as it needs to interrupt the traffic flow to send ACK packets so is potentially half the UDP speed. See this excellent desciption on TCP Vs UDP at http://www.skullbox.net/tcpudp.php

Try other protocol's - NV2 is better for an AP than a P2P (Hidden Node issues) - so try 802.11 as a comparison on your link.

If I am in a busy urban site I will still use NV2 for a P2P as the TDMA tends to keep the channel open and discourage 802.11 devices from using the channel (https://en.wikipedia.org/wiki/Carrier-s ... _avoidance) :-).

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Sun Oct 06, 2019 1:42 am
by mistry7
Welcome,

there is nothing wrong on your Setup, Limit transmit Power would possible gives a Bit more speed (you are about -34, this is to much, -50 ist ideal)

But for real ac Speed you Choose the wrong vendor
I agree on the TX-Power.

Also what method of speed test is being used and are you running it from devices either side of the radios that can generate a minimum of 500Mbps themselves without maxing out the CPU's e.g test from A->D where B&C are your radio link. A-->B<-->C--->D

Try testing both UDP and TCP. Remember TCP is always slower than UDP as it needs to interrupt the traffic flow to send ACK packets so is potentially half the UDP speed. See this excellent desciption on TCP Vs UDP at http://www.skullbox.net/tcpudp.php

Try other protocol's - NV2 is better for an AP than a P2P (Hidden Node issues) - so try 802.11 as a comparison on your link.

If I am in a busy urban site I will still use NV2 for a P2P as the TDMA tends to keep the channel open and discourage 802.11 devices from using the channel (https://en.wikipedia.org/wiki/Carrier-s ... _avoidance) :-).
I think your Knowledge with Mikrotik ARM is not real.... don’t compere it with mispbe

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Sun Oct 06, 2019 9:58 am
by 2jarek
@scampbell JUST test bridge ARM<=>ARM still works like crap even pure 802.11 in LAB. U need ONE MIPS BE MIPSBE<=>ARM for works but this speed :D.

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Mon Oct 07, 2019 12:28 am
by scampbell
@scampbell JUST test bridge ARM<=>ARM still works like crap even pure 802.11 in LAB. U need ONE MIPS BE MIPSBE<=>ARM for works but this speed :D.
I agree NV2 and ARM are not best friends but here is a quick Btest across a pair of SXTsq AC's running NV2:
sxtsqac.jpg

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Mon Oct 07, 2019 1:14 am
by SergeiF
So here is the config:

Far side
/interface wireless> print advanced 
Flags: X - disabled, R - running 
 0  R name="wlan1" mtu=1500 l2mtu=1600 mac-address=***** arp=enabled disable-running-check=no interface-type=IPQ4019 
      radio-name="*****" mode=station-bridge ssid="*****" area="" frequency-mode=manual-txpower country=new zealand 
      antenna-gain=0 frequency=5660 band=5ghz-a/n/ac channel-width=20/40/80mhz-Ceee secondary-channel="" scan-list=default 
      wireless-protocol=nv2-nstreme-802.11 rate-set=default supported-rates-a/g=6Mbps,9Mbps,12Mbps,18Mbps,24Mbps,36Mbps,48Mbps,54Mbps 
      basic-rates-a/g=6Mbps max-station-count=2007 distance=dynamic tx-power-mode=default vlan-mode=no-tag vlan-id=1 
      wds-mode=disabled wds-default-bridge=none wds-default-cost=100 wds-cost-range=50-150 wds-ignore-ssid=no 
      update-stats-interval=disabled bridge-mode=enabled default-authentication=yes default-forwarding=yes default-ap-tx-limit=0 
      default-client-tx-limit=0 wmm-support=disabled hide-ssid=no security-profile=default wps-mode=push-button 
      station-roaming=enabled disconnect-timeout=3s on-fail-retry-time=100ms preamble-mode=both compression=no allow-sharedkey=no 
      station-bridge-clone-mac=00:00:00:00:00:00 ampdu-priorities=0 guard-interval=any 
      ht-supported-mcs=mcs-0,mcs-1,mcs-2,mcs-3,mcs-4,mcs-5,mcs-6,mcs-7,mcs-8,mcs-9,mcs-10,mcs-11,mcs-12,mcs-13,mcs-14,mcs-15,mcs-16,
                 mcs-17,mcs-18,mcs-19,mcs-20,mcs-21,mcs-22,mcs-23 
      ht-basic-mcs=mcs-0,mcs-1,mcs-2,mcs-3,mcs-4,mcs-5,mcs-6,mcs-7 vht-supported-mcs=mcs0-9,mcs0-9,mcs0-9 vht-basic-mcs=mcs0-7 
      tx-chains=0,1 rx-chains=0,1 amsdu-limit=8192 amsdu-threshold=8192 tdma-period-size=2 nv2-queue-count=2 nv2-qos=default 
      nv2-cell-radius=30 nv2-security=disabled nv2-preshared-key="" nv2-mode=dynamic-downlink nv2-downlink-ratio=50 
      nv2-sync-secret="" hw-retries=7 frame-lifetime=0 adaptive-noise-immunity=none hw-fragmentation-threshold=disabled 
      hw-protection-mode=none hw-protection-threshold=0 frequency-offset=0 rate-selection=advanced multicast-helper=default 
      multicast-buffering=enabled keepalive-frames=enabled 
Near side
/interface wireless>  print advanced
Flags: X - disabled, R - running 
 0  R name="wlan1" mtu=1500 l2mtu=1600 mac-address=***** arp=enabled disable-running-check=no interface-type=IPQ4019 
      radio-name="*****" mode=bridge ssid="*****" area="" frequency-mode=manual-txpower country=new zealand antenna-gain=0 
      frequency=5660 band=5ghz-onlyac channel-width=20/40/80mhz-Ceee secondary-channel="" scan-list=default wireless-protocol=802.11 
      rate-set=default supported-rates-a/g=6Mbps,9Mbps,12Mbps,18Mbps,24Mbps,36Mbps,48Mbps,54Mbps basic-rates-a/g=6Mbps 
      max-station-count=2007 distance=dynamic tx-power-mode=default vlan-mode=no-tag vlan-id=1 wds-mode=disabled 
      wds-default-bridge=none wds-default-cost=100 wds-cost-range=50-150 wds-ignore-ssid=no update-stats-interval=disabled 
      bridge-mode=enabled default-authentication=yes default-forwarding=yes default-ap-tx-limit=0 default-client-tx-limit=0 
      wmm-support=disabled hide-ssid=no security-profile=default wps-mode=disabled station-roaming=enabled disconnect-timeout=3s 
      on-fail-retry-time=100ms preamble-mode=both compression=no allow-sharedkey=no station-bridge-clone-mac=00:00:00:00:00:00 
      ampdu-priorities=0 guard-interval=any 
      ht-supported-mcs=mcs-0,mcs-1,mcs-2,mcs-3,mcs-4,mcs-5,mcs-6,mcs-7,mcs-8,mcs-9,mcs-10,mcs-11,mcs-12,mcs-13,mcs-14,mcs-15,mcs-16,
                 mcs-17,mcs-18,mcs-19,mcs-20,mcs-21,mcs-22,mcs-23 
      ht-basic-mcs=mcs-0,mcs-1,mcs-2,mcs-3,mcs-4,mcs-5,mcs-6,mcs-7 vht-supported-mcs=mcs0-9,mcs0-9,mcs0-9 vht-basic-mcs=mcs0-7 
      tx-chains=0,1 rx-chains=0,1 amsdu-limit=8192 amsdu-threshold=8192 tdma-period-size=2 nv2-queue-count=2 nv2-qos=default 
      nv2-cell-radius=30 nv2-security=disabled nv2-preshared-key="" nv2-mode=dynamic-downlink nv2-downlink-ratio=50 
      nv2-sync-secret="" hw-retries=7 frame-lifetime=0 adaptive-noise-immunity=none hw-fragmentation-threshold=disabled 
      hw-protection-mode=none hw-protection-threshold=0 frequency-offset=0 rate-selection=advanced multicast-helper=default 
      multicast-buffering=enabled keepalive-frames=enabled 
5660 was chosen so it does not interfere with other access points.

The testing was done via the built-in bandwidth tester. TCP is about 30% lower than UDP test.
The 802.11 was chosen at the end because
1) Stability - no hiccups.
2) Latency - it dropped from 3-10ms to <1ms.
I don't think NV2 works properly on these devices.

What would be appropriate way to drop the transmit power?

The thing is when I tested these devices indoors (separated by couple of drywalls) and I got ~500Mbit/s TCP test. Perhaps drywall was attenuating power enough, as well as they were sub-optimally aligned?

Edit: I noticed it set to
frequency-mode=manual-txpower

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Mon Oct 07, 2019 11:45 pm
by SergeiF
Switching frequency-mode to regulatory-domain and running tests again (this time with UDP as well).
The results are:
~570Mbit/s unilaterally with UDP test.
~260Mbit/s bidirectionally UDP test.
~220Mbit/s unilaterally with TCP test.
~130MBit/s bidirectionally with TCP test.

This is using BTest on the device itself.
I suspect BTest is terrible at testing tcp, I should try iperf3 but I do not have a box on other end to test against.


That the devices are still "screaming" at each other:
Tx/Rx Signal Strength		-31/-31 dBm
Tx/Rx Signal Strength Ch0		-33/-33 dBm
Tx/Rx Signal Strength Ch1		-36/-35 dBm
Noise Floor		-105 dBm
Signal To Noise		74 dB
Tx/Rx CCQ		96/96 %
Overall Tx CCQ		96 %
Distance		1 km
I could add antenna gain in the interface setting, do you think it is a good idea?

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Tue Oct 08, 2019 8:50 am
by mkx
This is using BTest on the device itself.
I suspect BTest is terrible at testing tcp, ...

Bandwidth test is terrible when run on devices with less than fastest CPUs. If you do CPU profile while running bandwidth test, you'll most probably note 100% CPU load which means that CPU is (artificial) bottleneck in this case. If you really want to assess link speed, you just have to use external devices (iperf3 or iperf2).

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Tue Oct 08, 2019 11:36 am
by 2jarek
This is using BTest on the device itself.
I suspect BTest is terrible at testing tcp, ...

Bandwidth test is terrible when run on devices with less than fastest CPUs. If you do CPU profile while running bandwidth test, you'll most probably note 100% CPU load which means that CPU is (artificial) bottleneck in this case. If you really want to assess link speed, you just have to use external devices (iperf3 or iperf2).
U have old information now ARM easy close gbit port & now BT use multicore cpu well. Exacly same result if use external devices like 2x CCR.

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Tue Oct 08, 2019 11:38 am
by nickshore
Antenna gain on a Disc should be set to 21, you are running far too much power for a short link.

I could add antenna gain in the interface setting, do you think it is a good idea?

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Wed Oct 09, 2019 12:11 am
by SergeiF
Antenna gain on a Disc should be set to 21, you are running far too much power for a short link.

I could add antenna gain in the interface setting, do you think it is a good idea?


Sorry for being a noob, is the antenna gain subtractive to the TX power?
What I mean to ask is the antenna gain in dBm units? dB != dBm.
So if I wanted to get say from -30 to -50dBm I need to "add" (subtract) 20dBm?
Is the consensus that -50 to -60dBm is optimal target?


I shall experiment....

EDIT/UPDATE: it looks like antenna gain is subtractive to the TX power. I have added to both discs 20db and have now ~ -50dbm on both TX and RX.

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Wed Oct 09, 2019 12:39 am
by SergeiF
With -50dbm TX/RX power I get following results:

The results are:
~570Mbit/s unilaterally with UDP test.
~260Mbit/s bidirectionally UDP test.
~220Mbit/s unilaterally with TCP test.
~130MBit/s bidirectionally with TCP test.

Exactly the same as with -30dbm.

BTW running test leads to 50% CPU (I assume one core is pegged to 100%?).
Not sure if I can try with iperf as I can't willy/nilly go on to premises to plug random stuff in.

Since the antenna gain setting of 20 has no negative effect I will leave it in, just to be nice to the spectrum ;).

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Wed Oct 09, 2019 9:13 am
by mkx
Antenna gain on a Disc should be set to 21, you are running far too much power for a short link.

Sorry for being a noob, is the antenna gain subtractive to the TX power?
What I mean to ask is the antenna gain in dBm units? dB != dBm.

The big picture about Tx power is this: in most (if not all) countries the limit on Tx power of ISM devices is EIRP. Which is measured in point in space where signal strength is the strongest.

Real antennae are not isotropic (radiating in all directions equally which is physically impossible), the closest approximation being dipole antennae which don't radiate in direction of the dipole itself. Other antennae are more or less directional and thus (hopefully) have positive gain. And gain, by definition, is the ratio of radiated power (in "best" direction) of given antenna over radiated power of (hypothetical) isotropic antenna. Antenna gain is denoted in dBi. As with all deciBells, they can be summed even though they are not exactly same deciBells (e.g. Tx power is usually denoted in dBm but could be in dBW or some other logarithmic unit).

Now, if one wants to obey national regulations and limits EIRP to allowed level, the real Tx power has to be reduced proportionally to antenna gain. If EIRP allowed is, say, 30dBm and transmitter is connected to an average omni-directional antenna with gain of 2 dBi, then Tx power needs to be set to 28dBm to see EIRP at the upper limit (30 dBm - 2 dBi = 28 dBm). Likewise, when transmitter is connected to decent directional antenna with gain of 20 dBi, Tx power should be 10 dBm (30 dBm - 20 dBi = 10 dBm).

Now you might ask yourself why to use directional antenna if we have to reduce Tx power when using one? It's to enhance Rx. The full radio link budget equation is this: Rx_level = Tx_power + TxAnt_gain + RxAnt_gain - pathloss ... where Tx_power + TxAnt_gain = EIRP and pathloss is a function of physical space between both antennae (more or less proportional to more or less third power of distance) plus some other losses such as antenna feeder losses. Meaning that antenna gain (which is, in case of TDD networks such as WiFi, exactly the same in both directions) helps with link budget even if one sticks to allowed EIRP.
In addition to that, high gain antennae (being directional) help with interference coming from other directions ... and we all know that interference is killing SINR, which at the end of the day, defines achievable throughput (and we can't raise SI above certain level, so we strive to keep N as low as possible).
Too high Rx level causes some "artificial" noise inside receiver ... receiver features pre-amplifier (in the analogue part) with variable gain to adjust power level for DAC. If Rx level, fed by antenna feeder, is too high, PA gain can not be reduced enough and PA enters the non-linear range of operation ... it will clip the output signal to it's highest level, and for DAC this will seem lots of high-frequency noise. Due to extensive use of FFT this will map back into "usable" frequency space (Nyquist theorem covers this effect) causing receiver to perceive high noise.

Re: DISC Lite5 ac PtP NV2 Hickups and generally disapointing performance

Posted: Wed Oct 09, 2019 9:18 am
by 2jarek
With -50dbm TX/RX power I get following results:

The results are:
~570Mbit/s unilaterally with UDP test.
~260Mbit/s bidirectionally UDP test.
~220Mbit/s unilaterally with TCP test.
~130MBit/s bidirectionally with TCP test.

Exactly the same as with -30dbm.

BTW running test leads to 50% CPU (I assume one core is pegged to 100%?).
Not sure if I can try with iperf as I can't willy/nilly go on to premises to plug random stuff in.

Since the antenna gain setting of 20 has no negative effect I will leave it in, just to be nice to the spectrum ;).
Just ARM + ARM still works bad (2 years & MIKROTIK super slowly fix OWN CRAPPY DRIVER). ARM + MIPSBE from ONE SIDE works much better (125Mbit tcp traffic for each 20Mhz spectrum).