Epmp4600 to Force 4625 Signal Expectations

[Chris_L]

Hello All

I have currently set up a new AP and I am a bit confused on the signal levels I am seeing. I am in Canada, and this is how I have it configured:

• Model: ePMP 4600 4x4
• Antenna: AW3802-T2-H-6G
• Mode: WLR
• Channelwidth: 80Mhz
• Frequency: 6730 (AFC is clear, nothing within 100Mhz)
• TX Power: 17dbm
• Antenna: 18dbm (EIRP 35dbm)
• Firmware: 5.8.0-RC24

Based on link planer, using the stock antenna profile, expected signal is -63 tx to client and -70 rx to client +/-4dbm which makes sense to me. There is only one 1 db of additional gain with the antenna I am using so no big deal.

What doesn’t make sense to me is that I am currently getting roughly -67 tx to the client, and -64 rx from client. Quite literally the opposite of where the signals should be

I understand if I have the antenna misaligned or something, but usually both the TX and RX signals are affected if that’s the case, and not inversely of what is expected either.

Is the signal level based on the modulation being used or is there something else I am missing like mimo/beamforming gains on the upload side? I am running 5.8.0-RC24 while testing out web proxy functions, but 4.7.1 was about the same for disjointed signal levels.

We have a 4600 and a 4600L APs setup with Force 4625 clients and I am seeing the expected signal differences on those units, although the signals are quite low on those units as I didn’t calculate down tilt properly… 4 electrical down tilt plus 3 degrees of mechanical down tilt is a bad time when you only have 120 feet of height lol

[aka (Andrii)]

Hello Chris,

I appreciate your passion to test Beamforming! Could you send me techsupports from AP and SM on 5.8.0-RC26? It has all bells and whistles.

[Chris_L]

Just an update, added a second client to the AP yesterday, and seeing the same signal difference. Signal is poor at the second location as it has a pretty bad fresnel obstruction (shooting through a steel pole structure)

Performance is still pretty decent though for -80db on the downlink, pushing 213Mbps down and 134Mbps up.

[3-dBnetworks]

We are just starting to test 46xx and seeing the same thing. Our first test is 3.6 mi between 2 of our tower sites with good LOS. The uplink is 5 dB better than Linkplanner target but downlink is 7 dB worse.

[Lubomir_Zaciatocny (LubomirZ)]

a quick reply : 4.5km represents 122dB free space loss at 6.7GHz according to calculators. Because you have 35dBm EIRP (let’s call it TX side) and resulting signal of -67dBm, this means you only have 20dBm total gain on the client :

35dBm (TX Power + antenna gain) - 122dBm free space loss + 20dBm RX Client gains == -67dBm

Now, working with 20dBm antenna - because there is no other gain on RX side, just the antenna - is very little at these distances. You can’t do supertrooper radio link with 18dBm / 20dBm antennas at distances like this.

If my calculation is off and you use much stronger than 20dBm antenna at client, then it would mean you are heavily missing on the link in general. Improper aiming (horizontal / vertical), fresnel zone clearance, water in cables/connectors, improper cables and connectors (pigtails ?) used for this high frequency etc.

Next area to discuss : 6.7GHz is very high. I have no information about the analog part of Cambium devices, but suggestion would be to switch to lower frequency and test the behavior. Not only radios, but also antennas and their specific construction can significantly contribute to signal degradation and problems. This freq is on the outside boundary of allowed band and no single item in the whole chain is optimized for this frequency, leading to surprising snowball effect.

[Chris_L]

Hello LubomirZ, thanks for the reply!

Just for reference, it’s not the signal level that is the concern, but rather the difference between the TX/RX levels as it looks to be backwards.

Based on the data sheets for the Force 4625, the client radio has an antenna gain of 25DB and a frequency range of 5925-7125Mhz. The data sheet doesn’t provide attenuation over frequency, so there definitely could be a dead zone.

The AW3802-T2-H-6G antenna at the AP side is rated for 18.1dbi for 6525-6875 specifically and the ePMP 4600 is also rated for 5925-7125mhz

My math is:
Tower TX is 35 - 122 + 25 for client antenna = -62 for client RX but getting -67
(The tower side only lets me set 35db eirp)
Client TX is 36 - 122 + 18 for tower antenna = -68 for tower RX but getting -64

The issue is that in my experience, bad alignment, fresnel obstructions or cable damage typically hurts both TX and RX somewhat equally, and doesn’t add more gain then what should be possible on one side. That’s why I am wondering if there is beam forming or 4x4 mimo gains at play.

If I had a signal of -67 RX and -74 TX at the client, that would make sense to me and match link planner (+/- 4db)

Here is a screen shot of another sector we have where I know the antenna alignment is bad (waaaaay too much down tilt on the AP side, some clients are also off sector)

This sector uses the cambium 90 degree 4x4 sector and Force 4625 clients as well, but it is also running 5.7.0. This is also on 6685Mhz and 80Mhz channel width.
In this case, most clients are seeing higher downlink signals compared to uplink signals, which is what I am expecting.

In Canada, this frequency is legal with AFC, and currently this is the best 80Mhz channel I can use as this equipment is at a busy site. I will try and drop the frequency this week and see if that changes anything and report back. If this equipment has different performance characteristics in this frequency range, that would definitely explain it. I just haven’t seen anything on the data sheets to suggest this.

[Lubomir_Zaciatocny (LubomirZ)]

hi Chris,

The issue is that in my experience, bad alignment, fresnel obstructions or cable
damage typically hurts both TX and RX somewhat equally, and doesn’t add more
gain then what should be possible on one side. That’s why I am wondering if
there is beam forming or 4x4 mimo gains at play.

very correct.Everything you wrote is absolutely solid without any single flaw.

If this equipment has different performance characteristics in this frequency range, that
would definitely explain it. I just haven’t seen anything on the data sheets to suggest this.

absolutely correct again. I haven’t seen anything like that myself and I’m not saying / speculating this is the case. It should not be.

Your case indeed is curious.

[Eric_Ozrelic]

Any specific reason you’re using WLR mode? There is 3dB of beamforming gain while using WLR mode with the 4600 4x4. The new 5.8 firmware adds beamforming to TDD mode. One cool thing about the addition to beamforming on TDD is you have the ability to test with it on and off. This might give you a clues to the values you’re seeing.

At this point we’ve completely moved away from WLR mode for all of our e4k PtMP sites.

[Chris_L]

Very cool, I didn’t know about the TDD getting the beamforming toggle. I tested TDD before on a much older version and had a lot of issues with just getting stuff connected.

I ran into separate issue at an event that was supposed to connect to a different tower. They needed a solid backup connection for the weekend for live streaming, and I had no choice but to throw them on this sector with a -80 lol. The event will be done tomorrow so I can go back to fiddling with all the settings. I will definitely see if the latest RC + TDD + toggling beamforming changes the reporting.

I was under the assumption that beamforming would increase the link budget (like SNR) but the actual signal levels would stay the same.

Would be nice to see the signal levels of each chain in the ePMP interface. I am also looking at the 4500C and using a number of small 30 degree sectors to target specific subdivisions, and this could also be useful there.

[Chris_L]

Can confirm, RC26 to RC27, TDD and toggling beamforming doesn’t change RSSI or SNR, but modulations definitely get better with beamforming, so that’s fancy.
With Beamforming:

No Beamforming:

Would probably see a bit more modulation difference if I actually loaded up the clients. The -80 client tends to fluctuate quite a bit.

[Chris_L]

Just dropped frequency to 6345, which is the lowest frequency with 80Mhz channel width and full EIRP I can use, and signals are very similar. Gained a couple DB, but the difference between downlink and uplink rx is still there.

My only thought is if the client are exceeding the EIRP, because the downlink signals seem about right. It’s just that the upload modulations should be similar or better with increase in SNR. Alternatively, based on the fact that the Force 4625 doesn’t do MCS12 or MCS13, this could just be a wireless chipset difference.

[3-dBnetworks]

I am seeing RSSI differences across the band. At 6055, the downlink RSSI increased to -68 and uplink increased to -65. 5 and 3 dB down/up improvement from 6740. At 6300 RSSI is -71 down and -69 up. Link tests are reasonably close across the different channels and actually better on the lower RSSI, due to less noise.

[Lubomir_Zaciatocny (LubomirZ)]

@Chris_L

Just dropped frequency to 6345, which is the lowest frequency with 80Mhz channel width and full EIRP I can use, and signals are very similar. Gained a couple DB …

well, pure math says there’s less than one decibel difference between 6.9GHz and 6.3GHz in terms of free space loss at 5km link. Maybe the speculation about

• analog circuits being not able to behave exactly identically across the whole range and
• antenna design/construction doing the same

has some material merit.

Where I live, for the occasion you mentioned (backup connection for an event) I would not hesitate to crank up TX Power to gain a decibels or two above EIRP - thus, I would have -77 or -78dB. But I’m definitely not saying someone else should do that

[Chris_L]

Just wanted to give a quick update, we were running 5.9.0-RC13 and the signals flipped around

Then after running into issues, we reverted back to 5.8.0. The signal levels reported on 5.9.0-RC13 made a lot more sense to what I was expecting. Performance/modulation looked to be the same as well even though the signal changed. (750Mbps aggregate on 80Mhz channel width with fixed TDD)

My guess is the signal differences are just down to how the signal is reported rather than actual signal changes.

As we look at getting some 4500c equipment, is there anyway we could get signal per chain in the GUI? I know it would be a ton of info, but it would be useful for troubleshooting