You are here

Rocket M3: 3.1-3.3 GHz via firmware mod?

13 posts / 0 new
Last post
N6LDJ
Rocket M3: 3.1-3.3 GHz via firmware mod?
Our amateur group uses the Rocket M3 radio for two point-to-point links between high-elevation sites.  The Commission's new proposed rules, if implemented, would eliminate the current 3.3-3.5 GHz band with no replacement.  However, they may be leaving some (non-amateur) services in the 3.1-3.3 GHz portion intact for now.  If it was proposed to move the current amateur segment there, would there be any capable radio equipment?  Has anyone attempted to modify the band limits on the M3 hardware to have it operate in this band (ignoring for now any other equipment compliance issues)?

The Commission just released proposed rules for the upper 5.8 GHz band today which could impact (but not eliminate) a relatively clear portion there as well - which is why I ask the 3.1 GHz question. The ISPs have used all the current 2.4 and 5.8 at our tower sites so 3 GHz has become really important.  900 MHz is impacted by high-power geo-location services in out path as well.      

Greg
N6LDJ
 
AE6XE
AE6XE's picture
Modifying Ubiquiti 3GHz gear
Modifying Ubiquiti 3GHz gear to 3100 to 3300

See this forum post:  ​https://www.arednmesh.org/comment/13927#comment-13927

My testing shows there is a filter and the power drops to the floor outside the vendor's specification of ​​3370 - 3730.   I've reached out to some of the knowledgeable developers that maintain the wireless ath9k (802.11n) and ath10k (802.11ac) drivers.   One developer responded and is not aware of any way in the firmware to affect this kind of filter.  The current thinking is there is a hardware filter off-chip.

If anyone has a bad 3GHz Ubiquiti device (or is very careful), please pull off the case and possibly an RF shield cover and take pictures of any chips and circuitry.   The Rocket M3 has an Atheros AR9280 (from memory) chip we are looking for (ignore any chips that are AR803?, these are Ethernet chips).  We're looking for AR92?? chip.    Any chips or circuitry between this chip and the antennas is what we are looking for.

I did test on 3180 and made a link between two M3 devices, however they have to be right next to each other to link...

Joe AE6XE
N6LDJ
Thanks for the detailed
Thanks for the detailed response.   I have a M3 XM unit here but it is a *working* spare for our links that are tower mounted so I may not want to break the seal on this unit.   We will see if anyone else has one they can tear into.   

The potential for the ARRL to suggest, and the FCC to accept a secondary 3.1-3.3 approach likely remains low.   But if the devices are hardware limited then it may be a show-stopper.
       
I noticed UBNT has now labeled many products, including all of their 3 GHz products a "Legacy", and they have not gotten back to me on my question about future 3 GHz product development.   I guess 3.65 CBRS is turning out to be a private LTE band and not used for PtP or PtMP?

Greg
KE2N
KE2N's picture
Rocket M3 insides

It seems that one side of the board is sprinkled with Cirocomm dielectric resonator filters (a custom one for UBNT, no doubt).  There is another compartment on the other side of the board that was resisting my attempts to open - I suppose that side is the conversion to 3 GHz.  The filters on this side of the board seem to limit the pass band to the area around 5506 MHz - Which I suppose corresponds to 3.5.GHz (?)   If that is the case, then we are already down 100 MHz at 3.4 GHz and sitting on the shoulder of the filter(s).


Ken

Edit - looking at the board some more, I think the conversion to 3.4 GHz and the final amplifier are on this side of the board.  So no filtering once it is converted to 3 GHz. But lots of filtering on the 5 GHz path.



 

Image Attachments: 
AE6XE
AE6XE's picture
The vendor specifications for
The vendor specifications for the Rocket M3 is ​​3370 - 3730.  Thus the 5GHz chip-side would have to allow at least 5370 - 5730.   Do you have a Rocket M5 to compare with?   I do have a bad M5 to take apart, but limited time.   I'd speculate the circuitry for the 5GHz side of the fence is the same -- lowest cost to add the 3GHz downconvert and 3370 - 3730 filter on the other side of the fence?

The 5506 filter is near the breakpoint with a dry spot for u-nii allocation -- As I recall AirOS does not offer these channel options in this dry spot.    ch 100 is 5500 (5490 - 5510 @ 20Mhz), then the next lower u-nii unlicensed channel is 64 at 5320.     We are currently using, e.g. 3400 or ch 80 if my math is correct, but the key point is the current AREDN use of these devices is in this dry space below 5500. 

Joe AE6XE
KE2N
KE2N's picture
The center of 3370-3730 is

The center of 3370-3730 is 3550 although it should be 3560 if the center of a 20 MHz channel is +10 up from the nominal channel frequency.  5508 is a bit shy of 5560 (3560).   But the pass band response may not be symmetrical.

We do not know the bandwidth for that part number, it has to be 260 MHz at least (+/- 130 ? that would be the "13" number).  Just a guess.  If someone had a dead M3 we could remove one of the filters and plot the response.  I am not volunteering to do it with my shiny (almost) new one....   


Ken
 

KE2N
KE2N's picture
the M5

The M5 does not have all this filtering (probably an XM version, not sure),  There are two tiny filters visible (F12 and F15) on the input side of the final amps. It looks like two more filters F9 and F16 in the receive path.

It has occurred to me that the device closest to the antenna (S415) in all these radios must be a TR switch. 

A power-monitoring directional coupler is visible in this picture.  The design seems to have more isolation between chain 0 and chain 1 than the 3 GHz device.

Anyway - conclusion is that the (pre-conversion) 5.5 GHz signal generation involves a lot more filtering in the M3.    Either the spectra purity requirements are tougher and/or there is a lot of opportunity for spurious products with 2,3.5 and 5.5 GHz all in the same box.


Ken
 

Image Attachments: 
AE6XE
AE6XE's picture
Note, the worldwide freq
Note, the worldwide freq coverage in the specs for the M5 is  5170 - 5875.   A pass band center would be 5522.5, very close to 5506. 
K6AH
K6AH's picture
I'll offer an M3 Rocket to the cause...
I have an M3 that takes down the PoE power supply whenever you connect one.  What do you need me to do with it?
KE2N
KE2N's picture
m3

I had suggested that a defunct M3 could be used as a sacrificial source source for those DR filters. Take one SM device out and measure the frequency response.   My tracking generator only goes to 1.5 GHz, but I could use a signal generator and power meter that I have to manually determine the 5 GHz curve (the generator and power meter go to 18 GHz).  The expected result would be that there is a big drop off below 5.4 GHz.

This is presumes that Cirocomm would likely refuse to send us that info based on confidentiality agreements with UBNT.

However, in another posting, I note that the power amplifier in the M3 is only specified to go down to 3.3 GHz.  So it would seem there is more than one barrier to moving an M3 down to 3.1 GHz ... 
 
I am happy to try it anyway, if people are really interested. 

edit:  power amp info
https://www.skyworksinc.com/-/media/SkyWorks/Documents/Products/2401-2500/AWT6283R_204268C.pdf
 

K6AH
K6AH's picture
My Sacrificial M3...
Ken, my sacrificial Rocket M3 will arrive by end of day Tuesday, 12/24.

Andre
KE2N
KE2N's picture
it arrived

I removed the filter (a lot of heat so it dislodged a couple of other tiny SMD components) and I mounted it between two SMA connectors.  Testing showed that this is NOT a 5 GHz filter as speculated (based on part number) but, in fact, is a 3 GHz filter.
 
Looking at the response curve (which is relative to - 20 dBm) it would appear to be a 3.65 GHz filter (+/- 300 MHz).   I did not take a lot of points in the main pass band, but I took many points along the lower shoulder to define that closely, seeing as it falls off steeply. 

There must be some extra gain in the system to get things to work near 3300, because the filter is already down about 10 dB from the main pass band.



Ken
 

Image Attachments: 
N6LDJ
My - you guys work fast....
My - you guys work fast.....excellent info. 

Would it be worth an email to ​Cirocomm to see if they have any off-the-shelf devices in the same package size?  Ideally it needs to be ~3170-3300.  But I am 99% sure they will not as the 3.1-3.3 band is quite uncommon from what I see.   

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer