10GHz microwave link teardown

[mikeselectricstuff]

[T3sl4co1l]

Nice!

Going to guess that RF device is "pre tuned".  Looks like lots of matched length traces, stubs (pretty sure I saw a stub to GND, equal length as to the transistors, so that'd be a 1/4 wave bandpass style thing), all that.  I'd guess GaN, but it could really be any of those I think.

The diff pairs coming out of the QFPs are definitely balanced, which is funny because one takes an extra 3/4 wave zig-zag and just connects right back up.  Turns out they wanted them in parallel after all..

The fringed looking appearance of the dies on it make me wonder if it's a solder-bump style, like the EPC2000s you can get at Digikey (e.g. http://www.digikey.com/product-detail/en/EPC2010/917-1016-1-ND/2351742 ).

Also saw a zig-zag interdigitated filter with much longer lengths on there, perhaps a low band IF filter?  There should be preamp or mixer blocks along the RX path.

Tim

P.S. Everyone knows you're supposed to look away from the Ark when it's opened.

[Alexei.Polkhanov]

I think that cavities in enclosure that is attached to the board don't have any purpose besides just making a room for components. Some have square ferrite pads glued inside not for shielding but to prevent unwanted cavity resonances.

I made some attempts at explaining the microwave "woodu". Only thing I am not sure about is interdigital bandpass filters, seems to be capacitors missing. Perhaps someone can explain why they called interdigital if that in fact what it is.



What is the substrate PTFE (feels like soft-ish FR4) or alumina (should feel like ceramic)?

[bktemp]

My guess is, the asics are mixers. Therfore the same ic can be used for both transmitter and receiver.

The overall construction looks very similar to a 38GHz microwave link antenna I bought just out of curiosity. But mine seems to be a bit older:
It has a dedicated power supply board with a lot of similar looking inductors, a large mainboard with all the control an IF filter stuff and 2 metal cans for RX/TX with all the 38GHz voodoo.

The 38GHz signal is mixed down in the outdoor unit to an IF frequency somewhere between 100-150MHz and then processed by an indoor unit using 8 Spartan 2 FPGAs.

[Alexei.Polkhanov]

Awesome photos, thanks for sharing. I love looking at this RF/microwave porn 

I have not seen this type of stubs before.... I wonder if it is more common for >24GHz

[firewalker]

I really enjoyed this one!

Alexander.

[Alexei.Polkhanov]

I guess those little strips of green solder mask are used to protect microstriplines from being flooded by solder to preserve the impedance.

[PA0PBZ]

The asics must be oscillators, probably some kind of PLL also. One on the transmit frequency and one as a local oscillator to mix down the receiving frequency.

[bktemp]

The asics must be oscillators, probably some kind of PLL also. One on the transmit frequency and one as a local oscillator to mix down the receiving frequency.

It is hard to see in the video, but it looks like the asics does everything: LO, IF and other control stuff. Maybe it also contains a second mixer.
I have added a few lines how I think it works:
Blue: 10GHz RX (upper left) and TX (lower right)
Red: Mixer
Green: LO
Yellow: IF
Those 2 unpopulated ICs could be for additional LO signals.

And for all microwave porn fans: A picture of the 38GHz transmitter box.
Here the signal path is a bit more clear than in the more compact 10GHz transceiver:
The LO (~3GHz from an external PLL) enters at the top left. The E2 marked part could be a ERA2 0-6GHz amplifier.
The next IC is presumably a multplier generating the LO frequency near 38GHz.
The IC at the bottom (below the screw hole) looks different than all other ics. Maybe it is an amplifier or another multiplier (or both).
The IF enters at the bottom left and goes directly into the mixer, before the signal goes thrue a filter followed by an amplifier.
The next part looks strange. The signal splits and there are two diodes (?) mounted in each path. Since it is connected to an additional signal going to the top it could be a pin diode attenuator.
Then comes another amplifier and another attenuator.
The last two amplifiers are larger than the previous ones. Before the signal goes to the antenna, there is a small loop nearby to measure and control the output power (via the pin attenuator).

[Howardlong]

I really enjoyed this one!

Alexander.

Me too, and I'm and RF geek, but not quite at these frequencies.

Seriously entertaining.

[dave_k]

Hey Mike,

The BNC connector on the rear of the unit is used to aid lining up the link path. It provides a DC signal which corresponds to received signal level. Typically the installer would hook up a meter to the port and aim the dish for maximum level.

[mikeselectricstuff]

Hey Mike,

The BNC connector on the rear of the unit is used to aid lining up the link path. It provides a DC signal which corresponds to received signal level. Typically the installer would hook up a meter to the port and aim the dish for maximum level.

That makes sense, especially as it has a sealing cap.

[Dago]

Awesome photos, thanks for sharing. I love looking at this RF/microwave porn 

I have not seen this type of stubs before.... I wonder if it is more common for >24GHz

These are adjustable stubs. You just bridge the additional islands with solder to tune it.

I have a 26 GHz Microsource YIG-oscillator box which has the same type stubs. Curiously it also seems to use FR4 as the substrate. ~13 GHz YIG which then gets doubled.

[eb4fbz]

Interesting teardown. BTW: the power amplifier (the white rectangular IC) is an Eudyna ES/EMM5068VU.

[piranha32]

Awesome photos, thanks for sharing. I love looking at this RF/microwave porn 

I have not seen this type of stubs before.... I wonder if it is more common for >24GHz

My guess is LC filter:  https://www.jlab.org/accel/eecad/pdf/050rfdesign.pdf

[Koen]

Hello,

    about shielding : is a big block with cavities as seen here any different than multiple sparse cans as seen in smartphones ? Can one cavity "contaminate" the ones its sharing walls with for example ?

Thank you, Koen

[T3sl4co1l]

Hello,

    about shielding : is a big block with cavities as seen here any different than multiple sparse cans as seen in smartphones ? Can one cavity "contaminate" the ones its sharing walls with for example ?

Thank you, Koen

The shielding amount varies with thickness and distance.  In quantum mechanics terms, the wave tunnels through the barrier (or not).  The attenuation is the probability.

For most purposes, I think I would be surprised if the shielding were insufficient; machined cavities are probably better for structure and cooling.  And occasionally making resonators and filters.

Tim

[Crumble]

Sorry for kicking up this topic, but it looks like the PCB is made out of a material like Rogers 400 series, there is a datasheet of it here. I discovered it is being offered at my company to make PCB with low dissipation factors, and remembered Mike's clip about the transmitter. The material is unlikely to be PFTE or ceramic, because they are not compatible with the standard processes used for FR4. PFTE does not allow itself to be glued and will have a poor peeling strength (and multilayers will likely delaminate), making via's is going to be an issue too. Most ceramics are an absolute pain to mill, Al₂O₃ likely kills milling bits instantly and would at least slow the process down considerably, but there might be softer ceramics out there.

[Yansi]

Also I think the alumina plate that large is unlikely for manufacturing complications. Some kind of Rogers most likely. Also the alumina would not be etched, but likely have a screen printed and then baked traces using a suitable metal compounds, like for example this board I have found in a 12.6GHz DRO (Dielectric Resonator Oscillator).

However it may be PTFE - have seen such PCBs, however only 2 layer and extremely thin substrates for 24GHz operating devices. (The PCB bends just when slightly touched)