Starter for 10! (well 8.4)
|
Hi all,
Having looked at a few designs for LNA's I reckon it would be worth creating something based off of AN-MWO-LNA-XBND-2020.2.6.pdf in the files section. WG15 input, rather than WG16 so optimum frequency range would be 7.05 to 10 GHz. ? For the active device, the CE3512K2 looks excellent, documents at Spec: NF = 0.30 dB TYP., Ga = 13.7 dB typical @VDS = 2 V, ID = 10 mA, f = 12 GHz It may be possible to initially re-scale the wireframes from the AN-MWO-LNA-XBND-2020.2.6.pdf document from WG16 to WG15, but we probably need this simulating to ensure it will work. My guess would be to initially make a single stage to check the likely NF / gain, and if that works out, add a stripline BPF and a second stage etc. I like the idea of reduced height casing to avoid oscillations caused by waveguide modes, alsoo this removes the need to use bits of absorptive rubber which can contribute noise if incorrectly placed etc. From my perspective it would be great to have an optimised NF at 8425MHz, obviously with such a wide bandwidth design the NF +- a few hundred MHz will be excellent, and for local sources like LEO's its probably an overkill ;-) All ideas welcome! regards, Paul m0eyt. |
|
Hi Paul,
good idea regarding the LNA.? It's probably a good idea to start out with a single stage design to make sure that the performance of the FET is good enough on 8GHz. If it ends up being okay it might be a good idea to even add a third stage for those with long coax rungs. Of course it can be bridged if it's not required. I'm happy to help with CADing the PCB, I have a full license for AutoDesk EAGLE. Regards, Arved M0KDS |
|
Hi Paul,?
Endorse the suggested way forward- several building blocks here that can be worked on separately, WG port, matching section, bias arrangement, gate/drain bias supplies etc. I've always wondered about other scope for optimisation, so while there's nothing wrong with the ICL7660 charge pump for example, cleaning it up with a modern ultra-low-noise regulator can't help? Agree also that should probably shoot for a single fet stage "minimum viable design" - WG input, SMA out. Vincent Grigis, F1OPA's design (the AWR article) is a great starting point for the WG to microstrip transition. The California East Labs fets do look good - in addition to the low NF, there are S-parameters available plus they are inexpensive and readily available pretty much anywhere in the world, about 1GBP 1 off from Digikey. In terms of PCB fab, in my day job I've had good experiences with micron20.com in Bulgaria - they can make boards on Rogers 4003C cheaper than I can get the material. Arved: no idea how good Eagle is for RF layouts, personally I'd favour something like KiCAD, but at the end of the day so long as there's a gerber file! Once we get all that done, we'll start on the LN2 cryo-cooled version right? :) Regards, R. |
|
Hi Robert,
Thanks for the response. I'll have a go with Genesys over the weekend to see if I can replicate the WG input / impedance transformer of the F1OPA design at 10.36GHz first. I guess everything can be scaled once the design can be simulated. For the bias controller, I'd be in favor of using something like the ZNBG2001 which is an active controller that can support 2 fets, needs a single current setting R and has a Vds of 2V. I guess for a minimum viable design we can skip the integrated PSU and have appropriate external DC connectors so the board would be WG > microstrip, matching, fet, output etc. That would be simple to do and make a unit for optimizing. The mechanics would of course be easier since its a small unit, but with the cutouts for the tracks etc in the case lid, this will be a nice thing to machine ;-) I did see Iban EB3FRN on Twitter posted a video of LNA measurement with freezer spray applied, if we can make a nice enclosure that sandwiches the board properly we can try some peltier cooling before moving on to cryogenics, but lets get the room temperate design sorted out first ;-) regards, Paul EYT. |
|
Hi Robert, I've used both KiCad and Eagle for RF microstrip circuits in the past. Even though KiCad is slowly catching up with basic microstrip functionality Eagle still has much better support for custom shapes.? I think in this case Eagle would be more favorable than KiCad. Eagle is free for personal use but you are limited to a max board size and a couple of other things?if I remember correctly. It should still be fine for our?purposes I would?think. I was granted a full educational license without any limitations because I showed them school proof.? Kind regards, Arved M0KDS Am Fr., 24. Apr. 2020 um 12:32?Uhr schrieb Robert Watson via <robertwatsonbath=[email protected]>: Hi Paul,? |
|
Hi Arved,
No problem - previous Eagle user here, but migrated to KiCad for >2 layers and larger boards. Getting a bit OT, but likewise I've done some microstrip layouts in KiCad using custom footprints - mitres, tees, stubs, butterflies etc. That said, for a commercial project give me AWR or ADS every time :) I do want to give Daryll Harmon's openEMS work a try at some point??- be nice to have an easy end-to-end OSS solution for all. Since I'm here, back on-topic, I spent a couple of hours yesterday getting to grips with HFSS again for waveguide stuff. I made a simple section of WG15 in Inventor, exported as a .step file into HFSS. Just to check I can go from a mechanical CAD model and get a meaningful EM simulation. Hopefully useful to check designs. I also had a quick look at some academic papers on IEEEXplore to see if there's anything interesting. There are a number of options, might have a look at this one below with a stepped dielectric matching taper, might be easier for others to fabricate and assemble -? taken from "A Waveguide to Microstrip Inline Transition With Very Simple Modular Assembly", IEEE Microwave & Wireless Components, 2010, Zhang et al. Their design was optimised for 8.5-9.5GHz.? Regards, R. |
开云体育
Hi, i am very happy to see we will finally have a "universal" lna/downconverter for X band.
A lot of amateurs will benifit from your design.
(The final design)
Keep up the good work guys!
Jean-Luc Milette.
罢é濒é肠丑补谤驳别谤
From: [email protected] <[email protected]> on behalf of Robert Watson via groups.io <robertwatsonbath@...>
Sent: Sunday, April 26, 2020 8:01:26 AM To: [email protected] <[email protected]> Subject: Re: [xbandlna] Starter for 10! (well 8.4) ?
Hi Arved,
No problem - previous Eagle user here, but migrated to KiCad for >2 layers and larger boards. Getting a bit OT, but likewise I've done some microstrip layouts in KiCad using custom footprints - mitres, tees, stubs, butterflies etc. That said, for a commercial project give me AWR or ADS every time :) I do want to give Daryll Harmon's openEMS work a try at some point??- be nice to have an easy end-to-end OSS solution for all. Since I'm here, back on-topic, I spent a couple of hours yesterday getting to grips with HFSS again for waveguide stuff. I made a simple section of WG15 in Inventor, exported as a .step file into HFSS. Just to check I can go from a mechanical CAD model and get a meaningful EM simulation. Hopefully useful to check designs. I also had a quick look at some academic papers on IEEEXplore to see if there's anything interesting. There are a number of options, might have a look at this one below with a stepped dielectric matching taper, might be easier for others to fabricate and assemble -? taken from "A Waveguide to Microstrip Inline Transition With Very Simple Modular Assembly", IEEE Microwave & Wireless Components, 2010, Zhang et al. Their design was optimised for 8.5-9.5GHz.? Regards, R. |
|
Morning all,
I did learn over the weekend that doing anything sensible in Genesys is a nightmare on wheels. Basic bandpass filtering with microstrip coupled lines is fairly straight forward, but trying to model an impedance transformer in it is as far as I can see next to impossible. There is an S-parameter file for the proposed FET so I guess the first step is to import that into an EM package and see what the optimised input impedance is for the lowest noise figure, any idea how we do that? I've tried with Genesys can again can import the file no problem but cannot see how to plot NF based on swept input impedance. regards, Paul. |
|
Morning Paul,
Not had huge amount of free time to look at things, but one thing I did spot is that not all of the .s2p files have noise parameters - a smaller list of Fmin and Gamma opt at the end of the file typically. As far as impedance transformers, I'll have to dig out Pozar's Microwave Engineering book - there are some potted recipes in there for? multi-section Chebyshev quarterwave transformers. That'll generate a set of impedances which can be translated?into waveguide heights. Bit of a caveat here is that I've not looked at some of this stuff for a few years, so a bit rusty! Regards, R. |
|
Hi Chaps,
All excellent info. Could one of you simulate the CE3512K2 to see where the best noise figure is vs input impedance please. The F1OPA LNA uses RO4003C PCB with a thickness of 0.305 mm. I think once we know what is an optimised input then the probe / waveguide transformer can be designed around this. Once we have an idea of what the input will look like, we can build something as a POC. I'm happy to machine a few WG15 or WG16 transformer prototypes etc, and we can maybe get a few PCB's made to test this stuff out, and prove the input match etc, and once that's done we can add a 2nd stage and a BPF if needed etc? My idea is that we make the mechanics similar to the F1OPA design, with a reduced height enclosure to avoid waveguiding etc and prevent spurs / internal hooting etc. regards, Paul M0EYT. |
|
Hi,
toggle quoted message
Show quoted text
we tried some years ago a design with this FET as you said. Stefan simulated the whole signal chain in Agilent ADS and it looked promising. But we never get rid of the oscillations of the first stages. I would like to know what's wrong with this design. For kicad we wrote a footprint plugin to convert qucs microstrip schematics to an actual layout: I attached an example of a generated layout. There is also an optimizer for qucs but never used it. Best regards Sebastian DL3YC On 4/29/20 3:28 PM, Job PH4AS wrote:
Hi,For the active device, the CE3512K2 looks excellent, documents atI agree it makes most sense to use this FET. It is used in several |
|
on7kel
Putting my money where my mouth is: I quickly simulated the source stab circle, s11 and Nopt for 8.5 GHz using the 2V_10mA.s2p file in Qucs On Wed, 29 Apr 2020 at 15:05, on7kel via <kevin.voet=[email protected]> wrote:
|
Paul M0EYT
Arved Viehweger
Robert Watson
