Hi Stefan, Hi Job,

I looked again at? the 10.2-10.5 GHz F1OPA design which uses NE3515S02 fets). In the band of interest (and even partially beyond it) the stability circles for both input side and output side are all outside the unit reflection circle, ie is stable.

If you drop down to 8.4GHz however, the circles move in showing potential for instability.

So I guess one reason why some LNBs for the DBS 11-13 GHz look so simple is perhaps because there's not much required - the fets are stable on their own. The addition of extra inductance in the source terminals in the layout is likely to improve stability here at the expense of some gain.

The conclusion I've come to is that it's not straightforward to shift the F1OPA design down to 8.4GHz simply by appropriately scaling the WG transformer dimensions up from WR90 to WR112. However, all that said I'm going to give it a try :)
Regards,

R.

Hi Robert,

Indeed that also matches what I saw while simulating. I did spend some time testing source
inductance values back then, but concluded that any intentional inductance I placed
wouldn't globally help with stability.

Instead, I opted for simulating the layout geometry with the ADS field solver to make sure
I included the parasitic source inductance present in my layout - but it could be that I
was not so successful (evidenced by the fact that the design does not work). I have yet to
succeed replicating its instability in simulations...


Stefan

On 5/5/20 5:43 PM, Robert Watson via groups.io wrote:

Hi Folks,? Sadly not had much time to look at this lately. But can confirm experience of others - this FET really is on a knife edge. I added some degeneration source inductance to the bare FET in AWR to see what happens to the stability circles, it soon goes from oscillator to amplifier and back to oscillator again. Looks it will need some really quite careful analysis of the source via size, number and position.
R.

Hi Folks,? Sadly not had much time to look at this lately. But can confirm experience of others - this FET really is on a knife edge. I added some degeneration source inductance to the bare FET in AWR to see what happens to the stability circles, it soon goes from oscillator to amplifier and back to oscillator again. Looks it will need some really quite careful analysis of the source via size, number and position.
R.

Hi Paul,

I've started to look at this, but really only playing just to see if I understand the workflow.

I did harbour an idea of using a slightly different transition - I used it as a learning exercise just to check my code for the waveguide transformer calculations. I can pretty much replicate published results for a 60GHz, WR10 guide (see pic from paper below "A Simplified Design Inline Microstrip-to-Waveguide Transition ), but have now realised I it's probably not a great idea. The reason being is that the wave impedance is of the order of 300ohms and the dielectric waveguide wave impedance is very low, like 7 ohms. To match to the FET you'd need to get this back up with another matching network - defeating the objective of minimising losses. If I'd thought about a bit, it's obvious - the microstrip feed to the metal waveguide section is so wide (thus the impedance of the dielectric waveguide feed is obviously a lot less than 50 ohms).



Undeterred by this, I modelled a WR112 waveguide transformer along the similar lines, the holes in the face are a UBR84-PF flange.? I drew it up in Inventor using dimensions calculated using equations in Pozar and exported as a .step file. This was imported into HFSS and simulated...

? ? ?

I'm very rusty at this kind of stuff, I'm sure there's some seasoned HFSS veterans cringing. The input port 1 is the thin dielectric slot at the back (Rogers 4003C) and port 2 is a standard waveguide port. Waveguide body itself is modelled as 6061 aluminium. It gives vaguely sensible looking S-parameters although shifted off-frequency. This have given me a bit more confidence to give the F1OPA design a try. The equations for the WG transformer design are quite simple, few lines of Matlab.

Regards,

R.