Circuit Design at 30Ghz

Or even *aluminium* as we call it here in liddle ole Britland. ;-) But seriously, your answer is too parochial. I'd just like answers in general terms if possible.

I worked on a 30ghz downlink/20ghz uplink satellite receiver/xmt for JPL back in 2002 and we used open die in alumina substrate housed in gold-plated alumina fixtures. Nothing magic about it if you got the funding...

I meant aluminum fixtures... ;)

basically you use die chips instead of packaged IC chips. The cost is higher because you need wire bonding equipment instead of a soldering iron... ;)

That's as general as you can get in this subject.

30GHz and up is sometimes known a "millimetre wave", 'cause the free space wavelength is 10mm or less. I've never done any design work at that frequency, although I have designed (lower frequency) parts of mm wave systems (some of which are possibly still in orbit).

One of the non-obvious considerations that you might have missed is that dielectric loss is a bitch at these frequencies - you will need to use either waveguide or (if using microstrip) something like an alumina (Al2O3) substrate.

Regards, Allan

Apart from the dielectric losses, you have to take every geometric feature into account that is bigger than say 0.05mm, everything is causing a delay and/or a phaseshift. I'm not involved in these designs though.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net

1 to 2 GHz circuitry is well within the reach of homebrew gear, using regular FR4 circuit boards and, say, 0805 sized parts, or even tightly constructed haywire circuits over a copperclad ground plane.

Above, very roughly, 3 GHz, things start to change. The parasitic inductance of capacitors and chip wire bonds starts to make things difficult, so you have to start using tiny discrete parts. PCB trace losses start to get bad too, so you have to keep all traces short and/or go to low-K, low-loss pcb materials.

In the 10's of GHz, parasitics and board losses get crazy, so circuits get to be hybrids or monolithics. Eventually, for RF stuff, waveguides become the lowest-loss way to transport signals.

Coaxial cable and connectors have to keep getting smaller as frequency goes up, to keep the mode TEM. So ohmic and losses keep increasing, and eventually coax becomes useless. RG58 and BNC connectors still work at 2 GHz, SMA's and RG174 type stuff to at least 10 GHz, if it's short.

HP and Tek had a battle going on to make faster sampling oscilloscopes, but seemed to declare a truce at 70 GHz. I suspect that's because it gets impossible to accurately transmit a signal any practical distance (even a few inches) through coax at such speeds. LeCroy has a 100 GHz scope now (using the PSPL shockline sampler) but Tek and Agilent don't seem to be following.

When I started doing fast stuff, I went to flea markets and bought junk Avantek microwave assemblies, Tek/HP sampling gear and manuals, and studied them. There's a lot of that, along with cables and connectors, on ebay these days.

John

Hello Max,

Alumina substrate designs aren't that heavy on the funding. Hybrids were the first projects I had to do in my career. Basically good old CAD layout except that the CAD program must be able to handle more than two "component" layers because resistors and small caps are screened on, then laser trimmed. We kind of tossed the CAD aside because the horn was going all the time about design rule issues and all that.

The real nice feature is active laser trim where you can automatically fine-tune circuit performance while it is running. When I had to do a "normal" board after that I felt some withdrawal symptoms having to give up such luxuries.

GHz stuff is tough. But just imagine: You design inductors and other parts onto the substrate and then a laser tunes all those filters, phase shifters and what not to whatever precision you want. Bzzzzt .... done.

Regards, Joerg

Paul,

look up "distrubuted vs lumped elements.

At low freq we use lunped elements, i.e L and Cs that are actual components.

At 30 GHz everything is a distributed circuit, the Ls and Cs are made of various lengths and Zs of transmission lines.

Mark

To see how 10 GHz is done, open an old Ku band LNB.

LNBs seem to be seriously low noise designs - and they cost next to nothing.

Thomas

Yeah, but that's the case at much lower frequencies than 30Ghz. I was wondering what other - somewhat more unforeseen to the outsider - problems one might have to take account of at seriously high frequencies.

That's because half the parts are missing! There's just some funny shaped PCB tracks! And I'm sure my autorouter could do a neater job that that, they're all bendy...

--

John Devereux

It seems that almost every conference that even tangentially refers to microwave design always has someone talking about the manufacturing cost of consumer electronics involving them (and of course how to drive them down). Spun aluminum dishes like the satellite TV guys use (DirectTV & Dish Network) have been one of the big winners, and a current popular topic is building waveguides out of (metal coated) plastic or other cheap, moldable materials rather than having to machine something expensive like invar.

It's fascinating stuff!

---Joel Kolstad

Considering all these are offset-feed and many are optimized for multiple feeds (to see multiple satellites), I don't think they are "spun". I am assuming you were talking about how the dishes are manufactured, not "spin-off technology".

Surely all the consumer dishes one sees on houses and apartments are stamped or molded.

All the rest you mention like waveguide and lnb construction is no doubt accurate and into serious cost cutting while preserving performance efforts.

Well, the skin depth of electric conduction is generally smaller than the surface roughness at those frequencies, so the effective resistance of your transmission lines is anomalously high. You can also get into weird things like air absorption.

Cheers,

Phil Hobbs

Love it! Nice one, John.

Regards Ian