Mail Archives: geda-user/2016/09/02/06:01:43
On Thu, Sep 01, 2016 at 12:42:32PM -0400, Evan Foss (evanfoss AT gmail DOT com) [via geda-user AT delorie DOT com] wrote:
> On Thu, Sep 1, 2016 at 3:51 AM, Gabriel Paubert (paubert AT iram DOT es) [via
> geda-user AT delorie DOT com] <geda-user AT delorie DOT com> wrote:
> > On Wed, Aug 31, 2016 at 09:25:23PM -0700, Larry Doolittle wrote:
> >> Evan -
> >>
> >> On Thu, Sep 01, 2016 at 12:07:28AM +0000, Evan Foss (evanfoss AT gmail DOT com) [via geda-user AT delorie DOT com] wrote:
> >> > On Wed, Aug 31, 2016 at 9:39 PM, Larry Doolittle <ldoolitt AT recycle DOT lbl DOT gov> wrote:
> >> > > On Wed, Aug 31, 2016 at 03:53:25PM -0400, Evan Foss (evanfoss AT gmail DOT com) [via geda-user AT delorie DOT com] wrote:
> >> > >> I thought soldermask issues were why most of the microwave and other
> >> > >> RF hardware I have seen has no soldermask.
> >> > > For the long lines, yes, depending on the designer, but eventually
> >> > > those lines usually touch chips. And high-speed serial lanes (with
> >> > > the interesting frequency range now going up to 40 GHz) usually don't
> >> > > have soldermask stripped off.
> >> > At those frequencies I would actually think the chip would be supplied
> >> > as bare die and wire bonded.
> >>
> >> These days Xilinx BGAs have I/O rated for 25 Gb/s, with pretty broad spectral
> >> content up to 40 GHz. I'm used to microwave circuits mostly ranging from
> >> 1 - 3 GHz, and all the Mini-Circuits and Hittite chips that cover that range
> >> (and higher) are in relatively conventional packages.
> >
> > Even to higher frequencies, we've been building here amplifiers with
> > chips like Hittite's HMC462LP5 and using them up to 12GHz. They are
> > specified to up 20GHz, but I can't test that high because we reach the
> > limits of the connectors/launchers on the boards we've built.
>
> Cool.
>
> > At these frequencies, the choice of the substrate is important, we use
> > 20mil thick Rogers RO4350 (or RO4003, can't remember) but the transition
> > from coaxial cable to the controlled impedance track is the most critical
> > part: I've had once an enclosure with a mistake of 0.5 or 0.6 mm in the
> > depth of the hole in which a glass bead is inserted. The effect on the
> > SWR was enormous, catastrophic at 10GHz and higher.
>
> Glass bead?
Yeah:
http://www.fusite.com/en-US/Products/RF_Products/50_Ohm/Pages/RF50OhmFeedthrus.aspx
A long time ago, Thunderline Z was an independent company and they were
very nice people. It's much worse to my knowledge since Emerson bought
them. Fortunately at the time I bought several hundred of the glass
beads we needed and I still have stock.
Once the glass bead is inserted into the enclosure's wall, you can put
on top either male of female connectors (with care, the pin of the glass
bead is a bit fragile), for example:
http://deltarf.com/images/pdf/SMA%20-%2026.5%20GHz_cat.pdf
http://www.dynawave.com/products/display?product=connectors&seriesid=2&styleid=15&bsaseriesid=-1&typeid=24&usebsaseriesid=False
or even straight cable attachment, which saves connector and is very
useful intight spaces:
http://www.dynawave.com/products/display?product=connectors&seriesid=20&styleid=16&bsaseriesid=-1&typeid=24&usebsaseriesid=False
The female connectors I bought from the first reference above (Delta RF)
were not expensive (between $4 and $4.5 if memory serves) for model
1396000K921-023-EMI. Less than $5 for a good quality (not precision
either) RF connection with connectors easy to repace if they wear out
(happens in the lab when a board is connected and disconnected tens or
hundreds of times) or if it turns out to be better to change the
connector's sex.
Speaking of glass beads, most manufacturers of RF connectors only offer
a few glass bead models, only reselling a few selected references from
the humongous choice offered by Thunderline Z, and at an outrageous
markup (20x or so if I remember correctly!).
The first time I bought glass beads (1994, I'm getting old), I selected
18mil for the contact diameter because I did not think I would ever need
to measure circuits above 18GHz (4.6GHz at the time) and the smallest
hole when machining the connector is 1.1mm (the one traversing the
enclosure where the transmission line becomes an air line). I did not
want to use the smaller 15mil contact diameter because mechanical
dimensions become about 20% more critical.
I later learnt that 15mil is more popular and the day we ordered 18mil
SMA male connectors here in Spain, we had to send them back because they
actually provided us 15mil models. That's a frequent problem here, the
Spanish distributors think that they know better than the customers what
they need.
>
> > For the controlled impedance tracks, we typically use conductor backed
> > coplanar waveguide, taking into account the effect of the cover of the
> > shielded enclosure on the impedance.
>
> Do you know of a publicly available design guide for those? I was
> thinking about how pcb-rnd's new drc system should handle them last
> night.
No, I started a long time ago, I bought a book and wrote a few programs
(not at all designed for public consumption) that implement the formulas
published in:
Coplanar Waveguide Circuits, Components, and Systems
Rainee N. Simons
Wiley
ISBN 0-471-16121-7
These days you find interactive calculators on the web.
>
> > And no these boards don't have solder masks, for other boards, where I
> > also had to wire bond a diode, the solder mask was mostly removed along
> > the tracks and the edges of the ground plane. Some was left as solder
> > stops to limit solder flow along the track.
>
> I knew it!
>
> > Gabriel
>
> Thanks,
> Evan
Cheers,
Gabriel
- Raw text -