EMRFD Message Archive 8423

Message Date From Subject
8423 2013-03-28 14:34:43 ha5rxz Shielding Data Connections In RF Modules
We've covered the subject of screened modules before. Wes dealt with it when he detailed the construction of his spectrum analyzer and a few of us mentioned it again when the subject of Direct Conversion Receivers came up. The technology is known, seam soldered PC boxes or diecast boxes with decoupled power supply and control lines using feed through capacitors.

Easy.

Now what do we do with data connectors that enter an RF module? It's bad enough when we have a low speed parallel bus controlling something like a synthesizer, feed through caps might do the job but we're going to use a lot of them. Now, how about high speed connections such as SPI or I2C? If we use a 1000pF feed through this will load down the bus, cause timing errors and may even kill it altogether.

So what's the solution? We need to decouple the data lines to stop rubbish getting in and spoiling the phase noise figure, plus we need to stop local oscillator leakage getting out through everything except the output coax. I'm after some opinions here, especially from those with a few decades of experience in RF.

73

Peter HA5RXZ
8424 2013-03-28 15:30:14 William Carver Re: Shielding Data Connections In RF Modules
With SPI you can drop the clock rate to a shielded module. That's no big
deal if your digital signal is, for example, bandswitching signals. When
it's tuning numbers for a DDS you have to trade off your desired tuning
rate v.s. the amount of filtering you can do.

I use a tiny-scrap PCB on inside and outside of a box with SMD caps as
bypasses, and a -61 ferrite bead in the metal wall as a series element.
Then put a series resistor to set a known source impedance. You can
tailor this R and the bypass C's to produce a critically damped digital
waveform.

Actually, I make a string of holes. And although the PC material is
still a "tiny-scrap", I put as many SMD bypasses on it as I need. I wrap
copper tape around the edge of the board on the ground side of the caps
so when it's bolted to the box to reduce the ground impedance.

On the outside board I also put a strip of 0.1" spacing, .025" square
pins for Molex SIP conectors.

The only fly in the ointment: the connectors are 0.1" centers, you can
make the bypass caps on 0.1" centers, but the common size of ferrite
beads diameter is more than 0.1". DAMN! What I do is make two staggered
rows of holes for the beads.

To keep RF inside the enclosure you could put a "tiny logic" driver on
the outside board along with that series resistor. It has some amount of
reverse isolation that would help keep RF inside the box from leaking
out.

Bill W7AAZ





I've used -61 ferrite beads...lower u and HF...


On Thu, 2013-03-28 at 21:34 +0000, ha5rxz wrote:
>
> We've covered the subject of screened modules before. Wes dealt with
> it when he detailed the construction of his spectrum analyzer and a
> few of us mentioned it again when the subject of Direct Conversion
> Receivers came up. The technology is known, seam soldered PC boxes or
> diecast boxes with decoupled power supply and control lines using feed
> through capacitors.
>
> Easy.
>
> Now what do we do with data connectors that enter an RF module? It's
> bad enough when we have a low speed parallel bus controlling something
> like a synthesizer, feed through caps might do the job but we're going
> to use a lot of them. Now, how about high speed connections such as
> SPI or I2C? If we use a 1000pF feed through this will load down the
> bus, cause timing errors and may even kill it altogether.
>
> So what's the solution? We need to decouple the data lines to stop
> rubbish getting in and spoiling the phase noise figure, plus we need
> to stop local oscillator leakage getting out through everything except
> the output coax. I'm after some opinions here, especially from those
> with a few decades of experience in RF.
>
> 73
>
> Peter HA5RXZ
>
>
>
>
>
8425 2013-03-29 04:52:10 dixonglennb Re: Shielding Data Connections In RF Modules
Hi Peter, Bill,

Bypass caps to ground can often hurt more than they help as they cause current spikes on logic transitions that can be quite large with today's fast and beefy CMOS logic. Even though they slow down the voltage of the logic edge as viewed on an oscilloscope, it is often the currents in a circuit that nab you.

A solution to try is a series resistor right at the entry point into the shielded section. Because the input capacitance of most SPI logic is very small and there is hysteresis on the clock line, you can get by with quite large resistances. For example, in a recent PLL application that used a National (now TI) synthesizer chip with SPI input, slowing the SPI clock and using 22K series resistors on logic lines kept digital noise out of the synthesizer. You must make sure that whatever you use will still give good logic transitions and timing at the speed your SPI/I2C circuit is clocking at.

73,
Glenn AC7ZN
8426 2013-03-29 07:25:15 William Carver Re: Shielding Data Connections In RF Modules
EXACTLY what I was saying Glenn!

Not only does the resistor perform a current limiting function for the
logic chip, it performs a "damping" function: when you need the highest
possible data rate through the filter the series resistor value can be
adjusted so it, and the filter of series bead (R+jX) and bypass
capacitors form a critically damped network.

IE, adjust resistor and capacitor values and get the best signal
bandwidth for a given amount of filtering capacitance

In almost all cases today we are blessed with the MOS logic inside the
shielded box, drawing zero current, so we can add resistance without any
reduction in the final value of logic voltage at the destination.

W7AAZ


On Fri, 2013-03-29 at 11:52 +0000, dixonglennb wrote:
>
>
>
> Hi Peter, Bill,
>
> Bypass caps to ground can often hurt more than they help as they cause
> current spikes on logic transitions that can be quite large with
> today's fast and beefy CMOS logic. Even though they slow down the
> voltage of the logic edge as viewed on an oscilloscope, it is often
> the currents in a circuit that nab you.
>
> A solution to try is a series resistor right at the entry point into
> the shielded section. Because the input capacitance of most SPI logic
> is very small and there is hysteresis on the clock line, you can get
> by with quite large resistances. For example, in a recent PLL
> application that used a National (now TI) synthesizer chip with SPI
> input, slowing the SPI clock and using 22K series resistors on logic
> lines kept digital noise out of the synthesizer. You must make sure
> that whatever you use will still give good logic transitions and
> timing at the speed your SPI/I2C circuit is clocking at.
>
> 73,
> Glenn AC7ZN
>
>
>
>
>
8427 2013-03-29 09:25:19 Richard Johnson Re: Shielding Data Connections In RF Modules
Hello:

If the desire is for digital serial clocks, data streams or control signals to not generate interference with the analog world it should be know that it is the rise/fall times of these signals that makes a difference, not the absolute switching frequency. To keep a digital signal square the bandwidth needed is approximately 0.35 divided by the 10 to 90 rise time in seconds. So a 1ns rise time needs a bandwidth of 0.35/1e-10 = 350MHz. So the key is to use slow rise/fall time drivers. Another solution is to use differential drivers and twisted wire. I smaller differential delta is best so i would go with 3.3V LVDS drivers and receivers. However this does not work for I2C as those line need to be open collector/drain.

cheers,
rich
8428 2013-03-30 06:13:28 ha5rxz Re: Shielding Data Connections In RF Modules
Thanks for all the ideas so far.

One idea I thought of yesterday was to use a series of alternating grounded vertical elements similar to an interdigital filter and pass a ribb
8430 2013-03-31 18:16:12 Todd F. Carney / ... Re: Shielding Data Connections In RF Modules
I've dismantled a number of computer items such as printers and scanners,
and I usually find ferrites made to slip over ribbon cable. The cable
is enclosed
on all sides by ferrite at least 3/4" wide. Here's one of many sources:

http://www.lairdtech.com/Products/EMI-Solutions/Ferrite-Products/Ferrite-Cable-Cores/

Also Google "ribbon cable ferrites." Mouser has some.

I don't know whether or not this will keep trash out of the shielded RF
modules as well as feed-throughs, but it's worth a try. For high-speed 2,
3, or 4 wire serial signals, how well would twisted shielded pair (TSP)
work? Or RG-174 for each digital signal line, grounded on one end only?

Todd
------------------------------------------------------------------------------------
K7TFC / Medford, Oregon, USA / CN82ni / UTC-8
------------------------------------------------------------------------------------
QRP (CW & SSB) / EmComm / SOTA / Homebrew / Design


8439 2013-03-31 20:42:07 William Carver Re: Shielding Data Connections In RF Modules
Ferrites...beads or things made for specific size/shape cables...add a
SERIES impedance to RF and fast digital edged signals on the cable.

Feedthrough or SMD bypass capacitors produce low SHUNT impedance to
ground for RF and those fast edges.

A serious filter for a wire carrying control signals would use BOTH
series AND shunt impedances...series ferrites AND bypass capacitors...to
get the most filtering with the least degradation of the signals on the
wire.

W7AAZ