RE: [SI-LIST] : FW: signal isolation, RF board

Anthony Tse (anthony.tse@nrl.navy.mil)
Fri, 30 Jan 1998 15:00:38 -0500

-----Original Message-----
From: Fred Townsend [SMTP:fred@luxtron.com]
Sent: Friday, January 30, 1998 2:08 PM
To: Anthony Tse
Cc: 'si-list@silab.Eng.SUN.com'
Subject: Re: [SI-LIST] : FW: signal isolation, RF board

Anthony Tse wrote:
>=20
> Hello all,
> I am in the process of designing a RF board (160MHz) and I need to =
isolate a low > level inputs from the output of a high gain amplifier. =
I re-read all the article in > the archive on guard-banding and =
shielding and didn't really get a good feel on how > to go about doing =
this. =20

You probably will not have a good feel until you have done it a few
times.

> Here are some questions:
> Realistically, how much isolation can I get? The traces in question =
are about 2" > > long and 4" apart.

This sounds like plenty for electric fields which are much easier to
shield. Magnetic will depend on the strength of the fields. The
strength of magnetic field will depend on the ac current in the source
(field) trace.=20

> The stacking is 2 & 4 ground, layer 3 for RF signal, and layer 1 for =
video signal. > Since there are very few traces on layer 3, should I =
solder pour the whole layer and > stitch the 3 ground planes together =
around all the sensitive signal traces? Or am I > better of using fat =
guard traces and stitch them to the ground plane.

Can you do both?

If I do a solder pour, there won't be room for the guard traces, or do =
you mean do guard traces, and on top of that, do a solder pour. I can =
do both, there is plenty of room. Will It buy me anything?
> Does it make sense to separate the two signals in two different =
layers? Say one of > the signals in layer 2 with 1 & 3 as the ground =
while the other signal is in layer 3 > with 2 & 4 as ground. Or should =
I bit the bullet and go to a 6 layer board, with > 1,3,4,6 as ground and =
2 & 5 as signal (50 ohm trace width may be a problem with a 6 > layer =
board).

I fail to understand your statement about 50 ohms since impedance tends
to converge near 50 ohms. I would understand you were in deep yogurt if
you had said 300 ohms.

I am worry about the very very thin line required for a 50 ohm line. =
The more the layers, the thinner the dielectric, the thinner the lines.

You are in great danger of throwing out the baby with the bath water.=20
Video and RF are spectrally separated from each other so they do not
normally interfere with each other. Keeping the signal currents low and
matching impedances will do a lot to reduce coupling as does
separation. You have failed to mention anything about power here. Are
these circuits sharing power supplies or grounds? You will have major
problems with power and ground bounce if either video or RF have
transient conditions. (Video is almost by definition non-periodic and
therefore transient.) Better to go to 6 or even 8 layers and avoid
sharing any power or grounds.

They have split ground planes. I don't see any coupling between 160MHz =
and video in my current board and it's really not a concern since the =
video will be low pass filtered on the next board before ADC. The =
problem is, the IF (160 MHz) limiting amplifier puts out +10 dBm of =
power and has a +60 dB gain. The board seem to be setting up resonance =
because w/o any signal input, I get a perfectly clean loud sine wave =
coming out of the amplifier. When I do put a signal in at -30 dBm, all =
I see is the resonance and not the amplified version of my signal. So =
my problem is, how do I keep the +10dBm output from getting back into =
the input of the amplifier.

Fred Townsend
DC to Light Consulting Services
San Jose, CA
(408) 263-8768

-Anthony