Maker Pro
Maker Pro

Rules of thumb for inter-stage RF shielding?

A

Andy Iakovlev

Jan 1, 1970
0
Hello:

I am looking for some rules of thumb for RF shielding between the
different stages. In the device I am currently designing there will be:

1) a DDS chip, 100kHz - 175 MHz output;
2) a fully-differential elliptic low pass filter, with 175MHz passband
and 70dB rejection after 225 MHz; applied to the output of 1)
3) a VGA with 50dB control range, applied to the output of 2)
4) a fixed 25dB gain stage, applied to the output of 3)
5) output level detector stage, applied to the output of 4)
6) level-control integrator closing the AGC loop, between 5) and 3),

and I need to choose a minimum number of shielded enclosure cells, to
put the above into.

Thanks in advance for your advice.

-- Andy
 
J

Joerg

Jan 1, 1970
0
Hi Andy,

Much can be achieved using a full ground plane, clever parts placement
and the avoidance of parallel traces. You can also bury traces within
power and ground planes. My guess would be that shielding may be
necessary within the stages of your elliptic filter because of the 70dB
rejection requirement. The 50dB amp might also need a little shield
between input and output. Often this can be a simple metal piece that
acts as a wall. Full enclosure may not be needed although you can buy
the parts for that.

If you opt for little enclosures make sure to close them with a few
solder spots since the simple clamp-on lids tend to work themselves
loose on transport.

Regards, Joerg
 
J

John Woodgate

Jan 1, 1970
0
I read in sci.electronics.design that Andy Iakovlev <smdg_remove-
this@and-this_worldonline_this-also.fr> wrote (in <41361d75$0$29664$636a
[email protected]>) about 'Rules of thumb for inter-stage RF
shielding?', on Wed, 1 Sep 2004:
Hello:

I am looking for some rules of thumb for RF shielding between the
different stages. In the device I am currently designing there will be:

1) a DDS chip, 100kHz - 175 MHz output;
2) a fully-differential elliptic low pass filter, with 175MHz passband
and 70dB rejection after 225 MHz; applied to the output of 1)
3) a VGA with 50dB control range, applied to the output of 2)
4) a fixed 25dB gain stage, applied to the output of 3)
5) output level detector stage, applied to the output of 4)
6) level-control integrator closing the AGC loop, between 5) and 3),

and I need to choose a minimum number of shielded enclosure cells, to
put the above into.

Are you thinking of making the cell walls from 10 mm wide tinplate
strip? One strip around the edge, on each side of the board, with
internal partitions of the same material soldered across? Simple flanged
tinplate covers, probably soldered on unless you really want to give
easy access.

Assuming you also have to meet EMC requirements, you almost certainly
need six cells. You might be able to put 3 and 4 in one cell, but with
maybe 75 dB gain (you don't actually say what the maximum gain of the
VGA is), I wouldn't risk it for the sake of two bits of bent tin.

But the physical arrangement matters a great deal, Forget square; go for
a long, thin layout, so that the peripheral shielding acts as a
waveguide below cut-off. That is a big help in preventing the output
getting back into earlier stages.
 
A

Andy Iakovlev

Jan 1, 1970
0
John said:
Are you thinking of making the cell walls from 10 mm wide tinplate
strip? One strip around the edge, on each side of the board, with
internal partitions of the same material soldered across? Simple flanged
tinplate covers, probably soldered on unless you really want to give
easy access.

Do you mean that the partition walls would be necessary on both sides of
the PCB (most if not all the components will be SMD)? How big could be
the area of the opening in the walls?

I was to potentially go as far as splitting the board into the pieces,
and placing each in a cell of an enclosure like the one shown on the
pictures I've posted in a.b.s.e (Message-ID:
<[email protected]> , subject: "Shielded
enclosure"). The second picture shows the 2 available profiles of the
adjustable internal walls, and of the corresponding enclosure.


But then I would face the problem of properly interconnecting the
sub-boards. If less extreme measures were enough, that would be of
course better!
Assuming you also have to meet EMC requirements, you almost certainly
need six cells. You might be able to put 3 and 4 in one cell, but with
maybe 75 dB gain (you don't actually say what the maximum gain of the
VGA is),

The maximum gain of the VGA is 13dB only.

But the physical arrangement matters a great deal, Forget square; go for
a long, thin layout, so that the peripheral shielding acts as a
waveguide below cut-off. That is a big help in preventing the output
getting back into earlier stages.

Thank you!

-- Andy
 
A

Andy Iakovlev

Jan 1, 1970
0
Hi Joerg,


Hi Andy,

Much can be achieved using a full ground plane, clever parts placement
and the avoidance of parallel traces. You can also bury traces within
power and ground planes. My guess would be that shielding may be
necessary within the stages of your elliptic filter because of the 70dB
rejection requirement.

For the filter, do you think that the PCB - level shielding (groung
tracks between the stages) would be enough, or some tinplate walls would
be required?

The 50dB amp might also need a little shield
between input and output. Often this can be a simple metal piece that
acts as a wall. Full enclosure may not be needed although you can buy
the parts for that.

If you opt for little enclosures make sure to close them with a few
solder spots since the simple clamp-on lids tend to work themselves
loose on transport.

Thank you for the advice!

Regards,
-- Andy
 
J

John Woodgate

Jan 1, 1970
0
I read in sci.electronics.design that Andy Iakovlev <smdg_remove-
this@and-this_worldonline_this-also.fr> wrote (in <4136ebc0$0$8339$626a1
[email protected]>) about 'Rules of thumb for inter-stage RF shielding?',
For the filter, do you think that the PCB - level shielding (groung
tracks between the stages) would be enough, or some tinplate walls would
be required?

Start with the walls, and eliminate them later if you can. Then you get
two credits - a 'works first time' one and a 'cost improvement' one. The
other way, you get 'long development delay' and 'cost uplift'. It's a
no-brain choice.

I'm tempted to say ALWAYS start 'fireproof' and then cut back, but some
'fireproof' solutions won't cut back much, and the guy who sees that one
bipolar can replace four chips wins out.
 
J

John Woodgate

Jan 1, 1970
0
I read in sci.electronics.design that Andy Iakovlev <smdg_remove-
this@and-this_worldonline_this-also.fr> wrote (in <4136eb0c$0$8333$626a1
[email protected]>) about 'Rules of thumb for inter-stage RF shielding?',
Do you mean that the partition walls would be necessary on both sides of
the PCB (most if not all the components will be SMD)?

Yes. The electric and magnetic fields spread out from the surfaces. 10
mm depth is convenient if you have room for it, but you can cut it down.
Make sure the lid doesn't short-circuit anything, even if someone
presses it in.
How big could be
the area of the opening in the walls?

Slightly smaller than you need. (;-) Seriously, the openings act as slot
antennas, so provided they are small compared with the wavelength of the
highest likely harmonic of any signals, they won't have a significant
effect.
I was to potentially go as far as splitting the board into the pieces,
and placing each in a cell of an enclosure like the one shown on the
pictures I've posted in a.b.s.e (Message-ID: <4136e5ee$0$8333$626a14ce@n
ews.free.fr> , subject: "Shielded enclosure"). The second picture shows
the 2 available profiles of the adjustable internal walls, and of the
corresponding enclosure.

I stopped taking a.b.s.e, but I'm now getting it to look at your
graphic. I don't like the idea of sub-boards: they result in everything
being interconnected by variable inductors (the interconnections).
But then I would face the problem of properly interconnecting the sub-
boards. If less extreme measures were enough, that would be of course
better!
Exactly.

The maximum gain of the VGA is 13dB only.

So MAYBE you can combine 3 and 4, but is it worth it to save one strip
of tinplate that you can leave out WHEN/IF you find it's not needed.
 
A

Andy Iakovlev

Jan 1, 1970
0
john said:
Assuming your filter is a 'matched' LC design

Yes, the input impedance and the load impedane are 124 Ohm (or much so).
I'd have thought there's no
way you'll keep a 70Db rejection if it's followed by 50Db of VGA.

The VGA's maximum gain is about +13dB, the minimum - about -37dB.
The VGA
design would need to be very sophisticated to retain a firm resistive (say)
50ohm input impedance over that amount of gain and frequency variation.

Could you expand a bit why? I am planning to use LMH6502 (pushing it a
bit on the high freq. end).
The
filter itself should also be a problem. If the inductors can 'see' each
other then it's curtains.

I'll try to evaluate, analytically, to which point they will see each other.
Sound like you'll need lots of tin plate to
seperate the filter elements,
which in turn will act with the tinplate to
generate yet another filter response, which then can't be simulated and can
only be prototyped on the bench.

Yea... much sport!
Essentially, lots of screening, lots of buffers and *lots* of bench work are
called for.
regards
john

Thanks!

Regards,
-- Andy
 
J

Joerg

Jan 1, 1970
0
Hi John,
Start with the walls, and eliminate them later if you can. Then you get
two credits - a 'works first time' one and a 'cost improvement' one. The
other way, you get 'long development delay' and 'cost uplift'. It's a
no-brain choice.
I'll second that. Very good advice indeed.

Regards, Joerg
 
J

Joerg

Jan 1, 1970
0
Hi Andy,
For the filter, do you think that the PCB - level shielding (groung
tracks between the stages) would be enough, or some tinplate walls
would be required?

Take John's advice to heart, in previous post.

The enclosure you posted as a photo on a.b.s.e. looks very good but you
may be able to use tin walls on the board itself. You can, or at least
used to be able to, obtain walls with a row of pins that are soldered
into grounded vias. When adhering to a certain geometric pattern you
could even get matching lids. But it has been a long time since I used
these.

With the enclosure shown it seems you would have to split the circuit
board into section to fit into the individual compartments. That becomes
very tedious. Also, if you do use an enclosure like the one shown don't
rely on the snap fingers alone to hold the lid. Sometimes it is best to
provide a few solder spots after the board is tested and ready to go.

Last but not least: If you ever cut this tin stuff with metal shears be
careful. I have some scars to show from the deep cuts this can cause.
Also, be careful with beryllium copper fingers that some boxes may come
with. Beryllium oxyde can be poisonous and it is usually not necessary.

Regards, Joerg
 
J

john jardine

Jan 1, 1970
0
Andy Iakovlev said:
Hello:

I am looking for some rules of thumb for RF shielding between the
different stages. In the device I am currently designing there will be:

1) a DDS chip, 100kHz - 175 MHz output;
2) a fully-differential elliptic low pass filter, with 175MHz passband
and 70dB rejection after 225 MHz; applied to the output of 1)
3) a VGA with 50dB control range, applied to the output of 2)
4) a fixed 25dB gain stage, applied to the output of 3)
5) output level detector stage, applied to the output of 4)
6) level-control integrator closing the AGC loop, between 5) and 3),

and I need to choose a minimum number of shielded enclosure cells, to
put the above into.

Thanks in advance for your advice.

-- Andy
Assuming your filter is a 'matched' LC design I'd have thought there's no
way you'll keep a 70Db rejection if it's followed by 50Db of VGA. The VGA
design would need to be very sophisticated to retain a firm resistive (say)
50ohm input impedance over that amount of gain and frequency variation. The
filter itself should also be a problem. If the inductors can 'see' each
other then it's curtains. Sound like you'll need lots of tin plate to
seperate the filter elements, which in turn will act with the tinplate to
generate yet another filter response, which then can't be simulated and can
only be prototyped on the bench.
Essentially, lots of screening, lots of buffers and *lots* of bench work are
called for.
regards
john
 
J

john jardine

Jan 1, 1970
0
Andy Iakovlev said:
Yes, the input impedance and the load impedane are 124 Ohm (or much so).


The VGA's maximum gain is about +13dB, the minimum - about -37dB.


Could you expand a bit why? I am planning to use LMH6502 (pushing it a
bit on the high freq. end).


I was assuming discretes. My comment may not apply to the IC. (I'll grab a
pdf).
 
A

Andy Iakovlev

Jan 1, 1970
0
John said:
Start with the walls, and eliminate them later if you can. Then you get
two credits - a 'works first time' one and a 'cost improvement' one. The
other way, you get 'long development delay' and 'cost uplift'. It's a
no-brain choice.

What I think I will do: I will plan for the walls, and provide for the
necessary space & material, but won't install them in the first pass.

I'm tempted to say ALWAYS start 'fireproof' and then cut back,

At some point the effort starts to seem incommensurate, if not backed by
the past experience! That's why I was looking for the advice on a
reasonable starting point. There are 6 inductors in the filter (2 in
each stage, it's a fully differential filter). Create 6 cells for a 7th
order lowpass filter would seem to me too much... :)

Thank you.
 
A

Andy Iakovlev

Jan 1, 1970
0
Hi Joerg,

I have looked up such walls (tin plated steel with pins) across a number
of distributors - have not found yet. I would prefer to use some
standard product rather than cut the walls myself. Would you have any
pointer to such product?

I foresee this as a set of inside-board walls, together with the
perimeter walls, with a common cover. Would the contact of the inside
walls with the cover be a serious concern? I.e. if there are no fingers
there, the contact would be in a couple of points for each wall, with
long narrow holes between the wall's border and the cover - elsewhere.


Thank you.

Regards,
-- Andy
 
A

Andy Iakovlev

Jan 1, 1970
0
John said:
I stopped taking a.b.s.e, but I'm now getting it to look at your
graphic. I don't like the idea of sub-boards: they result in everything
being interconnected by variable inductors (the interconnections).

I see. I have abandonned the idea of sub-boards for the moment.
So MAYBE you can combine 3 and 4, but is it worth it to save one strip
of tinplate that you can leave out WHEN/IF you find it's not needed.

I agree. This would be a really very small overdesign.
 
J

John Woodgate

Jan 1, 1970
0
I read in sci.electronics.design that Andy Iakovlev <smdg_remove-
this@and-this_worldonline_this-also.fr> wrote (in <4139a01c$0$31383$636a
[email protected]>) about 'Rules of thumb for inter-stage RF
shielding?', on Sat, 4 Sep 2004:
At some point the effort starts to seem incommensurate, if not backed by
the past experience! That's why I was looking for the advice on a
reasonable starting point. There are 6 inductors in the filter (2 in
each stage, it's a fully differential filter). Create 6 cells for a 7th
order lowpass filter would seem to me too much... :)

I didn't suggest cells within each building block; just one cell per
block. but for the filter it would be wise to provide for one cell per
stage, not one cell per inductor.
 
A

Andy Iakovlev

Jan 1, 1970
0
Andy said:
I'll try to evaluate, analytically, to which point they will see each
other.


I have done some computation. The source inductor is 1mm long, 0.8mm
diameter. I have evaluated the ratio of the magnetic field strengths at
a given measurement point and that at the center of the source inductor.

The coupling coefficient between the source inductor and a similar
inductor placed at the measurement point would be equivalent to that
ratio (not bigger than 5 times that? I would need to do more computation
to get more precision).

For the measurement point at 3mm from the inductor's center axis and in
the plane orthogonal to that axis going through the center of the
inductor - the ratio is about 0.003. So looks like I should be able to
place the 2 inductors of each stage of the filter (it's a fully
differential version of a 7th order low-pass elliptic filter) as close
to each other as 3mm without significant harm to the filter's
characteristic.

For the distance between the inductors of the successive stages, 5mm
seems ok, as the ratio is then about 0.00083

Does this sound correct?

I will need to get the inductors and do some real measurement, of
course, to confirm/deny this forecast.


Thanks,
-- Andy
 
A

Andy Iakovlev

Jan 1, 1970
0
No, you did not suggest cells within each bulding block. I came from the
exchange with J.Jardine that I might need walls within the filter -
which provides kind of an extreme example (as I feel it). Sorry for any
misunderstanding.
 
A

Andy Iakovlev

Jan 1, 1970
0
No, you did not suggest cells within each bulding block. I came from the
exchange with J.Jardine that I might need walls within the filter -
which provides kind of an extreme example (as I feel it). Sorry for any
misunderstanding.

Regards,
-- Andy
 
J

Joerg

Jan 1, 1970
0
Hi Andy,
I have looked up such walls (tin plated steel with pins) across a
number of distributors - have not found yet. I would prefer to use
some standard product rather than cut the walls myself. Would you have
any pointer to such product?

Sorry but I don't since these were made in-house for my designs.
Basically they are stamped for mass production and the vias need to be a
bit larger than usual to accomodate the wider "pins", which really are
just extensions of the metal.

One source might be a ham radio supplier. Another solution that is a bit
cumbersome in production would be wire loops in the board that connect
to the ground plane and the shield pieces are then soldered to these.
But that's not a nice way to do it.
I foresee this as a set of inside-board walls, together with the
perimeter walls, with a common cover. Would the contact of the inside
walls with the cover be a serious concern? I.e. if there are no
fingers there, the contact would be in a couple of points for each
wall, with long narrow holes between the wall's border and the cover -
elsewhere.

Fingers are fine, personally I'd just stay away from copper beryllium
(the cheap versions won't likely have that anyway). But I'd add a few
solder spots just to hold that cover in place. You'd be amazed what can
work itself loose even from a tight fit while riding on a propeller
aircraft or on a rough truck journey over washboard roads. You never
know where your products will be shipped some day.

Regards, Joerg
 
Top