In article ,
David wrote:

On Sun, 04 Nov 2007 00:17:26 -0000, wrote:

On Nov 3, 5:49 pm, Telamon
wrote:
In article .com,

wrote:
If you need to prototype or build a lumped element RF filter, check
out the RF Filter PCBs at RF Bites.

http://www.rfbites.com

Those are not very good coax to PCB connectors. You can do better. Must
be pretty low frequency stuff ( 1 GHz ) if you are using FR4.

--
Telamon
Ventura, California

Connectors are rated to 12GHz.

FR-4 is good up to 3.5GHz, then it starts to fall apart.

http://en.wikipedia.org/wiki/FR4

There are much better places to go on the Internet to read about PCB
dielectrics then Wikipedia.

You will notice that FR4 has a high dielectric constant together with a
poor loss tangent means it is very lossy at high frequencies.

--
Telamon
Ventura, California

In article om,
wrote:

On Nov 5, 6:21 am, David wrote:
On Sun, 04 Nov 2007 00:17:26 -0000, wrote:
On Nov 3, 5:49 pm, Telamon
wrote:
In article .com,

wrote:
If you need to prototype or build a lumped element RF filter, check
out the RF Filter PCBs at RF Bites.

http://www.rfbites.com

Those are not very good coax to PCB connectors. You can do better. Must
be pretty low frequency stuff ( 1 GHz ) if you are using FR4.

--
Telamon
Ventura, California

Connectors are rated to 12GHz.

FR-4 is good up to 3.5GHz, then it starts to fall apart.

http://en.wikipedia.org/wiki/FR4- Hide quoted text -

- Show quoted text -

Hi David,

Good article. I partially agree with Talamon that FR4 is not ideal
over 1GHz, if you have the budget for Roger's material and the specs
that necessitate such performance in your substrate. In the commercial
world, you would have a hard time justifying such an expense, so most
designers have to eat the performance issues and work a little harder
to develop more robust designs to handle the gross tolerance swings. I
would challenge anyone to find a PCS cell phone (1.9GHz) or a
commercial GPS with Roger's material. When I said that the FR-4
material was good up to 3.5GHz, that means I am getting good rejection
in my filters up to that frequency. I wouldn't design my passband to
be anything much over 2 GHz. Isn't there a saying that goes something
like, "Any average Engineer can design a $50,000 car, it takes a good
Engineer to design a $15,000 car." Food for thought.

You always have two major ways to loose signal, which are conductor and
dielectric in PCB and coax. Either type of loss can dominate depending
on materials and construction.

--
Telamon
Ventura, California

On Mon, 05 Nov 2007 18:08:00 -0000, wrote:

On Nov 5, 6:21 am, David wrote:
On Sun, 04 Nov 2007 00:17:26 -0000, wrote:
On Nov 3, 5:49 pm, Telamon
wrote:
In article .com,

wrote:
If you need to prototype or build a lumped element RF filter, check
out the RF Filter PCBs at RF Bites.

http://www.rfbites.com

Those are not very good coax to PCB connectors. You can do better. Must
be pretty low frequency stuff ( 1 GHz ) if you are using FR4.

--
Telamon
Ventura, California

Connectors are rated to 12GHz.

FR-4 is good up to 3.5GHz, then it starts to fall apart.

http://en.wikipedia.org/wiki/FR4- Hide quoted text -

- Show quoted text -

Hi David,

Good article. I partially agree with Talamon that FR4 is not ideal
over 1GHz, if you have the budget for Roger's material and the specs
that necessitate such performance in your substrate. In the commercial
world, you would have a hard time justifying such an expense, so most
designers have to eat the performance issues and work a little harder
to develop more robust designs to handle the gross tolerance swings. I
would challenge anyone to find a PCS cell phone (1.9GHz) or a
commercial GPS with Roger's material. When I said that the FR-4
material was good up to 3.5GHz, that means I am getting good rejection
in my filters up to that frequency. I wouldn't design my passband to
be anything much over 2 GHz. Isn't there a saying that goes something
like, "Any average Engineer can design a $50,000 car, it takes a good
Engineer to design a $15,000 car." Food for thought.

FH
I cheat. Anatech makes my filters.

On Mon, 05 Nov 2007 17:40:58 -0800, Telamon
wrote:


You always have two major ways to loose signal, which are conductor and
dielectric in PCB and coax. Either type of loss can dominate depending
on materials and construction.

Series, shunt?

In article ,
David wrote:

On Mon, 05 Nov 2007 17:40:58 -0800, Telamon
wrote:


You always have two major ways to loose signal, which are conductor and
dielectric in PCB and coax. Either type of loss can dominate depending
on materials and construction.

Series, shunt?

That's the wrong way to look at it. Coax cable and a micro-strip-line on
a PCB are examples of transmission lines.

--
Telamon
Ventura, California

On Wed, 07 Nov 2007 02:04:57 GMT, Telamon
wrote:

In article ,
David wrote:

On Mon, 05 Nov 2007 17:40:58 -0800, Telamon
wrote:


You always have two major ways to loose signal, which are conductor and
dielectric in PCB and coax. Either type of loss can dominate depending
on materials and construction.

Series, shunt?

That's the wrong way to look at it. Coax cable and a micro-strip-line on
a PCB are examples of transmission lines.
Conductor = series
Dielectric = shunt

A transmission line is a bunch of capacitors and resistors.

In article ,
David wrote:

On Wed, 07 Nov 2007 02:04:57 GMT, Telamon
wrote:

In article ,
David wrote:

On Mon, 05 Nov 2007 17:40:58 -0800, Telamon
wrote:


You always have two major ways to loose signal, which are conductor and
dielectric in PCB and coax. Either type of loss can dominate depending
on materials and construction.

Series, shunt?

That's the wrong way to look at it. Coax cable and a micro-strip-line on
a PCB are examples of transmission lines.
Conductor = series
Dielectric = shunt

A transmission line is a bunch of capacitors and resistors.

A transmission line can be visualized as a series of LC not RC circuits.

--
Telamon
Ventura, California

Telamon wrote:

In article ,
David wrote:

snip

A transmission line is a bunch of capacitors and resistors.


A transmission line can be visualized as a series of LC not RC circuits.


Depending on physical scale, R or L may dominate in the models, please
see: http://sigcon.com/Pubs/news/7_01.htm "chip-scale transmission lines"

Regards,

Michael

In article ,
msg wrote:

Telamon wrote:

In article ,
David wrote:

snip

A transmission line is a bunch of capacitors and resistors.


A transmission line can be visualized as a series of LC not RC circuits.


Depending on physical scale, R or L may dominate in the models, please
see: http://sigcon.com/Pubs/news/7_01.htm "chip-scale transmission lines"

Classic transmission line theory is loss-less LC, which is a description
of constrained path propagation. R is a parasitic in real life models
not a part of the description of how the path theoretically functions.

Transmission lines in semiconductors and the packages they go into have
an extreme constraint placed upon them that does not occur any place
else.

As I stated earlier in the thread that was sniped out "You always have
two major ways to lose signal, which are conductor and dielectric in PCB
and coax. Either type of loss can dominate depending on materials and
construction."

The conductor losses are due to the resistance heating of the conductor
and the dielectric losses come from the electric field heating the
dielectric. Both loss types are dissipative and increase with frequency.

--
Telamon
Ventura, California

On Wed, 07 Nov 2007 17:51:20 -0800, Telamon
wrote:

In article ,
David wrote:

On Wed, 07 Nov 2007 02:04:57 GMT, Telamon
wrote:

In article ,
David wrote:

On Mon, 05 Nov 2007 17:40:58 -0800, Telamon
wrote:


You always have two major ways to loose signal, which are conductor and
dielectric in PCB and coax. Either type of loss can dominate depending
on materials and construction.

Series, shunt?

That's the wrong way to look at it. Coax cable and a micro-strip-line on
a PCB are examples of transmission lines.
Conductor = series
Dielectric = shunt

A transmission line is a bunch of capacitors and resistors.

A transmission line can be visualized as a series of LC not RC circuits.

My point being it can be analyzed as a network of discrete components.