[Joerg]

Nico Coesel wrote:
[...]


14 mil thick and 26 mils wide moves a LOT of field up into the air!
changes the 4.2 to 3.2 effective. Not sure it is good to have that
floating around above your circuitry.

What do you mean by 'serious mismatches in circuits'?
A slightly different subject: AFAIK there are at least two ways to
determine mismatch. TDR or by using a directional coupler and see what
is reflected back. I'd like to get a setup with a directional coupler.

The above is a slightly different topic. Once I get this setup going
I'd like to use it for testing some odd ways to connect ethernet. RJ45
is not always the most practical connector.

Measure the output without the filter. You can do that without building
or buying anything. Then if you have determined that it is really the
filter you can add 10-20% or so to a cap here and there and see which
Actually there is no filter at the output

I ordered some RF detector schottky diodes. If all goes well they'll
arive tomorrow. I really need to figure out whether the problem is in
the SA or in my circuit.

The spectrum analyzer can easily be tested for that. Hang a preamplifier
onto it with nothing connected to it's input. This will raise the noise
floor. Most spectrum analyzers are sensitive enough by themselves
though, you can test that by shorting the input and seeing if the noise
floor changes. A dip in the noise floor at 2.45GHz should be clearly

I tried that but the noise floor stays flat at -70dBm.

That would mean the spectrum analyzer should be fine.

Anyway, I've tried the simple RF detector (
http://1.bp.blogspot.com/_vfmOyxDCr...AGpc/IZAMFN1e-_I/s1600/RF_Probe_Schematic.jpg
with a HSMS-286B detector diode).
The output isn't flat but there is no extreme dip either. When I
change the output power by 6dB the detector output also changes
accordingly. The dip is about -20dB which should result in the
detector output going near 0.

Now it seems to be down to either an aliasing effect between ADF4350
stepping and the analyzer sweep, or plain old WLAN interference. Did you
try putting the board into a shielding can or wrap it with plastic and
then an aluminum foil grounded to the coax that goes to the analyzer?

So far I'm leaning towards the SA being bad. I'll wait for the RF
detector chips to arrive and put one on a PCB to (hopefully) get a
better measurement than the simple detector provides and reach a final
verdict.

I'll side the the spectrum analyzer on it's "Not guilty" plea

Any references? Google seems useless on this subject

I don't have any because I just design them from scratch whenever I need
one. In ultrasound I kind of grew up with those things. Essentially just
a glorified lowpass where the caps are varicap diodes.

 

[Nico Coesel]

Nico Coesel wrote:

Nico Coesel wrote:
[...]


14 mil thick and 26 mils wide moves a LOT of field up into the air!
changes the 4.2 to 3.2 effective. Not sure it is good to have that
floating around above your circuitry.

What do you mean by 'serious mismatches in circuits'?
A slightly different subject: AFAIK there are at least two ways to
determine mismatch. TDR or by using a directional coupler and see what
is reflected back. I'd like to get a setup with a directional coupler.

The above is a slightly different topic. Once I get this setup going
I'd like to use it for testing some odd ways to connect ethernet. RJ45
is not always the most practical connector.

Measure the output without the filter. You can do that without building
or buying anything. Then if you have determined that it is really the
filter you can add 10-20% or so to a cap here and there and see which
Actually there is no filter at the output

I ordered some RF detector schottky diodes. If all goes well they'll
arive tomorrow. I really need to figure out whether the problem is in
the SA or in my circuit.

The spectrum analyzer can easily be tested for that. Hang a preamplifier
onto it with nothing connected to it's input. This will raise the noise
floor. Most spectrum analyzers are sensitive enough by themselves
though, you can test that by shorting the input and seeing if the noise
floor changes. A dip in the noise floor at 2.45GHz should be clearly

I tried that but the noise floor stays flat at -70dBm.

That would mean the spectrum analyzer should be fine.

Unless the majority of the noise comes from the amplifiers after the
first few stages. When I attenuate the input signal, the dip gets cut
off by the noise floor.

Now it seems to be down to either an aliasing effect between ADF4350
stepping and the analyzer sweep, or plain old WLAN interference. Did you

The stepping is not the problem either. When I set it to a fixed
frequency I get the same effect.

try putting the board into a shielding can or wrap it with plastic and
then an aluminum foil grounded to the coax that goes to the analyzer?

I tried wrapping the board in aluminum foil. No effect at all.

I'll side the the spectrum analyzer on it's "Not guilty" plea

You don't know how much ethanol and soap it took to get it cleaned
It needed a few repairs as well so there is a chance the SA is bad in
more places. Fortunately I have a service manual. I should have bought
the Signal Hound :-(

I don't have any because I just design them from scratch whenever I need
one. In ultrasound I kind of grew up with those things. Essentially just
a glorified lowpass where the caps are varicap diodes.

Sounds easy enough The wide bandwidth will cause some problems.
But hey, this is a backburner project

 

[Joerg]

Nico Coesel wrote:

Nico Coesel wrote:
[...]


14 mil thick and 26 mils wide moves a LOT of field up into the air!
changes the 4.2 to 3.2 effective. Not sure it is good to have that
floating around above your circuitry.

What do you mean by 'serious mismatches in circuits'?
A slightly different subject: AFAIK there are at least two ways to
determine mismatch. TDR or by using a directional coupler and see what
is reflected back. I'd like to get a setup with a directional coupler.

The above is a slightly different topic. Once I get this setup going
I'd like to use it for testing some odd ways to connect ethernet. RJ45
is not always the most practical connector.

Measure the output without the filter. You can do that without building
or buying anything. Then if you have determined that it is really the
filter you can add 10-20% or so to a cap here and there and see which
Actually there is no filter at the output

I ordered some RF detector schottky diodes. If all goes well they'll
arive tomorrow. I really need to figure out whether the problem is in
the SA or in my circuit.

The spectrum analyzer can easily be tested for that. Hang a preamplifier
onto it with nothing connected to it's input. This will raise the noise
floor. Most spectrum analyzers are sensitive enough by themselves
though, you can test that by shorting the input and seeing if the noise
floor changes. A dip in the noise floor at 2.45GHz should be clearly
I tried that but the noise floor stays flat at -70dBm.

That would mean the spectrum analyzer should be fine.

Unless the majority of the noise comes from the amplifiers after the
first few stages. When I attenuate the input signal, the dip gets cut
off by the noise floor.

If a stage oscillates in the analyzer that could also explain it.

The stepping is not the problem either. When I set it to a fixed
frequency I get the same effect.


I tried wrapping the board in aluminum foil. No effect at all.

Hmm, that is indeed suspicious.

You don't know how much ethanol and soap it took to get it cleaned
It needed a few repairs as well so there is a chance the SA is bad in
more places. Fortunately I have a service manual. I should have bought
the Signal Hound :-(

Yeah, that Signalhound really rocks. If your analyzer was damaged when
you bought it then there is a chance that something else is broken in
there. Sometimes literally broken off.

Sounds easy enough The wide bandwidth will cause some problems.
But hey, this is a backburner project

Then you need several lowpass filters. A varicap can realistically not
do more than 5:1 in the cutoff frequency.

 

[josephkk]

Every now and then I spend some time tinkering with HF stuff.
I've put a board together with an ADF4350 fractional-N PLL chip to
serve as a simple RF generator. The ADF4350 can output a signal
between 140MHz to 4.4GHz. It seems to work but the output shows a
large dip at 2.487 GHz and small one at 2.287 GHz (see
http://img201.imageshack.us/img201/4512/img0498ry.jpg showing 2
traces). I can't test beyond 2.6GHz The board is a 4 layer FR4
PCB.

So far I tried to change the decoupling capacitors (0402), add extra
wires to the power supply pins, change the output termination, use
different cables (SMA) but this changes nothing at all.

Is there an easy way to pin this problem down without spending another
fortune on equipment? I guess there is a parallel resonator somewhere.
To make matters worse I'm not quite sure my SA is 100%. I'm thinking
about simulating the board using Sonnet or putting together a simple
power level meter.

Any other ideas?

Wild guess from the peanut gallery; so far it sounds like a layout based
problem.

?-)

 

[josephkk]

This is the HF part:
http://img96.imageshack.us/img96/1460/img0506bc.jpg

The next layers are ground and power. I already tried to make the 'S'
shorter with a wire but no difference at all.

Am i seeing through hole parts used in the GHz region?

?-/

 

[josephkk]

Robert Macy <[email protected]> wrote:

On Apr 11, 2:51=A0am, [email protected] (Nico Coesel) wrote:

Every now and then I spend some time tinkering with HF stuff.
I've put a board together with an ADF4350 fractional-N PLL chip to
serve as a simple RF generator. The ADF4350 can output a signal
between 140MHz to 4.4GHz. It seems to work but the output shows a
large dip at 2.487 GHz and small one at 2.287 GHz (see
http://img201.imageshack.us/img201/4512/img0498ry.jpgshowing2
traces). I can't test beyond 2.6GHz The board is a 4 layer FR4
PCB.

So far I tried to change the decoupling capacitors (0402), add extra
wires to the power supply pins, change the output termination, use
different cables (SMA) but this changes nothing at all.

Is there an easy way to pin this problem down without spending another
fortune on equipment? I guess there is a parallel resonator somewhere.
To make matters worse I'm not quite sure my SA is 100%. I'm thinking
about simulating the board using Sonnet or putting together a simple
power level meter.

Any other ideas?

Got a pic of the PC board?

This is the HF part:http://img96.imageshack.us/img96/1460/img0506bc.jpg

The next layers are ground and power. I already tried to make the 'S'
shorter with a wire but no difference at all.

Is that 'S' shape even 50 ohm trace?

It should be

I use femm 4.2 to calculate trace widths [very accurate], or an old
DOS tool called DIMSTRIP.exe that Engineers claimed better than 1%
accuracy to 1GHz. If you want a copy of the 50k program let me know.

Given a box lid 1 inch above surface of PCB, FR-4 dielectric 4.2 and
loss tangent of 0.06, 1 mil copper metallization layer [balance
between 0.7 and 1.4 mil], dielectric thickness of 20mils; Dimstrip
calculates the S curve trace width should be 38.9 mils to get Zo=3D50
ohms.

These are the specs from the board house concerning the layer
distances:
http://www.eurocircuits.com/index.php/technology-guidelines/pcb-multi...

The dielectric is 14 mils. The S trace is 26 mils wide.

Appcad puts that at 49 ohms. The trace is basically perfect, and is
short enough that it wouldn't matter much anyhow.

--

John Larkin                  Highland Technology Incwww.highlandtechnology.com  jlarkin at highlandtechnology dot com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

Not very perfect right at the IC and the run to the SMA is into the
wavelength size, which is significant length.

wavelength INSIDE dielectric is approx 300e6/2.4GHz/.0254/sqrt(4.2)/
0.0254 = .2.4 inches

that means the 1/4 wavelength where all the 'magic' rotations occur,
is around 0.6 inches!

Did you scale carefully for the nominal footprint size of an SMA connector
so visible in he layout?
?-)

 

[josephkk]

The above is a slightly different topic. Once I get this setup going
I'd like to use it for testing some odd ways to connect ethernet. RJ45
is not always the most practical connector.

Hmmm. Up to gigabit Ethernet i have to disagree. The 8P8C modular
connector (mistakenly called RJ-45, YCLIU) still works just fine
electrically and is adequate mechanically.
At 10Gb/s (Catagory 6) it gets ugly. Nothing i know if is any good, which
may be mostly my ignorance.

?-)

 

[Nico Coesel]

Every now and then I spend some time tinkering with HF stuff.
I've put a board together with an ADF4350 fractional-N PLL chip to
serve as a simple RF generator. The ADF4350 can output a signal
between 140MHz to 4.4GHz. It seems to work but the output shows a
large dip at 2.487 GHz and small one at 2.287 GHz (see
http://img201.imageshack.us/img201/4512/img0498ry.jpg showing 2
traces). I can't test beyond 2.6GHz The board is a 4 layer FR4
PCB.

So far I tried to change the decoupling capacitors (0402), add extra
wires to the power supply pins, change the output termination, use
different cables (SMA) but this changes nothing at all.

Is there an easy way to pin this problem down without spending another
fortune on equipment? I guess there is a parallel resonator somewhere.
To make matters worse I'm not quite sure my SA is 100%. I'm thinking
about simulating the board using Sonnet or putting together a simple
power level meter.

I've found the problem! It turned out to be the spectrum analyser
after all. I cleaned the thick film circuit with the directional
coupler at the input and now everything is fine!

 

[Nico Coesel]

Very Good

Unfortunately not so much. After closing it up the problem came
back... I have to look into it again. Like Joerg already noted: it may
be something 'mechanical'.

 

[Nico Coesel]

Hmmm. Up to gigabit Ethernet i have to disagree. The 8P8C modular
connector (mistakenly called RJ-45, YCLIU) still works just fine
electrically and is adequate mechanically. =20

Some of my customers disagree on the latter. If you need to run an
ethernet cable through a building and have someone crimp an RJ45
connector on it you are asking for major problems. I've seen lots of
problems with that. A simple screw or spring-clamp terminal is almost
idiot proof.

 

[Nico Coesel]

I've found the problem! It turned out to be the spectrum analyser
after all. I cleaned the thick film circuit with the directional
coupler at the input and now everything is fine!

Cheered too soon. It seems to be a temperature related problem in the
SA.

 

[Joerg]

Am i seeing through hole parts used in the GHz region?

It's an optical illusion because the picture is out of focus. Looks like
the only through-hole part on there is the SMA or RP-SMA jack.

But what's wrong with through-hole parts at GHz? They wouldn't have made
transistors like these if it wasn't done:

http://www.microsemi.com/en/sites/default/files/datasheets/MRF904_REV-.pdf

You can still buy them but at outrageous prices. I still have some one-
and two-layer phenolic boards in the garage, stuff that runs up to
around 800MHz. All through-hole.

 

[Joerg]

I've found the problem! It turned out to be the spectrum analyser
after all. I cleaned the thick film circuit with the directional
coupler at the input and now everything is fine!

Just cleaned it? Oh man, where did you get that thing from? Salvaged
from a barge that sunk on the Ijsselmeer a few years ago?

 

[Nico Coesel]

Just cleaned it?

Cleaning didn't do the job... It turned out to be a crack in the
coupler which is a thick film circuit on a metal carrier. The problem
showed with an Ohm meter. When I loosened the screws of the thick film
everything became fine, when I tightened the screws the contact opened
which meant the coupler wasn't terminated.

I managed to cover the tracks with a thin layer of tin. First tests
look promising...

Oh man, where did you get that thing from? Salvaged
from a barge that sunk on the Ijsselmeer a few years ago?

I got it from a guy in Singapore. He probably found it in the harbour.
And I mean in the water

 

[josephkk]

Some of my customers disagree on the latter. If you need to run an
ethernet cable through a building and have someone crimp an RJ45
connector on it you are asking for major problems. I've seen lots of
problems with that. A simple screw or spring-clamp terminal is almost
idiot proof.

This is a case where tools and training make a real difference. Crimping
works great done correctly, i know, i have done enough if it. The average
yokel cannot make good crimps without training and practice, which is
where most of the problems come from. Spreading your Cat5E cable out onto
screw or spring terminals is just about guaranteed to screw up the
transmission line properties. Just get 110 style punch down modular
jacks, even jerks can usually get those work ok. Then you can use factory
made crimped jumper cables.

?-)

 

[josephkk]

It's an optical illusion because the picture is out of focus. Looks like
the only through-hole part on there is the SMA or RP-SMA jack.

I suspected i was seeing things that weren't there.
Some 30 years ago i tried to make a pulse amplifier (10 x voltage) that
could do 25 volts either or both polarities into 50 ohms with 3 ns edges
using TO-39 bipolar parts. Ft's were about 2.0 and 2.5 GHz.

 

[Nico Coesel]

This is a case where tools and training make a real difference. Crimping
works great done correctly, i know, i have done enough if it. The =
average
yokel cannot make good crimps without training and practice, which is
where most of the problems come from.

But that means they have to send everyone on a crimp-tool training
course and provide proper crimping tools. Thats a big investment. I've
seen installations where about 75% of the connectors where crimped
badly. Over here they send 'jackass of all trades' kinda types for
installing stuff.

Spreading your Cat5E cable out =
onto
screw or spring terminals is just about guaranteed to screw up the
transmission line properties.

True. But me and my customer want to know how bad it is. So far
spring-clamp terminals seem to work OK for 100Mbit ethernet.

Just get 110 style punch down modular
jacks, even jerks can usually get those work ok. Then you can use =

Unfortunately punch down blocks can go wrong as well. People still
need special tools. And I don't like applying a lot of pressure to a
PCB with a lot of SMT parts on it. Nowadays if you want fool-proof
then make sure something can be installed with a blade shaped screw
driver which fits in the back pocket of a jeans.

 

[Joerg]

I suspected i was seeing things that weren't there.
Some 30 years ago i tried to make a pulse amplifier (10 x voltage) that
could do 25 volts either or both polarities into 50 ohms with 3 ns edges
using TO-39 bipolar parts. Ft's were about 2.0 and 2.5 GHz.

Could have done it with the 2N7000. I used the BSS123 on an ultrasound
machine design in 1986 because we were early SMT-adopters, no strictly
technical reason. But for the pulsers we used other transistors because
we needed up to 170V.

There were also a bunch of nice video-BJTs that could do it, blazingly
fast and cheap. those are gone now because of the demise of the CRT
technology for monitors and television sets.

 

[Nico Coesel]

Hire people who know what the hell they are doing.

Finding qualified people is very hard over here. If you know what you
are doing you are in management! The unemployment rates are still very
low over here. Low enough to keep loads of East-European people
working as well. Example: when they came to connect the optic-fiber in
my house they send a Polish guy and two Dutch guys. The Polish guy had
to explain to the Dutch guys how to drill a hole in concrete! The
Polish guy also welded the optic fibers and did the final testing.

 

[josephkk]

But that means they have to send everyone on a crimp-tool training
course and provide proper crimping tools. Thats a big investment. I've
seen installations where about 75% of the connectors where crimped
badly. Over here they send 'jackass of all trades' kinda types for
installing stuff.


True. But me and my customer want to know how bad it is. So far
spring-clamp terminals seem to work OK for 100Mbit ethernet.

Maybe only until you really need it.

Unfortunately punch down blocks can go wrong as well. People still
need special tools. And I don't like applying a lot of pressure to a
PCB with a lot of SMT parts on it. Nowadays if you want fool-proof
then make sure something can be installed with a blade shaped screw
driver which fits in the back pocket of a jeans.

If you have a circuit board just mount the modular connector
(bolted/riveted/glued/something mechanically strong) to it and solder it
in. Then all you need is fingers.

?-)