# Help needed for RF PCB design

Status
Not open for further replies.

#### rainlei

##### Junior Member level 1

I'm now designing PCB for test our VCO which is running on 5GHz.
The VCO's package is TQFP48. I decide to use a socket to mount the VCO chip.
Anyone has any idea about which socket is better in have less parasitic effects?
Can you recommend some sockets or where to find such kind of information? and can you recommend some book on RF PCB design?

Thank you my friend..

##### Member level 1
On 5G, Use socket? It is not a good idea.
Maybe you should solder VCO on PCB.

#### rainlei

##### Junior Member level 1
Thanks

But i have more than 40 VCO chips...I need to measure them one by one to investigate the process variation. If I solder them , then I need a lot of PCB board.....
That's one of the problems..

#### Spasomat

##### Junior Member level 3
Perhaps you can solder the VCO to small PCB pieces and then use a microwave test-fixture to connect to the transmission-lines of the PCBs.
e.g. Wiltron and Connexion Rosenberger are manufacturing universal test-fixtures. Although they are not cheap, they offer reproducable measurements.
Regards,
Spasomat

#### House_Cat

The sockets are expensive, but they are very low loss. See the following URL:

(substitute t for * in the following)

h**p://www.johnstech.com/products/index.html

#### Element_115

I have used sockets at 2.5GHz , the ones we used and or bought
were form :

Everett Charles
Synergetix

There are more like Oztek , Prime yield and others .
The price range was from $1K -->~$12K U.S.D.
But at 5GHz price and performance might be a
nightmare.

Look here --> www.chipscalereview.com/issues/0302/f1_01.html

Good luck

#### backdoor_

##### Member level 2
Synergetix makes good sockets.
For process variation, sockets are the way to go.
Make very detailed measurements with a socket, and then solder the same part to the PCB directly to repast the measurements.
In this way you will have a good idea how the socket impacts measurements.

bd_

#### rainlei

##### Junior Member level 1
it sounds good...

I have checked the website of Synergetix, but they really offer little information on there website.

Maybe I'll need to contact their sales.

Thank

And by the way, how much u pay for ur 2.5G socket?

#### rainlei

##### Junior Member level 1
BackDoor_::

Thanks, there is a problem, should we assume the socket has an linear impact on the measurement?
Does their product handbook give spec about the impact of measurement?

Thanks anyway

#### Spasomat

##### Junior Member level 3
About the price: a microwave test fixture with 2 launchers working up to 40GHz to contact microstrip lines will cost at about 20k USD....

#### rainlei

##### Junior Member level 1
Spasomat:

Thanks. the price is too expensive.......for me...seems i can not persuade boss pay for that.

I decide to buy sockets, if can not buy socket, I'll build lots PCB for the work....

anyway, thanks

#### Spasomat

##### Junior Member level 3
By the way: In the last microwave journal (june 03) is a longer article about test fixture basics. may be it might be worth a look....
Spasomat

#### artem

Hi

If VCO chip has only very limited number (1-3) of pins working on high freq may be it is better to dismount this part of cheap socket and use direct soldering but leave other pins connected through socket .

That you will need little bit time to connect and disconnect IC when changing for measurement

#### Mazz

rainlei
I've been using johnstech socket for years, but at lower freq (< 3 GHz)
Consider that normally we had to design new boards layouts when applying the ICs directly on boards.
They aren't cheap at all: 1 socket is about 8/10 k$, 2 12/15 k$ and so on.
They want you to send them some packaged samples before starting the design and they take some month.
I know that they have higher freq. series but I don't know them.

We usually use low freq. socket to verify functional samples and then solder on boards to final char.

In your case, as you need to verify process spread, I think you have the VCO tank fully integrated, so maybe also a lower freq. socket could be ok. But with ground and supply... ?!

My opinion is that, in this case, maybe solder samples on boards is cheaper and more safe.
Mazz

#### BigBoss

No need to solder the case of VCO..

Just create a test fixture that will press on the vco package onto pcb. Use hard plastics to design this test fixture. Yo can order it to any automated milling company with a feasible price..

I think it will have 4 hole crossing on the package and will be pressed on the pcb with a spring aligned caption.

Like this..

# #
----------
| |
| |
| |
---------
# #

In fact this needs some special tools but no way out.. You can not use sockets for 5GHz , you even use after some time socket will be deformed.
It's big trouble..

Don't forget to put some extra copper excess on the pads with 15-20 um gold plating..

Rgrds

#### rainlei

##### Junior Member level 1
Mazz, ur info is very helpful...

BigBoss, I'll try ur suggestion.

#### Mazz

I've also tried BigBoss solution, but with some difficulties:
first, to align correctly the package: maybe using something fix on the board in order to align it properly it would be nice.
second, how to be sure that each pin correctly contacts the pcb? Using the board in a lab, usually you will need a very robust design.
Maybe if the test fixture presses directly the pins instead of package body...
So take care of this, otherways you'll become crazy during measurements...
Bye
Mazz

Status
Not open for further replies.