Post subject: Noise figure degradation to chip to module
Posted: Tue Nov 07, 2006 4:40 am
Apr 25, 2006 2:01 am
I am noticing a 0.75
dB difference (degradation) between the noise figure measured on-chip
and the noise figure measured in module (after assembling the same chip
in said module with input and output microstrip transmission lines).
The chip is designed to work at 10 GHz and the length of the 50 ohm
input microstrip line (the loss of which, I would assume, would add
to the noise figure) is about 1.5 mm. The subtrate used is alumina.
It appeared to me that the NF deterioration was on the higher end and
I am presently looking at ways to minimize this. I am guessing that
the problem might be in the transitions (coaxial-microstrip and microstrip
line to chip). It would be helpful if someone could suggest practical
guidelines and tips that could help mitigate this problem.
Post subject: Posted: Tue Nov 07, 2006 7:01
Joined: Mon Jun 27, 2005 2:02 pm
When you add any length
of transmission lines (And especially in such high frequenices), you
can expect mismatches. I don't know what is the function of your chip,
but let's assume that this is an LNA. If this is an LNA for example,
then you have to match it to a slightly different input impedance caused
by the addition of transmission line. If the LNA was designed to Gamma
(opt) which resutls in the best NF, the now the LNA is matched to another
Gamma which degrades the NF. I suggest you to simulate the chip together
with the added transmission line and change the dimensions of the transmission
line if possible.
Anyway, from the simulation you will get an
insight of the problen and the needed changes to be done in order to
decrease the degradation.
Post subject: Posted: Tue
Nov 07, 2006 12:24 pm
Joined: Fri Feb 17,
2006 12:07 pm
Location: London UK
You might have
some error in measurement due to mismatch of the testgear.
you have an amplifier VSWR of say 2:1 and a testgear VSWR of say 1.6:1,
with a modest loss for the connectors plus 1.5mm alumina strip line,
this would account for about 0.7dB extra loss and therefore extra noise
factor (according to my software)
However I cannot simulate random
lengths of the interconnecting line to see what the max and min combination
of VSWR would be. Nonetheless you can see the order of magnitude. IR
is right: you need to simulate every component transition as accurately
as possible, then work a combination of inter-element spacings to see
if that equates to 0.75dB. If it does, then all you can do is use very
good connectors (which anyway degrade with use) and perhaps match the
testgear with a stub tuner or an isolator
It is notoriously difficult
(as you probably already know) to measure such small losses to an accuracy
In addition I agree with IR re the difference between
MAG matching conditions and Best Noise Factor matching conditions. They
are different and many workers in the field have noticed this, particularly
at the very low noise temperatures currently attainable.
Post subject: Posted: Tue Nov 07, 2006 12:49 pm
Joined: Mon Jun 27, 2005 2:02 pm
Just to emphasize what nubbage
You should perform simulation first before attempting
to do measurements. Doing measurements first will cost you in a lot
of time and effort. Simulation is the scientific and insightful way
to gain understaning over the root-cause of the problem and to be able
to quantize it.
When you get some results and hints from the
simulation, then you can do some tests and with the existing interface
between the simulation tools to test equipment you can see the differences
between the measured and simulated results. So at the end you can converge
into the final solution.
Good luck, if you need more help please
let me know!
Post subject: ThanksPosted: Tue Nov
21, 2006 2:23 am
Joined: Tue Apr 25, 2006 2:01
Thanks IR and nubbage for your help