LNA doubt ( help me )

[bcdeepak]

hi

previously i asked some doubts about the inductively degenerated LNA , which i designed

i got a doubt about the inductor value ...........

the LNA is designed such that the input impedance at the operating freq of 2.4GHz is 50 ohm
and S11 graph shows the matching curve .

S11 dip is occuring at freq 2.4GHz with a value of -25 db and i drew the input impedance curve Zin and the value @2,4GHz is 55 ohm .

the thing is ....... i am using an inductor of 0.05 nH ....... then only the input resistance is 55 ohm........ is 0.05nH inductor available in ICs ?????????

or am i doing wrong in seeing the input impedance ????

how to know whether the input impedance is 50 ohm or not @ the operating freq ?? just by plotting Zin

 

[khouly]

ohh it is very small , i don't know how to get it , normally in CMOS the valuse between 10 to 1 nH , used
but 0.05 , is very small , may be a just samll metal strip do the job

what is value of the Lg used ?

khouly

 

[dingjingfeng]

for inductive degenerated match, the key point is the value of ωT and Ls
becasue their product should be 50Ω.
for modern technology for example, 65nm CMOS, the fT is very high what make the Ls is too small if we choose the minimum Length.

because the ωT is independent of W, so the minimum Length is the only dominant parameter.

But never forgeting that we can choose a Transistor with non minimum L.
then the ωT is decreased.

and because fT=Gm/Cgs, we can also deliberately increase the Cgs by parallel a capacitor with the choosed Transistor .
then you can increase the inductance of the Ls.
but some other parameters are also sacrificed.

 

[bcdeepak]

hi kholy

my Lg is 5.6nH

 

[khouly]

the Lg is resonable , but the Ls is very small , can u desing this valuse with just a strip

khouly

 

[bcdeepak]

hi khouly,

can an Ls of about 0.3nH is reasonable(for fab) ??

if i use Ls = 0.3nH my input impedance is around 70ohms,instead of 50ohms

so can i use it ?? ( i know that Zin should be 50 ohms for LNA, but what should i do ,my inductor is becoming too less )

i am using hspice, is it reliable??

 

[khouly]

what about ur NF and S11 if u use this value ?

Hspice RF is 4 sure good simuator , but the main stream in RFIC use cadence SpectreRF and there also alot using ADS
khouly

 

[dingjingfeng]

we can realize an inductor for 0.3nH for all technique.
but the foundry doesn't offer a model for it and you should simulate it by youself
the specification decides whether you can use 70 ohm impedence match.
usually the requirement for VSWR is 2.0 for a reciever.

 

[bcdeepak]

hi khouly

with Ls=0.3nH ( Zin=70 ohm ) , S11 = -17.5db and noise fig = 1.2 db

( i thinks NF will not be effected much with Ls value )

Added after 1 minutes:


hi dingjingfeng

generally LNA designs are carried for Zin = 50 ohm but separately no specification is mentioned for
50 ohm

 

[khouly]

S11 is ok , in ur simulator u can also view the VSWR . so u can be sure it will not exced 2

khouly

 

[mirjam]

I have been recently designing LNA in AMSC35 from Austriamicrosystems with Cadence SpectreRF. The lowest inductor value they have modelled is 1.32n. Does that mean that they can not do it smaller I don't know but the truth is that they have given only 10-12 inductor models (values).

As for S11 and input impedance I have been thought that the impedance at the input of LNA must be 50 Ohms that would lead to very low S11 parameter. In case of low level signal (that are critical from one side) any reflection lead to possible signal sinking into noise.

In that case as ωTLs=50 Ω the only choice you have is to see how you can decrease ωT, but not too much as there is possiblity that some noise sources that have been neglected could play a part in the NF calculations (gate induced noise- blue noise). As suggested earlier increasing length could be one of the solutions.

another perhaps to use bondwire or something that could give the value you have precalculated.

 

[sharpsheep]

I have been recently designing LNA in AMSC35 from Austriamicrosystems with Cadence SpectreRF. The lowest inductor value they have modelled is 1.32n. Does that mean that they can not do it smaller I don't know but the truth is that they have given only 10-12 inductor models (values).

As for S11 and input impedance I have been thought that the impedance at the input of LNA must be 50 Ohms that would lead to very low S11 parameter. In case of low level signal (that are critical from one side) any reflection lead to possible signal sinking into noise.

In that case as ωTLs=50 Ω the only choice you have is to see how you can decrease ωT, but not too much as there is possiblity that some noise sources that have been neglected could play a part in the NF calculations (gate induced noise- blue noise). As suggested earlier increasing length could be one of the solutions.

another perhaps to use bondwire or something that could give the value you have precalculated.

mirjam is right.
You can use bondingwire. Usually the downbond inductor is about 1nH, so you can use three parallel downbond to get 0.3nH.

 

[bcdeepak]

hi mirjam,khouly

while calculating NF, S - parameters , we should define ports .

my doubt is at what position we have to define the input port .... i mean at the end of source resistance(where the actual ckt starts... blocking capacitor ) or at beginning of source resistance ( at the input voltage).

during noise figure should we remove the source resistance and directly connect the input ?? because finally on the LNA chip source resistance will not be there

Added after 53 minutes:


hi

is the source resistance should be defined in the LNA ckt ???????

i saw some papers where the simulation is done using ADS (schematic ) , so there the source resistance is used in the ckt.

 

[khouly]

always 50 ohm is the source resistance

khouly

 

[bcdeepak]

hi khouly

i wrote mistake in my last msg

what i want to tell is while simulating the LNA ..... is required to give the source resistance ???? ........ i saw some LNA papers where the simulation is done using ADS , so there they did not give the source resistance in the simulation schematic

my problem is when i include the source resistance in the ckt .... dip of the NF curve is not exactly not occuring at 2.4GHz ... it is occuring at 2GHz ... but when i remove the source resistance the dip is occuring @ 2.3GHz

so is it required to give Rs in the simulation ???

 

[khouly]

in ADS simulation when use the S parameter simulator and measure the NF , we use a terminations which are a 50 ohm by default

khouly

 

[dingjingfeng]

you can change the source restance I think
but the communication system is 50ohm usually
they will not change the specification because of your design!!

hi khouly

i wrote mistake in my last msg

what i want to tell is while simulating the LNA ..... is required to give the source resistance ???? ........ i saw some LNA papers where the simulation is done using ADS , so there they did not give the source resistance in the simulation schematic

my problem is when i include the source resistance in the ckt .... dip of the NF curve is not exactly not occuring at 2.4GHz ... it is occuring at 2GHz ... but when i remove the source resistance the dip is occuring @ 2.3GHz

so is it required to give Rs in the simulation ???

 

[bcdeepak]

hi dingjingfeng, khouly

i am uploading the S-parameters and Noise figure curves

plz verify them , tell me whether they are acceptable or not

i designed the LNA for 2.4GHz

awaiting ur reply

 

[totowo]

s22 is not perfect. Usually, -10 dB is necessary, <-15 dB is better.......
But it depends.
Your LNA is a stand-alone one or integrated with the RXs on chip?
If the former, you'd better improve your s22 because the output is provided to off-chip module, before which there is a transmission line.
But if the later, the s22 is not so important, because on chip we usually consider the voltage rather than power.


The NF is quite good. The inductor your used is ideal ones or on-chip inductors offered by foundry? you MUST use the real on-chip inductor since the ideal ones has few use in a design.

 

[bcdeepak]

hi totowo

thank you very much for verifying my results

i am using ideal inductors till now, but now onwards i will replace them with practiacal on-chip inductors, just to find the required inductor values till now i used ideal inductors

to improve S22 , i used a 20ohm resistor in series with the ouput inductor, so now S22 is -13.5 dB
but if use more resistance , S21 is becoming less

is S21 of 13dB acceptable( then only my S22 is -13.5 dB) ???