Designer’s Guide Community

eng

Community Member

Offline

Posts: 49
USA

Stability of Differential LNA
Jul 3^rd, 2007, 3:55pm

Hi all,
I'm designing a diffential LNA at UHF freq. I used well known diff architecture as seen in the attached figure1. The RF performance seems good however when I checked Stern stability criteria with Cadence after sp analysis, the delta (Bif) looks greater than 1 around the freq of interest which means the amplifier is potentially unstable. Both K and Bif vs. freq is seen on the attached figure. As far as I remember this architecture shouldn't have a stability issue.�I couldn't figure out why it is not stable. Could you please comment on why Bif > 1?

Also, to improve return loss I made the size of cascode transistor; (Wcascode) = (Wrfin)/2 and (Lcascode)=2*(Lrfin). Is this a good practice?

thanks
eng

Back to top

difLNA.JPG

« Last Edit: Jul 03^rd, 2007, 5:33pm by eng »

IP Logged

Visjnoe Senior Member Offline Posts: 233	Re: Stability of Differential LNA Reply #1 - Jul 4^th, 2007, 2:22am Dear, please provide your testbench: my guess is the answer to your stability issue is hidden in the testbench setup. Regards Peter
Back to top	IP Logged

eng Community Member Offline Posts: 49 USA	Re: Stability of Differential LNA Reply #2 - Jul 4^th, 2007, 10:34am Dear Peter, Thanks for the reply. Here is my testbench. The pads and ESD are supplied by pdk. eng
Back to top	TBdif.jpg IP Logged

eng

Community Member

Offline

Posts: 49
USA

Re: Stability of Differential LNA
Reply #3 - Jul 4^th, 2007, 11:08am

I removed the load tuning capacitor C2 & C3. B1f became less than 1. Good stability!

However this killed s22 response and ss gain s(21) is reduced 4-5dB as well. This LNA will be integrated with Mixer and I will add test pads between LNA and Mixer for individual testing. Do you think s22 response of LNA is important in this case namely are load caps necessary? Or should I just connect mixer & pads and care about overall front-end response?

thanks
eng

Back to top

IP Logged

didac

Senior Member

Offline

There's a million
ways to see the
things in life

Posts: 247
manresa,spain

Re: Stability of Differential LNA
Reply #4 - Jul 4^th, 2007, 11:56am

Hi,
Have you tried to remove C7 instead of C2 & C3? If the balun is a transformer I think that you will have a current loop with Vbias(also valid for a balun made with transmission lines if this isn't something like an hybrid).I think that S22 it's unimportant in this case because your Mixer will not have 50 ohms of input impedance(I suppose), i think you should check voltage gain rather than s-parameters, and you can also perform the optimitzation of the LNA after connecting the Mixer.
Hope it helps,

Back to top

IP Logged

eng Community Member Offline Posts: 49 USA	Re: Stability of Differential LNA Reply #5 - Jul 5^th, 2007, 9:33am Hi didac, I tried removing C7 but it did not fix the problem. When port2 is not matched the stability is ok, B1f is becaming greater than 1 if s22 is -20 etc. Same thing happens even if I do matching witout c2&c3. still couldn't find the reason
Back to top	IP Logged

didac

Senior Member

Offline

There's a million
ways to see the
things in life

Posts: 247
manresa,spain

Re: Stability of Differential LNA
Reply #6 - Jul 5^th, 2007, 12:12pm

Hi again,
There are a couple of things that I don't feel "right" in this testbench and I don't know if they could affect the simulation, here it goes:
1)As I said I think capacitors C7 & C6 are unnecesary.
2)You put the Load capacitance after the balun, I think this Cap should be attached to the differential outputs, it can change the impedance viewed by port 2.
3)Which differential impedance have you put in the balun?I think that it must be put equal to 100 ohms.
I will try to scratch my head a little more about this,keep in touch.

Back to top

IP Logged

eng

Community Member

Offline

Posts: 49
USA

Re: Stability of Differential LNA
Reply #7 - Jul 5^th, 2007, 12:56pm

Didac,
I removed C7 & C6 & C8 and placed 2 shunt load capacitors (input cap of mixer) right before balun. Balun is 50 ohm to 100 ohm. Right now, LNA is stable!!! Thanks a lot. Although s22 is around -6 dB I can put offchip matching network during the measurement. right?

The output dc blocking capacitors are 4 pF & at 400 Meg, they're 100 ohm. Do you think it is a good value to use. What should the impedance value be around when used as DC blocking?

thanks so much again for your helpful posts.

Back to top

IP Logged

aaron_do

Senior Fellow

Offline

Posts: 1398

Re: Stability of Differential LNA
Reply #8 - Jul 5^th, 2007, 11:23pm

Hi eng,

I think for unconditional stability you need Kf>1 and B1f>0 not B1f<1....may want to double check that though.

Also, you seem to be saying that your mixer's input impedance is capacitive. Since it is not 50 ohm, there is no need to match your LNA output to 50 ohm. i.e. ignore S22. You should simply ensure the LNA output node resonates at the operating frequency.

If you chose to do this, you need to change your testbench.
1) Remove the output balun and port, and simply run an XF and Noise analysis.
2) Keep C8 which represents your mixer input impedance. Also your mixer input impedance is not entirely capacitive. Since you have blocking capacitors, i guess you have biasing resistors. These appear in parallel with your mixer input cap.
3) Choice of C4 and C5 depends on how much loss you can afford...assume a resistor divider between C4,C5 and C8 (and biasing resistance). e.g. if C4 is 4 pF and C8 is 1 pF, the gain = 4/5 where the biasing resistance >> 1/(2pi*f*1p)

for actual testing you will need some kind of buffer for both LNA and mixer outputs since most lab equipment is 50 ohm.

cheers,
Aaron

Back to top

there is no energy in matter other than that received from the environment - Nikola Tesla

IP Logged

didac

Senior Member

Offline

There's a million
ways to see the
things in life

Posts: 247
manresa,spain

Re: Stability of Differential LNA
Reply #9 - Jul 6^th, 2007, 10:59am

Hi to all,
When I first read the post I didn't take much attention to the differentation between delta(as I did in microwave theory classes) and B1f in Cadence, after the post of Aaron I did a little google search and found two application notes by Cadence that relates directly to stability and I think that Aaron is right about his statement about stability,here it goes:
LNA Design Using SpectreRF Application Note and PA Design Using SpectreRF Application Note, in the first one they define the stability using delta(I remember to do this time ago) and define stability with K>1 and delta<1(nothing new here,stern definition) in the second one they define stability with Kf>1 and B1f>0(a parameter that is directly related with the delta factor), I also checked at the lab and at least in my current version of spectre the only parameters available in sp analysis(at least I didn't find the "old" delta parameter). In resume I think Aaron is right, I apologize to eng and other people in the forum for my previous posts(I assumed that it was just a different acronym).

For the other part I agree with Aaron, cap selection depends on how much loss you can afford,ideally a decoupling cap should be like a short for the RF signal.Regarding the gain I also agree that it's better to use voltage gain rather than s21,you won't have the mixer matched to the lna so s-parameters as a measurement for the gain can be tricky in this case.

Back to top

IP Logged

eng

Community Member

Offline

Posts: 49
USA

Re: Stability of Differential LNA
Reply #10 - Jul 6^th, 2007, 2:11pm

The assumptions I made due to my microwave background led me wrong for stability. Now things are more clear. Thank you Aaron & Didac.
thus,
- Since I need to match output of LNA only for measurement purpose I can just place ESD& pad on chip and put an off-chip buffer with the gain of 1 and output imp of 50. This way it will not disturb the output of LNA. Plus I can use same pads as input to measure Mixer. Right?
- Since nothing will be matched on receiver chain, should I always use xf and noise analysis to check the gain & NF? swept pss for linearity? pss + pac (or pxf) for gain when freq change occurs?

thanks
eng

Back to top

IP Logged

didac

Senior Member

Offline

There's a million
ways to see the
things in life

Posts: 247
manresa,spain

Re: Stability of Differential LNA
Reply #11 - Jul 7^th, 2007, 10:09am

Hi,
-Regarding the first question yes,although I think it's also possible to put something like a pi-match passive network to match LNA output if you want to characterize it using s-parameters and a VNA.For the mixer you can use the same pads due to the fact you will want to characterize it's gain conversion using voltage gain,also take into account the need of a Buffer for the Mixer measurement.
-Yes the analysis you said are the necessary,for mixer measurements refer to:http://www.eel.upc.edu/rfcs/Material/Cadence%202004/Mixer%20Design%20using%20Spe....
PS1:if you got a background in microwave engineering and want more information regarding the stability factor used in Cadence i did a little ieee search and found the following references(I'm curious about why it didn't use the classical definition):
Review and Discussion of Stability Criteria for Linear 2-Ports,RENE P. MEYS
Reappraisal of the Unconditional Stability Criteria for Active 2-Port Networks in Terms of S Parameters,DONALD WOODS
Stability and Power Gain of Tuned Transistor Amplifiers,Arthur P.Stern
Any comments regarding the use of B1f and K instead of delta and K will be appreciated.
Hope it helps, and sorry again for my error in identificating the different criteria used by Cadence.

Back to top

IP Logged