Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

Problem about Linearity of LNA.

Status
Not open for further replies.
D

dopradar

Member level 5
Joined
Oct 8, 2004
Messages
83
Helped
1
Reputation
2
Reaction score
1
Trophy points
1,288
Activity points
761
Hi all,

Recently I was tuning a LNA, But when I used two tones to test the 3rd intercept point, I found the intermodulated products are different, pls read the attached picture twotones.jpg. so I want to konow the reason.

Thanks a lot!
 

Check that all the intermodulation products are inside the LNA band.
I hope it can help.
Mazz
 

They are born to be different.
 

LNA uses matched pair of transistors, preferably bipolar technology. From the picture, your transistors are obviously mismatched.
 

That is uncorrelated to device matching.

The component (2*f1-f2) means the 3rd intermodulation term of (f1,f1,f2). And the component (2*f2-f1) means the 3rd intermodulation term of (f1,f2,f2). So they born to be different. By the way, Mazz is right, though all 4 tones are in the passband of LNA, slight deviation could be observed due to passband ripple.
 

Nope. In this case, your test tones are already same in power, your IMPs must also have the same in power. Not unless you have used test tones with different power.
Many people ignored this because they are only assumed that LNA should work fine as long as they get the w1-w2 and w1+w2, thinking that they both carry the same information, and using a filter to select one of it, usually the one with larger amplitude although has alittle more noise, and get done with it.

Where do you think the passband of LNA comes from? It has to be from the transistors and current sources. When interpreted in small-signal model, you will see everything in there contributes problems to linearity, and the main contributor is transistors.
 

SkyHigh said:
Nope. In this case, your test tones are already same in power, your IMPs must also have the same in power. Not unless you have used test tones with different power.
Many people ignored this because they are only assumed that LNA should work fine as long as they get the w1-w2 and w1+w2, thinking that they both carry the same information, and using a filter to select one of it, usually the one with larger amplitude although has alittle more noise, and get done with it.

Where do you think the passband of LNA comes from? It has to be from the transistors and current sources. When interpreted in small-signal model, you will see everything in there contributes problems to linearity, and the main contributor is transistors.
Click to expand...

Read this:

High-frequency analysis of linearity improvement technique of common-emitter transconductance stage using a low-frequency-trap network
Keng Leong Fong;
Solid-State Circuits, IEEE Journal of
Volume 35, Issue 8, Aug. 2000 Page(s):1249 - 1252

read the last paragraph on pp. 1250 to pp. 1251

And why do i need to care f1-f2 or f2-f1 when i am designing an LNA?

Because there is a praralleled LC network at output, so the overall S21 behaves like a bandpass filter. It peaks at designed central freq. with limited bandwidth. The higher Q of your loading inductor, the sharper it peaks. Such a network does not exhibit a flat band shape, so a little deviatioin could be observed.

By the way, i can not analyze the problem with a figure drawn by VISIO.
 

I checked the paper and have one question. The low frequency trap network to improve IP3 has only been verified for bipolar case, is it equally applicable to CMOS?

Thanks
 

Maybe. There is one paper doing the same thing as K. L. Fong but focusing on CMOS MIXER. You can follow his work.



But for CMOS, there is another linearization technique called Multigate Transistor (MGTR). It's better than low-freq. trapping network. And somebody apply the same method on HBT to prove that it is practical to HBT, too.
 

Hi Mazz,

Could you explain the reason for check all the IMPs?

You mean if testing the linearity in the low freq band or upper band will occur this case and not occur in center frequency band!

Added after 3 minutes:

Hi dsjomo,

Could you paste that mail here? Because I can't touch the IEEE papers easily. Thanks!

Keng Leong Fong;
Solid-State Circuits, IEEE Journal of
Volume 35, Issue 8, Aug. 2000 Page(s):1249 - 1252

read the last paragraph on pp. 1250 to pp. 1251
 

Sure, it is already uploaded.
 

Hi,

I have pasted the picture gotten from the spectrum analyzer. Pls help me analyze the reasons. Thanks a lot!
 

Thanks for the plot.
What I mean is that if the input tones spacing is too large, the intermodulation products can fall out the LNA bandwidth, so they can be filtered out in some way. Observing the graph this doesn't seem to be the case, unless you are measuring it with a filter.
To be sure check the meas with 200 kHz spacing, instead of 5 MHz.

Another question: what is the P1dB of your LNA?
Please remember that IP3 measurement should be done with small signal (rule of thumb, input power < Input P1dB - 15 dB).

I hope it can help.
Mazz
 

Hi Mazz,

Yes, I have ignored this problem. But could you explain the reason?
 

Can you post the S11 and S22 at 2*(operation freq.)?
 

i think 2nd intermodulator at f1-f2 or f1+f2 also contribute to 3rd intermodulater, if the spacing of the input tones is too large, then this signal at f1-f2 or f1+f2 fall out of the bandwidth of LNA ,so a little deviation can be observed.
i'm right??

to Mazz
when test intermodulation, the input level must be 15dbm below inputP1dB??

how do i set the extrapolation point?? because it is different, the result is different

thanks
 

Hi,

Pls see the pasted pictures!
 

tiger_ads said:
to Mazz
when test intermodulation, the input level must be 15dbm below inputP1dB??


thanks
Click to expand...

tiger_ads
15dB bolow IP1dB is only a rule of thumb.
If you use too low input level you'll not be able to measure the intermodulation products that will be under the output noise level. The IP3, by definition, is to be checked in linear condition, so well below the P1dB.
I hope it can help.
Mazz
 

Hi Mazz,

Thank you for your help. I have tried low input power, but it's no use!
 

dopradar said:
Hi Mazz,

Thank you for your help. I have tried low input power, but it's no use!
Click to expand...

Hi dopradar,

Please check this, you give a plot of IM3 measurement around 1.5GHz in the
former post, but you give us the S11 measurement in 2G and 1G. Could
you shrink the measured range around several ten MHz on 1.5GHz.

Thanks,
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top