A LNA's Linearity Problem!

[dopradar]

Hi all,

Recently I am designing a LNA work at L band ( from 900 to 2200MHz), the structure is shown as LNA.jpg.

It's gain and NF are good. S-params is shown as following:S11.jpg, S21.jpg, S12.jpg and S22.jpg.

but when I test the linearity, a problem occurs. at 900MHz IIP3 is 2dBm, but at 2000MHz IIP3 is 14dBm, the difference is so large, so what's the reason? how to sovle this problem?

Thanks a lot!

 

[dopradar]

here is S parameters.

 

[Grig]

Dear Dopradar

If your project is not commercial...
Can You upload your LNA ADS_prj
I want to study this structure more detail

Thanks

 

[vfone]

A poor selection of the by-pass capacitors (C6, C7, C8, C9) could degrade IP3 performance vs frequency (especially C6 and C7). As a rule of thumb, the impedance of bypassing circuit should be lower than 25% of the input impedance of the transistor at particular frequency spacing.

Any mismatch of the Noise matching components (C1, C2, C3, TL1) will increase IP3 but will degrade NF. Could be done a compromise vs frequency.

Increasing TL3/TL4 inductance reduces gain and improves IP3, but watch for microwave oscillations with excessive inductance.

R1 it provides stability, while reducing IP3.

 

[VSWR]

It's important that the decoupling capacitors are large enough to provide a low impedance for the beat tone created by the two input carriers, used for intermodulation test (and it's result used used for IP3 calculation). If the separation of the two carriers f1 and f2 are 1 MHz, a strong beat tone of 1 MHz is created in the bias circuits. If e.g. Xc=5 ohms is considered sufficient @ 1 MHz, the decoupling cap must have a value of 33 nF. If intermodulation spec shall be valid at 10 KHz carrier separation, the decoupling cap shall have a value of 3.3 uF. In this case, smaller capacitors shall be parallelled with this large 3.3 uF capacitor, for providing low impedance/wideband decoupling at UHF frequencies as well.

A good way of checking IP3 independance of carrier frequency separation is to simply change the separation frequency f1-f2 from 10 kHz to 10 MHz or so, while watching the intermodulation level. If the bias decoupling capacitors are sufficient, the inter mod level will not change with carrier frequency separation.

 

[dopradar]

Thank you！

But the confused problem is the IIP3 difference between low frequency and upper frequency is too large, who can tell me the reason?

 

[vfone]

Using only one stage would be hard to get a wide-band for IIP3. One reason could be the lower gain of the amplifier at high frequencies (resulting higher IIP3), and the second reason could be the reactance of the by-pass capacitors that is different vs frequency. Or a combination of both.

 

[dsjomo]

As the attachment.

 

[dsjomo]

solution

a. Improve S12 @ f2nd (if f1st = 900M, f2nd = 1800M, improve S12 @ 1800M)
b. IIP3 is related to I/O matching at f2nd, trap or match @ f2nd at output node prevents f2nd bouncing backward to LNA. Trap or match @ f2nd at input node to prevent the bounced f2nd bouncing again into LNA. Trap the f2nd on the biasing network to prevent the f2nd leakage to input node.

 

[dopradar]

Thank you, Mr. dsjomo.

And I understand what you mean, But how should I take effects to improve S12 @f2nd, and how to design trap or match @f2nd.

 

[dopradar]

Hi, all.

You can analyze this LNA topology (as shwon by LNA2.jpg). if we use such topology, the IIP3 difference between low and upper frequency band is smaller than the upper one (as shown by LNA.jpg). but the NF of such lna is not very good.

So pls help me analyze both LNA topology and give the comparison of such two LNAs.

And mention the confused problem again:what is the reason why the IIP3 difference between low and upper frequency is so large for the first LNA topology? and how to solve it?

Thanks!

 

[g579]

About the Noise Figure.
No wonder to increase, because the R2 resistor is in parallel to the transistor input,
so the some power is dissipated in it, and that loss will increase the NF.
Would comment more easily, if element values would be seen.
About IP3. Perhaps the 2nd amplifier is a wideband one, so the harmonic products are more similar to each other.

 

[dopradar]

Hi g579,

yes, the resistor loss really causes the larger NF, so I set R2 and R3 as large values to weaken its influence. is that ok? Thanks!

 

[g579]

Resistors make matching easier, amplifier more stable.
Should analyze the circuit and find an optimum trade-off between the more and less important things.

 

[dopradar]

But the 2nd amplifier has another problem, because it uses the feedback, so at high frequency band the stability factor < 1. For such topology which methods should be used to ensure LNA stable at the whole frequency band?

Added after 8 minutes:

Hi all,

For such lna design:low noise figure and good linearity at such wide band. Is there another topology to realize such lna? Are there some papers to describe new lna topology?

Thanks!