voltage gain variation over PVT corners

Hi, all.
I am now designing the RF front-end for 2.4ghz band, including a LNA and IQ mixers. By using gm-constant biasing for LNA and mixers, there is still about ±1.5 dB variation in voltage gain over -45~85 ° temp variation in a given process corner, and can be up to about ±3dB gain variation for fast-best, typical and slow-worst corners.

fast-best corner is the combination of -45° temp, highest VDD and ff process corner, typical corner is the combination of 45° temp, normal VDD and typ process, while slow-worst corner is the combination of 85° temp, lowest VDD and ss process corner.

My LNA's gain is provided by the product of the gm and LC tuned tank at 2.4ghz, while mixer's gain is also given by gm*R. The gm-constant biasing using on-chip resistors, and there is about ±12% variation over the 3 corners.

Did anyone have senses or experiences in the pvt variations? What is about the typical value of the gain variation for RF front-end? Or, do you have any good ideas to decrease these numbers?

I will try to answer on generic terms as I don't know what technology you have or what other blocks you use.

-What kind of resistors ( if any) you are using. Is this the best tolerance you can use?

- Can you use Bandgap voltage in parts of your design as VCC?

- Have you tried to compensate process variations with circuitry technics?

D.

dkace said:

I will try to answer on generic terms as I don't know what technology you have or what other blocks you use.
D.

Click to expand...

-What kind of resistors ( if any) you are using. Is this the best tolerance you can use?
I am using TSMC18rf, and rphres as the resistor. This type is the best tolerance over temperature (within 5% over -40~85°), but still have about ±12% variation over process.

- Can you use Bandgap voltage in parts of your design as VCC?
Yes, and also an internal LDO is used to regulated VDD (3.3v with 2.1~3.6V variation) to internal 1.8V for the front-end. The internal VCC is almost constant at 1.8V over external VDD variations.

- Have you tried to compensate process variations with circuitry technics?
gm-constant biasing is used to bias the LNA and mixer. PTAT current from bandgap reference was also tried, but is worse than the gm-constant bias.