unstable due to substrate grounding with bondwire

[wizardyhnr]

during simulation of a lna operating from 3~3.5 GHz, i find that the stability of the lna is very sensitive to the inductor of bondwire which ground the chip substrate. if there is no inductor of bondwire, then the lna works perfect. But if the bondwire inductor reaches 0.4nH, then the lna doesn't maintain unconditional stable and power gain drops.
so it is very important to minimize the bondwire inductor of the substrate ground. a bigger pad can attach several bondwires and provide small bondwire inductor. Yet i think this is not enough. how can i minimize a smaller bondwire inductor? What if i add several substrate ground pads, each of which can attach three bondwires(then three pads can attach nine bondwires)?

 

[Ruritania]

Hi, wizardyhnr,

Yes, the bondwire could be one of the reasons introducing unstability, and it is easy to explain: at certain frequency, the souce impedance of the input transistor becomes capacitive, thus the real part of the input impedance of the LNA becomes negative correspondently. However, in some design, the bondwire is indispensible, for example, for source degeneration. Which means, such an issue CAN be resolved at all.

Normally, a bondwire with around 1nH is acceptable. Since you mentioned that even 0.4nH is too large for you, you may need to check the substrate parasitic capacitance and make some simple calculation to verify.

Wish this could help you some,

Ruri

 

[VSWR]

Adding severalbond wires in parallel is standard solution for minimizing the inductance. A bond wire has an inductance of about 1 nH/mm length, so four bondwires in parallel will give about 0.25 nH.

 

[wizardyhnr]

Hi, wizardyhnr,

Yes, the bondwire could be one of the reasons introducing unstability, and it is easy to explain: at certain frequency, the souce impedance of the input transistor becomes capacitive, thus the real part of the input impedance of the LNA becomes negative correspondently. However, in some design, the bondwire is indispensible, for example, for source degeneration. Which means, such an issue CAN be resolved at all.

Ruri

thanks, i have re-checked the circuits. The lna consists of three stages. each stage is unconditional stable and not sensitive to the substrate ground bondwire inductor. yet since the process rules require all substrates of transistors should be connected to a ground pad. After all substrates of transistors of the three stages are connected together, the circuits become very sensistive to substrate ground bondwire inductor. Some people say this is because every stage has feedback to each other. i wonder why. it is a terrifying problem!