Impedance matching in RFIC

If I am designing a CMOS class E power amplifier (IC, not PCB)working at 433MHz, do I have to match the impedance to 50 ohm between driver stage and output stage for maximum power delivery? How about, say 40 ohms or 60 ohms or what it is between the two stages?

If my PA receives RF signal from a mixer in a transmitter, do I have to make sure impedance between Mixer and PA blocks is matched to 50 ohms? How about other values, say 60 ohms or 40 ohms, or what it is between the two blocks?

As for as the output of my PA, it directly drives an antenna which is 50 ohms. So I know that I have to match the output impedance to 50 ohms because of the antenna.

Thanks for your help.

The classical impedance matching techniques can also be applied for RFIC circuits with some constraints.
Generally bonding wires can be used for "shunt L component", and series MIM capacitors complete the matching circuit.But as I said before, matching will have some
constraints and might not be applied to all kind of circuit to be matched.Parallel L and Series C configuration is frequently preferred but if you have place, some
on chip inductros can also be convenient.

You have to match the driver to the PA input impedance, whatever it is. The value will be determined by circuit analysis.

Regarding matching of external input and output ports, you can't build effective passive matching elements for 433 MHz in chip design. Particularly output amplifiers will require external matching networks to achieve high efficiency.

Maybe this tool for the matching can help you:

Tools & Converters

Usually interstage matching is NOT used in RFIC. In particular,for low frequency such as 433MHz, there is no advantage to do it.
The only disadvantage you will have if you match your circuit is to have more power consumption.

The reason is really simple, and is related to signal wavelength related to circuit phisical dimentions. It is obvious that, for 433MHz, the IC will be many order of magnitude smaller than wavelength.

A different matter is related to interface with external circuits, such as bondings, passives and transmission lines. Here the assumption of small (lumped) circuit is no more valid, and all distributed parameters design techniques have to be used (let me tell this very generally, considering power matching only as one possible choice).

I'm saying this because for PAs, this is not (usually) a strong requirement, but more important are power delivered to the load and efficiency. And you will probably discover, during your design, that the best efficiency is NOT when you are power matched.

For PAs in class E, a good procedure is to find the optimum load (with load pull) and try to reproduce it (adding harmonic filters) in real application.

I hope it can help.

Mazz

Yes. But I don't see a contradiction. Load-pull is just an effective method to achieve impedance matching of a non-linear source.