Impedance matching inside the IC

viperpaki007

Full Member level 5
Joined
Jul 2, 2008
Messages
274
Helped
11
Reputation
22
Reaction score
8
Trophy points
1,298
Location
Finland
Activity points
3,437
Hi,

As far as I know, impedance matching is done in order to transfer maximum power from source to the load and also to avoid signal reflections back to load. Reflections are problem at IC-PCB interface because dimensions of the interface are comparable to wavelength of signal. However, when we go inside the IC and we are operating at lets say 1GHz frequency, IC dimensions are much smaller than signal wavelength. Therefore, reflections should not be a problem. Therefore, I think that impedance matching between LNA and mixer may not be necessary. I have seen people designing cascaded amplifier stages inside the IC in voltage mode (i,e. low source impedance and high load impedance). My question is do we really need impedance power match between the amplifiers inside the IC or can we just design amplifiers in voltage mode interface? signal can be transferred in voltage mode as well so why to do power match?
 

Let's say a block A is driving block B.

Given an input impedance of B, the swing at the input of B will be maximised if there is conjugate matching of output impedance of A and input impedance of B.

So, yes, impedance matching always helps to improve voltage swing which helps everything (except linearity lol).

Do you have to match them to 50 ohms inside the chip? The answer is no. I have designed chips where the internal impedances used was 75 ohms as opposed to 50 ohms. Only at the external interfaces, 50 ohms was used.
 

Thanks for the reply. However, I do not fully get it. How can the voltage swing be maximized for conjugate matching? It will be maximum if source impedance is zero and load impedance is infinity (In ideal case). In conjugate matching, voltage at load will be half of source voltage.

I have seen chip designs around GSM frequencies where interfaces inside the IC are not matched at all. For example, LNA-mixer interface is designed with voltage mode interface. i.e. low LNA output impedance in comparison to high mixer input impedance.
 

I'm no RFIC designer but have participated in RF chip designs and I saw the RF guys pick a 500 ohm impedance for on chip, 50 ohms was just going to eat power and help none.

Something about distance less than tline electrical length by some factor = "fuggedaboudit". 2Ghz GPS band was not HF enough to get fancy over.
 

Conjugately Matching promises Maximum Power Transfer not Voltage.
Matching Circuits are not practical in RFIC Circuits due to space constraints. In additional to this, Impedances are not very reasonable to be matched for certain circuits.
 
Thanks BigBoss. So does this mean that impedance matching is not done inside the IC at low frequencies because it is not feasible to do impedance matching because of impractical matching components size and higher current consumption? Lets suppose, we would not have these constraints. Will we still do matching then? Can't the signal be transferred in the form of voltage?
 

As I've said before. Matching is not practical inside RFIC Circuits due to very short distances.
But Matching is somehow possible in Very High Frequencies like mmWave or Microwave.These are seen in MMIC Processes.
I didn't understand the last question. What does it mean " transferring in voltage form" ?
 

I mean from "transferring the signal in voltage form" that source impedance is ideally zero and load impedance is ideally infinite between the cascaded amplifiers. This means that there is no current transfer but only voltage signal transfer from one amplifier stage to another. i.e. signal can be said to be in the form of voltage.
 

Cookies are required to use this site. You must accept them to continue using the site. Learn more…