The Designer's Guide Community
Forum
Welcome, Guest. Please Login or Register. Please follow the Forum guidelines.
Mar 28th, 2024, 11:48am
Pages: 1
Send Topic Print
50 Ohm match in IC? (Read 2813 times)
RC
New Member
*
Offline



Posts: 9

50 Ohm match in IC?
Oct 20th, 2002, 6:20pm
 
For on chip IC, we have LNA at front-end, which needs input 50 Ohm match, and Power-amplifier for output 50 Ohm match.
However, in between, in CMOS, we always have high impedance input/output, and we don't need to match them to 50 Ohm. Hence, when charaterize them, should we still use 50 Ohm, get the value in power and convert it back to voltage, dbV etc.?
In ADS(advanced design system) from Agilent, and spectreRF from Cadence, when we doing simulation for finding IIP3, NF, etc., I always confuse whether what value of the source impedance or load impedance need to be specified?
For IF block, eg. filter, can I use 50 Ohm for input (just to convert the voltage to power), and use a high impedance load at output(as the output impedance of the filter is high) ? But, how high? and will it affects the results?
For RF block, should we still doing input/output matching for the internal blocks for maximum power transfer (I mean the blocks after LNA and before mixer)?
For high frequency, we always talk about matching, and for low frequency, voltage is always as references. How about mixer (from high frequency to low frequency, in low-IF topology), its input is at high frequency and output at low frequency.

Thanks.


Back to top
 
 
View Profile   IP Logged
Eugene
Senior Member
****
Offline



Posts: 262

Re: 50 Ohm match in IC?
Reply #1 - Oct 20th, 2002, 10:43pm
 
I can sympathize with your question. When I first started analyzing RF systems I struggled with the 50 ohm issue too. As long as you are aware of the differences between your the quantities you measure and how you specify your models, you should not have a problem. However, that is harder than it sounds. There are factors of two lurking around every corner.  

As for how to measure/simulate noise figure and iip3, I think the trick is to use the port that simplifies the issue. For example, when talking about the receiver, I use input referred quantities. Then the issue of whether to use output dB or dBm is irrelevant. Everything is referred to input dBm,which are the units that matter. Similarly, for transmitters I use output referred quantities. Then the input units become irrelevant. All that matters are the output dBm.

If you are working with ICs, chances are you can only measure end-to-end performance unless you have dedicated test circuits on the chip. In any event, at some point you will probably want to compare simulation or calculation against measurement. Let's take noise figure for example. You design each block for a specific gain and noise figure. For the LNA, the 50 ohm source and load is physical. For the mixer, perhaps the load is a high impedance.  I wanted to compare end-to-end noise figure measurements and/or simulations to what I calculated using the Friis formula. I had to look at the chain very carefully with a manual calculation to realize that the Friis formula requires a slight modification when the chain switches from 50 ohm loads to high impedance loads. I had to draw the linear circuit out with voltage sources, source impedances, and input impedances then derive the composite noise figure. An extra factor of 4 appears in the later stages.

If you are using the Cadence RF behavioral library, I can say a few things about the 50 ohm assumption as it applies to the model specifications. Are you using that library?
Back to top
 
 
View Profile   IP Logged
Chi
Guest




Re: 50 Ohm match in IC?
Reply #2 - Mar 3rd, 2003, 10:17pm
 
Dear all,

   I have the same problem. Actually, in the RF circuit, all the measurement is in the power term. However, in the analog circuit, all the measurement is in the voltage term. Which is make me so confused that as the next loading in the interstage circuit (such as the output of the LNA to the input of the mixer) is high enough, then much of the voltage will be loaded to the next stage (by the voltage divider calculation). However, the power transfer is the lowest at this condition. Therefore, in the RF IC circuit, I should use the power transfer or voltage transfer for consideration? If using Power transfer as the consideration, WHY? Is it due to the all measurement apparatuses are in power term? So, in RF circuit are using power for calculation?

Best Regards,
Back to top
 
 
  IP Logged
rf-design
Senior Member
****
Offline

Reiner Franke

Posts: 165
Germany
Re: 50 Ohm match in IC?
Reply #3 - Mar 10th, 2003, 2:23am
 
Hi,

I would suggest to use a different terminology for internal rf stages.

First it is important to differentiate between voltage and current terminals. A terminal is a voltage terminal if the terminal voltage relative change is less than the current relative change if you attach the next stage. For current it is vice verse. If you have identified the type of terminals you can define output and input resistance. So for a pair of stages you can estimate the attachment loos. So I would not say coupling loss. For a classic receive path analysis including noise, distorsion and blocking be best describing value is dBVeff. So it refers to a 1.414 Volt amplitude sinewave. If the type of port changes from one stage to another the gain should be expressed as transimpedance or transconductance. The measure is then dBOhm of dBS. For equivalent noise calculations you should use the equivalent noise resistor. If all stages are defined by there  equivalent input noise resistor then a voltage input port have an series resistor and a current input port have a parallel resistor. Via that way IC receive path calculation is easy as classic ones. The difificult analyis parts are noise filtering, blocking effects and real to complex path split situations.
Back to top
 
 
View Profile   IP Logged
Eugene
Senior Member
****
Offline



Posts: 262

Re: 50 Ohm match in IC?
Reply #4 - Mar 10th, 2003, 1:55pm
 
The equivalent noise resistance idea is interesting but doesn't that make it hard to specify terminal impedances and noise figures independently? Or do you simply assume the equivalent noise resistance and terminal impedance are the same?
Back to top
 
 
View Profile   IP Logged
Asad Abidi
New Member
*
Offline



Posts: 3

Re: 50 Ohm match in IC?
Reply #5 - Mar 25th, 2003, 11:00am
 
I don't think the noise resistance idea will work very intuitively if, for instance, an LNA whose output impedance is mainly inductive, connects to a mixer whose input impedance is mainly capacitive. This is the usual case in integrated CMOS receivers.

The simplest way to deal with this, or any non-50 ohm situation, is to use the equivalent input noise voltage and current of each stage. After all, this is the only way noise is calculated in baseband circuits, such as the equivalent input noise of op amps where all internal nodes have a very high impedance. Cascade calculations then involve normalization by stage voltage or current gain, as appropriate.

However, in the RF context eventually all the noise is referred to the 50-ohm source connected to an LNA whose input impedance is also roughly 50 ohms. It is very easy to make an error in gain of 2X when referring noise to the source. The correct way to calculate noise is specified in equation 3 and the accompanying text of:

A. Rofougaran, G. Chang, J. J. Rael, J. Y.-C. Chang, M. Rofougaran, P. J. Chang, M. Djafari, J. Min, E. Roth, A. A. Abidi, and H. Samueli, “A Single-Chip 900 MHz Spread-Spectrum Wireless Transceiver in 1-µm CMOS (Part II: Receiver Design),” IEEE J. of Solid-State Circuits, vol. 33, no. 4, pp. 535-547, 1998.

which can be downloaded from:

http://www.icsl.ucla.edu/aagroup/PDF_files/33ssc04-rofougaran2.pdf
Back to top
 
 
View Profile   IP Logged
yorande
Guest




Re: 50 Ohm match in IC? About the paper
Reply #6 - Nov 30th, 2003, 8:35pm
 
Asad Abidi,

thanks for the link of the paper.
Could I ask a question of that paper?
In page 541(the marked page), it presents another way for noise figure calculation besides what is been shown in page 539.
It uses the output power desity.
And I think it could be expressed in a formular:
NF=[inputRFpower-(outoutpower-outputnoise)]-174dbm
But dosen't it dependent on the input power?
As I simulate with Hspice, the (outoutpower-outputnoise)  dosen't change in range where the circuit could work. Then the inputRFpower seems decide the NF.  But it's not true, isen't it?
So I wonder why could the paper get NF in that way?
And I don't quite understand about the part of limiting amplifier.
Back to top
 
 
  IP Logged
Feng
New Member
*
Offline



Posts: 2

Re: 50 Ohm match in IC?
Reply #7 - Jan 25th, 2004, 12:34pm
 
So, 50 ohm is not necessary for the inter stage interfaces, is my understanding right? for example, the output of an LNA is not necessary to be matched to 50 ohm since the input of the mixer is of high impedance. is that is true ? thanx
Back to top
 
 
View Profile   IP Logged
Cao-Thong Tu
Guest




Re: 50 Ohm match in IC?
Reply #8 - Jan 29th, 2004, 2:14am
 
Feng,

from my believe impedance matching is necessary
for power transfer. This comes from transmission
lines theory. In an IC, you usually have much
smaller length than the wavelength of the signal you may be using. For example : a 5 GHz has a wavelength
in the AIR of 6cm but your IC CMOS works in the range of several microns max. So any transfer of current
or voltage can be more or less considered to be local and the propagation term may be ignored. (Of course
this assumption is not totally correct because sometimes you may want to model long connection with
RLC distributed elements)

So now, where and why do we need to do impedance matching ? We need impedance matching anytime a signal has to be transfer on a distance which value is approaching its wavelength, because otherwise terminal reflection and losses occur. For example
a 50 Ohms RF generator may indicate -40dBm, but it implicitly means that it provides -40dBm on a matched
50 Ohms load.

So for your case impedance matching has to be done at the LNA input, so as to make sure the RF source will correctly provide the signal at the IC input. Then for LNA and mixers that depends if your using a let's say SAW filter at the LNA output. If yes, than impedance matching is required because SAW filters are also working on transmission lines mode. If not, then integrated LNA and mixers do not require matching at all, because an the IC, your information signal is now either LOCAL current or voltage. So assuming that in your case, the LNA output is a voltage then you might to have a high impedance at the mixers.

Just for info 50 Ohms value is more an historical
aspect than anything else, coming from the cable and
measurement instruments suppliers. Some ceramic filters
require higher source impedance such as 200-300 Ohms.
The choice of your LNA input impedance has to be done depending on current consumption and linearity requirements. Low impedance levels always require high current values ...

regards,

Cao-Thong

Back to top
 
 
  IP Logged
Cao-thong Tu
Guest




Re: 50 Ohm match in IC?
Reply #9 - Jan 29th, 2004, 2:17am
 
Yorande,

I'd rather use output noise to do any calculation.
Input noise is a mathematical view, but the only thing
you can really measure is output noise.

regards,

Cao-Thong
Back to top
 
 
  IP Logged
Chong Chen
Guest




Re: 50 Ohm match in IC?
Reply #10 - Jan 29th, 2004, 7:21am
 
thanx Cao-Thong, u made good comments on matching.

I am afraid my question is that 50 ohm is necessary for output of LNA or not?

I've been designing CMOS LNA, mixer. I designed each circuit in a way of 50 ohm in and out for faciliating the testing of the individual ciruit using network analyzer and RF wafer probes which are all 50 ohm instruments.

But in a real system, I think the output of LNA and input of mixer do not need to be matched to 50 ohm, as long as they are conjuate matched, say output of LNA is 200 + j100, and input of mixer is 200 - j100.

Then arises is the problem that I should design it for testing , or for real system ? and I think it is easier not to match output of LNA to 50 ohm. 50 ohm matching on-chip is hard. If I donot match LNA output to 50 ohm , how should test the LNA using 50-ohm network analyzer? Using the equations given by Dr. Abidi? Thanx a lot for clarifying these confusions I have.

Back to top
 
 
  IP Logged
Byron
New Member
*
Offline



Posts: 1

Re: 50 Ohm match in IC?
Reply #11 - Mar 2nd, 2004, 6:19pm
 
Let's continue the discussion.  At the input of the LNA the 50ohm   impedance match is necessary , and then some problems still need to be clarified.

1) How to deal with the ouput impedance of the LNA? For a low-IF or zero-IF topology, which is very popular in fully-integrated receiver design, 50ohm impedance match is not necessary between the LNA output and the mixer input. But do we need to match them to other impedance levels?

2) The cascade noise figure can be calculated  according to the Friis formula as long as we know the impedance of the LNA output and the mixer input. But there is always obviously difference between the hand calculation and the simulation from spectreRF.
Back to top
 
 
View Profile   IP Logged
engrvip
Junior Member
**
Offline



Posts: 12

Re: 50 Ohm match in IC?
Reply #12 - Jun 20th, 2017, 10:22am
 
Hi

So in case of a standalone lna in cmos to packaged, how shoul i use port in my ADE spectre rf testbench.

Input port should be 50ohm, but what about output port, what impedance value should be used here?

Also with VNA measurements, what port impedance should be used at port 2 of VNA, should it be kept 50ohm or something else?

Regards
Back to top
 
 
View Profile   IP Logged
Horror Vacui
Senior Member
****
Offline



Posts: 127
Dresden, Germany
Re: 50 Ohm match in IC?
Reply #13 - Jul 20th, 2017, 1:22am
 
If stand-alone then you will have no interstage matching on your die. You will interface it with the off-chip world. In this case if you do not know what impedance to use, just stick with 50ohm.
Back to top
 
 
View Profile   IP Logged
Pages: 1
Send Topic Print
Copyright 2002-2024 Designer’s Guide Consulting, Inc. Designer’s Guide® is a registered trademark of Designer’s Guide Consulting, Inc. All rights reserved. Send comments or questions to editor@designers-guide.org. Consider submitting a paper or model.