Continue to Site

# Is IP3 always 10dB higher than P1dB?

Status
Not open for further replies.

#### cmosbjt

##### Full Member level 5
p1db ip3

10dB is a theoratical number, from what I've seen (testing) so far, this is true.

But I am wondering if you have seen, expecially from measurement data, that this rule of thumb doesn't hold anymore? In what circuit? Passive or active? And do you know why?

Thanks a lot

#### khouly

ip3 p1db

there are 2 test cases with 2 different values of the difference between 1 dB and IP3

case 1
=====
make single tone test and plot the Pin Vs Pout ,
make 2 tone test and in the same axies plot the Pin Vs IM3
this will result 10 dB difference between 1 dB and IP3

Case 2
=====
make 2 tone test to plot the fundametal by this will plot the 1 dB compression point
and plot the IM3

this will result about 15 to 16 dB difference between 1 dB and IP3

"check the book radio frequency intergrated circuit" by Calvin Plete

khouly

#### Element_115

what is ip3

If you have one device there are equations that show IP3 and P1dB are
~9.6dB different, but if you have a system this might not be the case.

#### byteptr

##### Member level 4
p1db iip3

"If you have one device there are equations that show IP3 and P1dB are
~9.6dB different, but if you have a system this might not be the case.
"

Yes, since you can add more devices via splitters-power combiners

#### gszczesz

##### Full Member level 2
iip3 p1db

I have made a document that explains the 10dB rule and how it can be increased or altered:

Greg

V
Points: 2

#### stober

##### Junior Member level 3
iip3 and p1db

khouly said:
Case 2
=====
make 2 tone test to plot the fundametal by this will plot the 1 dB compression point
and plot the IM3

this will result about 15 to 16 dB difference between 1 dB and IP3

khouly

This is because each tone works as a blocker to each other. In fact, I always think the case 2 is a wronge way to test the 1dB CP, though it even occasionally shows up in literatures.

Stober

#### khouly

yeah , i agree with u stober
the 1st case is the main one

Khouly

#### RF-OM

ip3 plot

Theoretically it is not difficult to show that IP3 is 9.54 dB higher than P1dB point. This is why 10dB is considered as rough estimate. But such analysis is done with suggestion that coefficients into polynomial members are constant. But when signal exceed P1dB point it may be not true anymore. Some power amplifiers have IP3 just about 1 dB above the P1dB level. For broadband amplifiers third order is not always well enough represent actual IMD. Higher order products may contribute a lot and reduce IP3 level.

#### iaman

##### Junior Member level 1
input p1db how to make measurement

RF-OM said:
Theoretically it is not difficult to show that IP3 is 9.54 dB higher than P1dB point. This is why 10dB is considered as rough estimate. But such analysis is done with suggestion that coefficients into polynomial members are constant. But when signal exceed P1dB point it may be not true anymore. Some power amplifiers have IP3 just about 1 dB above the P1dB level. For broadband amplifiers third order is not always well enough represent actual IMD. Higher order products may contribute a lot and reduce IP3 level.

yes, I think so. Acturally I think Case 2 test is more accturater than Case1. From my experince , i think IIP3=IIP1+13

#### humbleRF

##### Newbie level 4
iip3 p1db 10 db

10dB rule only applies when the power coefficient is constant, which is roughly true in block with single transistor and without feedback. In more complicated case, the difference is usually larger.
By the way, IIP3 is a virtual number (intecept point of fundamental and 3rd order curves extrapolated from linear region); it indicates how linear is the linear region. when the circuit is saturated (close or beyond P1dB), the calculated Pi+(Po-Pim3)/2 (dB) is not IIP3, it doesn't mean anything. one block only has one IIP3 number. In many cases, when IIP3 is not efficient way to calculate system performance, then THD is used at specific input level.

#### RF-OM

p1db vs ip3

It looks like our discussion went far from initial question. We are talking about some issues that actually very rarely rise. The engineering theory is just a tool which serves to get answers to common questions and allow engineers to do the job. This is why Frees introduced Intercept Point conception in July of 1944 (if I remember it right). It is convenience way to describe system non-linearity, but not so much more. In majority of cases it is good enough, but not always. Again, I need to state that IP conception was derived in suggestion that all coefficients in polynomial approximation of analyzing device are constant. But it is definitely not the true for all cases. Let’s consider simple, but rarely taking into account, soldering point problem. Usually we suggest that this contact is linear, but often this is not true. RF engineers sometimes have a hard time to investigate and fix the problems originated by poor contacts.

We may think that soldering point is good contact and has constant coefficients in polynomial that describes it transfer function. Yes, it is acceptable for some level of signals, but more often than not bad solder point works as non-linear element. And to make things worse it is very broadband element. If you try to do IMD analysis for broadband element you probably will very surprise with the effect from high order products. IP3 often not a major contributor in this case, and depends on system parameters and bandwidth 5th, 7th, and so on order products dominate. I investigated cases when 15th order was the major contributor to the system non-linearity and I believe that this is not the last order. Probably even higher order products contribute as well, but we cannot measure such products with contemporary spectrum analyzers in average lab.

By the way, the very first demonstration of radio on May 7, 1895 used bad contact device named “koherer” as a detector. Therefore, bad contact’s non-linearity is old as the radio itself. By the way, this very first usage of radio actually was UWB!

I think that this is really not a good place to such a discussion. For majority of RF engineers everyday work it is enough to know that 10dB rule is good for almost any case, but sometimes it is necessary to be very careful. There is no one rule for everything.

Best regards,
RF-OM

#### biff44

ip3 polynomial

Just a rule of thumb. IP3 is usually specified as a measure of two-tone intermod products, in dBc. As such, there is variation in IP3 level as the two tone input power is varied. Also, there are other effects, like bias circuit capacitance vs. two-tone frequency spacing. Finally, there is the "sharpness" of the P1db knee.

All in all, I have seen numbers from 3 to 20 dB.

Here are some with 18 dB:

#### RF-OM

ip3 p1db relation

There even 19 dB difference for PCS band amplifier! RFMD is the best for cell phone amplifiers, so it is not very surprising that their MMICs are so good. I am proud that I worked for this company, but for another design center.

BR,
RF-OM

#### mouloud08

##### Junior Member level 1
why is ip3

Dear co,

In pozar, IP3 is between 12 and 15 dB bove 1 dB compression point

In Gonzalez, IP3 is only 10 and 15 dB bove 1 dB compression point

So, commonly, you will find IP3 is taken 10 dB...

Mouloud

#### Jim cage

##### Full Member level 2
ip3 pozar

No. Its only theoretical point. In practice when you are designing circuits you will encounter a phenomenom called "Sweet Spot" where in fact, the IIP3 is better then 10dB.

#### DDavid

##### Full Member level 5
plotting ip3 point

Hi,

this paper a little difficult but will explain what i need!

David

Points: 2

### John Miles KE5FX

Points: 2

#### RF-OM

what is p1db

Interesting presentation (it is not a technical paper), but it is more for academia than for engineering and may be confusing for this topic. Major drawback is that this analysis was done for narrowband systems. Also some assumptions were made that further restrict this text usage. It is definitely good for some applications, but not for this discussion. Broadband system linearity is different.

Best regards,
RF-OM

John Miles KE5FX

### John Miles KE5FX

Points: 2

#### shabzbd

##### Newbie level 3
what p1db

iaman said:
RF-OM said:
Theoretically it is not difficult to show that IP3 is 9.54 dB higher than P1dB point. This is why 10dB is considered as rough estimate. But such analysis is done with suggestion that coefficients into polynomial members are constant. But when signal exceed P1dB point it may be not true anymore. Some power amplifiers have IP3 just about 1 dB above the P1dB level. For broadband amplifiers third order is not always well enough represent actual IMD. Higher order products may contribute a lot and reduce IP3 level.

yes, I think so. Acturally I think Case 2 test is more accturater than Case1. From my experince , i think IIP3=IIP1+13

this is exactly what i came across. during taking some practical results..

Cheers!

#### fujian

##### Newbie level 3
iip3 vs. p1db

these days i'm simulating OIP3 and 1dB compression point of pa @2.4GHz , i found when p1dB equals 23dBm, OIP3 only about 25~26dBm, I don't know why

#### lkjhgf

##### Junior Member level 3
p1dbå’Œiip3

For more details read Practical RF System Design by William Egan

Status
Not open for further replies.