FET Resistive Mixer, how to improve the linearity?

[cmosbjt]

resistive mixer

I am designing a FET resistive mixer using E mode pHEMT.
The initial design has good matching and port to port isolation.
But it has poor IIP3 (only 1 dBm).

So how to choose the device size and bias voltage for better IIP3?
The attached image shows the DC I-V curve of the device I use.
I am biasing it with 0.5V Vgs. Is there something wrong?

thanks a lot
cmosbjt

The latest measurement result is updated here in this post.

Hi my friends,

I finally have my mixer fabricated and tested. Here are some of the testing results. I'd like to continue the discussion on how to improve the mixer IP3.

I have 2 mixers tested.
The 1st one has a channel width of 250 X 5 um
The 2nd one has a channel width of 250 X 8 um
Both mixer has similar conversion Gain, about -7 dB.

You can see from the attached pictures, the IIP3 of both mixers is in the range of 20 to 30 dBm.

Here are some observations:

* The LO to RF isolation is not very good. Since the IF Freq is low, RF and LO Freq are close. It is hard to improve L2R iso by filtering.
* I do see the IIP3 is about 6 to 10 dB higher than P1dB(in) for this topology (Resistive FET mixer)
* I don't see a bigger (larger channel width) device has better IIP3.

** Does a FET device with longer channel length has better IIP3 ???

I don't think you can get IIP3 of 30 dBm of a resistive FET mixer, having only 0 dBm of LO drive level. You neet at least 15-20 dBm LO drive.

You are right. I am using 17 & 20dBm LO in my testing, but how can I further improve the IIP3? Thanks.

the IP3 is related to the 1 dB compression point , it should be about 10 dB higher
Khouly

Yes, the measurement result does shows about 7 to 10 dB difference between IIP3 and P1dB (IP3 is higher). I don't have a plot here, but I have about 18~21 dBm P1dB from my mixers.

But I do want to point out that this is not always true. For some of my LNA designs (SiGe), IIP3 is more than 15dB higher than P1dB(in).

20 dBm of LO power is very high , i think 20 dBm IIP3 , may be the limit of your transistor , try to use larger device , with 350 um gate width
khouly

Definetely 20dBm IIP3 is not a problem for my device as you can see from the testing result. I have a very similar mixer (different device) from other company, it have more than 30 dBm IIP3. And the spec lower limit is 30dBm.

cmosbjt,

When creating a 20dBm IP3, you will need to generate an P1dB of 10dBm. This is a large power and will require a large device, which means the input impedance into the device on the LO and RF side will be low. This will require to have a large matching network and will force the design to be more narrow band.

Obviously you acheived 1dBm IP3 with a -9dBm P1dB. Just scaling everything to be 10X Larger, re-matching the circuits and increasing the LO drive by 20dB will give you the IP3 you require since the P1dB will be increased to be over 10dBm.

Remember that you also need to scale down the IF impedance equally. When you scale up the circuit, you are preserving the voltage swings and increasing the current levels. So if you increase your circuit by 10X in size, the IF section needs to drop in resistance by 10X. Everything needs to scale.

Greg

Hi gszczesz,
From the measurement result, I don't see a bigger device (with 8 fingers) has better IIP3 than a smaller device (with 5 fingers).

 

[khouly]

resistive fet mixer

for a good performance , u need to bias the gate at the threshold voltage

about the IIP3 , it mainly depend on the device size , so as thedevice gets wider the better IIP3 , also in FET mixers the IIP3 depend on LO power , so u need to optimize it

khouly

 

[cmosbjt]

fet resistive mixer

for a good performance , u need to bias the gate at the threshold voltage

about the IIP3 , it mainly depend on the device size , so as thedevice gets wider the better IIP3 , also in FET mixers the IIP3 depend on LO power , so u need to optimize it

khouly

I am biasing it with 0.5V. As you can see from the image, 0.5V is very close to the threshold voltage.

I also did a IIP3 vs LO power simulation, I got the best IIP3 of 10dBm at 0dBm LO voltage. But the spec is 30 dBm minumum. I have a long way to go.

About the device size, I will run some experiment on different size. But the problem is, as the device size getting biger, it is harder to do the matching. The size I am currently using is 250um/0.5um (W/L), which give me a impedance very close to 50 Ohm at the RF/IF port.

 

[khouly]

high level mixer fet

the 30 dBm IIP3 is very high. i think it is the OIP3 of the mixer

which device r u using .or u use a GaAs process ?

within my master thesis , i have desigened Resistive FET mixer

khouly

 

[cmosbjt]

masters thesis mixer

the 30 dBm IIP3 is very high. i think it is the OIP3 of the mixer

which device r u using .or u use a GaAs process ?

within my master thesis , i have desigened Resistive FET mixer

khouly

The 30 dBm is definetly IIP3 because the resistive FET mixer has about 7 dB conversion loss. The OIP3 will be 7 dBm lower, which is 23 dBm.

I tried a bigger device size, 6x the original one. The bias voltage is set to 0.3V.
It doesn't do any good to the linearity. IIP3 is still around 0 dBm. I think bigger
device can handle higher power and thus improve P1dB. But IP3 test is all about
small signal. Device size doesn't play much role on it. Am I right?

 

[VSWR]

mixer dbm project

I don't think you can get IIP3 of 30 dBm of a resistive FET mixer, having only 0 dBm of LO drive level. You neet at least 15-20 dBm LO drive.

 

[cmosbjt]

fet mixer experiment

I don't think you can get IIP3 of 30 dBm of a resistive FET mixer, having only 0 dBm of LO drive level. You neet at least 15-20 dBm LO drive.

In simulation, I do use 20 dBm LO initially. I got the similar IIP3 when I change the LO to 0 dBm. BTW, the spec does mentioned a 20 dBm LO.

 

[khouly]

fet mixer

20 dBm of LO power is very high , i think 20 dBm IIP3 , may be the limit of your transistor , try to use larger device , with 350 um gate width

khouly

 

[cmosbjt]

resistive mixer

20 dBm of LO power is very high , i think 20 dBm IIP3 , may be the limit of your transistor , try to use larger device , with 350 um gate width

khouly

Already tried. As said in my previous post, larger device doesn't improve IIP3 according to my simulation. Not sure if that's real.

 

[khouly]

how to improve rf mixer ip3

the IP3 is related to the 1 dB compression point , it should be about 10 dB higher

which device do u use ?

Khouly

 

[cmosbjt]

x band mixer fet

the IP3 is related to the 1 dB compression point , it should be about 10 dB higher

which device do u use ?

Khouly

I have the same question on the IIP3 simulation result and I think the accuracy of the model is skeptical. But is the rule of thumb that IIP3 is 10 dB higher than P1dB always true?

I am using a E mode pHEMT model from RF*D

 

[khouly]

fet size ratio to current oip3

As far as i know , yeah
i have tried it , it is always like this
i have used the NEC NE4210 Heterojunction FET , it gavve nice results

is it a packaged device , or a judt die

Khouly

 

[cmosbjt]

how to choose fet

As far as i know , yeah
i have tried it , it is always like this
i have used the NEC NE4210 Heterojunction FET , it gavve nice results

is it a packaged device , or a judt die

Khouly

From the project and product I've done, I also get not exactly but about the same result: 10 dB difference between IP3 and P1dB. But sometimes people have trick to short circuit the low frequency and harmornic to boost up the IP3. Well that means the IP3 is more than 10 dB higher. So ... even harder to explain ...

That's die level, simulation only. We are going to do some experment later.

BTW, what's the IP3 and other performance of your mixer using NE4210?

Thanks a lot

 

[khouly]

boston fet

the mixer followed by the upper band select filter give
-9dB conversion gain

1 dB compressoin point (IF input Power) 7 dBm

IIP3 16 dBm

khouly

 

[gszczesz]

ne4210

cmosbjt,

When creating a 20dBm IP3, you will need to generate an P1dB of 10dBm. This is a large power and will require a large device, which means the input impedance into the device on the LO and RF side will be low. This will require to have a large matching network and will force the design to be more narrow band.

Obviously you acheived 1dBm IP3 with a -9dBm P1dB. Just scaling everything to be 10X Larger, re-matching the circuits and increasing the LO drive by 20dB will give you the IP3 you require since the P1dB will be increased to be over 10dBm.

Remember that you also need to scale down the IF impedance equally. When you scale up the circuit, you are preserving the voltage swings and increasing the current levels. So if you increase your circuit by 10X in size, the IF section needs to drop in resistance by 10X. Everything needs to scale.

Greg

 

[cmosbjt]

high ip3 fet mixer

The updated testing result is in the original post on the top

 

[kspalla]

bias fet

Hi,
From the calculations, the LO power will be around 20 dBm.
To avoid such medium power input requirement for LO for the proper operation. internal amplifiers are the choice.

 

[lladnar23]

improve ip3 linearity fet

The "rule-of-thumb" relation between P1dB and IP3 holds when the nonlinearity is dominated by the third-order product.

 

[cmosbjt]

fet resistive mixing

The "rule-of-thumb" relation between P1dB and IP3 holds when the nonlinearity is dominated by the third-order product.

I think you are right, but in what case the 3rd order product dominate the nonlinearity or in what case it doesn't?

Thanks

 

[sozkarabacak]

narrow band fet mixer

HELLO,

This image NFM receiver for antenna amplifier+impedance coupling+mixer+local osc+if filter 10m7

This project link : **broken link removed**