passive mixer----help me

Hi, friends

I am designing a passive mixer.

When I stimulate, I follow the pdf file” LAB2_Mixer_Tutorial”

I have two questions.

First:

As you can see in the schematic I have uploaded, RF PORT and LO PORT are all paralleled with a 50 Ohm resistor.
Why?
It is a necessary to add the 50 Ohm resistor?

In my opinion, a circuit actually does not contain the resistor. But when I stimulate noise figure, I find the resistor contribute noise to noise figure. Then it is the noise figure correct?

Second:

A off-chip bandpass filter will followed the mixer. The bandpass need 50 Ohm source and 50 Ohm load. Considering the isolation and driving capacity, I add a source follow buffer. The buffer has gm=20 ms.

In the circuit, the 50 Ohm resistor at the output of mixer represents the bandpass. Is it right?


Thank you all~~~

Hello

they used this resistor to provide a wideband matching for both RF and LO.
it is correct , but it is not necessary but u can use them as long as ur designing wideband circuit ,

4 sure this resistor will contribute to the NF , coz it will attenuate ur RF signal.

yes u can add the 50 Ohm load , which will be ur BPF

Khouly

khouly said:

Hello

they used this resistor to provide a wideband matching for both RF and LO.
it is correct , but it is not necessary but u can use them as long as ur designing wideband circuit ,

4 sure this resistor will contribute to the NF , coz it will attenuate ur RF signal.

yes u can add the 50 Ohm load , which will be ur BPF

Khouly

Click to expand...

Thanks for reply.

I read a post from https://www.designers-guide.org/Forum/YaBB.pl?num=1237656418/0

The post says add 50 Ohm to PORT will not get correct CG NF and IIP3 ?

WHy?

YES,my mixer is a wideband mixer.
Does it mean that i have to add the 50 Ohm resistor in the real circuit?
if a LNA (50 Ohm input and output match) is behind the mixer or if i test the mixer(mean it connect to 50 Ohm instrument port), is it necessary to add the 50 Ohm resisitor?


When i add a 50 Ohm at buffer out, the noise is about 15 dB. But at the net MIXO_N MOXO_P, NFis 10dB. It seems that 50 Ohm at buffer out contribute huge noise.

However, when i add a bandpass which is genarated by filter tool, the circuit NF is keep 10dB at buffer out. Why?

I stimutate the generated filter with SP analysis(of course port is 50 Ohm and load is 50 Ohm). At IF frequency, the filter is about 0dB loss, and input impedance is 50 OHm.

Why NF becomes bad, when adding a resistor directly?

And why add the bandpass in the stimulation is not correct?

Another:
https://www.designers-guide.org/Forum/YaBB.pl?num=1231854969/5#5
says :noise figure is defined as the ratio of signal-to-noise at the input to signal-to-noise at the output. The output would not be a problem if it was not terminated with a port, because all that happens is that the noise gets excluded from the load anyway.
I think 50 Ohm at buffer out is not port. So 50 Ohm resistor should not contribute noise.

And the 50 Ohm resistor is not a real resistor.

steven_l said:

I read a post from https://www.designers-guide.org/Forum/YaBB.pl?num=1237656418
The post says add 50 Ohm to PORT will not get correct CG NF and IIP3 ?
WHy?

Click to expand...

You are misunderstanding.
If added 50ohm are truely part of circuit, it will give you correct CG, NF and IIP3.
See https://www.designers-guide.org/Forum/YaBB.pl?num=1277737192/1#1

steven_l said:

YES,my mixer is a wideband mixer.
Does it mean that i have to add the 50 Ohm resistor in the real circuit?
if a LNA (50 Ohm input and output match) is behind the mixer or if i test the mixer(mean it connect to 50 Ohm instrument port), is it necessary to add the 50 Ohm resisitor?


When i add a 50 Ohm at buffer out, the noise is about 15 dB. But at the net MIXO_N MOXO_P, NFis 10dB. It seems that 50 Ohm at buffer out contribute huge noise.

However, when i add a bandpass which is genarated by filter tool, the circuit NF is keep 10dB at buffer out. Why?

I stimutate the generated filter with SP analysis(of course port is 50 Ohm and load is 50 Ohm). At IF frequency, the filter is about 0dB loss, and input impedance is 50 OHm.

Why NF becomes bad, when adding a resistor directly?

And why add the bandpass in the stimulation is not correct?

Another:
https://www.designers-guide.org/Forum/YaBB.pl?num=1231854969/5#5
says :noise figure is defined as the ratio of signal-to-noise at the input to signal-to-noise at the output. The output would not be a problem if it was not terminated with a port, because all that happens is that the noise gets excluded from the load anyway.
I think 50 Ohm at buffer out is not port. So 50 Ohm resistor should not contribute noise.

And the 50 Ohm resistor is not a real resistor.

Click to expand...

I can't understand meaning of your questions.

Describe in detail and correctly using correct terminologies.

Sorry,.
i describe my quqetion in detail.

My mixer is a broadband mixer. And a off-chip bandpass filter will follow it. The filter needs 50 Ohm source and 50 Ohm load.

Considering the filter requirement, I add a source follower buffer. The buffer has gm=20ms.

You can see in the uploaded picture file “sch without bandpass”and “sch with bandpass”.

In the file “sch with bandpass”, a bandpass is inserted between buffer out and 50Ohm resisitor.

And at the mixer out, I label “MIXO_P” and “MIXO_N”. Also, at buffer out, I label ”BUFFER_P’ and “BUFFER_N”.

The bandpass generates from the “Filter Solution Tool”.
And when I stimulate the filter with SP analysis, I find the input impedance is 50Ohm at 200MHz


My stimulation follows the file “Lab2_Mixer_Tutorial”

When I set up the stimulator such as PSS, PAC and so on, I just change the RF frequency, LO frequency, IF frequency, RF power and LO power.
Other setup are the same to pdf file “Lab2_Mixer_Tutorial”
Basic parameters are :

RF=900M
LO=700M
So: IF=200MHz
Prf: -30dBm@ 50 Ohm
Plo:10 dBm @50Ohm

Because RF signal is wideband, so I try several different frequency in order to confirm my mixer can work well.

Mainly, I use PSS+PAC for CG, PSS+Pnoise for NF QPSS+QPAC for IIP3.

One difference can be found in the two picture file “me_CG_Dircec_Plot” and “pdf_CG_Dircec_Plot”

In the pdf file “Lab2_Mixer_Tutorial”, the author plot in the method: pac---voltage---differential—sideband—peak---db20---output Sideband -1 100M-300M.

My method: pac---voltage gain--+ -Output and + - inputNets---Sideband—dB20---Output Sideband -1 100M-300M Input Side band 0800M-1G.

I find in my method, the voltage gain of PAC is the same to transient. In transient analysis, I plot differential signal @+ -Output and + - inputNets. Then do dft analysis to find RF and IF signal magnitude. Finally calculate 20log(Vout/Vin)


The +-input net ate RF_N and RF_P. The output net are MIXO_N and MIXO_P or BUFFER_N and BUFFER_P.


Now here are questions.

When doing stimulation “sch without bandpass”, I plot NF, CG.
See picture file
“CG without bandpass @ mixer out”,
“NF without bandpass @ mixer out”,
“CG without bandpass @ buffer out”,
“NF withoutt bandpass @ buffer out”.

Also, I stimulate “sch with bandpass”.
And plot NF, CG.
See picture file
“CG without bandpass @ mixer out”,
“NF without bandpass @ mixer out”,
“CG without bandpass @ buffer out”,
“NF withoutt bandpass @ buffer out

You can see: CG changes huge at differential output. Also NF changes huge, too.

1
NF changes because the load resistor @ buffer out or at filter out contributes noise.
But you say https://www.designers-guide.org/Forum/YaBB.pl?num=1231854969/5#5
The output would not be a problem if it was not terminated with a port, because all that happens is that the noise gets excluded from the load anyway.
I think 50 Ohm at buffer out can be viewed as a port. So 50 Ohm resistor should not contribute noise.


How can I reduce NF?

Adding buffer and 50Ohm load, I need CG and NF are all bellow -10 dB.
Can this performance be reasonable?

2
CG curve at mixer output has gain above 3.9 dBb. It is impossible. The maximum CG is 2/pi, or -3.92dB.
What is wrong with my circuit? The method I plot CG curve is wrong?

According to theory, NF shouleg be equal to CG.
But in my circuit, it is not ,Why?[color=darkred][/color]

steven_l said:

NF changes because the load resistor @ buffer out or at filter out contributes noise.
But you say https://www.designers-guide.org/Forum/YaBB.pl?num=1231854969/5#5
The output would not be a problem if it was not terminated with a port,
because all that happens is that the noise gets excluded from the load anyway.

Click to expand...

You don't understand an operation of your mixer.
Load impedance value affects an operation of your mixer, so it could degrade CG, NF and IP3.

Even if Load impedance value does not affect an operation of your mixer, Noise characteristics after Mixer could affect total NF, because your mixer is gainless and there is no LNA before Mixer.

Simply consider NF for cascaded system.

steven_l said:

I think 50 Ohm at buffer out can be viewed as a port.
So 50 Ohm resistor should not contribute noise.

Click to expand...

Right.

steven_l said:

Adding buffer and 50Ohm load, I need CG and NF are all bellow -10 dB.
Can this performance be reasonable?

Click to expand...

Negative NF value is impossible physically.

steven_l said:

According to theory, NF shouleg be equal to CG.

Click to expand...

I don't know such theory.
I think NF_dB > abs(CG_dB) for mixer with CG_dB <= 0dB.