OIP3 & OP1dB values of an active mixer

[hrkhari]

Hi Guys:

I have a doubt here. When it is mentioned in the journals that the performance of the active mixer results in an OIP3=18dBm and OP1dB=8dBm, respectively. Are they referring to the modulus value of it, since I find it quite difficult in obtaining a positive value for the OIP3 and OP1dB, whereas a comparable positive value could be achieved for IIP3 and IP1dB. Can anyone shed some light on it. Thanks in advance

Rgds

 

[toonafishy]

oip3 rule of thumb

If the IIP3 is positive and the mixer has gain, then the OIP3 must also be positive since OIP3 = IIP3 + gain.

 

[Mazz]

oip3

Simply it seems a circuit that has high linearity, it can deal with large signals, so it has positive OIP3 and (as usually happens and it's theoretically predicible) OP1dB=OIP3-10dB.
I hope it can help.
Mazz

 

[g579]

oip3 dbm

Just a comment to Mazz's equation above.
The equation is a rough and good approximation for practice, a so called "rule of thumb", and not a general exact equation.

 

[hrkhari]

pasar de oip3 a iip3

Hi Guys:

Thanks for your beneficial inputs, I had obtained a Voltage Conversion Gain (VCG) of around 0.3dB, with an IIP3 of +15.8dBm and IP1dB of 4.25dBm, with regard to the estimation OIP3=IIP3+gain my OIP3 would be +16.1dBm and with regard to OP1dB=OIP3-10dB, my OP1dB would be 6.1dBm, but from the plot I obtain a OIP3 of around -15.1dBm and OP1dB of around -28.65dBm, I had attached the output spectrum plot of the VCG, the plot of IIP3 and 1dB compression point for your refeence. I really hope you could clear out my doubt. Thanks in advance.

Rgds

 

[Mazz]

impedance mismatch iip3

to g579: I agree completely.
From first plot (P1dB), if you see Pout(-30dBm)=-62 dBm -> Gain=-32dB; second plot confirm this.
Third plot shows the voltage gain, and is around 0dB. There is obviously a difference between the test bench used to evaluate the gain and the one used for linearity. Infact you have to use port to evaluate linearity in Spectre. The solution is in the different test benches (or at least in the different nets you use to evaluate gain and linearity), that you should post to let us understand something...
I hope it can help.
Mazz

 

[hrkhari]

oip3 rf

Hi Mazz:

Thnx for your reply, I had used PSS and swept PSS analysis to obtain the VCG and 1dB compression point plot, respectively. A combination of PSS and PAC analysis results in the IIP3 plot, the two tone test is performed with an input frequency of 10MHz and 11MHz. I had used an external ideal balun and 50 Ohms termination port both at the inputs and outputs. I still can't seem to get the idea behind this problem, Is there any error in the way I simulate stuffs or else. Hope you could shed some light on it. Thanks in advance

Rgds

 

[Mazz]

p1db op1db

The ideal balun (what is the impedance ratio?) and 50 Ohm source/load works if the Circuit under test is matched both at input and output.
If not you may see big differences plotting voltages (what you have done in the case of Gain) and powers (in case of IP3 and P1dB). From your simulations it seems that it is the case (voltage gain measured from first plot is really different from voltage gain measured in third plot). This is maybe the problem...
Try to plot the single end to single end gain, maybe you'll see something different, or with the same bench try a PSS+PSP (s-parameter analysis, that takes into account power and not voltage), plotting S11,S21, S22 (input match, power gain, output match).
I hope it can help.
Mazz

 

[dsjomo]

equation relating ip1db to op1db

The power gain and voltage gain are equal if and only if the Zin equals Zload. And both IIP3 and 1dB point are measured as power in those graphs.

Further more, the voltage gain is a little bit low for a 1MHz IF downconverter.

 

[Mazz]

oip3 op1db

dsjomo
You're right, but what do you mean?
Mazz

 

[dsjomo]

why oip3?

To Mazz,

From P1dB figure, we can find that the POWER gain is roughly -32dB, but the VOLTAGE gain figure shows that the VOLTAGE gain is 0dB, so the difference between these two values is resulted form the difference of Zin and Zload, or, impedance mismatch.

OIP3(dBm) = IIP3(dBm) + "POWER" Gain(dB) ;
OIP3(dB) = IIP3(dB) + "VOLTAGE" Gain(dB).

The problem hrkhari made is that he uses voltage gain to calculate OIP3(dBm).

But i think, if anyone wants to get a thorough explanation of the problem he or she encountered, it should be better if the schematic is uploaded for us to diagnose.

 

[Mazz]

pin oip3

Fully agree.

Mazz

 

[hrkhari]

iip3 conversion oip3

Hi Mazz/dsjomo :

Thanks for your feedback, I really appreciate it and had learnt much. I had attached the schematic of the complete external test bench circuit and the type of input matching utilized. I had utilized an integrated source follower to drive and output 50 Ohms load (see attachment).I hope you could go through the attachment and feedback on the drawback of this matching scheme and test bench and hopefully you could help out in suggesting an alternative. I had also attached the rectangular plot of the S-Parameter for your constructive comment. Thank You

Rgds

 

[dsjomo]

op1db 75 ohms

(1) S11 and S22 should be matched to 100ohm, at least lower than -10dB. Your S11 is obviously not matched, and S22 is poorly matched.

(2) 10nH is too small to be an RF-choke.

(3) The voltage of output node of the transconductance stage is not well defined, you'll suffer a serious problem of process variation.

 

[Mazz]

what is oip3

As usual agree with dsjomo.
The Gilbert cell in PMOS is just a mistake of the drawing or I'm not understanding something?
If LO is from ic circuit you don't need to be matched on it, simply drive with ideal voltage source in order to be sure to control the wanted voltage swing on it.
I hope it can help.
Mazz

A note: is it clear why we use S parameters and speak about power in RF?
If not, please advice.
If you have a low freq. output, you can avoid to load it with 100 Ohm (equivalent load of the balun), because most probably in real circuit you will sense the out voltage with an high impedance load. To simulate it you can use a VCVS simply to convert the differential voltage to single end, and connect it to a port to properly read the "equivalent" power.

 

[cmosbjt]

oip3 vs op1db

1. Agree with dsjomo: your RF port matching is too bad, you will lost a lot of input power;
2. Since the load of the Gilber cell are just resistors, I think the IF is low, just like what Mazz said: no need to use balun and matching for it as we usually need a voltage output rather than power.
Refer the the following picture:
**broken link removed**
If the input has a perfect matching, then Pin = Psource.
CG=Pout/Psource=Pout/Pin=(Vout²/Zload)/(Vin²/Zin)=(Vout/Vin)²×(Zin/Zload)=VG²×(Zin/Zload);
thus CG(db)=VG(db)+10log(Zin/Zload).
Normally Pin=(1-Γin)×Psource, then CG=VG²×(Zin/Zload)×(1-Γin)
Since your S11 is too bad, Γin is close to 1, that is one of the reasons why there is a large discrepancy between your CG and VG even after the factor of (Zin/Zload) is considered.

BTW, can you check the output buffer to see whether M2 and M4 work properly? Where do you get this schematic?

Note:
CG : Conversion Gain
VG : Voltage Gain