Help me bias and get suitable w/l ratio for a mixer design

hi.. i'm a undergraduate student and now i'm working on designing a downconvert mixer for my project. for the purpose of learning, i'm working on single balanced mixer. i have the circuit. but i really dont no how to bias it and determine suitable w/l ratio. i have no experiance in rf analog ic design. i have read a lot of book like "rf microelectronic" from behzad razavi, "analyisis and design for analog intergrated circuit", and a lot more IEEE paper. but still i can't find how to calculate the w/l ratio using hand calculation. one more thing, i have read the "gilbert cell tutorial" at rfic.co.uk and it use ads software to simulate the design. I only have Tspice from tanner. can Tspice do the job? please help on how to begin couse my lecturer also dont know how to design it. i really have to work on myself.

I don't know if you use diodes or FETs. Unfortenately the theory of designing a good mixer seems to be complicated. One book that maybe can help you is microwave mixers by stephen A Maas. However it seems hard to get design guidelines.
The bias point should be somewhere around the turn-on point of the device, the optimal bias points seems to depend on both LO power and w/l ratios. Soo sweeping the parameters in some circuit simulator is a good idea.
It also seems that a larger area is good for Conversion Gain and compression/IP3
at least for my own simulations of a resistive mixer.
As you can see, I am not a expert in this area, but I hope it can help you a littel bit.
I am also interesting in what your project is?

/Joakim

download the thesis i have uploaded here

i think you will find it useful

https://www.edaboard.com/viewtopic.p...os+cable+tuner

I agree with joakim, it is hard to write down a design guideline for mixer design even for single balence mixer.

For the Gm transistor in the bottom, it is said that the larger overdrive (CMOS), the higher linearity. You can verify that by plotting the VIP3. But higher overdrive means larger tail current for the same device. That may not memt your spec. If you use bipolar, then you need to decide the current density. Ploting the VIP3, you can see a peak follows by a valley, after that the curve increase again (with respect to the overdrive (CMOS) or current density (BJT)). You should not choose the peak because any parameter variation will kill your design. So it is a common practice to choose somewhere on the right side of the valley. But working in this part require more power. Actually I cannot see much help using this VIP3 curve if your design requires low power.

For the switching pair. As joakim said, "the bias point should be somewhere around the turn-on point of the device", but the current of this device (CMOS) is small at that point. In order to have half of tail current, the device should have a relative large W/L, in this case, be careful that your LO may not be able to drive this device. So don't always use ideal LO in your simulation. Try an LO buffer to see whether it has the ability to drive according to your spec.

I also have many questions in designing mixer especially low power one. Correct me if i am wrong.

Thanks for your reply. actually my project is to design a mixer for rf transceiver. As i said before i'm a student. so the perpose of this project is to verify the behavior of mixer. So what i'm planning to do is, i will set my conversion gain and after that i will calculate the w/l ratio to archive that conversion gain. But i dont know how to bias it. I'm using MOSFET in my mixer design. At the end of my project i will have the layout for the mixer. to simulate the layout i must simulate it using Tspice. This project is required for me in order for me to get my degree. thanks for all your opinion i will study more how to design. if anyone out day have any opinion, please reply to me i really appreciate any comment.

hye... i have started on design the mixer. for the porpose of understanding the mixer behavior i have design single balanced mixer first.(using MOS) i have the output but the output is in time domain. i dont know how to plot in freq domain. Iam using spice for simulation. can somebody suggest me a way how to prove my design is true.

this is what i have understand about mixer. please correct me if i'm wrong

the above part (LO) is the switching part. just one transistor is switch on in the
same time. so to prove that my design is correct i've do the transient analysis
in spice and plot Id of both transistor. yes it show that when one transistor has
max current the other transistor has 0 current [about 10^-8]. is my analysis
correct?

my LO is 2.39Ghz and my RF is 2.4Ghz. i'm using 0.18μ from mosis. For high frequncy mixer, what should i consider in designing a mixer for this type of frequency?

i'm still confuse about biasing a circuit with dc voltage. for the single balanced mixer i have bias all the NMOS gate with some dc value to make the NMOS transistor operate in saturation region. is this correct?

i have tried the command .four to do fourier analysis to my IF output. the reason i do this couse i want to know the amplitud of my signal in certain frequncy. but i got the amplitude is so high [10^8]. why?

please help me. i know somebody my think my question is some basic stuff. but i'm still learning. i will ask somebody if i dont understand about something. thank you in advanced for anyone who willing to help me.

p/s: the next step i will proceed to my actuall project to design the Gilbert cell mixer but only if have enough knowledge about the single balanced mixer first. thanks to Ashtaroth. i've read the tesis. it really help me alot in understanding mixer bahavior

There are two good articles here. Search for Steve Long in the Design Seminar section.


http://eesof.tm.agilent.com/library....ished_articles

Hi evilguy, what is the tool you do the simulation? SpectreRF? HSpiceRF?

sadly i only have T-Spice. Can i simulate my design using this software? i just read the thesis that i've mantion. he used Spectre tu simulate the design. i dont have the software. So any suggestion for me to simulate it?

is there any free software to for the purpos of rf simulation? i'm cant affort to buy software like cadence spectre like i have mantion before my university only have Tspice. from what i have read, spice is not capable to do the test for linearity. if i not mistaken the "two tone test" correct me if i'm wrong.

i've design my gilbert cell mixer. but i've some problem in determine the conversion gain. as i said i only have t-spice in my uni. so when i read the thesis that Ashtaroth has suggested to me, the author determine the conversion gain using the fft. that mead he do the fft for the output signal and input signal after he do the transient analysis.

i have done exactly the same thing. my LO is 2.39ghz and my rf is 2.4ghz. so i should get the IF = 10mhz. my rf amplitude is 0.2V and my LO amplitude is 0.4V. after doing the fft, my IF value at 10mhz is 0.15V. it mean that my RF signal is not amplified but the signal seam to lost. why this is happening? i can't determine what is wrong. can somebody help me? fell free to ask me anything if you not understand my explanation. i'm really bad in explaning things

Evilguy, are you sure the deivce models are you using in the design are good at the frequencies you are working? I suggest you an approach in the frequency domain this is better to understand the circuit you are working and the books written from Steve Maas are very good for this purpose and for learning the mixer world.

I hope this can help.

Bye

hello,
I'm also interested in developing mixer and local oscillator. If you want to plot in T-Spice your input/output signal in the frequency domain, first plot in the time domain then select your plot and choose in menu Chart -> FFT and you will get the frequency components. So, if you do that in the output signal, you will be able to see the additon and subtraction of both frequencies (RF and LO). Hope it helps you.

thanks for replying. i have discover the fft function in T-spice recently. mine is t-spice version 11. previous version dont have that fuction. as for now i can calculate the conversion gain of my mixer using the fft analyisis.

can someone here suggest me a method on how to calculate linearity in my mixer design using spice? i think it is imposible for me to do the two tone test for my mixer using spice.

thanks STOIKOV. hopefully we can exchange knowledge about designing mixer at later time

Evilguy, I′m simulating in T-Spice a Gilbert cell all MOS in saturation region and seems its output frequency spectrum is ok. It's not clear for me the concept IP3, IIP3 and I guess it is an important feature of mixers, you have information about it. Regards.

my gilbert cell design is also ok. thanks to Ashtaroth. The thesis really help me. i'm using 0.18μ and the rf freq is 2.4. i think my mixer is operating fine. i just read somewhere that gilbert cell mixer need at least 2.7V vdd in order to make all the transistor in saturation. may i know what is your vdd voltage?

ok the concept of 1db compresion point and iip3 is it determine how linear our mixer is. to obtain the iip3 we need to do the two tone test to our mixer. linearity is also the important concept in designing mixer beside conversion gain. i cant eloborate more because i still dont know much about it.

today i've presented my mixer design to my lecturer. he seem satisfied. beside transient analysis, fft, and dc analysis, what kind of analysis should we do to our mixer design? is is important to do ac analysis?

I'm working with 5V and 0.5μ (Gilbert topology, see picture). Already calculated geometries and DC bias, and I'd like to know :

- how to calculate conversion gain of mixer ?
- if fabricate chip, how to apply RF signal in differential mode (RF port) ? need a balun or a RF generator can provide differential signal ?

Thanks and regards,

ok i would like to share something. to calculate conversion gain, let say voltage gain, in tspice, u can plot the output spectrum using fft. plot in in magnitude. the fft graph will show the vrms (output) versus frequency. so at the IF frequency determine the magnitude of the spectrum. voltage conversion gain is:

C.G = vrms(out) ÷ vrms (rf)

C.G dB = 20 log C.G

i still trying to do simulation to calculate C.G using spice and plot the gain using W-edit in tannner. not succesfull yet.

for the second question. i cant answer that. sorry. hopefully someone here can answer it.