The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> RF Design >> current-mode passive mixer problem
https://designers-guide.org/forum/YaBB.pl?num=1275897849

Message started by kibby on Jun 7^th, 2010, 1:04am

Title: current-mode passive mixer problem
Post by kibby on Jun 7^th, 2010, 1:04am

I am designing a current-mode passive mixer which consists of gm,mixer core and tia stages. The mixer core is ac coupled to the first stage. I wonder about the input offset voltage of tia, is it feedback by the resistor or other ways?
I tried to use the feedback resister. The input and output common mode voltage is vcc/2 at most times. By as I tried to enlarge the ratio of input and load mosfets to mitigate flicker noise. The cm voltage becames vcc. The input mosfets are pmos. Does it means that a start-up circuit is needed?
Thank you for your help :)

Title: Re: current-mode passive mixer problem
Post by Mayank on Jun 8^th, 2010, 12:24am

Quote:

I wonder about the input offset voltage of tia, is it feedback by the resistor or other ways?

Yes, generally it's a feedback by a resistor..Unless you wanna plunge into complexity of SC circuits.

Quote:

By as I tried to enlarge the ratio of input and load mosfets to mitigate flicker noise. The cm voltage becames vcc.

Seems like a Loop Stability Problem to me.
Your loop(check both CM loop & DM loop) might be going unstable because of the excess gain introduced by increasing MOS W/Ls & because of Shifted Poles due to larger parasitics from increased sizes.

--
Mayank

Title: Re: current-mode passive mixer problem
Post by kibby on Jun 8^th, 2010, 5:54am

Thank you, Mayank~
The PM of both DM(open loop) and CM loop are 90° and 60°, I think it`s enough. But the problem still exists. I don`t know how does spectre do dc analysis, but I don`t think the problem relates to ac response.
Since there is no dc voltage or current after mixer core. If the input node of tia begins with a random voltage like vcc. Then it seems that the whole circuit can`t work. Why not just bias the input node (vcc/2) of tia with a large resistante?

Title: Re: current-mode passive mixer problem
Post by rfcooltools.com on Jun 8^th, 2010, 3:53pm

Sounds like your circuit has a common loop has more than one solution, the desired and either a transient or permanent trap state when the input and output cmode is vcc, if its a transient in nature you may not be seeing it resolve itself due to the length of your simulation time. I would suggest finding the root cause because these issues are usually a clue to other potential problems. A lot of times this can be solved by making the cmode current transistor to provide a maximum of 50% of the total bias current. If you are short on time take your own suggestion and bias it to vcc/2.

http://rfcooltools.com

Title: Re: current-mode passive mixer problem
Post by aaron_do on Jun 9^th, 2010, 9:36pm

Hi,

I think you were right in that your circuit is not starting up. If you show us your schematic it would help but I can understand if you don't want to. You can ensure start-up by either adding some start-up transistors, or by making sure your common-mode feedback loop allows the circuit to start-up.

Out of curiousity, why did you AC couple the mixer core to the op-amp? Are you trying to get a high-pass function? If so, the corner frequency of your HPF depends on the average resistance of the mixer core (for a double balanced mixer core). The smaller the on-resistance, the larger the capacitors you need.

Also, if your mixer core is a double-balanced type (as is normal), you should look out for your TIA loop gain which will depend on the average resistance of the mixer core. Your loop gain will affect the TIA linearity.

cheers,
Aaron

Title: Re: current-mode passive mixer problem
Post by kibby on Jun 11^th, 2010, 5:58am

Thank you Aaron!
//Out of curiousity, why did you AC couple the mixer core to the op-amp?
May be you misunderstand my meaning. Here is the system and block schematic.
Now I want to bias the input node of tia with a large resitance(bias at vcc/2) so that the start-up problem can be solved. As far as I know, the large resistance will not add additional noise, maybe a little input offset voltage. But if I use a start-up circuit, it will also add noise and additional complexity. Could you give some suggestions?
Thank you for your help~
kibby

Title: Re: current-mode passive mixer problem
Post by Mayank on Jun 11^th, 2010, 6:17am

Quote:

Now I want to bias the input node of tia with a large resitance(bias at vcc/2) so that the start-up problem can be solved. As far as I know, the large resistance will not add additional noise, maybe a little input offset voltage. But if I use a start-up circuit, it will also add noise and additional complexity. Could you give some suggestions?

AC Couple the Mixer's Output to the input of TIA through a decoupling cap.
DC Bias the input at TIA by dropping mixer o/p current into a resistor to VDD/GND.
This is what is done generally. Rest of your circuit looks fine to me.

Title: Re: current-mode passive mixer problem
Post by rfcooltools.com on Jun 11^th, 2010, 12:15pm

if IFin is high Node A can be undefined so you will need to define A by putting two current sources one in parallel with M3 and one in parallel with M4 which are always on. This will solve your start up issue.

http://rfcooltools.com

Title: Re: current-mode passive mixer problem
Post by kibby on Jun 11^th, 2010, 8:23pm

Happy to see you Mayank~ The receiver is designed to directly down-convert the rf signal. As the IF frequency is zero, the mixer core cannot ac couple to tia or the cap may be too large.
Thank you rfcooltools.com, the suggestions are good. However, when I use the tool initial condition in spectre simulation to do transient simulation, I find that the ota with resistance feedback can start if the initial voltage at input node is 0 to near vcc. Though the dc simulation remains wrong...

Title: Re: current-mode passive mixer problem
Post by RFICDUDE on Jun 12^th, 2010, 6:32am

Do you have a circuit on the input of the TIA (or at the input to the mixer) to set the CM voltage at the input of the TIA to be the same as the output CM voltage?

The CM voltage can be connected to the TIA or mixer input with large resistors since you are just trying to establish the DC CM voltage.

All this will do is to allow the input CM to be the same as the output CM such that the output doesn't need to supply any current through the feedback resistor to establish a stable input CM level.

The circuit may not be necessary, but perhaps it is related to your issue?

Title: Re: current-mode passive mixer problem
Post by kibby on Jun 12^th, 2010, 7:12pm

Thank you RFICDUDE!
The CM of input of TIA should be feedback by the resistance so it is the same as output CM voltage.I use a large resistance to bias the input of TIA just to avoid start-up problem.
kibby

Title: Re: current-mode passive mixer problem
Post by RFICDUDE on Jun 13^th, 2010, 4:35am

Well then you should take rfcooltools advice and see if the CM of the OTA is stuck in a metastable state.

If node "IF" starts out at Vdd then the CMFB circuit should make Vfb fairly high. If Vfb is high and "IF" at Vdd and the volt"A" is low then the OTA is stuck because the input PMOS pair is off.

I think, normally, there would be some leakage currents that would help charge node A up, but maybe not since the NMOS loads are sinking any leakage currents that maybe present.

You might try what rfcooltools suggested (a circuit to pull node A up) or lower your CM reference voltage such that the input PMOS pair and current source will be in saturation.

You should analyze the input CM range of the diffpair and current source and then determine if changes need to be made to align the input and output CM requirements.

Analog design is 10% inspiration and 90% bias robustness.

Title: Re: current-mode passive mixer problem
Post by aaron_do on Jun 13^th, 2010, 5:53am

Hi Kibby,

looks like you already have plenty of good suggestions. Anyway from your schematic there is clearly a start-up issue. I think a good option would simply be to add a couple of transistors in the right places to start-up the op-amp. They can safely turn off when the op-amp is on and may not affect the operation much. Alternatively, you could change your CMFB arrangement so that the op-amp is able to start up.

If you add biasing resistors, you may run into a couple of problems. The biasing resistors would have to be small enough to ensure the op-amp is able to start, and large enough not to add noise to the circuit. If you can live with the small capacitance they add, start-up transistors are a better idea IMO.

BTW that is an interesting arrangement you have. I guess your loop gain is now affected by the transconductor's output impedance. Looks like the loop gain will be difficult to calculate...

cheers,
Aaron