Designer’s Guide Community

Mayank

Community Fellow

Offline

Posts: 334

Gm-C / Active RC / S-C
Oct 15^th, 2009, 2:10am

Hi all,
I have done quite an extensive reading on design of analog fitlers. But anyways, before starting the design, i wanted to take a counsel...
Application : Analog Base-Band for WLAN RF
Intended Use : For Digital Video Broadcasting
Filter Cut-off : 80 MHz
Major Concern : Power Savvy Design.
Others Concerns : Low-Noise, Area Savvy.

Which Topology should i go for -- Active RC/ Gm-C / S-C ??

regards,
Mayank.

Back to top

IP Logged

aaron_do

Senior Fellow

Offline

Posts: 1398

Re: Gm-C / Active RC / S-C
Reply #1 - Oct 15^th, 2009, 4:57am

Hi,

just to get the discussion going, I think switched-cap filtering is out since it requires anti-aliasing and a high clk frequency.

With that kind of cutoff frequency, you would need a pretty high UGB for an active-RC filter. For example, for a loop gain of 20 dB at 80 MHz, you would need a UGB of 800 MHz. You also need to consider the out-of-band linearity and the highest interferer that you want to filter.

For gm-C your linearity will not be as good as the active-RC, but you should be able to get low power consumption. The linearity will be most important for the first stage of course...

Feel free to correct me if i'm wrong. BTW what topology did you think was the most suitable?

cheers,
Aaron

Back to top

there is no energy in matter other than that received from the environment - Nikola Tesla

IP Logged

buddypoor Community Fellow Offline Posts: 529 Bremen, Germany	Re: Gm-C / Active RC / S-C Reply #2 - Oct 15^th, 2009, 5:00am Hello, Mayank ! I suppose you need a LOWPASS, right ? For my opinion, RC-active as well as S/C are not very well suited to your requirements (80 MHz). As a result , only the gm-C techniques remains as an alternative, which in addition is suited for IC technology.
Back to top	LvW (buddypoor: In memory of the great late Buddy Rich) IP Logged

Mayank

Community Fellow

Offline

Posts: 334

Re: Gm-C / Active RC / S-C
Reply #3 - Oct 15^th, 2009, 5:16am

Hi all,
I am sorry i forgot to mention -- I am talking about a Low-Pass Filter.
@ aaron_do : I already did the math for Active-RC filter. But the main fight, as you said and as i experienced in 65nm, was high BW Product.
High Speed Clock for S-C is not a problem, I can get it from RF LC-VCO---Did some math for it also,but not the whole architecture.
My Choice :--- I want to go for Gm-C coz of Power Optimization and (OTA vs Opamp advantage) BUT am a bit skeptical how it will turn out in the end in terms of linearity and calibration for Cut-Off Frequency Tuning. I was currently studying some papers on linearized Gm Cells and high speed Gm cells (such as Nauta Cell).
How should i proceed with the Architecture for Gm-C ?? Any further reading material on Higher Order Gm-C Filter Design would be helpful.

@ Buddy : Yeah, i am sorry not to mention it's a Low-Pass Filter.

Back to top

IP Logged

buddypoor

Community Fellow

Offline

Posts: 529
Bremen, Germany

Re: Gm-C / Active RC / S-C
Reply #4 - Oct 15^th, 2009, 6:27am

Mayank wrote on Oct 15^th, 2009, 5:16am:

...........................
How should i proceed with the Architecture for Gm-C ?? Any further reading material on Higher Order Gm-C Filter Design would be helpful.
.........................

Mayank, it�s a rather challenging task to choose an architecture which best serves all needs and requirements. Mainly, the answer depends on the required filter order which directly is connected with complexity.
For orders of four or higher I would suggest a topology which is based on a passive ladder structure (leap frog structure). This topology has the advantage of being least sensible with respect to passive element tolerances.

Back to top

LvW (buddypoor: In memory of the great late Buddy Rich)

IP Logged

Mayank

Community Fellow

Offline

Posts: 334

Re: Gm-C / Active RC / S-C
Reply #5 - Oct 15^th, 2009, 8:26am

Hi,
Thanx Berti for your advice...I was thinking on the same lines...Actually, i did the math for Active RC implementation of 4th Order Chebyshev LC Ladder. I think most Probably i will be needing a 4th Order Cheby OR else a 5th Order ButterWorth Filter...

Question : When i designed Active RC Impl. of 4th Order Cheby LC Ladder, I designed a fully optimized (for noise/sensitivity) Ladder, where each L & C value were different, resulting in different R & C values.
Can we design a chebyshev LC Ladder where all L & C values are identical respectively ??

--Mayank.

Back to top

IP Logged

Mayank

Community Fellow

Offline

Posts: 334

Re: Gm-C / Active RC / S-C
Reply #6 - Oct 15^th, 2009, 8:30am

Hi,

1 more question : The Primary Resonator Block PRB Approach used in FLF(Follow-the-Leader Feedback) Topology is useful because i want tuning capability for Cut-Off Frequency....
Can i do something of that sort in LeapFrog ?? I also have to keep in mind the calibration circuitry required for cut-off tuning. That's the reason i wanted equal L & C ladders....Pls help.

regards,
Mayank.

Back to top

IP Logged

buddypoor

Community Fellow

Offline

Posts: 529
Bremen, Germany

Re: Gm-C / Active RC / S-C
Reply #7 - Oct 16^th, 2009, 12:40am

Mayank wrote on Oct 15^th, 2009, 8:30am:

Hi,

1 more question : The Primary Resonator Block PRB Approach used in FLF(Follow-the-Leader Feedback) Topology is useful because i want tuning capability for Cut-Off Frequency....
Can i do something of that sort in LeapFrog ?? I also have to keep in mind the calibration circuitry required for cut-off tuning. That's the reason i wanted equal L & C ladders....Pls help.

regards,
Mayank.

1.) No, I don�t think that the leapfrog architecture allows easy tuning of the cut-off. In this respect , PRB is more flexible.
But I would say that for a lowpass of (only) 4th order a classical cascade approach would also be appropriate. Don�t you think so ?
And it requires less amplifiers.

2.) To your former question:
[i]Question :
...................
Can we design a chebyshev LC Ladder where all L & C values are identical respectively ??[/i]
There are some filter programs available (partly for free via internet) which can be used to tune LC laders and to modify parts values.

Regards
LvW

Back to top

LvW (buddypoor: In memory of the great late Buddy Rich)

IP Logged

loose-electron

Senior Fellow

Offline

Best Design Tool =
Capable Designers

Posts: 1638
San Diego California

Re: Gm-C / Active RC / S-C
Reply #8 - Oct 16^th, 2009, 5:39pm

IMHO at this bandwidth its going to be gmC, that said, the problem really has 2 big parts - Splane definition of the filter and transistor level definition of the gmC sytem itself.

I have done the Splane in behaviorals in Cadence (Vlog AMS), and Matlab (Simulink), and Excel (!! using phasor methods)

The you got to define the gm stage, and get the linearity right, and then define the filter stages themselves.

I prefer cascaded bi-quads, easier to understand whats going on. Uses 4 gm stages.

Dont forget to define how you are going to tune the thing. Adjusting either Cap size or gm value. If the linearity is an issue, I prefer switching the capacitor values, (digital switching) and running a tuning adjust cycle fairly often. (thermal variance) On that, I like to adjust the -6dB point of the filter, because it is easy tune with an input signal, and compare it in amplitude with a Vsignal/2 version.

What else? Hmmmm.. may want to go one order higher on the filter (5 pole instead of 3 for example) because the rejection characteristics may not be quite as good as the ideal math model.

My thoughts and "learning experiences" as they say.
Cheesy

jerry

Back to top

Jerry Twomey
www.effectiveelectrons.com
Read My Electronic Design Column Here
Contract IC-PCB-System Design - Analog, Mixed Signal, RF & Medical

IP Logged

Mayank

Community Fellow

Offline

Posts: 334

Re: Gm-C / Active RC / S-C
Reply #9 - Oct 21^st, 2009, 7:58am

Hi folks,
I was on a vacation...Festival of Diwali in India...Now am Back..Saw your replies..
@ buddypoor : Thanx for ur reply...But in Biquad Cascading...each RLC network is separated by active gain stages...resulting in larger uncorrelated variation in poles of each stage with component variation...That's The reason i dnt wanna use it...
BottomLine -- Linearity can be best attained in Ladder Simulated structures--LeapFrog serves good..So wanna go with Leapfrog...You are right...Tuning in LeapFrog is tough...but any ideas on how on efficient calibration in Ladder structures ??

@ Jerry : I like your style, Jerry...Gotta know you better ....

1. So Gm-C is Final...But i saw you quoting on some other topic that high o/p resistance of Gm Cells increase noise of the cell...Now that seriously scared me Sir[u seem quite senior].... :( My Noise Specs are tight too...esp on the RX side...

2. Quote:

(thermal variance) On that, I like to adjust the -6dB point of the filter, because it is easy tune with an input signal, and compare it in amplitude with a Vsignal/2 version.

Could you please elaborate more on it ??? As to How do you do that...

3. In accordance with my Specs, I arrived at either a 4th order Cheby OR 5th order ButterWorth...Do i need to go for even higher orders considering the non-ideality effects as you indicated ??

Back to top

IP Logged

loose-electron

Senior Fellow

Offline

Best Design Tool =
Capable Designers

Posts: 1638
San Diego California

Re: Gm-C / Active RC / S-C
Reply #10 - Oct 21^st, 2009, 9:39am

"seemed quite senior" - LOL! Grumpy old dude eh? Well Here:
http://www.designers-guide.org/experts.html
Bottom of the page, and that's a pretty recent picture (last summer)

Noise is an issue, no doubt, and the only way you are going to know is to put it together and model it, and run some noise analysis.

As for the tuning?
1. Create a sinusoid wave (heavily BW limit a square wave, it doesnt have to be perfect.) The sinusoid gets created at a frequency that the filter has as its properly tuned -6dB point. (Vsig)

2. Put Vsig into the filter to get an output (Vsigfilt)

3. Take Vsig and voltage divide it with some resistors. (Vsig/2)

4. Compare the amplitude of Vsigfilt and Vsig/2 (you need some peak detection and averaging circuits to create DC equivalents.)

5. Adjust your "tuning knob" until Vsigfilt = Vsig/2 and you have tuned your filter.

Back to top

Jerry Twomey
www.effectiveelectrons.com
Read My Electronic Design Column Here
Contract IC-PCB-System Design - Analog, Mixed Signal, RF & Medical

IP Logged

Mayank

Community Fellow

Offline

Posts: 334

Re: Gm-C / Active RC / S-C
Reply #11 - Oct 22^nd, 2009, 8:01am

Hi Jerry,
Naah man !! Tongue

dint mean an old dude...Just respect...saw the pic...looks pretty cool man.sailboating..shall i mail u sometime if thats nt a problem ?? its good to know great designers...

As for Tuning, Chebyshev shows 1 dB PassBand Ripple. So isnt the 1dB point the filter cut-off Point ?? I thought i would tune the 1dB point for it..i wanted to know a bit more about noise analysis of filter...Wanted to arrive at a noise spec for Gm-Cell for designing the OTA..How can i do that ??

thanx,
Mayank.

Back to top

IP Logged

loose-electron Senior Fellow Offline Best Design Tool = Capable Designers Posts: 1638 San Diego California	Re: Gm-C / Active RC / S-C Reply #12 - Oct 22^nd, 2009, 8:54am If you tune the -6dB point its a lot simpler. Everything shifts together, and doing V/2 is easy. Yeah, I take email, no big deal. Busy as h@ll right now, and coming to this forum is my coffee break entertainment. jerry (at) effectiveelectrons(dot)com
Back to top	Jerry Twomey www.effectiveelectrons.com Read My Electronic Design Column Here Contract IC-PCB-System Design - Analog, Mixed Signal, RF & Medical IP Logged

duncandu

New Member

Offline

Posts: 6

Re: Gm-C / Active RC / S-C
Reply #13 - Oct 24^th, 2009, 9:48am

I don't think SC is suitable here, because you must put another high-frequency anti-aliasing filter prior to it, also with the problems of clock feedthrough and high-speed clock generation.
It seems Gm-C would be more suitable for ~80MHz application rather than active-RC, since it is open-loop system which has high-speed potential. Of course, as we all know, it is tricky to solve the linearity problem and parasitic load.
However, it is worthy mentioning you can really design an active-RC filter for such a high frequency with a good compromise between power consumption, speed and linearity, as long as you know how to cleverly design the operational amplifier in it.
If you are an experienced designer who knows the details of operational amplifiers, just choose active-RC.

Back to top

IP Logged