| Author |
Message |
yixiusky
Joined: 08 May 2008
Posts: 11
|
 13 May 2008 11:46 ask for help about gm-c filter ^^ |
|
|
|
hello everyone, i am really want to ask for help about gm-c filter, thank you very much! 
here are my questions:
1. what's the requirement of OTA in gm-c filter? phase margin and GBW? some senior told me, for gm-c filter, the phase margin of OTA must be 90 degree. Is that true ?
2. I am really not clear about the process to design gm-c filter. I read Martin's book, but i still feel very confuse. I would like to know is there any book teach the step of designing gm-c filter?
3. For active RC filter, the filter solution software can calculate the coefficient values , but for gm-c there is no software to help. 
sorry to ask so many questions, i really appreciate your help. thank you very much ^^
|
|
| Back to top |
|
 |
ohoh
Joined: 11 Mar 2008
Posts: 4
|
| 15 May 2008 10:28 Re: ask for help about gm-c filter ^^ |
|
|
|
I think you can read this book < continuous_time active filter design>. it can show you how to design a gm-c filter & things you should consider during your design.
you can also get some papers from ieee ,they are very helpful
|
|
| Back to top |
|
|
LvW
Joined: 07 May 2008
Posts: 449
Helped: 84
Location: Germany
|
| 15 May 2008 15:22 Re: ask for help about gm-c filter ^^ |
|
|
|
For your research it may be helpful to realize that gm-C-filters are identical to OTA-C-filters, since gm is the transconductance of an OTA.
Here is an introductory article with basics to simple gm-C-circuits:
R.L. Geiger and E. Sanchez-Sinenzio: Active filter design using operational transconductance amplifier: A tutorial.
I am sure you can download it from the net. Look at Sanchez-Sinenzios home page.
LvW
|
|
| Back to top |
|
|
sutapanaki
Joined: 02 Nov 2001
Posts: 473
Helped: 14
|
| 15 May 2008 19:34 Re: ask for help about gm-c filter ^^ |
|
|
|
There are many things to the design of gm-c filters. You don't say for what frequency is your design. I'd assume it is high frequency since gm-c filters are usually used for high frequency applications. Also, what kind of filter are you targeting - ladder or cascade of biquads?
There are at lest two things one should do in the design, according to me. First, create a model using ideal gm components. If you use ladder filter, start from a passive prototype to get the C and L for the frequency you want. Convert it to Gm-C architecture. Then you will know what are the Gm values you need and also the capacitances. Then you have to design your transconductor. Things to look there are speed, linearity, noise, tunability. And to answer one of your questions, yes, the 90 deg. phase for the integrator is important since if it is not 90deg, it will alter the Q factor of your filter - either increase or decrease, depending on the sign of the excess phase. To get those 90 deg you'll need pretty high voltage gain of the transconductor and a single pole roll-off. Not always easy to achieve, especially for high frequency designs and the latest technologies.
May be a good source of information are the first 10 lectures of EE247 form Berkeley which are available online.
Also search the forum for Nauta's book "Analog CMOS filters for very high frequencies"
Last edited by sutapanaki on 15 May 2008 22:26; edited 2 times in total |
|
| Back to top |
|
|
LvW
Joined: 07 May 2008
Posts: 449
Helped: 84
Location: Germany
|
| 15 May 2008 21:56 Re: ask for help about gm-c filter ^^ |
|
|
|
Just for clarification, since there are some confusions :
1.) PHASE MARGIN: An OTA itself cannot have any phase margin. A phase margin is a stability measure which is defined for an open loop system with the intent to close it later. However, normally OTAs are used without feedback.
This has nothing to do with the 90 deg phase shift which belongs to an integrtator stage.
2.) Quote: To get those 90 deg you'll need pretty high voltage gain of the transconductor ad a single pole roll-off.
Voltage gain of a transconductor stage ??? This causes confusion.
3.) However, as said in the last reply, OTAs are used normally for high frequencies and when their special features are needed (tunability via gm).
LvW
|
|
| Back to top |
|
|
sutapanaki
Joined: 02 Nov 2001
Posts: 473
Helped: 14
|
| 15 May 2008 22:23 Re: ask for help about gm-c filter ^^ |
|
|
|
| LvW wrote: |
Just for clarification, since there are some confusions :
2.) Quote: To get those 90 deg you'll need pretty high voltage gain of the transconductor ad a single pole roll-off.
Voltage gain of a transconductor stage ??? This causes confusion.
|
Yes, this indeed causes confusion. However, thinking about the simplest possible filter - 1st order integrator, ideally having a -20db/dec roll-off starting from 0Hz, one can realize it with a single Gm-cell loaded by a capacitor. And then you'll have a voltage gain. In real integrators the roll-off doesn't start from 0Hz, because of the finite output resistance of the Gm cell and in this case you'll have a dc voltage gain and a dominant pole. Preferably the phase at 0db cross-over should be -90deg, but because of the finite dc gain it is not and this causes problems.
|
|
| Back to top |
|
|
LvW
Joined: 07 May 2008
Posts: 449
Helped: 84
Location: Germany
|
| 15 May 2008 22:36 Re: ask for help about gm-c filter ^^ |
|
|
|
Mmmmhhh, I see what you mean......
However, to define something like "voltage gain" for an OTA without any load (capacitor at 0 Hz) seems - at least for my opinion - questionable.
But anyway.....it is not so important.
LvW
|
|
| Back to top |
|
|
sutapanaki
Joined: 02 Nov 2001
Posts: 473
Helped: 14
|
| 15 May 2008 22:46 Re: ask for help about gm-c filter ^^ |
|
|
|
| Ok, forget about the capacitor. Suppose you have a fully differential OTA and a CMFB to fix the output common mode. You can not expect to have an infinite voltage gain out of it even for dc, because the OTA is loaded by the rds of the transistors. In case you are only interested by the Gm of the cell you'll have to short the outputs and see what is the deltaI caused by deltaV at the input. Because you're shorting the outputs now, it is useless to talk about voltage gain.
|
|
| Back to top |
|
|
yixiusky
Joined: 08 May 2008
Posts: 11
|
| 19 May 2008 4:16 Re: ask for help about gm-c filter ^^ |
|
|
|
Thank you very much for all of your answers ^^ Have a nice day.
I would like to add more information
My targeting is the RF tracking filter in Ditital TV Tuner.
The roles of this filter are (1) harmonic rejection. (2) If possible, unwanted signal rejection.
The working frequency of tracking filter is from 48MHz-860MHz.
Considering the working frequency and fine tuning, i decide to use Gm-C type
filter. The RF tracking filter is sharp, narrowband, tunable fitler.
(A) For ladder case :
When i simulate with "filter solution software". I can get the gm value, but I feel there are many inductors, if I replace them with gyrator, the circuit will become large. Do you have this kind of problem when u design Gm-C filter?
(B) For biquads case :
When I design active-RC filter, I can get the value of R,C of each biquads from software , but for Gm-C, I can not get that value. So i don't know how to get the value when design the biquads gm-c filter.
By the way, I would like to know "create a model using ideal gm components. "
what's meaning of ideal gm components?
I am sorry to ask so many questions ^^
Thank you very much Have a nice day ^^
|
|
| Back to top |
|
|
LvW
Joined: 07 May 2008
Posts: 449
Helped: 84
Location: Germany
|
| 19 May 2008 8:08 Re: ask for help about gm-c filter ^^ |
|
|
|
(A) For ladder case :
When i simulate with "filter solution software". I can get the gm value, but I feel there are many inductors, if I replace them with gyrator, the circuit will become large. Do you have this kind of problem when u design Gm-C filter?
The number of inductors depends on the selected low pass approximation resp. the order of the filter. If you prefer active realization of the ladder topologie you cannot reduce the nunmber of active elements - opamp or OTA.
(B) For biquads case :
When I design active-RC filter, I can get the value of R,C of each biquads from software , but for Gm-C, I can not get that value. So i don't know how to get the value when design the biquads gm-c filter.
Thatīs the disadvantage of filter software; only a limited number of topologies is available. But itīs not too complicated to calculate by yourself. Use Kirchhofs law and the OTA-equation Vout=Vin*gm*Rload.
By the way, I would like to know "create a model using ideal gm components. "
what's meaning of ideal gm components?
Thatīs nothing else than an ideal OTA with a transconductance gm and infinite input/output impedances. But realize: gm is a finite value, not gm=∞ like opamps.
|
|
| Back to top |
|
|
sutapanaki
Joined: 02 Nov 2001
Posts: 473
Helped: 14
|
| 19 May 2008 19:20 Re: ask for help about gm-c filter ^^ |
|
|
|
| Just to add ot the last post. Ideal gm means voltage controlled current source. Should be present in any simulator. Once you do your filter with ideal components, you can start adding to the ideal gm a finite output impedance, additional poles and zeroes and see how all these affect your response.
|
|
| Back to top |
|
|
yschuang
Joined: 01 Feb 2004
Posts: 139
Helped: 10
|
| 20 May 2008 3:43 Re: ask for help about gm-c filter ^^ |
|
|
|
Scroll down to read spice file.
http://www.ecircuitcenter.com/Circuits/smps_buck_vm1/smps_buck_vm1.htm
There is an ideal g_err element with output resistance "rgain" in paralle with capacitance "cgain".
|
|
| Back to top |
|
|
