|
The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> Mixed-Signal Design >> The phase error of Integrator in Gm-c filter https://designers-guide.org/forum/YaBB.pl?num=1216301872 Message started by yixiusky on Jul 17th, 2008, 6:37am |
|
Title: The phase error of Integrator in Gm-c filter Post by yixiusky on Jul 17th, 2008, 6:37am Hello, I would like to ask one question ^^ in Gm-c filter design, we must make sure the phase margin of intergrator is equal 90 degree. if phase is lower than 90, the Q factor will reduce. If phase is larger than 90. the Q will increase. I wonder the phase error only can change Q factor? Is there any other disadvantage of phase error ? thank you very much ^^ |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by buddypoor on Jul 17th, 2008, 7:43am As far as I understand your question you speak about the Q factor of a filter stage containing among other circuitry an integrator, correct ? (it would be helpful to see the whole filter circuit). It is very important that each stage - and of course also the integrator - exhibits the specified phase characteristic as the whole circuit always is more sensitive to phase errors than to gain errors. But, of course, phase errors influence not only the Q factor which is responsible for the shape of the transfer curve, but also the pole resp. corner frequencies. Does this answer satisfy you ? |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by yixiusky on Jul 17th, 2008, 7:46pm thank you very much ! I would like to discuss more ^^ I attach the filter and formular ^^ I am designing 2nd-order gm-c bandpass filter now. I want to design the center frequency of bandpass filter is 860MHz, and the bandwidth is 20MHz. According to the transfer function, center frequency, Q factor value, I have some questions: (1) after calculation, i set the gm3,gm4,c1,c2 to get the center frequency 860MHz, but it is different from the theroy. Also. if we want to change the center frequency, which value should we change? Gm3,Gm4 or C1, C2. and when I change anyone, the phase of integrator will be change, not 90 degree anymore :( (2) use B=Gm2/C1, i set the bandwidth of the filter. But it is also not exactly same. I guess because of parasitic capacitance, is that true? (3) From the formular, after we choose the very high value Gm3. Gm4, and very small value Gm2, we can get very high frequency and very narrow bandwidth. But in practical, when I desing, it is not ture :'( Sorry to ask so many questions I am designing a RF tracking filter now. So i need to tune the center frequency from 50MHz to 860MHz, and keep narrow bandwidth about 20MHz. when i tune the circuit, i can not make sure the specification. Thank you very much ^^ have a nice day ^^ |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by buddypoor on Jul 18th, 2008, 3:57am yixiusky wrote on Jul 17th, 2008, 7:46pm:
Well, a lot of questions - not easy to answer. Letīs try it: 1.) That there is a difference between calculated and measured/simulated values is quite normal. It is important how big the difference is. Do you know the frequency limits of the OTA ? And its output impedance ? No OTA behaves like an ideal current source. For w tuning you should NOT use Gm2 or C1 as the BW is required to be constant. More than that, a good integrator does NOT change its phase by tuning the frequency. 2.) Of course, a lot of parasitics are always present. Not only capacitances; again check the OTA output impedance. 3.) What means "in practical" ? Simulation or measurement ? Of course, the calculated frequency goes high when Gm2 becomes small, but only under the assumption that the OTA behaves ideal. Summary: At first, verify the OTA parameters (bandwidth, output charateristics). And when you communicate results: Be more exact; just to say "there is a difference" or "specification not met" is not enough. Good luck and regards Buddypoor |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by yixiusky on Jul 20th, 2008, 8:46pm Buddypoor, thank you very much for your valuable advice, I am really appreciate it ^^ I follow ur advice and check my OTA 1) the OTA parameters (bandwidth=2GHz, output charateristics=2K) 2)about the bandwidth: When I tuning the center frequency, I tune Gm3, Gm4. I didn't use Gm2 and C1 as the BW is required to be constant. But the bandwidth still change :( I guess because of the paracitic capacitance, they are difficult to control, isn't it? So it is very difficult to me to keep the constant 15MHz bandwidth. 3)about the central frequency From the formula we can see that, if we make the Gm3 and Gm4 very large, we can get high frequency. But when I simulate, I make very large Gm3, Gm4 (about 15mS), but the frequency still not as high as I expect :( I guess because of phase error or paracitic capacitance :( 4) about phase error: since i should tune my tracking filter from 50MHz to 860MHz, I use parallel Gm3 and Gm4 (2Gm,4Gm,8Gm.. as shown in attachment) to tune the center frequency. Sure thing the phase can not be 90 degree for every case. Should I compensate the phase every time? If that, the tuning will become very complex:( I am so sorry to ask u so many questions again :( thank you very much ^^ have a nice day ^^ |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by buddypoor on Jul 20th, 2008, 11:58pm yixiusky wrote on Jul 20th, 2008, 8:46pm:
At first, you shouldnīt forget that all design formulas you are using apply only for idealized conditions (OTA input/output impedance as well as transit frequency infinite, no parasitics in hardware realization). Therefore, assuming all given formulas are correct (I didnīt recalculate yet) all the deviations from design values must be attributet to these non-ideal influences. As the OTA bandwidth is 2 GHZ it is really no surprise that you cannot tune the filter up to 860 MHz (corresponding to a ratio of only 2,3). Normally, at least a factor of 20 is required between opamp transit frequency and filter pole frequency. Another problem - especially for OTA bandpass filters - can be the dynamic range, which is rather limited for OTAs. Do you have access to a program for symbolic circuit analysis ? In this case you could try to derive formulas for non-ideal conditions in order to see the influence of some unwanted parameters. |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by yixiusky on Jul 21st, 2008, 8:04pm thank u very much ^^ sorry to ask u so many questions ^^ i think u have many experiences to design filter, am I right ? ^^ |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by Tlaloc on Sep 8th, 2008, 1:56pm One other fundamental thing that influences phase error is the finite gain of the amp. This can have a significant affect on the phase error, especially as the DC gain varies a function of pvt. |
|
Title: Re: The phase error of Integrator in Gm-c filter Post by loose-electron on Sep 11th, 2008, 11:45am A good guideline - the parasitic poles of any of the system parts and pieces need to be at least a decade away from the controlled filter poles that you are trying to create. So - If you want a gmC filter with 100MHz rolloff, your OTA (or gm stages) need to nave negligible phase response up to 1GHz over all process corners. That 20X figure mentioned above is probably pretty close. |
|
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |