help~~~~~~~ about GainBandWidth of telescopic OTA

[wdd]

hi all,
In my design, the TSMC CM025 process is used.
GBW need to be higher than 500MHz, so the gm of input NMOS is fixed to 2*phi*Cl*700MHz(Cl is the load capacitance, which is 3pF).
AC and PAC simulation shows the GBW is 220MHz(phase margin is 85), much less than 700MHz,then I checked the nondominant pole(located larger than 4.5GHz, is enough i think). When the cursor is moved to phase margin= 70 point ,the frequency is about 900MHz.

So how can I modify the circuit to increase the GBW (also reduce the phase margin to about 70)? The topology is telescopic OTA with NMOS transisitor as input, the OTA is used in pipelined ADC.

Thanks a lot!!
wdd

 

[ezt]

Hi.
I don't know if the gm of input devices is large enough, but in this case. by reducing Cl (load cap.) phase margin will decrease and GBW will increase. But if you should use a specified Cl and if the whole power consumption and low gain of your opamp is not important, then I suggest increasing input gm. But usually this is not a good way, so I recommend using a two-stage structure with a specified Cl. Then GBW will mostly change by Cc (compensation cap.). By using a two-stage structure, you will have both good output swing and high gain at the same time which are two important specs of opamps in ADCs.

Regards,
EZT

 

[wdd]

I'vd checked the gm of input device who is large enough.
when reduce the value of Cl, the GBW is larger, but the phase margin has little change, and in fact the Cl shouldnot be changed.

Intuitively, with high enough nondominant pole, the GBW should be defined by the gm of input devices and load. So what's wrong with it, i mean maybe i lost something.

Regrads.

 

[ezt]

Hi.
gmi = 2×pi×ωu×Cl ≈ 13.2 mA/V
and if A0 is the DC-gain of opamp then we have
f (dominant-pole) ≈ f-3dB = fu/A0 = 700MHz/A0
Please check if it is true in your circuit. By the way, if the non-dominant pole has located in very high frequencies, then PM (Phase Margin) of your circuit won't be very less than 90 deg. However, 4.5 GHz is high enough I think. Because as you know we consider the effect of each pole (or zero) on phase-freq response from 1/10 of its magnitude, which is about 450MHz here. If we assume there is just one non-dominant pole, then each decade PM will decrease by 45 deg. So in 700 MHz, PM would be
90 - log10(700/450)*45 deg = 90 - 8.63 = 81.37 deg
Which is much larger than 70 deg. If it is important (e.g. due to speed requirements) to have less PM, then you should bring the non-dominant pole a little closer to the origin. But ofcourse, all these calculations are being held for a simple structure with a low-freq dominant pole and just one non-dominant pole (However, these approximations are highly true for single-stage simple cascode opamps without gain-boosting implementation. I think your topology is like that, am I right? because you have used Cl in fu calculation).

Regards,
EZT

 

[noiseless]

One thing should be reminded you is that what is the ft of the 0.25um. I know there is on the process manul. However, that is digital ft. For the analog design, ft is much smaller than that because of your small effecitve overdrive voltage. The question is are you sure 0.25um can run so fast?

 

[wdd]

Hi.
gmi = 2×pi×ωu×Cl ≈ 13.2 mA/V
and if A0 is the DC-gain of opamp then we have
f (dominant-pole) ≈ f-3dB = fu/A0 = 700MHz/A0
Please check if it is true in your circuit.

in my design, the gmi is choosed a little higher than 13.2mA/V for safety,
about 14.5mA/V. f(dominant pole)≈220MHz/A0, not 700MHz/A0.
Do you have some suggestions? Thx a lot, I have no idea now.
For the phase margin, later a boosted amp will added to this OTA, so nondominat
pole will reduced a little for the new parasitic capacitor, that's why a little higher
phase margin is selected.

Best Regards,
wdd

 

[nxing]

Hi wedd,

Can you post your simulation result online? If you want, you can post your the schematic and bias condition as well. Maybe that will help!