Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

GBW id different from -3dB frequency

Status
Not open for further replies.

Junus2012

Advanced Member level 5
Joined
Jan 9, 2012
Messages
1,552
Helped
47
Reputation
98
Reaction score
53
Trophy points
1,328
Location
Italy
Activity points
15,235
Hello

My operational amplifier Open loop frequency responce showing first order performance with GBW=48.66 MHz and phase margin about 58 Deg as you can see below

openloopAC.png


After running the closed loop AC response with G= 1 I was expecting to get f-dB = GBW, but it is not the case, kindly see the below image

closedAC.png

What does this mean ?

Thank you
 

GBW is gain times bandwidth. At 48MHz you've got A=1, and that's where it starts to rolloff. (Well, not exactly, since you've got some peaking). GBW has nothing to do with the -3dB point. It simply identifies the frequency point where, if the opamp where open loop, the gain drops to 1.
 
Your gain peaking is hiding the true rolloff you'd see in
A=10, A=100, AVOL gain vs frequency plots. You might
look at whether more or less peaking gives the best
bandwidth and stability margin.
 
Open loop frequency response showing first order performance
No it's not first order. First order means phase margin of 90 degree. There's considerable phase lag from additional poles, it causes gain peaking and the increased -3dB frequency in closed loop.
 
Thank you friends,

I just want to say that gain bandwidth product is not equal to unity gain frequeny,

in my amplifier the GBW equal to 43 MHz and as I presented before the unity gain frequency is 48.668 MHz. However, most of the designers they treat it the same which is confusing for me and might be the answer of my problem
 

Hi again,

Below you find please in swquence the closed loop ac response with gain of 2, 4, 8, 16 and G=128

Four your further discussion

thank you

G2.png

G4.png

G8.png

G16.png

G128.png
 

GBW has nothing to do with the -3dB point. It simply identifies the frequency point where, if the opamp where open loop, the gain drops to 1.
Don't want to decide about right or wrong, but there are at least other commonly used definitions. According to Analog Devices and TI application notes and datasheets, GBW is the product of (non-inverting) gain and small signal bandwidth. It's usually measured at higher gains (e.g. 10 or 100). Particularly for decompensated OPs, the GBW is considerably different from the open loop unity gain frequency, but smaller differences occur as soon as the phase margin is different from 90 degree.
 
Don't want to decide about right or wrong, but there are at least other commonly used definitions. According to Analog Devices and TI application notes and datasheets, GBW is the product of (non-inverting) gain and small signal bandwidth. It's usually measured at higher gains (e.g. 10 or 100). Particularly for decompensated OPs, the GBW is considerably different from the open loop unity gain frequency, but smaller differences occur as soon as the phase margin is different from 90 degree.

Dear FvM. I have also read this from TI, they stated as you said that for uncompensated GBW and the unity gain frequency are different because there are different slope region when the gain is rolling down, mean transition from -20 dB/decade to -40 dB/decade and so on... I would expect that if I have in my open loop response such different gain slop in the region up to unity gain frequency, you saw in my first post how my amplifier is rolling with one slope only,

Can you please give me your comment on the last results I uploaded,

Thank you once again
 

I assume you are simulating your loop gain with already compensated amplifier and if it is compensated well, meaning the non-dominant pole is about 3x farther in frequency than the UGBW, then closing the loop should have a -3dB frequency pretty close to the GBW.
However, from your original plot of the loop gain I see that right about 0dB crossing there seems to be some slight flattening of the magnitude response and at the same time the phase drops. This suggest for RHP zero. If your amplifier is Miller compensated, did you take care for mitigating the effects of the RHP zero that usually appears in that kind of compensation? Also, it will be better if you simulated your loop gain to higher frequency, this way you can see what happens after it crosses 0dB and where the non-dominant poles are.
 
Last edited:
I assume you are simulating your loop gain with already compensated amplifier and if it is compensated well, meaning the non-dominant pole is about 3x farther in frequency than the UGBW, then closing the loop should have a -3dB frequency pretty close to the GBW.
However, from your original plot of the loop gain I see that right about 0dB crossing there seems to be some slight flattening of the magnitude response and at the same time the phase drops. This suggest for RHP zero. If your amplifier is Miller compensated, did you take care for mitigating the effects of the RHP zero that usually appears in that kind of compensation? Also, it will be better if you simulated your loop gain to higher frequency, this way you can see what happens after it crosses 0dB and where the non-dominant poles are.


Dear Suta,

Thank you for your reply

I am using fully differential folded cascod amplifier with class AB output buffer stage , as you can also see the second stage of it in my below image, I used Ahuja or indirect compensation, the capacitor is fed to the cascode transistors, where it is supposed not needed to add resistor with it in series for the RHP compensation, as far as I know

rhp_circuit.png

Here I have simulated for you at higher frequency

rhp.png
 

OK, I see. If you do Ahuja compensation, you don't need the resistor in series with the Miller capacitor, that's true. However, Ahuja compensation needs careful design. It can lead to complex conjugate poles in your loop gain if not designed correctly. I think you have a bit of this showing in your plots - the slight bump in the Bode plot near crossover frequency. And it, of course will roll-off your phase faster. Did you check the stability of the internal feedback loop of the Ahuja compensation?
 
OK, I see. If you do Ahuja compensation, you don't need the resistor in series with the Miller capacitor, that's true. However, Ahuja compensation needs careful design. It can lead to complex conjugate poles in your loop gain if not designed correctly. I think you have a bit of this showing in your plots - the slight bump in the Bode plot near crossover frequency. And it, of course will roll-off your phase faster. Did you check the stability of the internal feedback loop of the Ahuja compensation?

Thank you Suta for your reply,

Indeed I didnt get what you mean by "Did you check the stability of the internal feedback loop of the Ahuja compensation?",
 

There is an internal loop when using the Ahuja compensation, which needs to be stable. You probably forgot but we have already discussed this point before.

https://www.edaboard.com/showthread...ded-OPAMP-with-nulling-resistor-or-transistor

Thank you Suta for reminding me with that post

You have presented a solution as shown below,

Ahuja.PNG

To increase gm M10 I have to increase the current and or W/L, since I dont want to disturb the current in the output branch I would go to increase W/L, however this will increase my design area, I am already using 20 µm for the NMOS and 60 µm for the PMOS

I read back that post and your discussion with Frankrose, looks like your prediction was right since that time, Ahuja is creating an issue of peeking in the closed loop configuration

One more question please, how can I see my former posts in the forum ?

Thank you once again
 

how can I see my former posts in the forum ?

Many ways:
1. Use advanced search
2. Click on user name in a post, select View Forum Posts

view1.PNG

3. In top menu, select Important Links/My Posts or /My Posts as Threads
The latter shows only the threads you have started
 
Do you use min L for M10? Also, in that older post I pointed to a reference where there are other ways to compensate for the Ahuja loop.
 
Do you use min L for M10? Also, in that older post I pointed to a reference where there are other ways to compensate for the Ahuja loop.

Dear Suta,

I am not using minimum length since the gain was dropping severely,

for other methods, I tried to follow those links but was not working
 

Your gain shouldn't really depend strongly on the cascode device. You should use small L if you want to push the zero ot increase gm10. Maybe not the min L but close to it.
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top