Designer’s Guide Community

vivkr

Community Fellow

Offline

Posts: 780

Miller compensation with leaky caps
Aug 18^th, 2009, 11:23pm

I have to work with capacitors that are a bit leaky (advanced process) and nonlinear with voltage. The nonlinear part may be manageable (somehow) but the leaky part is casuing trouble with compensation.

Does anyone have experience and/or ideas for dealing with leaky capacitors for compensating 2-stage amps? Maybe there is a better architecture to handle this.

Thanks,

Vivek

Back to top

IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: Miller compensation with leaky caps
Reply #1 - Aug 20^th, 2009, 5:55pm

Hi vivek,

Several times I had to use leaky caps for compensating opamps. Specially when working at high temp (> 100 degrees).
I never found major dificulties with using these caps (poly on one side, n+ on the other plate). What I particularly took care of is connecting the implantation or diffusion side on the output node when used as a Miller cap. In this way the leakage flows through a low impedance node having gained a lot, thus its effect is negligible.
Did you already considered this?

Regards
Tosei

Back to top

Keep it simple

IP Logged

loose-electron

Senior Fellow

Offline

Best Design Tool =
Capable Designers

Posts: 1638
San Diego California

Re: Miller compensation with leaky caps
Reply #2 - Aug 22^nd, 2009, 2:49pm

Vivek:

Here is an idea - before closing the feedback path - take the output of the amplifier and go thru a source follower - which then drives the capacitor (leaky or not now does not matter) as the BW limiting feedback path.

This gets rid of a HF feedforward path, and works pretty well. Personally, I have been using this method for many years, instead of a resistor in series with the feedback capacitor. It tends to be less process dependent in the corners as well.

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

vivkr

Community Fellow

Offline

Posts: 780

Re: Miller compensation with leaky caps
Reply #3 - Aug 24^th, 2009, 4:34am

Dear Tosei, Jerry,

Thanks for your replies.

Tosei: I think you are assuming leakage from one terminal to bulk. I am looking at caps which are leaky from terminal to terminal, i.e. the cap actually has some finite shunt resistance in parallel. Otherwise, your suggestion would work.

Jerry: I thought of this too. I think this falls in the same category as the Ahuja compensation. I did not try it out because I could not see how the Miller compensation would work correctly still. Of course, the DC behavior would be much improved due to the additional source follower, and if one is lucky, then the impedance presented by the Miller cap is much smaller than the parallel shunt resistance (to be expected) in the critical frequency range around the unity-gain crossover. Will try it out, although I may have found an alternate solution in form of a dense enough cap which leaks only from one end to the ground instead of carrying a shunt resistance. I can then connect the leaky end to the opamp output as Tosei suggests.

Best regards,

Vivek

Back to top

IP Logged

cczhang

New Member

Offline

Posts: 9

Re: Miller compensation with leaky caps
Reply #4 - Sep 9^th, 2009, 1:06am

vivkr wrote on Aug 18^th, 2009, 11:23pm:

I have to work with capacitors that are a bit leaky (advanced process) and nonlinear with voltage. The nonlinear part may be manageable (somehow) but the leaky part is casuing trouble with compensation.

Does anyone have experience and/or ideas for dealing with leaky capacitors for compensating 2-stage amps? Maybe there is a better architecture to handle this.

Thanks,

Vivek

Dear Vivek
could you tell me what's the meaning by leaky cap.
thanks

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: Miller compensation with leaky caps
Reply #5 - Sep 9^th, 2009, 6:49am

cczhang wrote on Sep 9^th, 2009, 1:06am:

vivkr wrote on Aug 18^th, 2009, 11:23pm:

I have to work with capacitors that are a bit leaky (advanced process) and nonlinear with voltage. The nonlinear part may be manageable (somehow) but the leaky part is casuing trouble with compensation.

Does anyone have experience and/or ideas for dealing with leaky capacitors for compensating 2-stage amps? Maybe there is a better architecture to handle this.

Thanks,

Vivek

A leaky cap is one which has some DC current flowing between the plates. In other words, there is a shunt resistance in parallel to the cap.

Regards,

Vivek
Dear Vivek
could you tell me what's the meaning by leaky cap.
thanks

Back to top

IP Logged

wave

Senior Member

Offline

Posts: 117
Silicon Valley

Re: Miller compensation with leaky caps
Reply #6 - Sep 9^th, 2009, 6:42pm

[quote author=vivkr link=1250663035/0#0 date=1250663034]I have to work with capacitors that are a bit leaky (advanced process) and nonlinear with voltage. The nonlinear part may be manageable (somehow) but the leaky part is casuing trouble with compensation.

Just curious of the construction of the Cap?
� � � � �Poly - LeakyThin Oxide - Nwell perhaps?

The idea floated of a Src Follower would be a good current blocker (both for DC and AC currents) if you have the headroom.

Wave �:)

Back to top

IP Logged

RobG

Community Fellow

Offline

Posts: 570
Bozeman, MT

Re: Miller compensation with leaky caps
Reply #7 - Sep 13^th, 2009, 12:05pm

This is an interesting problem that I didn't know existed. What dimensions does it start to show up? Is it really resistive (i.e. signal dependent)?

If it is a resistor in parallel with the cap I'm not sure that the SF in feedback loop would work. The miller effect will still decrease the resistance seen by the first stage, lowering the overall gain.

If sampled you could use a double sampling method to boost gain. That way the first stage gain could be low enough that the resistance isn't an issue.

Is there a reason why metal/metal caps will not work? You could control the spacing using side-by-side metal paths.

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: Miller compensation with leaky caps
Reply #8 - Sep 18^th, 2009, 1:14am

RobG wrote on Sep 13^th, 2009, 12:05pm:

This is an interesting problem that I didn't know existed. What dimensions does it start to show up? Is it really resistive (i.e. signal dependent)?

If it is a resistor in parallel with the cap I'm not sure that the SF in feedback loop would work. The miller effect will still decrease the resistance seen by the first stage, lowering the overall gain.

If sampled you could use a double sampling method to boost gain. That way the first stage gain could be low enough that the resistance isn't an issue.

Is there a reason why metal/metal caps will not work? You could control the spacing using side-by-side metal paths.

Hi Rob,

Firstly, as I noted somewhere above, I have managed to avoid this problem for the moment by using other caps which are slightly less dense but not leaky. Metal-metal caps would be too area-inefficient, and MIM were not available.

Basically, the problem appeared when I tried to use some MOS-based-caps. These are a bit leaky in the technology I am working with. That would be at 45 nm.

Regards,

Vivek

Back to top

IP Logged