The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> Analog Design >> super source follower in LDO https://designers-guide.org/forum/YaBB.pl?num=1226495598 Message started by trashbox on Nov 12th, 2008, 5:13am |
Title: super source follower in LDO Post by trashbox on Nov 12th, 2008, 5:13am Hi all, As a common frequency compensation method, super source follower is used as a buffer between error amplifier and power stage(pass gate) in linear regulator such as low-drop out regulator.I have a question about the stability issue about the super source follower itself as attachment. The super source follower itself is a negative feedback system(break at Va): Va to Vb via M2(common source) with a negative gain -gm2*Ro_vb and a pole 1/Ro_vb*C_vb, Vb to Va via M1(common gate) with a positive gain gm1*Ro_va and another pole 1/Ro_va*C_va. In LDO, generally, M3 is very big so that the C_vb is large, Ro_vb is small because M1 is a source follower, on the other hand, C_va is very small and Ro_va is large. It results in the stability issue about this super source follower itself, my simulation shows phase margin is about 25 degree. How can anybody solve this problem? Or would you please give me any advice? Thanks in advance! |
Title: Re: super source follower in LDO Post by thechopper on Nov 12th, 2008, 8:33am Hi Trashbox, If I understood correctly, it is the super-buffer structure the one that shows stability issues (PM 25deg) when included in the LDO loop. If that is the case I think you hit the nail when referring to the large gate capacitance of the pass transistor affecting the super-buffer stability. What I see happening here is the super-buffer output pole being pushed to low frequency and therefore getting closer the the super-buffer loop dominant pole. If I was correct one potential solution I guess would be to use a compensation capacitor at VA. In this way I imagine you could push further downwards the SB loop dominant pole. By the way, I'm not totally clear on what the real advantage of using this super-buffer is for driving the pass transistor's gate in your LDO. Could you please clarify this a little bit more? Thanks & Regards Tosei |
Title: Re: super source follower in LDO Post by buddypoor on Nov 12th, 2008, 12:52pm Hi trashbox, do you really need a buffer stage between the error amp output and the pass element input ? Does the error amp has no low impedance output ? Regards LvW |
Title: Re: super source follower in LDO Post by buddypoor on Nov 13th, 2008, 12:53pm Hi trashbox, here are some additional points which just have come to my mind: The internal feedback of the super source follower is in fact a sort of local feedback - if compared with the large overall feedback loop of the whole LDO circuitry. Such a local feedback loop normally is uncritical - even when its phase margin is as low as 25 deg. as long as the time constant (response time) of the overall loop is much larger (which means its bandwidth is much lower) than the corresponding parameter of the local loop. Tests have shown that a factor of 50...100 is appropriate. Perhaps this helps a bit. Regards. |
Title: Re: super source follower in LDO Post by trashbox on Nov 14th, 2008, 7:06pm To Tosei(thechopper): Because this LDO output current range is very large(0-300mA), the pole at pass element gate also has a very large position range, which results in the diffficult frequency compensation for whole LDO loop. If the buffer is added with a very small output impedance 1/gm, the pole at pass element gate is pushed out and becomes the third pole (1st pole is LDO's output because of large offchip cap, 2nd pole is error amplifier's output because of its large output resistor)even if the output current is changed from 1 to 300mA. This is why I need this buffer. To LvW(buddypoor): Thanks very much for you advice. It seems reasonable. :- ) |
Title: Re: super source follower in LDO Post by thechopper on Nov 15th, 2008, 5:06am Hi Trashbox, Now I see the purpose of the super-buffer. Thanks for the clarification. In case you still do not feel confortable with this circuit inside your loop (altough buddypoors suggestion indicates you should not have any problem) an alternative solution (instead of using a buffer at the error amplifierīs output) would be to have a transconductor error amplifier and copy the output current into a PMOS diode that generates the bias for the pass transistor PMOS. In this way the impedance at the pass transistor gate would still be 1/gm. I guess you could use the same amplifier you are currently using but just add an output pass transistor so you can get the currrent at its drain. Such current should be proportional to the error you are trying to compensate. I used this scheme successfully before. Regards Tosei |
Title: Re: super source follower in LDO Post by Tareeq on Nov 15th, 2008, 5:34am Hi Chopper, I appreciate these intesresting insights. HdrChopper wrote on Nov 15th, 2008, 5:06am:
Do you mind drawing a picture of the modification with respect to the original circuits ? Thanks. Tareeq. |
Title: Re: super source follower in LDO Post by fonseca.ha on Nov 16th, 2008, 10:55am Hi. A way to get better stability for this circuit seems to be by increasing the consumption of the first stage. Is that an option? Cheers, H |
Title: Re: super source follower in LDO Post by thechopper on Nov 17th, 2008, 2:28am Tareeq wrote on Nov 15th, 2008, 5:34am:
Hi Tareeq, Please see the attached figure. Regards Tosei |
Title: Re: super source follower in LDO Post by Tareeq on Nov 17th, 2008, 12:29pm Hi, Thanks Tosei for the illustration. if a simple low impedance is desired at the gate of the pass transistor, the 1/gm solution (chopper style ;-)) could be probably enough for stability. My feeling is that there is something else at stake: a better transient response. Although the gm of the pass transistor is high and could potentially lead to a fast response, things are slowed down by the Cgs. In that case a very low impedance (that the super source follower bring about) lower than 1/gm could be achieved ?(since the actual output impedance of the source of the common drain is divided by the loop gain= Gain of the first stage roughly) Tareeq. |
Title: Re: super source follower in LDO Post by Pravesh Kumar Saini on Oct 22nd, 2014, 7:11pm What about using the CS-CG SSF instead of CS-CS based SSF? |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |