ESR value for LDO and tempco

[ee_cchac]

I am now working on LDO, seesm to be almost finish.

But i got a problem.

When i was working on my LDO design, the loading cap 's esr value is 1 ohm. Then, when i almost end, i heard that 1 ohm is too large, and 20m ohm is suitable.

I was then super down, since it seesm to me i need to back to first step to tune the size of LDO.

Questions would like to ask:

1> i heard that 1 ohm is suitable to use, While small value say, 20m ohm is also good. It is application dependent ? Whcih value i should use ?

2> my tempco result is evry bad, i guess i should be "thanks" to the bandgap(* BJT*) circuit, since time is runnign out, any expert give me hints to design a simple and low tempco (50ppm/oC) bandgap ?

 

[fredflinstone]

Hi,
Though high ESR values are good from stablility point of view. It will badly affect your PSRR. So its good to use cap loads having low ESR. I think you should be designing the LDO for 20m Ohms or lower.
I think a simple constant Gm bias circuit with first order temperature compensation would give you < 50 - 60 ppm tempco. You can also do some trimming if its not possible to meet this spec. But I think you should be able to achieve this without trimming.

regards
fred

 

[ee_cchac]

Thanks for reply

i now got another problem, i find that LDO tempco not follow bangap tempco.

my bandgap tempco is done shape, but LDO is a line with a increasing slope..

it seems my circuit should correct in connetion.

anyone know what's wrong ?

 

[cinch]

Hi,
If u want a stable LDO with lower ESR, then u have to go for frequency compensation with Gm enhancement. There is a paper by "Chaitanya Chava" on this.

Also, tempco with increasing slope can be made dome shaped by varying the resistances. One thing to be remembered is the resistors in feedback loop must be kept in multiples. For example, R1=20k and R2=37.5k, then choose unit resistor as 5k. For R1, use 4 unit resistors in series. For R2, use 7 unit resistors in series (35 k ohm) + 2 unit resistors in parallel (5k/2 = 2.5k). To summarize, whatever the resistance value u need has to be derived from series-parallel combination of unit resistors. This will reduce variation of different resistance values across PVT corners.

Hope u understood...

Regards,
cinch