Choice of error amplifier for LDO

Thread Starter

Franco Maggi

Joined May 6, 2021
2
Hi There!
I am, for several reasons about building a adjustable LDO. Some of the specifications are:
- min input voltage 4.4V
- max output voltage 3.6V
- 500mV dropout, due to sense resistor
- wide current output range
- at least 20kHz bandwidth @ 10uA load, ideally 100kHz

Pass element shall be a PMOS, not only due to the low dropout, but I need to separate the ground currents from the sensing element.
Performance/stability at light loads is the key aspect for the application.
I have sketched a discrete solution with a bunch of transistors, which is working in simulations, but is quite stuffy and its outcome uncertain when the parasitics of a real pcb add upp, heat generated on pcb. I also doubt the real bandwidth of the transistor models I found for the matched pairs.
So I would like to select an IC as EA.
What I am looking for is a wide band, mid-range gain amplifier. From the gain-loop analysis I have been simulating, I see I need the dominant pole of the EA to be at least 10kHz to avoid relative stability, given the desired bandwidth.
Any parts in mind?
 

Irving

Joined Jan 30, 2016
3,843
If you go to the Analog Devices (AD) or Texas Instruments (TI) websites, amongst many, you'll find parametric search tools that will allow to search on GBW, supply volts, common mode rejection, etc etc. That will give you many options. In particular LTSpice has many of AD's LT & LTC device models installed as does TINA for TI's offerings. Only you can pick an appropriate device based on your overall design requirements.
 

Ian0

Joined Aug 7, 2020
9,667
This is a circuit I've used many times. I was just curious to see how it behaves, so I simulated it.
The most important thing it needs is an electrolytic on the output with a fair amount of ESR, with just a ceramic capacitor it will oscillate in the MHz region.
It's not a perfect circuit, but it was adequate for what was required at the time. I used the BCM547B (the matched pair version)
I suspect that some emitter degeneration resistors in the BC547s and the BCV62 enough to drop about 100mV, and probably a Wilson current mirror would improve the accuracy.
If you are trying to replicate this with a 4.4V supply voltage, then choice of MOSFET would be important, as too high Vgs(th) would mean it runs out of drive.
Another thing - there is no current limiting!

It's much better behaved with a bipolar transistor on the output - no sign of oscillation even with a zero ESR capacitor* - and wouldn't suffer at low supply voltages from lack of drive.
It would also be possible to effect some crude current limiting by choosing a higher value of R3 to limit the base current, but relying on a particular value of Hfe is never a great idea.LDO.png



*of the type that exists in SPICE but not in real life.
 

Thread Starter

Franco Maggi

Joined May 6, 2021
2
Thanks Ian0,
I am happy to see I am not the only guy thinking discrete components at this point.
my main challenge finding an IC, is the phase drop that comes along with opamp's internal compensation. So I started designing a differential pair, not far from the one you sketched.
ESR is a must, agree. I got a current mirror in place of R3, to control the bias of the transistors above.
Now, do you think the BJT models in LTspice might have too optimistic bandwidth? I see NXP specs 100MHz GBP for BCM547B. I would expect these to roll off at some hundreds kHz when they are put together in the loop.
How is your build working at light loads? Any need for additional compensation?
 

Ian0

Joined Aug 7, 2020
9,667
You don’t need much bandwidth for a voltage regulator. Bc547 types should do the job, they manage perfectly adequately in audio amps where they have to be well behaved up to about 1MHz. Your guess at roll off would be about the same as they behave in a compensated audio amplifier.
My circuit was mainly a battery charger, so spent most of its time in current limiting (part of the design I didn’t draw). When it got to float voltage it was stable.
 
Top