LDO oscillating -- gives me the sads

[djacobow]

I have a simple new design (Arduino-ish) that uses a 3.3V MIC5205 LDO. I get a 25-50mV sawtooth shaped oscillation. There appear to be two frequency components, one of which is reasonable low frequency, and is actually audible as a very faint chirp.

Loading down the 5205 with a 100 ohm resistor stops all the oscillations entirely. Similarly, adding a boatload of output capacitance also stops the oscillation.

However, the circuit already has all the capacitance recommended in the datasheet, and then some. There is a 470pF bypass cap on pin 4 as well as a 22uF 1206 MLCC on the output. The output also goes to a bunch of chips that are all bypassed by 0.1uF MLCCs. Maybe there are 15 of those. The datasheet recommends at least 2.2uF of output capacitance (I have 10x that) and that you use a tantalum. But my understanding was that the MLCC would have lower ESR and would perform as well as a tantalum.

What else might I have done wrong?

If I were making one or two of these boards, I'd just leave the fat electrolytic on there and not worry about it, but this is going to production in the 100s at least. Really want to understand what I've done wrong.

Regards,
Dave J

[ovnr]

Try reading the datasheet, it's right there on page 1:

"For low-dropout regulators that are stable with ceramic output capacitors, see the µCap MIC5245/6/7 family."


Or, in short, your MLCC cap has a too low ESR, and causes oscillations. Insert a series resistor, or switch it out for a tantalum. In the datasheet (page 8, "Output Capacitor"), Micrel suggests a capacitor with an ESR of 5 ohms:

"The output capacitor should have an ESR (effective series resistance) of about 5? or less and a resonant frequency above 1MHz. Ultra-low-ESR capacitors can cause a low amplitude oscillation on the output and/or underdamped transient response. Most tantalum or aluminum electrolytic capacitors are adequate; film types will work, but are more expensive."


Of course, you could replace the regulator with something made for ceramic caps.

(Edit: Formatting & stuff.)

[djacobow]

I guess that's what confused me. It says "5 ohms or LESS" which to me implies that adding a resistor in series won't help.

Try reading the datasheet, it's right there on page 1:

"For low-dropout regulators that are stable with ceramic output capacitors, see the µCap MIC5245/6/7 family."

But that part is well taken. I'll try a tantalum and/or the other regulator.

[ovnr]

I guess that's what confused me. It says "5 ohms or LESS" which to me implies that adding a resistor in series won't help.

Yes, but right after that it said "Ultra-low-ESR capacitors can cause a low amplitude oscillation on the output and/or underdamped transient response.". Ceramic caps are classified as ultra-low-ESR.


Realistically you'd probably be best off choosing a different regulator; tantalums are annoying in their own special ways, like catching on fire. Plus it'd probably be more expensive.

[djacobow]

Yup, thanks for the help. This was definitely my mistake. I adapted something I did not understand and got slightly burned.

Realistically you'd probably be best off choosing a different regulator; tantalums are annoying in their own special ways, like catching on fire. Plus it'd probably be more expensive.

[LukeW]

You only need a tiny touch of resistance, you could try putting say 1 ohm in series. Or, some people suggest using a 0-ohm resistor, which has just a tiny touch of resistance which is sufficient.

[djacobow]

Hey, the worked a treat! I didn't have room to bodge in a series resistor, so I just soldered my cap and resistor vertically and soldered a wire on top to make a little stone henge. Not pretty, but worked. And lo, the oscillation was gone.

But this is good, I just need to change my BOM and don't need to spin the board. (At least for this issue...)

You only need a tiny touch of resistance, you could try putting say 1 ohm in series. Or, some people suggest using a 0-ohm resistor, which has just a tiny touch of resistance which is sufficient.

[amyk]

I guess that's what confused me. It says "5 ohms or LESS" which to me implies that adding a resistor in series won't help.

Yes, but right after that it said "Ultra-low-ESR capacitors can cause a low amplitude oscillation on the output and/or underdamped transient response.". Ceramic caps are classified as ultra-low-ESR.

In this case I think the datasheet needs to be clearer on what range of ESRs are allowable - what exactly is "ultra-low"? I've seen LDO datasheets that specify e.g. 0.5 - 10 ohms ESR for the output cap.

[Mad ID]

For cases like this I always put a 1R series resistor in series with the ceramic. Much better option than a tantalum IMHO.

[T3sl4co1l]

Eugh, that's one of the best specified "awful LDO" I've seen.

Usually, if they're bad, they just don't specify anything at all.  Mystery box.  Cheap Chinese things, often.

Micrel isn't a bad brand for the most part but this isn't one of their better examples.

What I'm looking for, and what it's missing:
- Is it CMOS or bipolar design (yes: bipolar)
- Supply current under different operating modes, esp. if bipolar
- Compensation, discussion of load (or source) capacitor ESR
- Graphs of operating conditions: load/line regulation, bias current, temperature variation, transient response, stability (C + ESR) ranges, etc.

This part has none of the important factors, and therefore fails my conditions.  I would not design this into a new product, and I would strongly recommend designing it out of an existing product.

The reason bipolar is important is, a naive regulator will simply apply current to the PNP base as needed to maintain output (that's what the schematic says: error amp into base).  When it goes into dropout, the bias current explodes as the amp tries to account for something which is no longer possible (i.e. the pass transistor is saturated, and more base current will not raise output voltage any further).

There are good bipolar LDOs, which specify that supply current (current into the ADJ/GND pin) does not increase significantly in dropout, and that remains roughly proportional to load current (being essentially the pass transistor's hFE and little more).

The disadvantage to CMOS LDOs is, the pass transistor has poor gain (near dropout, it's in the triode region where gain is low, output resistance is low, and therefore the loop characteristics change, usually to the mushy side of things), and the current density sucks (so it takes a bigger die for the same pass current and dropout voltage, often increasing cost).  Poor examples still suffer from bad compensation, but in general, CMOS devices rarely suffer from anomalous current draw as the poor bipolar devices do.

Tim

[Mad ID]

The reason bipolar is important is, a naive regulator will simply apply current to the PNP base as needed to maintain output (that's what the schematic says: error amp into base).  When it goes into dropout, the bias current explodes as the amp tries to account for something which is no longer possible (i.e. the pass transistor is saturated, and more base current will not raise output voltage any further).

One engineer tried to persuade me that it's OK to use an LDO when input/output differential is not good enough and I could not give him an argument against it when he showed me that it works. The regulator was MIC5205 and I have a design when in some cases the input voltage is too low and we use the LDO as a "pass element".  Can you comment on this issue and if it is OK practice, just with a "smarter" regulator?

How big can the extra power draw get? The system I'm referring to is running from batteries..

[T3sl4co1l]

Obviously, we can't tell in this case... you'd have to measure and characterize.  I'd assume, the worst (>mA?).

There are good regulators, I don't have a comprehensive list or anything.  One that I've seen recently, that looks good: LP2950, a 100mA bipolar LDO.  Headlining the graphs section, you see quiescent current vs. input voltage, and you clearly see it does draw excess current in dropout, but hardly more than operating levels -- this would be fine for battery power applications.  They also discuss capacitance and ESR (1-4.7uF, 0.1-2 ohm), though they do miss on saying "see typical requirements" when there's no graphs relating to ESR at all.

Tim

[Chupacabras]

Eugh, that's one of the best specified "awful LDO" I've seen.

I know this is old thread, but I must comment on this statement.
Actually I'm thinking to use MIC5205 and it has some good parameters and I have problems to find low-ESR replacement LDO.

Input voltage can be up to 16V,
pretty good line (0.004%/V) and load (0.02%) regulation,
ripple rejection PSRR 75dB,
low output noise (density 260nV/?Hz)

I find this combination quite impressive.
In the same time I agree that datasheet could contain more details.