Tant. cap ESR

There are two kinds of LDOs, ones that need ESR and ones that hate ESR. This one hates it, so use big ceramics.

It should be possible to design an LDO that doesn't care; there are c-load opamps like that.

No, not as long as the regulator doesn't require a minimum ESR value. If one does I usually ditch the regulator.

Why would anyone in their right mind use a 4.7u tantalum? Besides ESR issues they can also turn into fireworks.

If you encounter one that needs ESR my advice would be: Run!

A few good ones can but they are also expensive. Unless there is a really, really compelling reason I'd never spend $2 on an LDO. Increasingly I am seeing cases where engineers use LDOs where they are not required, for example to regulate 12V down to 5V or 3.3V. That's LM317 turf.

Lots of regulators, probaby most, need some minimum ESR. Classics like LM1117s do. Maybe an LM337. Older parts like this often don't say, since they were designed back when big ceramic caps weren't common.

Buck switchers especially need some ESR. The output inductor and a low-ESR cap make a 180 degree phase shifter.

It's scandalous how many data sheets seem to deliberately avoid dealing with ESR/stability issues.

I like LM1117 in dpak, which is a cheap MDO version of the 317. But it needs ESR. I usually use a 22 uF tantalum, with suitable anti-detonation precautions.

I use a lot of LP2985's, which are specifically stable with ceramic caps. We pay about 35 cents.

I like to use the cheap LM1117 regulators. I put a 10uf ceramic with a

Negative regulators are often of the same architecture as LDOs. Best to stay away if possible. With switchers it is different, there we (usually) have unfettered access to the loop compensation. Not having such access is what's the problem with msot LDOs.

Mostly that is the case with LDOs. The most you often get is a graph with a shaded area in there. But guaranteed limits in the datasheet table section? Zip, zilch, nada. When designing something where a failure would make it into court or into the evening news it's not so cool to use that architecture.

I am cured of the 29-series after an LM29-something went unstable when the DC source (!) impedance went above a certain limit. No mention in the datasheet. A lengthy phone conference ended with an "Oh s..t!" kind of discovery when they looked at the chip design.

But their LM317 is a great part. It's like the Willys Jeep, nearly indestructable.

And then lots of circuitry is fed by it, the bypass capacitors which don't have series resistors add up and up and up and ... tssst ... *PHUT*

Increasingly, _I'm_ seeing the abbreviation "LDO" used where the author means "plain old regulator".

Drop-out voltage seems to be a hard effect for folks to remember. I only needed one circuit with 6V to an LM78L05 to cure me of that, but it did come as a surprise at the time.

I used LM1117s as power amps, inside an opamp loop, to drive some load cells. Just in case, I tried all sorts of voltages on the in/out/adj pins to see if some unusual condition could damage the regulator chip, like shorting the output and yanking adj way up. I wasn't able to break it.

Oh yeah, I've seen that a lot as well.

My comeuppance came as a kid when I built a frequency counter. At 230V line voltage it worked nicely, at 220V it started to have issues when a lot of display segments came on. But I could not afford a bigger transformer so I hung a resistor across the regulator, hoping that the current consumption would never drop below half an amp (this was in the SN7490 days) and that the line voltage never goes above 235V or thereabouts.

Heck, Digikey doesn't even have a "plain old regulator" category, they're all mixed into the same one.

Tim

A lot of the old Tek tube scopes had a power resistor across the series voltage reg tubes, to share the load current. What was the number of that gigantic low-mu triode they used to use?

I've done that trick myself. If you use a c-load opamp as a regulator, you can pass 100% of the expected load current through the resistor and let the opamp push or pull the remainder.

Was that the 12B4?

In my frequency counter that would begin to glow. The whole thing consumes way more than an amp at 5V. Much of that goes towards the big

I still have that counter but now that I bought a Chinese arb generator which to my surprise has a pretty good frequency counter in there it has outlived its useful life. So far I couldn't bring myself to scrapping it but I know that some day I'll have to do it. A counter that I painstakingly build by hand when I was a kid .

Actually thats not a problem when the capacitors are further away.

Awww, that just breaks me heart.

We have been using old HP5370Bs for ages. It does single-shot time interval measurement at 20 ps resolution. But eventually all old equipment dies. We just ordered an Agilent 53230A counter which, after all these decades, finally gets back to 20 ps time-interval resolution. At least it has a TCXO standard; on their $2000 counter, they want another $1000 for a TCXO!

That depends on the definition of the word "farther". In well-engineered designs they'd be right there on the same board and the 3.3V is not a rail but a power plane. Then with an LDO all hell can break loose.

Yeah, absolutely! I've had to repair a few things where a Tantalum burned up. Not really fireworks, but definitely damage to the PC board.

Also, in gear that is stored for a couple years, you often have a Tantalum fail when it is finally turned on again.

I never design tantalum caps in, but have had good luck with OxiCaps, which are guaranteed to not burn. If you put them in backwards, they do get hot, but no flames and no serious board damage. But, they are only available up to 6.3 V in most capacitance ranges.

Jon

Time for my LDO rant.

Not on all boards. Most microcontroller designs I make use a 2 layer board. The trace impedance is high enough so the downstream capacitors don't screw up the LDO.

:-)

Notice to IC-designers: Please pipe out the compensation node so us circuit designers can reach in. Like it's done on nearly all PWM chips. Why can't LDO designers get that into their heads?

Yup, a non-linearity that reaches into the time domain is plain ugly.

I wish universities would teach that stuff. I mean, what kind of effort would it take? Two minutes of a whole semester?

Careful. Even ye olde TL431 will do a mean Tarantella dance if the load capacitance is within a certain range.