LM317L 20% out of spec

Seem to remember seeing something like with not enough minimum load

-Lasse

Yep, You must draw much more than the ADJ pin current spec with your divider, otherwise "weirdness" ...Jim Thompson

Never, and it's odd. The data sheet

does say 1.20V to 1.30V. Lasse suggests that you might not have been drawing the 10mA minimum output current

but everybody designs their voltage-setting resistive divider to soak up that 10mA.

Are you sure you don't have something backing up to the rail from elsewhere? THe 317 is a sourcing device the last time I looked, it won't sink..

Jamie

Well, if they're thinking. I'll admit that I've failed at that rather basic test on occasion, and I've debugged circuits for folks that did the same thing.

The ADJ pin current and min load are not really related. ADJ is only

Which just means that even if you have a guaranteed minimum load you still need to pay attention to the ADJ current.

You missed my point. If your divider, by itself, pulls only 200uA, the output voltage will be...

Assuming the reference potential stays put. ...Jim Thompson

Well, sure, at least you have to calculate that into the resistor ratio.

So, guys, be honest, who else has forgotten the min load in the heat of the game at least once?

It's not like a gear-up landing but it sure is something where you'd think "I hope nobody saw that".

up

ther

he

My wife studied speech errors at one stage in her career, and I got exposed to some of the literature on "errors of action".

Basically it's typo's - you intend to do something and actually do something else. Sometimes you omit a planned action (omission), sometimes you do it twice (perseveration) and sometimes you do something familiar but inappropriate (deviation).

Programmers typically make one mistake per thirty lines of code, and people drawing circuit diagrams and making up parts lists get stuff wrong pretty regularly, though - sadly - not at predictable intervals.

Lots of engineering is simply checking for errors and correcting them. Programmers have walk-throughs. We have design reviews.

I always use a 240 ohm resistor from Vout to Adj, which never fails to use up the required current, regardless of the output voltage, but hey, s**t happens, and I've similar things from time to time.

Best regards, Spehro Pefhany

On a sunny day (Thu, 04 Oct 2012 16:31:28 -0400) it happened Phil Hobbs wrote in :

No

The last time I saw this effect (on a LD1117S12TR) my FPGA was getting

Michael Kellett

Yup, pilot error and not monsters under the bed --I was just under the

Cheers

Phil Hobbs

This board is full of MSOPs and SC70s and 0402s, which is _not_ my natural habitat, so debugging it is irritatingly slow. It's that 1 nA/100 MHz biochip front end gizmo with the Really Triangular Triangle calibrator. The layout guy did a nice job, especially an array of 39 pF

So far I found a voltage reference oscillating because its bypass was too small. Fussy modern thing, the LTC6655--a 1N823 or LM329 wouldn't do that. ;) It seems to have an LDO-type architecture, with a high-Z output that's sensitive to the bypass value. (Of course in a precision voltage reference, I'd far rather have that than an internal charge pump.)

Cheers

Phil Hobbs

AFAIR they need somthing big (>3uF?). Comes in 0402 even with 6.3V rating. But ... no X7R, yet. So you have to take capacitance derating into consideration, to make sure things aren't "riding on the edge". Also, mind the microphonics and piezo effects if this is a super-low-noise app. Then ceramic caps may not be so cool.

I stuck in a 10 uF 0603 X5R, which seems to be okay. Since the shot noise limit is equal to half the number of electrons you collect in one inverse bandwidth

SNR_max = n/(2*B)

the shot noise limited SNR of 1 nA in 1 MHz is

SNR_max = (6.2E9 electrons/s)/(2.0E8 Hz) = 31.

That's not 31 dB, it's 31, i.e. 15 dB.

I don't know how close I got, yet. We shall see, assuming the rate of beginner mistakes drops to something reasonable. ;(

Cheers

Phil Hobbs

LDOs often has a requirement for both the minimum and maximum ESR of the ceramic cap to be unconditionally stable

Cheers

Klaus