another LDO

down

I have used opamp-servoed 1117s to power load cells, and I did some serious-abuse tests to make sure I wouldn't fry them. I wasn't able to find a case where abuse of the ADJ pin hurt the chips. Nice part.

Ok, if and extra two or three bucks don't matter then depletion mode can be your friend. Makes life much simpler in many situations.

Most every thing we do is low count. We only make devices for our own (my employer) business operations, which are world wide. But R&D is done to design new products and new ways of making the product.

After the chemist are done doing their work and our group has satisfied their needs, then it's on to the production area. In this area, machines need to be built, modified and or new devices made to make it work in a cost effective way.

Jamie

Indeed- the older LM317 did require a diode (adj to Vout) when you bypassed the ADJ pin to ground. The LM1117 does not.

It's exactly the same situation every time you power down with a light load and ADJ bypassed with a largish cap as driving the adj pin with an op-amp above Vout. Note that the older LM117/317 incorporated a 6V zener and series 50R resistor, which would break down well before the ~9V breakdown of a reverse-biased NPN BE junction. Maybe they increased the 50R in the LM1117-- anyone have one handy to curve trace?

Best regards, Spehro Pefhany

Not having looked at the datasheet... is that path by any chance a lateral / PNP junction? Those typically go for 30V rather than 5.

Tim

work down

especially

occasionally

I always liked the idea of the LM10, and I have a dozen or so in a drawer someplace, but I never used it for anything in real life. Multiplying up a 200 mV reference is fairly silly from a noise and drift POV--unless you need it to run at very low VCC, there are lots better parts available, cheaper, from multiple sources. (Using a vreg chip as a bulletproof series pass device is even older than the LM10, but it works great.)

Cheers

Phil Hobbs

It's NPN devices. There are resistors up front so BE-zenering will not be catastrophic. However, typically this changes performance parameters unless they have some magic tricks in there. Such change are usually irrecoverable except probably via annealing, which is kind of tough to do on a plastic package. Plus smoking has been banned for most places, so no more extinguishing cigarette butts on TO cans :-)

down

Maybe they do have some magic trick in there. If not BE-zenering can change parameters over time that may or may not matter to you.

You will need to prevent the output from overshooting if you use the depletion device since it is naturally in an ON state. This seems like a bad scheme to me.

Here is what you get with a well designed COTS LDO.

The device will behave under all input voltage conditions up to the maximum input voltage. That is, the design is either inherently controlled in low voltage conditions (i.e. too low to operate as a regulator) or it has an undervoltage lockout.

The device has some scheme to determine when the internal reference is ready, and/or the reference works in a manner where it won't cause the output to overshoot under odd input start up conditions.

The device has some sort of current limiting. It may not be specified in the datasheet. Much protection circuitry isn't specd since nobody wasn't to lose yield on the parameter. [If you state the limit, you have to test it.]

The device has thermal limiting. This is important since combined with current limiting, the prevents the part from frying due to the fat fingered engineer or tech probing around with a scope probe and causing a short.

The device has been evaluated under a range of input start up conditions. This includes bat out of hell ramping of the input, often using custom built pulsers, and dog slow input ramps of 100ms. Of course, the part can be held at any DC input and be well behaved.

Given that LDOs often follow switchers, the DUT input voltage will be DC with swept AC, insuring a known PSRR.

The LDOs with P-fet pass devices are quite stable. The PNP pass is plain ugly due to the sat-caching schemes which add an additional feedback loop.

Why would it overshoot? If the opamp holds the gate down, the source will only go a volt or two above the gate. It's a source follower. It will ramp up to 10 volts. Even if it did overshoot to 12 volts, which it won't, no harm would be done.

As I noted, I need very low TC and really good long-term stability. I can't get that from any standard LDOs that I know of.

p.

But you aren't enough of an engineer to put a number on the "very low TC" that you need, nor the "really very good long-term stability", even though representative numbers can be mined from data sheets.

-- Bill Sloman, Nijmegen

Your mind is becoming increasingly feeble through disuse, and maybe too much wine. Just three posts up in this sub-thread I said

"A commercial LDO won't be as accurate as I need; I'd like to hold a few hundred PPM over time and temperature, and tracking the DAC reference is good here, too..."

Specifically, I need five or six current sources to charge caps to make linear ramps at better than 10-bit accuracy, so a few hundred PPM, or better, for the reference is all I'd like to give up. There are lots of other sources of error here, and a 10 volt power supply is easy, low-hanging fruit in my error budget.

Have I noted lately that you are an idiot? For roughly the same reasons that JT and JF are idiots.

Moron. ADR44x parts are voltage references, not 10 volt regulators. We use them already, and I showed an ADR363 in my sketched LDO schematics.

Maybe I should use a string of Susumu thinfilms as the feedback voltage divider. They would match to a few ppm. I could use four resistors and a 2.5 volt reference, or maybe three and a 3 volt ref to get 9 volts.

ain

loop.

It's not my mind that's feeble.

How much temperature and how much time? "A few hundred PPM" over an unspecified time and temperature range isn't any kind of numerical specification, even if it has a vague number in it.

So what's your temperature range? Air-conditioned lab, or arctic shed?

Because we haven't noticed that you are an intellectual giant?

What kind of moron would think it necessary to remind me that the ADR44x parts are voltage references, rather than voltage regulators. I used them as an example of parts with specified temperature and time stability.

Why not. The rest of us have been using them for years. I got enthusiastic about them back in 1979, but the nitwit who took over the project didn't notice that the resistances tracked, and specified the discrete resistors that he'd always used before.

-- Bill Sloman, Nijmegen

How about the commercial temperature range and a reasonable calibration interval? I don't suppose you are familiar with numbers like that.

Because you use your emotions to supress your intelligence. Because you are more interested in scoring ego points than designing electronics. The result is not being very good at either.

Wait: you whine that I didn't specify time or temperature stability, then brag that "I used them as an example of parts with specified temperature and time stability."

I need a word stronger than "idiot" to address that.

Or, not using them for years in your case.

Go away!

Jamie

Of course not. Has Larkin ever really _specified_ sufficiently enough that his claims can be checked?

And Larkin has no clue about loop dynamics. He can't cope with analyzing a 3-pole+delay feedback... excuse me, for simpletons like Larkin, let's fall back to the simpleton concept of OpAmps being two poles... so we have a 4-pole loop, two from the OpAmp, one from Cgs of the NMOS, and one from the NMOS in its resistive region working against the filter caps.

You can insert a zero... cancel one pole.

Add feed-forward... cancel another pole.

You're left with two pole system that hunts (small amplitude due to the large filter caps... which he can't see on a scope in his noisy environment).

(Nevermind that the resistive pole moves with load.)

Larkin doesn't want to actually brainstorm electronics, he just wants to appear the prima donna.

He will respond shortly with more "Tourette's by keyboard" ;-) ...Jim Thompson

Why does it hunt? Your sim shows few-microsecond-wide rectangular pulses with vertical edges, appearing across over 20 uF of capacitance. What does the gate waveform look like? What is the nonlinear mechanism that keeps your pulses so small? My sims show nothing like that, even if I crank the time steps down to 1 ns, or go to zero ESR on the caps.

Not that those tiny spikes would matter. My regulators work fine. But the spikes look like model/sim artifacts to me.

Sure. That's common enough in source/emitter follower regulators, or even in p-types at low drop. How would you avoid that with 300 millivolts across a mosfet? Not easily. But my poles don't move much because my load is pretty constant.

You sure brag a lot without showing any real work or actually explaining anything. You even claim that waveforms of undisclosed origin are "Patent Pending" which is ignorant on several levels, and illegal besides.

I don't think you have ever patented anythying yourself. All your patents seem to be assigned to corporations. You seem to not understand the patent process.

That's just childish. At that kind of headroom and load current, a cheap bipolar transistor will outperform a depletion mode FET.

plain

ck loop.

Bit of a waste if they are going to be stuck in a rack in an air- conditioned test lab - but if you are prepared to let us to guess your operating conditions for you, you really don't have a clue.

But it's going to go in rack, with other electronics, which may make the immediate environment a little warmer.

Really? It's your emotions that are perceptible here - you seem to be excited enough to have forgotten how to spell "suppress".

I'm scoring ego points by posting here? I do that in peer-reviewed journals ... You seem to be the one that carries on about designing "insanely good" electronics at the rate of one world-shattering invention every two weeks. The rest of us - with the possible exception of Jim Thompson - seem to have a firmer grasp of reality.

...

Competent? You did - and still do - seem to need an example of a stability specification that actually species something. Pulling ne off the web made the point a little faster - and with less typing - than concocting an example for you.

I've got a couple ordered from Farnell for the low distortion oscillator. They should have arrived today ...

-- Bill Sloman, Nijmegen