another LDO

Somewhat complex stability conditions though:

Best regards, Spehro Pefhany

You don't get it. You don't know how a chip will start up. All you know, that is all that is guaranteed to work, are the electrical specifications with min and max limits. That is it. Anything else can change in the future. Processes drift, masks get replaced, fab equipment replaced, wafer vendors changed, blah blah blah. All that matters in shipping a chip is that it pass the electrical limits. What hack you found today may not be there tomorrow. Once the chip is in production, it is probably tested off shore. The designer isn't looking for changes. The designer could be at a new job, dead, etc. The only reality check is the test program.

The test program has all sorts of wait states in it. Relays are switched, meters need to settle, etc. The tests, especially for a LDO, are all DC, so any dynamic behavior isn't monitored. If you don't spec it, you don't test it.

Here is how undervoltage lockout generally works. You use something in the process to hold a master control signal down. This could be an epifet, or the DS connection of a mosfet with the gate going to the supply rail, with the assumption that nothing happens until at least the supply voltage is at least VTH. The pull down could be a very weak Nfet, with the pull up being a somewhat stronger Pfet. There are many simple lockout schems.

You might want to insure the bandgap is awake. They don't turn on instantly. This could be as simple as insuring the bandgap output exceeds a VBE. Generally at that point the bandgap voltage will rise and settle, but it won't likely cause the LDO to overshoot.

Look at it this way. Think of how many power amplifiers make thumps when you turn them on or off. This is clearly due to some asshole designer not considering start up or shut down.

The COTS LDO works as a system. Start up and shut down are tested like the shit will hit the fan if they glitch. You can't chance zapping the circuitry following the LDO. Electrical overstress is quite easy to verify, and if your chip is the only thing between the raw power and the circuitry, it is pretty obvious who blew stuff up.

I don't think there is a chip company around where the QA VP isn't at the same level as the design engineering VP. You have stuff blowing up in the field, you can expect to be getting shit from multiple directions. Parts failing in the field are worse than losing a design-in. You lose future business.

In volume, most LDOs are about a quarter. At cell phone volume, maybe a dime. You don't buy enough parts I guess.

If you can do UVL, reverse power, thermal shutdown, current limiting and voltage regulation for less, be my guest.

Regarding stability, as I've said many times, use the P-mos pass devices. For any project I build for my own use, that is what I do. Generally TI parts. Cheap, stable, and well stocked.

You are arguing that I can't design anything at all using ICs. Or that I can't design a voltage regulator that will start up. But I do it all the time.

Or maybe you are arguing something else that doesn't make sense to me.

output into the error amplifier:

Seems like the ideal error amp for that is just a gain of maybe 500, wideband, uncompensated. The dominant pole is the load.

output into the error amplifier:

Close, but no c-e-e-e-gar. Sounds like you have been reading P-E-T Magazine but not understanding the subtleties ;-) ...Jim Thompson

You have no sense, so why are you surprised ?:-) ...Jim Thompson

So the question remains, what does that make Joerg? Because, he's stated on numerous occasions that they *don't* work. That's by his standards, which as a potential customer, are the only standards that matter.

Apparently, products he has been involved in number in the millions, so if someone actually made a product that met even just his requirements (let alone anyone else's, which likely would), it would be worthwhile and cost-effective to do so.

So either there is a fundamental lack of communication between customers like Joerg and the manufacturers, or the manufacturers really don't give a shit (which you deny as above), or Joerg really isn't a customer at all, and somehow doesn't matter (which seems somewhat insulting, and unlikely given the quantities). So which one is it?

Tim

Years ago when I was just a kid at 13 I think It was, when I really found some one that knew his shit. My mentor, a retired navy radio technician and later on became a EE for some one, told me the best design is one that includes all the crutch patches. You can remove what is not needed, but the design is made to accept it if so.

He was mostly referring to making boards with bypasses, phase error correction networks. etc. To him, it was a good thing to design in all possible parasitic events, just in case.

I suppose the same can happen with IC's from one vender to another. I have worked with IC's like the old 723's that had serious issues, depending on who made them and which lot.. Problems that can be worked out. It just makes it kind of hard to fix in production if you never made any previsions for it.

Jamie

At 10V you'd have to get above 0.01uF gate capacitance for that to happen, and then there is still the option of a phase lead cap. Of course that would add one cent back into the cost sheet :-)

See figure 24, the circuit I meant is pretty standard except we'd use a depletion mode device here for the desired LDO performance.

Exactamente. This has happened with LDOs from reputable manufacturers. Where I found out stuff they clearly did not know could happen. So far for testing rigor.

You bet it is. Plus being able to supply large production quantities on time, which some companies are unable to do.

Usually not millions but there are some that have production volumes in the four digits per month.

Easy: I generally roll my own LDOs and all is well :-)

The design of those "exotic" IC's is driven by nutty customers (and marketing promising anything to make a sale) insisting on no external components, even though performance would be dramatically better. So you get LDO's which depend on ESR for stability. That's plain crazy. ...Jim Thompson

I think some designers are getting fixated on using IC's for the simplest of circuits.

IC's are a wondering component, it has save lots of money etc how ever, IC's also come with lots of problems on top of that.

If there is a need for the IC over transistors etc., then I see no problem working out the issues but, when all you need is a simple average gain loop that only requires marginal components which can occupy a very small foot print, why not use them? They most likely will be carrying less parasitic baggage with them over IC's

I always try to solve with simple low count components before jumping to the IC

SOme of these IC's are so exotic they may only be in production for a few months. Building around them kind of makes the whole product EOL when that happens. But, if you are into the chinese way of disposable products, then it is an attractive scene.

Just a thought.

Jamie

Yup. It's the same in other areas. Peopel pop a pill instead of trying to see what sort of exercise or diet can help with a particular ailment. What drives me almost mad is when doctors prescribe antibiotics for every little flatulence that won't come out.

Other can't cook anything from raw materials anymore. "What? You make pizza dough from scratch??". Heck, we even bake it over charcoal or wood fire in the Weber.

But with engineers this leads to a major problem for our country: We have a total lack of engineers who can design discrete circuits on the transistor level. So far we can import them from places like Russia but it's only a matter of time until that supply dries up as well. Plus there's a H1B-phobia going on.

Usually also less purchasing risks.

Old rule here: I try never to use any ICs starting with TAA, TBA, TCA, TDA and the like. Because many are for consumer electronics and when a key account stops using them they can go unobtanium in no time.

Yep. They want everything on one chip... driven by consumers who want everything smaller and smaller and...

I was recently involved with a chip with ELEVEN power domains :-(

I think there's a complete lack of engineers who studied actual circuit design... every degree is now in "Computer Science and Underwater Basket Weaving".

This group is a hang-out for pompous prima donna hackers ;-)

On a sunny day (Fri, 01 Jun 2012 09:56:24 -0700) it happened Joerg wrote in :

But some are very very good and have no alternative,

One possibility is that you design your LDOs knowing what the input, output, loads, and caps will be, like I did with the dual LDO that I posted. Its arguable that mine's not an ideal general-purpose regulator, but it doesn't have to be.

Commercial chips are spec'd (and often not well spec'd) to operate over a huge envelope of voltages, loads, capacitances, and ESRs. And they have current limits, thermal limits, ESD parts, all sorts of extra stuff to go wrong. And, once it's in production it has masks and test sets and all that, all hard to fix if there are bugs. I can write an ECO to change a resistor or a cap or an opamp in about 10 minutes, if I have to.

At our design reviews, it's common to close out an issue with "well, it's just some parts values here. We can fix that if we have a problem". So we concentrate harder on things that might require a board spin.

Even a board spin is pretty easy for us, far simpler than making a new mask set. We do PCB revs to fix minor manufacturing issues.

I like LDOs that are designed to work with lots of low ESR ceramics. No problems so far.

I know, they are tempting. But for me no dice. Back when I was living in Europe they did bring me the occasional design-out job, just like Maxim chips do 8-)

And move Rb from gate to output. Then the TL431 runs at almost constant current, and you get the Idss current limiting.