another LDO

is plain

dback loop.

ty. I

low

ing

is

So how much warmer is the inside of the box than ambient? If you are dissipating any power in a closed box it's got to be warmer than ambient to get rid of the heat.

-- Bill Sloman, Nijmegen

There are some good ones now, like LP2985, that are stable with low-ESR ceramic caps. The 2985 data sheet is quite good, unlike older LDO data sheets than tended to hide the warts.

plain

feedback loop.

I haven't built it yet, but I'd expect the pcb surface temp to be around 15C above ambient. So the actual operating board temp might be

Once I get the references and resistors under control, the next problem is the base current/beta variation vs temp in the PNP transistors current sources. A BCX71K is beta-graded, 380 to 630, so that's not too bad. We could software futz the cal dacs to zap the median beta-driven TC error. Mosfets have no such error but tend to have a bunch of drain capacitance compared to bipolars.

I've played with tricky circuits to actually sense base current and compensate, but that's a lot of work.

Speaking of which, I just kicked one of those out of a design I reviewed.

I don't like phrases such as, quote "It is critical that both the minimum capacitance and ESR requirement be met over the entire operating temperature range".

On one of its older siblings (could have been the LM2931 but it's a bit long ago now) I had it go into serious oscillation and heat up like crazy if the source impedance of the supply became too high. Nowhere was this mentioned in the datasheet. Hence I prefer to stay away from this sort of architecture where possible. So far it was always possible.

How about we both agree to proceed henceforth thru life as gentlemen?

Then I could show you some tricks for easy beta compensation.

What kind of tightness do you need? A compensation targeted at specific beta grading can be easily calculated. ...Jim Thompson

My first exposure to Boston traffic was in one of these...

My buddy from Minneapolis was fond of driving down sidewalks at high speed :-( ...Jim Thompson

After you apologize for insulting my wife. No gentleman would ever do something so coarse.

You know, I figured that would be how you responded.

I didn't insult your wife, I insulted you. You're coarse, and should be treated thusly.

Accept my offer or not, I don't give a damn.

If you insist it be a war, it will be, and you'll lose. ...Jim Thompson

You don't know how the discrete chips will start up. That is the problem with rolling your own design. Worse yet, even if the design works at present, you have to hope none of the individual components are modified in future designs in a manner that might effect your start up. COTS LDO are tested as a system. They are sold by the millions.

LDO are designed to power circuits comprised of chips that have a reasonable operating voltage range. When it gets to the point that the COTS LDO accuracy is not sufficient, I claim the task is something that doesn't require a LDO. If the parts downstream are that sensitive to voltage, I claim the design is not very good. That is, maybe you need a buffered reference rather than a LDO. For one thing, LDOs only source current. For precision, you probably need source/sink capability.

Dammit, don't make me agree with you.

Less is more when it comes to the error amp in a LDO.

plain

loop.

Ever thought about using a P-channel for your LDO ? Take a look at this. It's a 1. 2 amp pulse load, the diode is there to fix the current source idiotic behavior. I attached the file, Don't know if it will work though.

Jamie

Of course I do.

That is the problem

That's what I do, design circuits.

Worse yet, even if the design works at

Amd some still do weird things.

I have no idea who you are responding to, but if it's me, all I can say is that I know what I need and you are mostly wrong.

I have a simple and elegant circuit that needs a very stable +10 volt supply, around 50 mA, and I have +12 available to power it.

Of course you do. I keep pronouncing you to be an asshole, and you keep trying to make cozy. Who woulda thunk you had feelings?

But you call me "coarse"? Make up your own insults, you moron, and quit copying mine.

Lose what? The "brotherly love" of another redneck geezer who can't remember high school physics? I don't think so.

I was only aware of the supertex ones before, but there are these others too. Digikey has a category for them, some look like they will be ~10c each, a genuine jellybean part. They deserve to be used more :)

I hope they continue to be widely available. There are a few applications (particularly current limiting and processs current loop control) where they make life a lot simpler, and are almost enabling technology. Those current mirrors looked pretty good too, but they're no more. 8-(

Best regards, Spehro Pefhany

They make great high-voltage current limiters. To protect the front end of an instrumentation amp, for example. Or in series with an LED, across a power supply output.

The little SOT23s, like LND250, are great for signal-level stuff. Idss is around 1.5 mA and Ron around 1K.

How to easily make mirrors beta-insensitive :-) ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

plain

loop.

Seems to work.

Using a pfet does require an extra inverter if one is to keep a conventional inverting error amp. Of course, a non-inverting amp can be made to work... it's just a tad harder to think about.

I used a nfet partly because I needed two regulators, and we have the cool Zetex dual fet in stock. We try to keep the BOM short.

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

Dang, for a moment there I thought we were going to have (partial) peace on SED. Nice thread though... just catching up after the Memorial Day hiatus.

George H.