high voltage LDO

Figure 5 of the datasheet is enough to generate considerable heartburn

...Jim Thompson

I tested one, and it works fine all the way up to 30 volts, where the OVP thing cuts it off. It does NOT work much below +5 out.

Here's a homebrew LDO:

The nfets are seriously ohmic here, which makes math analysis not worth the effort. It Spiced, and worked, fine.

I breadboarded it. It seems happy with a 220 uF aluminum output cap and a bunch of paralleled ceramics. That combination, an alum or a tantalum plus the various ceramic bypasses, is usually stable, and has good transient holdup.

One can hope the 30V cutoff is a predictable value. I imagine it's a kind of foldback limit set by resistors ratios and with the aid of some series 7V zener Veb breakdown. If so it may reliably stay far away from say 25V in. If not ...

If you have a lot of bypass caps (something vaguely comparable to the electro's value) and not a lot of series resistance, you could get into trouble, I suppose. A bigger electro will fix that, because the closed-loop BW drops like 1/sqrt(C), and its ESR is less as well, so it gets a lot harder for the ceramics to cause enough phase shift across ESR to destabilize the loop.

The sheet resistance of 1-oz Cu is about 0.5 ohms per square at 25C, so if the minimum ESR is 30 milliohms it doesn't take too much trace length to get you out of the danger zone.

It isn't just LDOs that have this problem--even my fave LM2594 switcher will oscillate with 20 uF ceramic on its output. 10 uF ceramic and 68 uF aluminum works fine, even with very little trace length.

Cheers

Phil Hobbs

milli

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

Den tirsdag den 20. september 2016 kl. 01.39.52 UTC+2 skrev John Larkin:

What's the transformer?

There's no transformer. The funny looking TP things are PCB-foil current shunts, for measuring load currents. I usually use 0.1 ohm resistors, but I didn't want the voltage drop here.

Note the Italian opamp.

No, I was referring to T1 in the T900 pdf circuit schematic. It's a 12-pin six-coil transformer.

Oh, different schematic. That's a custom pot core design.

(I wound that prototype myself)

Nothing remarkable. I recently redesigned it; the original was bifalar with litz wire, which was overkill. The new version is just #28 solid Beldsol, wound any which way.

Allow me to plug Newava, the outfit making the next batch of 300 transformers for us. They do magnetics and cable assemblies and are great to work with.

Nice. Too bad they don't have a larger portfolio of standard parts.

It's a shutdown feature to protect the internals from load dump. The origin al NS datasheet has two graphs "output voltage at the extremes" for Vout at 5V and 15V with 100R loading both cases, Vout vs input voltage, both cases have very sharp Vout nose dive to 0V at exactly 30V. They also have in the tabulated specification, Maximum Operational Input Voltage, Typ is 31V at room temp, with 26V guaranteed over "Full Operating Temperature Range", for the T, S, J, WG and LD part suffixes, there is no such full temp range gua rantee for the CT and CS suffixes, just a 26V at room temp. So, assuming th e Vin=Vout + 5V does not make a difference, he should be safe with a 24V+ /-5% input.

I can only speculate that Vout>5 is an internal error amp feedback stabilit y requirement. As for the Vout, max at 20V I am now certain that has to do with the original NS datasheet specifications. All the wonderful regulation , output impedance, noise, ripple rejection, and stability, specs that powe r supply designers look for, are only good for Vin= Vout + 5V. So that wi ll naturally limit Vout to 20V to keep Vin with the maximum input voltage o perational region

Something like this...

...Jim Thompson

That's nice. A TL431 regulator mandates another inverter stage, and two transistors serve well. They can be in a single sot-23-6 package, e.g., types as shown in AoE-III, Table 8.1b, page 502. Maybe a DMMT3906 or a BCM857DS.

What happens to the stability of your circuit with a humongous output capacitor load?

Yeah, it is cute but I figure you don't even need the two transistor mirror - a single pnp would do the job? Jim is a chip designer so is just so habituated to tossing zillions of current mirrors around :)

piglet

Humongous is good...

Below that you run out of phase margin as the Cgs of the FET snags you. ...Jim Thompson

Smoke and (habituated) mirrors are cheap ;-) A single PNP provides too much loop gain... we already have more gain than we really need. ...Jim Thompson