Negative LDOs with high PSRR?

[SiliconWizard]

Looks like those are pretty rare and expensive.
I'm looking for something with <= -16V input voltage, -12V output (can be adjustable of course), PSRR >= 65 dB @ 1 kHz, output current max ~200 mA.
Very few or pretty expensive (> $6). Anything that would be more around $2 with these specs? Or any idea?

[magic]

What do you mean by "<= -16V input voltage"?

Literally, -16V or less means -16V or -17V or so on. That's 4V headroom, definitely not requiring an LDO.

It seems that LM337 with ADJ bypass could just meet your 65dB@1kHz spec.

edit
OK, on second reading I guess what you meant is inputs between -12V and -16V.
Well, nevermind then

[spostma]

TPS7A3001 would do IMHO, but costs about 3,50 (25+)
the datasheet says that when regulating at -5V @200mA, PSSR is 68dB @ 1kHz.

https://nl.mouser.com/c/semiconductors/power-management-ics/voltage-regulators-voltage-controllers/ldo-voltage-regulators/?q=TPS7A3001&sort=pricing

[SiliconWizard]

Well, <= -16V literally means what it says. I know negative bounds can look confusing sometimes.
That means the input must be able to handle -16V or lower (negative bound, not in absolute value.)
Which doesn't mean that I can tolerate a large dropout voltage. But I didn't mention it. Typical input would be somewhere between -16V and -13V, so it must handle -16V or lower, but max dropout should not exceed about 1V.

That, and the requirement of the LM337 for a min. load current of 10mA is not going to fit my requirements. Otherwise, albeit old, the LM337 is pretty good in terms of PSRR.

I had seen the TPS7A3001  and while a bit expensive, it should be ok, so it could be a match. Nice specs. If only it was available. Earlier availability is feb 2023, and I don't even trust that.

Note that the LM337 in SMD package is also currently unobtainium as far as I've looked.

Now the LM337 approach for getting a high PSRR tells me that I might be able to achieve the same using some adjustable LDO with lower PSRR and improve it bypassing one feedback resistor?

[magic]

It's better than not bypassing it if you choose an adjustable regulator, but also no guarantee that a well designed fixed regulator couldn't do better.

The benefit is of course that the regulation loop essentially runs unity gain at AC, and whatever gain selected by resistors at DC. More feedback, better performance. You will never get better than the same adjustable regulator set for its minimum voltage, 1.2V or whatever it is.

[SiliconWizard]

Yeah, but bypassing will also make the regulation loop slower, won't it?

Otherwise, for a dual positive/negative supply with good PSRR and low noise, the LT3032 could be nice. It's fricking "expensive" though. (As with most LT and AD parts, the unit price per 1000 that AD shows on their website is nowhere near what is shown through usual distributors, which prices especially on LT parts can be as much as 50% higher or more that the suggested price. Not sure how easy it is to buy from AD direct now? I used to buy direct from LT back when it was still LT.)

[magic]

It makes startup slower because the cap needs to charge. There are potential gotchas when input/output gets shorted and the bypass cap discharges through protection diodes in the regulator.

Otherwise it's faster. The cap is basically a constant voltage, and the regulator runs like a 1.2V regulator regulating with respect to the cap voltage rather than ground. Unity gain.

[mzzj]

Room for capacitance multiplier "preregulator" ?

Quick fooling around on ltspice gives 60db attenuation at 1khz and about 0.5v voltage drop with circuit built from 2SAR574 PNP, 470uF 16V electrolytic and 700 ohm resistor, total BOM <1 USD

[langwadt]

how bad is an LM7912?

[Someone]

how bad is an LM7912?

A *79*12 part can meet the requirements but outside the guaranteed specifications (and is model specific) so it is an option but needs validation testing.

As usual, too many undefined dimensions/requirements to make any real suggestions.

[SiliconWizard]

how bad is an LM7912?

Depends too much on vendor. It's from pretty mediocre to good in terms of PSRR. The SMD variants in smaller packages (thus rather the L versions) tend to be worse, and are only rated for up to 100mA which isn't quite enough. Otherwise you get pretty large packages.

[SiliconWizard]

Room for capacitance multiplier "preregulator" ?

Quick fooling around on ltspice gives 60db attenuation at 1khz and about 0.5v voltage drop with circuit built from 2SAR574 PNP, 470uF 16V electrolytic and 700 ohm resistor, total BOM <1 USD

Something to consider, but area cannot be wasted too much so large capacitors are not that great.

[magic]

Hmm, at -13V input you could afford to use an emitter follower, and place it in the feedback loop of... something?

7912 won't work from -13V, forget it.

[Swainster]

Am I going mad or was Magic correct with his original interpretation of "<=", and SiliconWizard the one with it back to front?

[SiliconWizard]

Am I going mad or was Magic correct with his original interpretation of "<=", and SiliconWizard the one with it back to front?

Uh. If you need some part that can accept down to -16V input, it must handle a minimum of -16V. So <= -16V.
If it were a positive voltage, I would have said >= 16V.
It's basic maths and basic physics.

[langwadt]

Am I going mad or was Magic correct with his original interpretation of "<=", and SiliconWizard the one with it back to front?

there was no difference both were correct, just wasn't clear whether the <=-16V was the available supply or the minimum rating of the regulator

[SiliconWizard]

magic's point was about not necessarily requiring an LDO, and I gave complementary info afterwards.

Anyway, even if not strictly requiring a low dropout, that wouldn't help much as the offering is very limited: 29x12 are not a match as I explained above, and the LM337, beside its highish dropout voltage, is unobtainium in SMD packages at the moment (or the larger DPACK stuff). So that's pretty much it.

Good matches have been mentioned: TPS7A3001, LT3032 (I could use the positive regulator in it to replace one other regulator in my design), but they are either unobtainium or pretty expensive. Now given the LT3032 would be two regulators, its price is not that awful though, so that might be my pick. Still a bit pricy. The LT1964 could also be a match, but it's expensive and its PSRR is not that great. Now I could add some pre-filtering to improve the PSRR if I find a good match otherwise, as there aren't many at all anyway.

There are tons of matches for positive LDOs with similar characteristics (and even more with lower positive input voltage, in terms of PSRR and low noise), but extremely few for negative LDOs. I guess there is little market. And the current state of shortage doesn't help.

[spostma]

Maybe you could make a discrete regulator with a LM3411 and a transistor?
The LM3411 is a kind of 'negative TL431' with reference tied to V+.

[spostma]

or a LM4051 (adjustable 3-pin negative shunt regulator)

[nctnico]

Am I going mad or was Magic correct with his original interpretation of "<=", and SiliconWizard the one with it back to front?

Uh. If you need some part that can accept down to -16V input, it must handle a minimum of -16V. So <= -16V.
If it were a positive voltage, I would have said >= 16V.
It's basic maths and basic physics.

IMHO the wrong way around. For positive voltage you'd say supply voltage is up to 16V. So the regulator would need to be able to deal with +16V or less (<= +16V). Or Vmax = 16V. Same for negative. Down to -16V so -16V or higher ( >= -16V). Or Vmin = -16V. OTOH <=-16V means -16V or less (-20V? -100V?). How much less is unclear in that case which has thrown quite a few people off including me.

Anyway, it looks like component shortage is causing you trouble finding a good part.

LCSC seems to have some LM337 in stock but not many and quite expensive. I ordered parts from them before and received real parts so at least they are legit.
https://www.lcsc.com/product-detail/Linear-Voltage-Regulators-LDO_Texas-Instruments-LM337KVURG3_C2871286.html

[langwadt]

Am I going mad or was Magic correct with his original interpretation of "<=", and SiliconWizard the one with it back to front?

Uh. If you need some part that can accept down to -16V input, it must handle a minimum of -16V. So <= -16V.
If it were a positive voltage, I would have said >= 16V.
It's basic maths and basic physics.

IMHO the wrong way around. For positive voltage you'd say supply voltage is up to 16V. So the regulator would need to be able to deal with +16V or less (<= +16V). Or Vmax = 16V. Same for negative. Down to -16V so -16V or higher ( >= -16V). Or Vmin = -16V. OTOH <=-16V means -16V or less (-20V? -100V?). How much less is unclear in that case which has thrown quite a few people off including me.

is the right way around, just some confusion on whether it was max rating for the regulator or supply

for negative you need a regulator that can handle <= -16V when running on supply that is >= -16V

for positive you need a regulator that can handle >= 16V when running on supply that is <= 16V

[SiliconWizard]

Am I going mad or was Magic correct with his original interpretation of "<=", and SiliconWizard the one with it back to front?

Uh. If you need some part that can accept down to -16V input, it must handle a minimum of -16V. So <= -16V.
If it were a positive voltage, I would have said >= 16V.
It's basic maths and basic physics.

IMHO the wrong way around. For positive voltage you'd say supply voltage is up to 16V. So the regulator would need to be able to deal with +16V or less (<= +16V). Or Vmax = 16V. Same for negative. Down to -16V so -16V or higher ( >= -16V). Or Vmin = -16V. OTOH <=-16V means -16V or less (-20V? -100V?). How much less is unclear in that case which has thrown quite a few people off including me.

is the right way around, just some confusion on whether it was max rating for the regulator or supply

for negative you need a regulator that can handle <= -16V when running on supply that is >= -16V

for positive you need a regulator that can handle >= 16V when running on supply that is <= 16V

Yep.

[SiliconWizard]

Maybe you could make a discrete regulator with a LM3411 and a transistor?
The LM3411 is a kind of 'negative TL431' with reference tied to V+.

Given the problematic offering, I might go this route of rolling my own. As long as it doesn't take up too much space.

[edavid]

Maybe you could make a discrete regulator with a LM3411 and a transistor?
The LM3411 is a kind of 'negative TL431' with reference tied to V+.

Given the problematic offering, I might go this route of rolling my own. As long as it doesn't take up too much space.

If you have a positive supply, you don't need a negative reference, just an op amp and a PNP.
(Assuming you don't need fancy current limiting/thermal protection, which you shouldn't at 200mA.)

[SiliconWizard]

If you have a positive supply, you don't need a negative reference, just an op amp and a PNP.
(Assuming you don't need fancy current limiting/thermal protection, which you shouldn't at 200mA.)

You mean by using the positive rail as a reference?