Undervoltage lockout for LM2594

Den mandag den 4. juli 2016 kl. 21.41.14 UTC+2 skrev Phil Hobbs:

itcher around, or the cheapest, but if you use it with a toroid it doesn't cause a lot of EMI problems, and if the output cap has some reasonable amou nt of ESR, it Just Works.

and won't take the full input voltage range. That means that you can't impl ement UVLO with a series zener and a resistor to ground--you need some more parts, e.g. a comparator or a FET and a couple more resistors. Not too bad .

the usual output, and letting it pump its own ground pin negative. Works g reat at moderate voltages, but complicates the UVLO problem. It's irritatin g to have to use 8 or 9 parts in the UVLO when the regulator itself only ne eds 6 (or 4 for the fixed-voltage versions).

d be a win, but due to its 6V abs max, for me it needs enough bandaids that it's back up in the 9-10 part range.

random thought, figure 17 from

pre-biased npn, zener, pull-up, and an LED to limit the on/off voltage

-Lasse

The pre-biased NPN is an idea I hadn't thought of, thanks. In my case I'm m aking a larger negative voltage (about -9V) so the hysteresis could get inc onveniently large. I'm coming in with +18V or so, which means I can't just drag the enable pin up to V_in. That costs an extra resistor.

Using a PNP with its base grounded is one way to transfer the current to Vm inus-land, but the component count keeps creeping up.

There has to be some simple hack for this, but I'm not finding it.

Cheers

Phil Hobbs

The data sheet shows a circuit, fig 15. The enable pin pulls enough current that there's no danger of overvoltage there. Fig 11.

The inverting case looks to need two transistors, a zener, and a quad resistor pack, to do it officially right. Maybe one zener and one NPN and an r-pack, depending on what your limits are.

You might save a lot of parts by using a different switcher chip.

We use a lot of LTM8023s. It's about $10, but it has the inductor and stuff inside, and it's very quiet. The UVLO works with just two resistors. There's an LT3480 inside.

Den mandag den 4. juli 2016 kl. 23.16.41 UTC+2 skrev John Larkin:

witcher around, or the cheapest, but if you use it with a toroid it doesn't cause a lot of EMI problems, and if the output cap has some reasonable amo unt of ESR, it Just Works.

and won't take the full input voltage range. That means that you can't imp lement UVLO with a series zener and a resistor to ground--you need some mor e parts, e.g. a comparator or a FET and a couple more resistors. Not too ba d.

g the usual output, and letting it pump its own ground pin negative. Works great at moderate voltages, but complicates the UVLO problem. It's irritati ng to have to use 8 or 9 parts in the UVLO when the regulator itself only n eeds 6 (or 4 for the fixed-voltage versions).

ld be a win, but due to its 6V abs max, for me it needs enough bandaids tha t it's back up in the 9-10 part range.

that's figure 17, less one resistor

with Vin=18V, Vout=-9V it could exceed the on/off absmax 25V, though wi th full input voltage the transistor should clamp it so maybe not

the big problem is the hysteresis with be equal to the output voltage

-Lasse

What's with the Google Groups post ??? My filters dispatched your original post. ...Jim Thompson

Posting from the hammock via my phone. :)

That's interesting, thanks. I'm not sure it works with a vanilla TL431 due to the minimum current requirement, but I'm using a TLV431 elsewhere in the same board, so something along that line might do.

Cheers

Phil Hobbs

Interesting, thanks. I'll revisit the overvoltage thing. Annoying that you can't just connect it to V_in to turn it off.

The current design has a TLV431 and two resistors instead of the zener, and a PNP current conveyor to get rid of most of the hysteresis. (The TLV431 i s already in the P&P, so I just have to figure out whether there's a combo of existing values that will get close enough.)

I'll get a few to play with, just in case it turns out to be instrument-qui et and not just smps-quiet. ;)

It took me a few tries to get even the LM2594 down to where it could live w ithin a couple of inches of a 20 fA/sqrt(Hz) TIA.

Cheers

Phil Hobbs

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

Hmm. Not guaranteed to work, as V_IL is 0.6V, not TTL's customary 0.8V. Plus tempco of the diodes, so it would be worse at high temperature I suppose.

I rather like the V_OH clamping method.

Also, I don't recall that TL431 is capable of pulling down very far. Uh, it's a Vbe below Vref, isn't it? So 1.8V, so the sim should say 1.1V "on"... (That said, your simulation model is probably better.)

TLV431 would give good enough saturation (and low bias), but would break the V_OH clamp. Maybe use an LED?

Tim

Datasheet says threshold is 1.3V

Tempco of diodes doesn't matter that much.

That's TI's model, not one generated by me.

...Jim Thompson

Just do it this way...

...Jim Thompson

The chip loads the on/off pin, so it's safe with the proper pullup resistor.

That's not a bug!

Ok, if a somewhat soggy ill-defined threshold about 12V is OK then how about this:

If you have a 5V rail that appears quickly during power up then the zener and its resistor can go to reduce BOM.

piglet

If hysteresis is needed to cope with Vin dipping as the converter fires up then here are two simple possibilities:

piglet

Cute, thanks.

Cheers

Phil Hobbs

witcher around, or the cheapest, but if you use it with a toroid it doesn't cause a lot of EMI problems, and if the output cap has some reasonable amo unt of ESR, it Just Works.

and won't take the full input voltage range. That means that you can't imp lement UVLO with a series zener and a resistor to ground--you need some mor e parts, e.g. a comparator or a FET and a couple more resistors. Not too ba d.

g the usual output, and letting it pump its own ground pin negative. Works great at moderate voltages, but complicates the UVLO problem. It's irritati ng to have to use 8 or 9 parts in the UVLO when the regulator itself only n eeds 6 (or 4 for the fixed-voltage versions).

ld be a win, but due to its 6V abs max, for me it needs enough bandaids tha t it's back up in the 9-10 part range.

Figure 20, GND referenced shutdown, is the most adaptable scheme. He does n eed to ensure his load is resistive enough to sink the /ON current in all t he inverting configuration schemes. Then the datasheet limits on the /ON/OF F thresholds are 0.6/2V worst case over temperature. He can use an active l ow POR /RESET type driving the transistor base or an active high RESET driv ing the resistor input to the emitter. Some of these parts like the Diodes Inc APX with built-in delays, hysteresis, and temperature stable thresholds are just a dime in quantity.