ZXTR2112F-7 "regulator transistor": Any traps for young players?

[AnasMalas]

Hello!
Summary: I want to use the ZXTR2112F-7 (https://www.diodes.com/assets/Datasheets/ZXTR2112F.pdf) which is a very simple regulator. Are there traps that someone inexperienced could fall in?

Information and steps taken
In a design I am currently working on, I need two voltage rails: one "12V" (for nmos gate drive), and one 3.3V. The input voltage range is 6-30V, with the understanding that at 6V input the "12V" rail may be an unregulated voltage of 5V depending on the regulator. This affects the MOSFET Rds(on), but thats taken into consideration and a "low performance" mode is activated. The intention there is to allow the use of 2 to 7 cell Li-Po battery to power the module (2*3V to 7*4.2V).

The only load on the "12V" rail is a gate driver (MIC4606), and thus the gates (4x 109 nC at 10V, driven at 25kHz). I used the driver's datasheet to try to find the absolute worst case current consumption which came up to 8.5 mA, with 6 mA being how much I anticipate drawing at the way I want to run the gates. So say 0.15W power dissipation.

My issue was that the adjustable regulator I ordered from digikey was lost, courtesy of our great customs inspection. Shipping new parts is prohibitively expensive, so my only option now is to order it along with the PCBA. ZXTR2112F-7 seems to be my best option from JLCPCB.

I expect the leads of this regulator to be subjected to a maximum of 80C (its thermal path is in parallel with a 6W MOSFET), so with a junction to lead thermal resistance of 197 C/W (and excluding other paths), lets say the regulator's junction may get to 110C. As the temperature range is upto 150C, its not ideal but I think it might be safe.

This regulator loses regulation at 15V, under that they say that "VOUT maximum = VIN – 1V", but no minimum is provided.

Have I done my homework correctly, or am I in for a rude surprise? I realize there are probably a thousand considerations I dont know about, for one, I have no idea if this has good enough regulation for my project. I dont see why this would have a maximum capacitance, but what if it does?

I will use a second ZXTR2112F-7 to be connected to a 3.3V regulator. The 3.3V rail includes a microcontroller and sensors, a rough calculation tells me that rail will draw somewhere around 1-3 mA max. Here's a dumb idea: why dont I put both in parallel with series resistors to ensure they balance themselves out? would that be safer temperature wise?

[T3sl4co1l]

Don't forget the gate charge: 4 * 109 nC * 0.025 MHz = 10.9 mA.  It will be less at low voltage, but high input is worst case, so that doesn't matter.

The ZXTR thing is literally what it says, a zener and transistor; if you're having trouble getting it, why not use the components instead?  You can then use a larger e.g. SOT-89, SOT-223, etc. part for higher dissipation.

Likewise, the output voltage tolerance will be gross, too much to use parallel ones with sharing resistors.

If you need lower dropout, consider using a depletion MOS such as CPC3701CTR, BSP149H6327XTSA1, DN1509N8-G, etc.  Same circuit but less zener current is required (no gate current!).  Mind the looser tolerance (Vgs will likely be in the -2 to 0V range, versus the BJT follower's ~0.2V range, with respect to load current).

It's also not a bad argument to add a small SEPIC, and get stable control/drive voltages even at low input; sounds like it's maybe on the margin whether I would bother to, personally.  Also, a multi-winding inductor could potentially combine the 3.3V output with it, from a single switching regulator.

Tim

[AnasMalas]

Hello Tim!

I did include gate charge, but only for two of the MOSFETs with the thinking being that I won't use lock anti phase drive (where all 4 are always switched), instead I plan on using sign magnitude drive, in which only one is switched at a time, and I may alternate the bottom and the top switching to even out the thermal dissipation. But youre right, its not quite the worst case if I dont consider that I may code it differently...

why not use the components instead?

I 98% understand how to pick out an enhancement MOSFET, I know the gotchas, and it took me a while to get to this stage. The zener should be a piece of cake, but a depletion MOSFET or a BJT I dont quite know how to work with correctly (you can say I dont know the traps for young players, lol ).

You're entirely right, I didnt find a good online resource, but I just saw that The Art Of Electronics explains this well. Depending on my confidence I may just build a regulator to go straight into the gate driver, or I may step down to something like 15V and use a second regulator to get me to 10 or 12V.

As for the SEPIC, I never implemented one. Do you think a SEPIC with all its components would be easy to design and under $1 in 1k qty? I was looking at implementing a synchronous buck converter (synchronous to deal with the low load when on standby) for this project but it just gave me a headache trying to learn its gotchas.

[T3sl4co1l]

Ah, that's what you meant with "how [you] want to run the gates".  So, inverter, class D amp, something like that.

Depletion is no different from enhancement besides Vgs(th) being offset below rather than above zero.  Put a coin cell in series with the gate and you have a fake depletion MOS, it's that trivial.  Well, they can make 'em without having to use an external source of charge, which is rather convenient, so there you go.

BJT, I suppose if you're not familiar with their behavior in general (Ic ~ Ib*hFE ~ exp(Vbe)) it won't be obvious, but on the upside there's only one free variable in the circuit (the bias resistor) and component ratings have to be chosen for operating area regardless so you aren't going to hit a problem there.  Be mindful of SOA, which is necessary for any type, both BJT and MOS exhibit 2nd breakdown.

Since BJT draws base current, that, with maximum load current, sets minimum resistor value at minimum Vin.  Zener and resistor dissipation are then set by Vin(max).  Vout is just whatever's left.

MOSFET doesn't draw gate current so just bias the zener within tolerance and go.  There are families of zeners specified for rated voltage at low currents (and maybe have lower noise in that condition? I'm not sure), which may be helpful here.  Vout is somewhat above Vz (or below for enh. mode), and by how much, depends on Iout.  See the transfer curve (Id vs. Vgs).  Of course with enh. mode, this affects dropout, but depl.'s dropout is limited by Rds(on) only.

SEPIC is no different from buck or boost, you just need a dual winding inductor, an extra bypass cap (across the inductor), and output rectifier(s) for multiple outputs (with more taps on the inductors, or multiple windings wired appropriately e.g. Coilcraft Hexa-Path).  Getting all of those under a buck is probably a challenge from western sources, but China, should be able to find things easily enough.

Tim