Solved: 50nA low-leakage FETs that don't use 'satanic tiny little DFN packages'?

[incf]

For what it's worth, Philips NXP Nexperia has a series of "low-leakage" FETs. The specs aren't particularly impressive (25 nA guaranteed) but, hey, that's something. I seem to recall another manufacturer also has such a line but at the moment I can't remember which one.

Nexperia's catalog lists these parts as PMZ600UNEL / PMZB600UNEL (N-Ch), PMZ950UPEL / PMZB950UPEL (P-Ch), PMDXB600UNEL (Dual N-Ch), PMDXB950UPEL (Dual P-Ch), and PMCXB900UEL (dual complementary).

Oh, and they're in one of those satanic tiny little 1mm by 0.6mm DFN packages. Just in case you need an excuse to buy a JBC Nano station....

I've got a commercial application where would like to use a low leakage P-Channel FET like the PMZ950UPEL, but I "absolutely hate" the package it comes in.

Does anyone make a manufacturer friendly version in a leaded package like SOT23 / SOT323 / SOT523 / etc. (SC-74/SC-70/etc.) ?

All I'm looking for is a P-FET with less than about 50nA of leakage at room temperature with a ~3.3V signal (drain to source voltage) and a ~2.5V gate drive (gate to source drive voltage).

edit: I also feel like tiny leadless packages are not great for leakage, clean-ability, or inspect-ability. A combination of moisture and flux residue under the chip is liable to misbehave in the long term.

[Zero999]

ZVP4525E6
https://docs.rs-online.com/b73b/0900766b814f370d.pdf

[incf]

ZVP4525E6
https://docs.rs-online.com/b73b/0900766b814f370d.pdf

That's interesting, it is rated for 500nA at Vds=250V

Is there any sort of universal law that dictates the scaling of leakage current vs voltage? I know it's definitely not linear. I seem to recall certain datasheets for non-FET parts having a constant leakage current below a certain voltage threshold.

Can we really be certain (or even "reasonably certain") that its leakage current is guaranteed to be 50nA or less at Vds=3.3V?

edit: threads like https://electronics.stackexchange.com/questions/224056/mosfet-drain-to-source-leakage-current-over-voltage-and-temperature make me think a 250V part might leak nearly "just as much" at 3.3V as at 250V
edit 2: chapter 14 leakage of small signal MOSFETs https://assets.nexperia.com/documents/user-manual/Nexperia_document_book_MOSFETGaNFETApplicationHandbook_2020.pdf is interesting


It shows a sample 12V FET having a relatively flat leakage curve.
Not sure if that behavior extends to 250V FETs at 3.3V.

edit 3: the formula for that FET's leakage using that graph at 5V ends up being I=I_Leak * 2.15 * 2 ^ ( delta_T / 10)
given Ileak=0.2nA
and being pretty much independent of voltage

[Zero999]

ZVP4525E6
https://docs.rs-online.com/b73b/0900766b814f370d.pdf

That's interesting, it is rated for 500nA at Vds=250V

Is there any sort of universal law that dictates the scaling of leakage current vs voltage? I know it's definitely not linear. I seem to recall certain datasheets for non-FET parts having a constant leakage current below a certain voltage threshold.

Can we really be certain (or even "reasonably certain") that its leakage current is guaranteed to be 50nA or less at Vds=3.3V?

edit: threads like https://electronics.stackexchange.com/questions/224056/mosfet-drain-to-source-leakage-current-over-voltage-and-temperature make me think a 250V part might leak nearly "just as much" at 3.3V as at 250V
edit 2: chapter 14 leakage of small signal MOSFETs https://assets.nexperia.com/documents/user-manual/Nexperia_document_book_MOSFETGaNFETApplicationHandbook_2020.pdf is interesting

More often than not, manufacturers don't bother to test their parts for leakage. They just slap an arbitrary figure on the data sheet. My advice is to buy a range of, high voltage, low current parts and test them.

[PartialDischarge]

In your case an analog switch may be better suited as they test for lower leakage currents, TMUX1101 is an example, there are parts for higher voltages, etc ,etc

[SiliconWizard]

In your case an analog switch may be better suited as they test for lower leakage currents, TMUX1101 is an example, there are parts for higher voltages, etc ,etc

If it doesn't need to pass significant current, then yes, absolutely.

[incf]

In your case an analog switch may be better suited as they test for lower leakage currents, TMUX1101 is an example, there are parts for higher voltages, etc ,etc

If a suitable "jellybean" "chinesium" switch exists - maybe

It need to cost less than about $0.10 (US), and have a bunch of suppliers so that it can't easily go out of stock.

That TI part cost $0.80 in volume - 8 times the budget.

In your case an analog switch may be better suited as they test for lower leakage currents, TMUX1101 is an example, there are parts for higher voltages, etc ,etc

If it doesn't need to pass significant current, then yes, absolutely.

It only passes a microamp.

[Zero999]

Is a higher voltage available to switch it off? If so, then how about a J-FET?
https://www.nxp.com/docs/en/data-sheet/J174_175_176_177_CNV.pdf

[incf]

Is a higher voltage available to switch it off? If so, then how about a J-FET?
https://www.nxp.com/docs/en/data-sheet/J174_175_176_177_CNV.pdf

No, I can't generate higher voltages due to the battery life constraint.

It's a device with a ~3V battery that need to last 15 years. It samples an analog value several times per minute and needs to self calibrate/compensate for leakage current / moisture absorption by the PCB using the FET/analog switch each cycle.

I like JFETs, they are great switches, but I don't think I can make them work.


 

[PartialDischarge]

In your case an analog switch may be better suited as they test for lower leakage currents, TMUX1101 is an example, there are parts for higher voltages, etc ,etc

If a suitable "jellybean" "chinesium" switch exists - maybe

It need to cost less than about $0.10 (US), and have a bunch of suppliers so that it can't easily go out of stock.

That TI part cost $0.80 in volume - 8 times the budget.


It only passes a microamp.

Then it is up to you to look for those parts, but what you need in this case is an analog switch

[SiliconWizard]

A low-leakage analog switch in this price range, that will be tough.

Edit: TS5A9411 maybe?

[incf]

A low-leakage analog switch in this price range, that will be tough.

Edit: TS5A9411 maybe?

That fits the bill. There surely must be a few compatible clones out there if I look hard enough.

I still wish I could find a FET like the one I originally specified, but the analog switch will work. I know my device will have to operate at 70C. But there is some chance of intermittently going into ~125C environments (it happens...). It's easier to predict how a FET might behave at high temperatures.

But that is a concern for the future I suppose.

[jbb]

I expect you’re already aware of it, but not many cells will last for 15 years. Have you got a source for long life batteries (perhaps lithium thionyl chloride chemistry?) and a specification for how much they might self discharge?

[incf]

I expect you’re already aware of it, but not many cells will last for 15 years. Have you got a source for long life batteries (perhaps lithium thionyl chloride chemistry?) and a specification for how much they might self discharge?

We are well aware, yes, we use lithium primary cells of various chemistries that are rated by the manufacturer that sort of lifespan. Somewhat expensive, and a pain to deal with, but they work.

[PCB.Wiz]

I still wish I could find a FET like the one I originally specified, but the analog switch will work. I know my device will have to operate at 70C. But there is some chance of intermittently going into ~125C environments (it happens...). It's easier to predict how a FET might behave at high temperatures.

The problem is most vendors spec FETs for the mass market uses, so they only give a very conservative MAX and little else.

With analog switches, you have more chance of finding specs for leakage, because that matters for some users.
eg I see Nexperia XS5A1T4157 gives typical of 20pA at 25°C and max of 320nA at 125°C

A somewhat similar NMUX1237 specs typ 2nA (25°C) and MAX of 55nA at 85°C and 490nA at 125°C, so you can see the temperature effects.
 
Others you could test.
I see a cheaper GT4157S6 from GTIC at lcsc that specs < 1uA at 125°C

[cortex_m0]

Diodes Inc has a line of low leakage MOSFETs. I didn't a PMOS one rated as low as 50nA, but I didn't check the whole family. This one is rated 100nA max.

https://www.diodes.com/datasheet/download/DMP2900UW.pdf

[incf]

Diodes Inc has a line of low leakage MOSFETs. I didn't a PMOS one rated as low as 50nA, but I didn't check the whole family. This one is rated 100nA max.

https://www.diodes.com/datasheet/download/DMP2900UW.pdf

Good enough. I appreciate you finding that.

I ended up looking at the datasheets for all of their single channel SOT-23 PMOS devices. The one you gave plus the '1013 appear to be the only ones with ratings of 100nA or less.

[Terry Bites]

Idss, don't you just hate it. I know I do.

So what's all this about "measured at Vds=0 about.
How do they do that? Not likely in a real circuit is it. A useless spec.

[cortex_m0]

So what's all this about "measured at Vds=0 about.
How do they do that? Not likely in a real circuit is it. A useless spec.

You're right, that would be nonsense. Perhaps you can show is a data sheet with measured current at zero voltage.

I will note that the Diodes and Zetex datasheets discussed in this thread have Idss specified at the rated working voltage of the part.

[Smokey]

...I know my device will have to operate at 70C. But there is some chance of intermittently going into ~125C environments (it happens...). ...

Battery operated thing at 125C?  That sounds sketchy.  Lithium Thionyl Chloride is the only thing I could find that had that high a rating.

[Zero999]

Is a higher voltage available to switch it off? If so, then how about a J-FET?
https://www.nxp.com/docs/en/data-sheet/J174_175_176_177_CNV.pdf

No, I can't generate higher voltages due to the battery life constraint.

It's a device with a ~3V battery that need to last 15 years. It samples an analog value several times per minute and needs to self calibrate/compensate for leakage current / moisture absorption by the PCB using the FET/analog switch each cycle.

I like JFETs, they are great switches, but I don't think I can make them work.

What do you require in terms of power consumption?

Note that analogue switches do have low leakage current, but have a higher power supply current, compared to a single transistor.

Looking at the data sheets of a couple of the suggested parts:

90μA @ 25 °C
https://assets.nexperia.com/documents/data-sheet/XS5A1T4157.pdf

Much lower, at 0.5μA, but only rated to 85 °C
https://www.ti.com/lit/ds/symlink/ts5a9411.pdf

[PCB.Wiz]

Looking at the data sheets of a couple of the suggested parts:

90μA @ 25 °C
https://assets.nexperia.com/documents/data-sheet/XS5A1T4157.pdf

That 90uA is only for logic level shifting, when the VIS is not VCC CMOS levels. See fig 4.
If you drive to GND and VCC, there is no 90uA effect.

[voltsandjolts]

...I know my device will have to operate at 70C. But there is some chance of intermittently going into ~125C environments (it happens...). ...

Battery operated thing at 125C?  That sounds sketchy.  Lithium Thionyl Chloride is the only thing I could find that had that high a rating.

Lithium sulfuryl cells are available to 165C, and thionyl to 200C

[Zero999]

Looking at the data sheets of a couple of the suggested parts:

90μA @ 25 °C
https://assets.nexperia.com/documents/data-sheet/XS5A1T4157.pdf

That 90uA is only for logic level shifting, when the VIS is not VCC CMOS levels. See fig 4.
If you drive to GND and VCC, there is no 90uA effect.

Oh, I see, that's 8000nA or 8μA, which is much better but still more than the leakage specified by the OP, assuming it's a power consumption issue.