High side power FET switch ideas needed. (maybe charge pump/gate driver?)

[Odd]

The mission is to switch 3-40v  , up to 10A.
The signal is a MCU output, between 0 and 3v3.

I could always do something like the traditional high-side PFET design (black attachment)
it would perform poorly if the voltage to be switched is low.


What I would really like is the hand-drawn thing, where the questionmark is some charge pump with very few external components, and I can control the gate of the NFET from logic level.

[PGPG]

About your traditional PFET design.
What is PFET type that accepts 42V gate-source?

Using NMOS at high side you need voltage higher than high switched voltage.
You said no word for how long the switch have to switch the load on. If it is like in DCDC converter only several us than there is no special problem with supplying with capacitor, but if it have to be switched on for hours you have to made some permanent source to drive NMOS gate.
You also said no word why load can't be switched on the low side making the NMOS solution much simpler.

[langwadt]

how fast, what duty cycle?

[Odd]

very slow, basically on/off in a per minute basis, so no need to make the transition very fast.
As for PFET and the GSvoltage, well, this is why I seek advice 

I do not wish to switch it on the low side because the different outputs will supply different subsystems, which have common ground and ground wired in data-connections.

[Benta]

In that case, the simple solution is a photovoltaic optocoupler.
Like this one:
https://www.vishay.com/en/product/83469/

[BrianHG]

Just use an N-Channel Mosfet with a photo-voltaic.

~20ma into the LED, and the photovoltaic will output ~18v for the Gate regardless of any other system voltages.

This one has fast turn-off circuitry built in, so no external jfets needed:

https://www.mouser.ca/ProductDetail/Toshiba/TLP3910C20-TPE?qs=TuK3vfAjtkUvVtweYOqjVw%3D%3D&srsltid=AfmBOorsWvMG9Vb9B8uyFlksa3zT4Ro7OeIp8r8kHHBu1gWuYg8y3fQ0

[BrianHG]

In that case, the simple solution is a photovoltaic optocoupler.
Like this one:
https://www.vishay.com/en/product/83469/

Warning, that photovoltaic only delivers 8.9v.  The one I recommended delivers at least 14v, normally 18v.

[Benta]

"Warning"? Thanks a lot.

But yeah, the TLP3910 is nice. I thought Vishay was state-of-the-art here, but am mistaken.

Downside to the TLP3910 is that you'll need a voltage booster circuit for driving the LED. But that's pretty simple.

[BrianHG]

"Warning"? Thanks a lot.

But yeah, the TLP3910 is nice. I thought Vishay was state-of-the-art here, but am mistaken.

Downside to the TLP3910 is that you'll need a voltage booster circuit for driving the LED. But that's pretty simple.

Sorry, I didn't mean to offend.  Just in case you miss the 9v and do not use a logic level mosfet, you may potentially have a partially turned on mosfet baking circuit if you are driving a huge amount of current.

I do not think that it is if Vishay is state-of-the-art or not, they must have a real reason for their 9v we just do not know about.  The toshiba part probably just has 2x in series compared to the vishay part.

It could be something as annoying as a patent rights issue.

If you have a 5v rail somewhere, then all you need is an open-collector output to drive the LED.
I would personally recommend a transistor driver as ~15ma on an MCU IO is doable, but I know something like a 2n3904 or 2n2222 will always hit the LED at my prescribed current through a series resistor.

[artag]

I have a similar problem.
A PMOS transistor switches on a battery (3.7 - 4.2V) supply to a 61088 boost converter generating 5V. The PMOS transistor fails to achieve it's best on-resistance, probably due to the low battery voltage. At lowest battery voltage I have highest current drain and worst sensitivity to on-resistance. Like the OP, a common ground makes it difficult to switch the negative rail.

I'm considering whether to attempt to provide a supply for an NMOS fet, bootstrapped by using a capacitor to initially turn it on and then maintained by the boost converter's output.

Does this sound feasible, or overcomplicated ?