High-side N-channel FET Switch (100% duty cycle ?).

[Dannyx]

Good day folks. I'm currently working on building a kind of ghetto/poor man's UPS for a 12v DC system. Yes, yes, there ARE commercially ones available, but that's not the point, is it ?  I like to tinker with stuff and build my own junk

TL;DR: can a IR2117 drive a high-side N-channel FET with 100% duty cycle (DC only) ? To tell you the truth, the datasheet doesn't make this clear.....to a newbie like me. This is the first time I attempt to use an N-channel FET to switch the HIGH side of anything, so I figured it's good place to learn.
I kind-of understand why most boostrap circuits CANNOT stay 100% on all the time and I assumed that using a purpose-built IC for this would solve this problem by default, but not quite so. From what I've read, even ICs require a minimum "off" time to recharge the bootstrap cap. I was hoping I'd find one that has its own "charge pump" (?).

Granted, the datasheet for the IR2117 doesn't SAY "charge pump" anywhere, BUT the block diagram does show a "pulse generator" and what appears to be a flip-flop (?)....so does this mean it takes care of the required "switching action" internally, even if the "in" pin is pulled high indefinitely
If it turns out this one is not suitable, what would you recommend for the job....if such thing even exists ? Not many get mentioned out there.



THIS is how I imagine this project should work. I only built it up to the batteries and I'm working on improving it. Schematic is far from "finished", so please bare with me.

I found a 53v PSU in the trash for some reason. Still works though, so it's going in my project   The DC/DC converter at the top-left is a random CV/CC module I picked up for 2 bucks off Aliexpress (because it arrived busted and I requested a refund. The fix cost me nothing ). This guy acts as my "charger". When the 53v supply is powered by mains, the DC/DC converter also powers a second DC/DC converter (which I do not yet have) to supply the load with 12v, while also charging the batteries. Of course, the catch is that if the LOAD requires a higher current than the batteries, the CC aspect goes out the window, but through basic experiments I found it's not the case - the batteries ARE the highest load.

Now for the fun part: I'm thinking of designing a crude cut-off for when the batteries are close to fully discharged. I thought: op-amp and MOSFET as switch. Still a lot of work to do there, but hopefully it makes some sense as to what I was picturing. There are probably MANY things wrong there - no caps, no protection mechanisms, op-amp is wired incorrectly for the job, I know  .....I'm just pounding out the details for now. Thanks.

[langwadt]

TL;DR: can a IR2117 drive a high-side N-channel FET with 100% duty cycle (DC only) ?

no ..

[DavidAlfa]

Just use p-CH fet?

[Zero999]

A bootstrap circuit works by charging a capacitor via the load, whilst the driver transistor is off, then connecting the capacitor to the transistor's gate to turn it on.

If the transistor is turned on for too long, then the capacitor will discharge.

It's possible to get quite a high duty cycle, but never 100% with a bootstrap circuit. Here's an example of a circuit I designed to give a high duty cycle but M1 still needs to be turned off to charge C1, via D1 and RL. Note it's inverting, so the transistor turns off, when the input goes high.

[Benta]

If it's just static switching (and it seems to be) where you can live with some milliseconds of switching speed, use a photovoltaic optocoupler.
Eg, https://www.vishay.com/docs/49262/49262_pt0314.pdf

[DavidAlfa]

Also there're integrated charge pump gate drivers, like the LTC7100 LTC7001(135V).
Or use an auxiliar charge pump:

[jmelson]

Yes, it can, if you provide a separate, floating supply to keep the bootstrap capacitor charged.  I built a full-bridge motor drive this way, it had a small, transformer coupled charge pump that created the bootstrap supply.
Jon

[BrianHG]

Also there're integrated charge pump gate drivers, like the LTC7100 (135V).
Or use an auxiliar charge pump:

Wrong part number...

High side gate driver with integrated boost converter:

https://www.analog.com/media/en/technical-documentation/data-sheets/ltc7001.pdf

[BrianHG]

If it's just static switching (and it seems to be) where you can live with some milliseconds of switching speed, use a photovoltaic optocoupler.
Eg, https://www.vishay.com/docs/49262/49262_pt0314.pdf

This one has an improved gate voltage of 14v compared to the older 7-9v photovoltaics I used to use.
Thanks.

[Kim Christensen]

If the 24V to 12V convertor will be the only load, then simply buy one that already has a low voltage cutoff feature.
You can also get convertors with a remote on/off feature that you could control with your own low battery detector.

[ArdWar]

If the 24V to 12V convertor will be the only load, then simply buy one that already has a low voltage cutoff feature.

Or use converter with ENable input, driven by the comparator.
Tho a bit tricky if you really want to get rid of the quiescent draw.

Also what's with the fuses? At what point it start to be counterproductive?

[BrianHG]

If the 24V to 12V convertor will be the only load, then simply buy one that already has a low voltage cutoff feature.

Or use converter with ENable input, driven by the comparator.
Tho a bit tricky if you really want to get rid of the quiescent draw.

Also what's with the fuses? At what point it start to be counterproductive?

The aforementioned ltc7001 has a built in UVLO circuit.
Need to double check the quiescent current draw when it is in shutdown mode.

[Dannyx]

Thanks for all the replies.
Let's review then:

I WAS going to go with a P-channel FET, but I thought I'd try to improve my "design" by keeping the On resistance as low as possible, plus N-channels are cheaper AND I have a box full of them

Someone mentioned the IR2117 can be used after all, but I'm not sure I understand what would classify as a "floating" supply in this case. The datasheet shows the IC sharing a common GND with the load, so I assume that is NOT a floating supply, nor is my LDO idea which I featured in my crude schematic. To me, a floating supply sounds like it involves a separate GND entirely, so the "COM" pin of the IC would be tied to that and the GND of the load tied to the rest of the system. I don't think I want to start designing such a supply from the ground up, so.....

Yes, it can, if you provide a separate, floating supply to keep the bootstrap capacitor charged.  I built a full-bridge motor drive this way, it had a small, transformer coupled charge pump that created the bootstrap supply.
Jon

The LTC7001 seemed to show up the most during my searches indeed. The supplier I tend to get parts from doesn't carry it unfortunately, so I'll have to find something equivalent. It seems to make the most sense at the moment.

High side gate driver with integrated boost converter:

https://www.analog.com/media/en/technical-documentation/data-sheets/ltc7001.pdf

Not to sound like a smart@$$ - just askin' - but the internal block diagram of the LTC looks similar to the one of the IR2117 (save for the charge pump block in the middle, which might be the key difference - makes sense, since I need to drive the gate of the FET above the input voltage). It seemed to me like the IR2117 would also "pump" the external cap via the diode using those two transistors in the block diagram, hence why I thought I was all set and rolled with it for a while until I decided to slow down and check. Unfortunately, the datasheet provides no details on this or what happens to the "VCC" pin inside the IC and whether it gets "boosted", so there's probably a good reason why it wasn't suggested anywhere......

[Andy Chee]

For a brief glance at various high-side gate drive topologies, refer to the table on page 29.

[Dannyx]

Or use converter with ENable input, driven by the comparator.
Tho a bit tricky if you really want to get rid of the quiescent draw.

Also what's with the fuses? At what point it start to be counterproductive?

I was planning on using THIS as the convertor on the far-right, which does actually have a "CTL" input and claims to have a current draw of only 120uA in standby......using its own pin to switch it off might not be a bad idea actually, but I still want to learn about driving a high-side MOSFET, even after the fact

The fuses:.......well I just threw one wherever I felt was necessary. A TVS diode on the outputs might also prove necessary. I'm still working on that part  Not sure if I should use a zener or a TVS as a means of blowing the fuse. Poor man's OVP protection, I know.....

[Zero999]

Thanks for all the replies.
Someone mentioned the IR2117 can be used after all, but I'm not sure I understand what would classify as a "floating" supply in this case. The datasheet shows the IC sharing a common GND with the load, so I assume that is NOT a floating supply, nor is my LDO idea which I featured in my crude schematic. To me, a floating supply sounds like it involves a separate GND entirely, so the "COM" pin of the IC would be tied to that and the GND of the load tied to the rest of the system. I don't think I want to start designing such a supply from the ground up, so.....

The IR2117 has two power supplies. One which powers the logic, which is connected between V_CC and COM and the MOSFET driver V_B and V_S. In the circuits given on the data sheet, V_S is connected to the load, which connects it to COM, hopefully via a fairly low resistance. V_B is connected to V_CC via a diode and to a capacitor. When the transistor is off, the capacitor charges via the diode and when the transistor turns on, the voltage on V_S rises, because current flows through the load. which pushes the voltage up,

The idea is to connect an isolated power supply, for example a DC:DC converter between V_S and V_B, so it can turn the MOSFET on all the time, without having to rely on the energy stored in a capacitor.

[Dannyx]

To make it easier to grasp, I like to think that "Isolated DC:DC converter" you added is a battery. Now, if this calls for an "isolated" DC-DC supply, it means a plain-@$$ buck converted won't work, because the "input" GND is the same as the "output" GND, so it looks like THIS, if I understand correctly. When the FET closes, the "load power" would get shorted to GND through the buck converter or LDO, whatever...

In the meantime, I managed to locate what on first glance appears to be a charge pump-based driver. An LM9061. I haven't gone through the entire datasheet - just scanned my eyes across it

[Doctorandus_P]

N-channel MOSfets, are in some qualities "better" then P-channel MOSfets. (Cost, Rdson), but they are easier to use as a high-side switch.

VO1263 costs EUR5 at digikey in low quantities, I don't know the cost of IR2117 (Digikey lists them as obsolete) But I guess you can buy a really fancy P-channel MOSfet for that price, or put a bunch of them in parallel to each other.

But you do not need to switch fast, (I don't know how slow you can tolerate) One option is to use an isolated DC-DC converter, connect the secondary side to the gate & source of your N-channel MOSfet, and then switch the primary side on or off.

Another option is a charge pump. Just a few diodes and capacitors. As soon as you start putting an AC voltage (uC PWM pin, beefed up by an NE555 or similar) into it, it will start pumping the voltage above your output rail.

[Postal2]

... can a IR2117 drive a high-side N-channel FET with 100% duty cycle (DC only) ? ...

If one month is enough, then yes. Just increase the capacity of the capacitor. I can't say for sure about IR2117, but I saw such a solution on a chip from the same series. Just check the discharge rate of the capacitor, most likely it will work.

[Dannyx]

I came across the LM9061 which seems to fit the bill. People seem to have missed it

This is how the "new" schematic would look. One slight disadvantage is that the maximum VCC for the LM9061 is given as 60v, while the recommended operating voltage is 26v, which is at the very top of my "charger's" output for those 12v batteries is series, so I hope the IC doesn't complain too much.

[Doctorandus_P]

Your application is a battery protection / switch circuit, so using an IC that is designed for that seems to be a better option then an IC designed for motor controllers. But I have not looked into details.

[Kim Christensen]

One slight disadvantage is that the maximum VCC for the LM9061 is given as 60v, while the recommended operating voltage is 26v, which is at the very top of my "charger's" output for those 12v batteries is series, so I hope the IC doesn't complain too much.

As per the datasheet:

7.3.6 Overvoltage Protection
The LM9061 remains operational with up to +26 V on VCC. If VCC increases to more than typically +30 V the
LM9061 will turn off the MOSFET to protect the load from excessive voltage. When VCC has returned to the
normal operating range the device will return to normal operation without requiring toggling the ON/OFF input.
This feature allows MOSFET operation to continue in applications that are subject to periodic voltage transients,
such as automotive applications.

You're 3v away from it going into overvoltage protection, so as long as there is not too much ripple coming from your first DC-DC, or spikes from other loads on the 27v line, it's marginal but probably OK for a hobbyist project.

[Dannyx]

Yes, that is true. I did miss that part. I'm not patient at reading through datasheets I also learned that the time required for the protection to trip can be altered by the cap on pin 8, which is required to begin with, so I guess I should make it so that transients don't trip it. The load is rather light in this case: a DVR, (so a HDD), a NAS (2 more HDDs, but this will soon be removed), 2 cameras and 2 network switches. The highest load will be at startup, when all the caps charge and the hard drives spin up. The NAS appears to be smart enough to power up the drives one after the other, instead of both at the same time, so I estimate the maximum current draw to be around 4 A MAX.

[Dannyx]

Moving away from driving the FET for a moment, I started thinking about how it would be best to power my crude "logic", namely the op-amp I'll be using and reference.

The most crude (and far from ideal) way I can think of is to go with an LM393 - it can technically run straight off the 24v rail. To generate a stable reference for it, I could either A) use a zener diode (and use a voltage divider from it to reference one of the inputs ?) or B) use a voltage reference like a TL431 (and use a voltage divider from it). I assume both are pretty horrendous in terms of temp-co and current draw. Probably equally as bad would be to power the op-amp itself FROM the TL431.

Alternative C): an LDO to power the op-amp, with the reference also derived straight from it.

Alternative D): and LDO to power the op-amp AND a TL431 to generate the reference separately (that's what the 431 is actually called: "voltage reference").

Alternative E): any other suggestions for either of these parts, as always.

[Kim Christensen]

A) Zener is fine for this. You really don't need much precision. You can get ones with built in temperature compensation or DIY in your own with an additional diode. ( See 14.2.1 Zener diodes )
B) TL431 would be more than good enough.
C) You don't need a LDO... A plain old jellybean voltage regulator IC will work since you'll have lots of headroom.
D) Using both a regulator and TL431 would be overkill in my opinion.