High power high side power switch

[elektrinis]

Hello,

I am doing a project, where I need to switch DC line on or off, using a N-MOSFET. Up to 30A 80V. One more requirement: very low quiescent current (below 50uA).
Since this is being switched under capacitive load, a traditional resistor divider mosfet driver is not a good idea due to slow switching and high current draw. Also it will not work with N-fet. So I need a proper driver with boost voltage...

After a long search I came up with only one IC: Vicor's PI2061.


But it is not very easy to buy (and not very cheap, even if you buy 3000 of them).

Are there more alternatives? Or maybe someone has a discrete solution on hand? There is a MCU onboard, so I could generate some PWM for charge pump. But would rather avoid intense software support to keep things from blowing up.

Thanks.

[rbola35618]

You maybe able to use gate transformer to drive the high side. Try to use as low of coupling capacitance as possible because when the gate drive is turned off, the energy in the transformer might turn on the Mosfet. So I would add a resistor arosss the gate and source.

Hope this helps.

Robert

Below is a video that I made of a gate driver used to drive a two switch forward converter. I used it to drive both a high and low side MOSFET.

[JohnnyBerg]

Have a look at the IR2110

[elektrinis]

I think he is looking for constant on controller, a load switch.

What you could look into is a photovoltaic fet driver if you omit the passive turn-off resistor and substitute an active turn off circuit you may be able to hit the 50uA quiescent current. You could maybe pulse the input LED to maintain the needed Vgs at a reptition rate that gets you down to 50uA or better. I haven't tried this but with just a resistor for turn-off they take on the order of 100's of nS to a couple uS  to turn a fet off depending on the size of the fet. So if you completly omit the discharge resistor the fet would only have leakage discharge so you may be able to get away with just pulsing the SSR input at a low repition rate, use a low leakage transistor for active turn-off under a fault.

http://www.mouser.com/ds/2/427/lh1262ca-254015.pdf

Thanks, this is some fresh thinking. Little expensive though.
Will investigate further.

Yes, I meant constant-on solution, so IR2110 and similar solutions are not an option.

[rx8pilot]

I have used Linear's LT1910 high-side N-FET controller with success - 1000's of them. It can be configured for capacitive loads and current limited. Logic level on/off.

Just pick a FET that can deal with the current and the 1910 will drive it. I think it's good to about 60v or so. I am driving a low-ESR 5000uf capacitive load up to 20A. I have configured it to deal with the inrush gently so the FET does not fail. Then it limits the current at a pre-set value and also has a programmable current transient timer for brief transients.

[elektrinis]

LT1910 is a nice chip, I missed it somehow. Very similar to Vicor's solution. More elegant, but even more expensive and only 48V. Also quiescent current is 2mA...
I'm thinking about my own charge pump now... Seems much cheaper and possible to do much lower current consumption. I'm monitoring current separately, on low side.

[eneuro]

...
http://www.mouser.com/ds/2/427/lh1262ca-254015.pdf

Thanks, this is some fresh thinking. Little expensive though.

Not only expensive, but if you look into its datasheet you'll findout that it has very low output gate current, so it might be useless unless you calculate needed gate current in your application, while mosfet total gate charge will determine this frequency limit with this IC and do not fall into the trap from its datasheet, while turn on/off times given there are for... 15pF only at 20mA input diode current and then you have ONLY ~7uA for mosfet gate    

[elektrinis]

I have notices that figure. Did not calculate what I get yet, but looks like not enough.
It is too expensive and energy-hungry to be seriously considered...

[elektrinis]

What I am afraid of is slow turn on, which would cause mosfet damage during switch-on with capacitive load. Since my high side is floating in reasonable levels (0-80V), I can use simple PNP to pulse discharge the gate quickly. But fast turn-on still remains an issue.
I was hoping that there are standard parts available for such 'standard' tasks, but looks like I thought wrong.

[eneuro]

What I am afraid of is slow turn on, which would cause mosfet damage during switch-on with capacitive load.

Maybe solution could be bypass this switch with resistor and another smaller cheap N-channel mosfet in series than even photovoltaic driver or other discrete mosfet driver could be fine at small currents just to precharge caps bank to DC voltage before switch and while those bulky caps will be pre charged turn on main bypassed switch with load switched off-another switch after those bulky caps 

It depends what this caps load does-is it a filter to another inductive/resistive load or something else?

[elektrinis]

Maybe solution could be bypass this switch with resistor and another smaller cheap N-channel mosfet in series than even photovoltaic driver or other discrete mosfet driver could be fine at small currents just to precharge caps bank to DC voltage before switch and while those bulky caps will be pre charged turn on main bypassed switch with load switched off-another switch after those bulky caps 

That is actually an interesting alternative to floating supply - to have it done with photovoltaic device and cap buffer. Still somewhat relies on software - a delay is required to precharge before any switching could take place. And probably even cheaper than transformer solutions.

It depends what this caps load does-is it a filter to another inductive/resistive load or something else?

It's a motor controller. My source is battery (ebike application).

[eneuro]

It's a motor controller.

This is on top of  google'd Precharge & PostDischarge circuit, but many other possible:
http://forums.aeva.asn.au/forums/precharge-postdischarge_topic2715.html

It can be done even better and we can implement current limited charger instead of resistor, so endless posibilities to improve precharging process depending on caps bank size, but in e-bike you'll save a lot of energy if... you will drive not to fast, because of you have very bad aerodynamics parameters, so energy needed to turn ON/OFF this main switch and control precharge is negligible.
Probably your e-bike main controler monitors voltage on those caps, as well as battery voltage, so main switch (relay) can be turned on when those voltages are similar or it can be done in precharge controler easy-when those voltages are the same we turn off precharge while it is not needed.
Endless posibilities to make it smart 

[elektrinis]

This is on top of  google'd Precharge & PostDischarge circuit, but many other possible:
http://forums.aeva.asn.au/forums/precharge-postdischarge_topic2715.html

A soft precharge is actually another topic and I am solving other issue now. However I would be interested in an elegant solution for active constant-power precharge. Was looking for ideas, but did not find any that are cheap and elegant.

Wouldn't a Relay reated for the DC current and voltages be a little pricey?
I don't quite understand why the 50uA quiescent current requirement. With a fet rated for say 150V and 3.8 moHms 30A countinous is 4.5w or so power dissapation,this makes the minuscule q current seem a bit odd.

In this application, relay is not the best solution, because it will be damaged if switched at high current: contacts will weld when closing, and they will melt when opening. And of course no-load current will be a huge issue. Also relays for such current and voltage are ridiculously hard to find and expensive. Latching relay is an option in terms of idling current, but all other aspects of it are even worse.
Low no-load current is required, because the battery should be internally protected and still compatible with old lead-acid batteries, I mean output contacts should always be live, unless there is a good reason to switch it off. Battery could be stored for months or even years in such "live" state and it should not self discharge. Take a power tool battery for example. We all hate it when you decide to drill a quick small hole and your battery is empty.
Same with ebikes. Battery does not have an external "enable" input. It is simply switched on or off with a switch. All current goes through that switch. And no, it is not possible to add protection after switch due to compatibility issues.

[eneuro]

Battery could be stored for months or even years in such "live" state and it should not self discharge.

Probably not too many years even with pefect circuit
http://en.wikipedia.org/wiki/Self-discharge
Typically, among rechargeable batteries, Li-ion absorb the least amount of self-discharge (around 2–3% discharge per month,) then lead-acid at 4-6%, while nickel-based batteries are more seriously affected by the phenomenon (NiCad, 15–20%; NiMH, 30%,) with the exception of Low self-discharge NiMH batteries (2-3%.)

[elektrinis]

eneuro, what you are quoting is far from being true. NiMH and especially NiCd, when not new, can self-discharge fully within few days. For example AAs from older digital cameras. Even well reviewed eneloops suck at this badly after year or two of mild usage. NiCds, used in power tools (Makita, I'm looking at you), also self-discharge within several days.
Lithium is ~2-3%, but per year, not month. If higher numbers are given, probably it contains internal electronics and protection circuitry (poorly designed one or on very small pack, like laptop), but not battery itself.
I don't have any good papers on hand that could be quoted, so I will not edit the wiki page. Just use that mainstream info with caution.

But enough about that, it's offtopic. Requirements are what they are.

[eneuro]

Requirements are what they are.

Energy harversting might be best solution for tricky requirements 
Energy Harvesting - Overview by TI
From low power to no power, TI is the driving force behind today's energy harvesting applications.

[elektrinis]

I'd call it non-mainstream, but certainly not tricky... Few years ago there were serious problems in battery management industry, but now many lessons were learned and we have several good choices from off the shelf ICs. I'm sure someone is working on that.

[ben7]

I have used Linear's LT1910 high-side N-FET controller with success - 1000's of them. It can be configured for capacitive loads and current limited. Logic level on/off.

Just pick a FET that can deal with the current and the 1910 will drive it. I think it's good to about 60v or so. I am driving a low-ESR 5000uf capacitive load up to 20A. I have configured it to deal with the inrush gently so the FET does not fail. Then it limits the current at a pre-set value and also has a programmable current transient timer for brief transients.

I've had trouble with the 1910, it wouldn't turn off the FETs correctly. The FET would stay partly ON and it would overheat. I came up with my own circuit, using an op-amp, and it works perfectly

This is a tricky one, because the FET is switching on into a low impedance load (capacitors).
If the FETs are appropriately rated, you could do a linear ramp up of the gate voltage, that will cause a relatively constant charge current into the capacitor load.
Or you could do that pre-charge thing.

What capacitance is the capacitor bank/load?

[eneuro]

If the FETs are appropriately rated, you could do a linear ramp up of the gate voltage, that will cause a relatively constant charge current into the capacitor load.

Yep, but external motor driver shoudn't switch on load at this stage else we'll have a huge power loss in this switch and probably "magic smoke" 
So, in custom controlers rather basic stuff that we never connect battery directly to capacitors bank and it could be part of motor controler itself, so I have no idea where the problem is-in bad motor controler design? 
Even amateur made @ home controlers have ANOTHER battery to power driver eelctronics and separate from main battery, so no problem to take care of this small cheap 12V a few Ah battery, which can power up controler electronics and we can connect battery without any builtin precharge ^ postdischarge circuits while those subsystems are integral part of controler powered from ANOTHER battery and take care to limit this current to main capacitors bank 

Never buy commercial products unless you saw its @Dave teardown 

[elektrinis]

Usually capacitor bank is around 1-22mF. Can be more on higher powered controllers. By "usual" I mean ~1kW ones and by "higher power" - 20kW and more.
And yes, such systems often run on 48V. 400 amps is a normal figure in such systems, even on DIY ebikes.
And no, separate batteries are not used in >99% 48V ebike systems.  Except some niche applications. And if it is used, it is always 12V (of 24V in marine applications).

[Siwastaja]

http://en.wikipedia.org/wiki/Self-discharge
Typically, among rechargeable batteries, Li-ion absorb the least amount of self-discharge (around 2–3% discharge per month,)

... which is complete and utter bullshit, an urban legend probably originated from the infamous "Battery 'University'" website so long ago that it has been quoted and quoted again for more than a decade, by people who don't have a freaking clue about what they are talking about. This is why there are a lot of sources stating the same. Unfortunately, having sources seem to be what matters over reality. I have tried cleaning up Wikipedia on li-ion stuff, but all kind of misinformation and legends are so deeply rooted that it is very difficult; and, random internet forum myths are preferred over scientific research because the Wikipedia standard heavily discourages "own research", and all research is someone's own until it becomes quoted widely enough, which doesn't happen because the "generally accepted" (wrong) figure is already there. But this is only in the world of those who talk; everybody actually dealing with the li-ion technology (manufacturers, analysts, heck, even the hobbyists!) know that the real-world rate of self-discharge of most types of li-ion batteries is practically negligible, around 1/10th of the quoted number in most cases, unless the cells have quality problems. 2-3% per month may be some kind of guaranteed worst-case acceptance number for some Chinese manufacturers, but it doesn't reflect the reality.

Minimizing "sleep" mode consumption to about 2-3% of the battery capacity per year makes a lot of sense, as does getting rid of mechanical relays. The most elegant way would be designing this at the motor controller level as it already has the required power switches anyway with very little leakage current - you also get rid of precharge (which is always at least bit nasty - no perfectly elegant way!). Our 3-phase electric vehicle induction motor VFD drive consumed only 1 µA through the IGBT & bypass cap leakage. Only the low current logic side, which worked from 12V, needed switching. If you keep all of this in mind, you can design a very good motor controller that "does everything" in one package with the least cost, elegantly and reliably.

But if you cannot alter the motor controller, you'll need the power switch and some precharge system.