How to quickly turn on/off N-FET switching high side of 55V?

[Michael]

---
He said: "The OPA445's 15V/us typical slew rate should probably be
fast enough."

Here's something just a little faster than that:

Version 4
SHEET 1 1220 1156
WIRE 400 304 32 304
WIRE 736 304 400 304
WIRE 32 384 32 304
WIRE 400 384 400 304
WIRE 736 384 736 304
WIRE 560 432 512 432
WIRE 672 432 640 432
WIRE 32 528 32 464
WIRE 880 528 32 528
WIRE 736 560 736 480
WIRE 400 608 400 464
WIRE 512 608 512 432
WIRE 512 608 400 608
WIRE 880 608 880 528
WIRE 736 688 736 640
WIRE 832 688 736 688
WIRE 400 752 400 608
WIRE 736 752 736 688
WIRE 880 752 880 704
WIRE 224 800 160 800
WIRE 336 800 304 800
WIRE 512 800 512 608
WIRE 560 800 512 800
WIRE 672 800 640 800
WIRE 32 832 32 528
WIRE 160 832 160 800
WIRE 32 960 32 912
WIRE 160 960 160 912
WIRE 160 960 32 960
WIRE 400 960 400 848
WIRE 400 960 160 960
WIRE 736 960 736 848
WIRE 736 960 400 960
WIRE 880 960 880 832
WIRE 880 960 736 960
WIRE 32 1008 32 960
FLAG 32 1008 0
SYMBOL pnp 672 480 M180
WINDOW 0 75 83 Left 0
WINDOW 3 53 48 Left 0
SYMATTR InstName Q1
SYMATTR Value 2N5401
SYMBOL npn 336 752 R0
WINDOW 0 71 15 Left 0
WINDOW 3 45 48 Left 0
SYMATTR InstName Q2
SYMATTR Value 2N5550
SYMBOL voltage 32 816 R0
WINDOW 3 40 81 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
WINDOW 0 40 53 Left 0
SYMATTR Value 55
SYMATTR InstName V2
SYMBOL res 864 736 R0
SYMATTR InstName R4
SYMATTR Value 1.8
SYMBOL voltage 160 816 R0
WINDOW 3 24 104 Invisible 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
WINDOW 0 39 57 Left 0
SYMATTR Value PULSE(0 5 0 1e-6 1e-6 0.001 0.002)
SYMATTR InstName V3
SYMBOL res 320 784 R90
WINDOW 0 -46 56 VBottom 0
WINDOW 3 -38 58 VTop 0
SYMATTR InstName R9
SYMATTR Value 1000
SYMBOL nmos 832 608 R0
SYMATTR InstName M2
SYMATTR Value Si4470DY
SYMBOL voltage 32 368 R0
WINDOW 3 36 73 Left 0
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
WINDOW 0 36 44 Left 0
SYMATTR Value 10
SYMATTR InstName V1
SYMBOL res 384 368 R0
SYMATTR InstName R2
SYMATTR Value 10k
SYMBOL res 656 416 R90
WINDOW 0 -46 56 VBottom 0
WINDOW 3 -38 58 VTop 0
SYMATTR InstName R1
SYMATTR Value 50k
SYMBOL npn 672 752 R0
WINDOW 0 72 17 Left 0
WINDOW 3 47 50 Left 0
SYMATTR InstName Q3
SYMATTR Value 2N5550
SYMBOL res 656 784 R90
WINDOW 0 65 58 VBottom 0
WINDOW 3 73 54 VTop 0
SYMATTR InstName R3
SYMATTR Value 50k
SYMBOL res 720 544 R0
WINDOW 0 41 43 Left 0
WINDOW 3 34 74 Left 0
SYMATTR InstName R5
SYMATTR Value 1000
TEXT 54 984 Left 0 !.tran .01

Hi John - looking at the simulation of your circuit, I am not sure if
your circuit is fast enough. If you look at the power dissipated by
the FET ((V(n004)-V(n007))*I(R4)), it peaks at about 420W when
switching. Wouldn't that kill the poor bastard? Additionally, I think
it'd be even worse in my application, as the switched load is very
inductive (about 80% of the load is motors). It looks like it takes
about half a microsecond, so perhaps is that short enough that the FET
would survive? How can one decide if this is safe or not? Is there a
rule of thumb that I don't know?

Thanks!

-Michael

 

[Michael]

John Fields wrote:
---
View in Courier:
.10-50DC>--------------+-------+------+
. FWB | | |
. +----+ | | |
.120AC>-+ +--|~ -|---+ | |
. P||S | | | | |
. R||E | | [BFC] [R] |
. I||C | | |+ | |
.120AC>-+ +--|~ +|---+->12V | |
. +----+ | | |
. E | |
. PNP B-----+ |
. C | |
. | | D
. +-------|----G NCH
. | | S
. | [R] |
. [R] | |
. | C |
.TTL>---[R]------------|-----B NPN [LOAD]
. | E |
. | | |
.GND-------------------+-------+------+
There's something wrong with your drawing, John.
SFAICT, you've added a 12V dc source riding on top
of the 10-50V 30A power source, let's call that 62V,
but you switch it with a PNP, from a signal connected
to 50V rather than 62V. That can't work. I'll fix
that, and I'll change the level-shifting signal into
a current source, as you know I prefer to do when the
destination voltage can vary. And when it doesn't.
Also note the separate grounds and the output diode.
Note the MOSFET zener, and R4 to help damp the RF
oscillation possible during slow on/off switching.
. FWB
. ,----, +12V
. 120AC >-, ,--|~ +|---+-----+-----,
. P||S | | | | |
. R||E | | [BFC] 3*R1 |
. I||C | | |- | E 150V pnp
. 120AC >-' '--|~ -|---+ +---B
. '----' | | C
. | | |
. 10-50DC >==============o=====================,
. | | |
. 150V npn | | D 100V
. C +--R4-- G n-CH
. TTL >------- B | 18V S
. E | zener |
. | +---|<|---o=o===(O)
. R1 | |
. | R3 _|_ LOAD
. | | /_\
. GND >----------+-----' |
. 50V RTN >=====================================o===(O)

You all are making an assumption that isn't correct: I don't have AC
available to me. This is a battery powered application. Everything
is running off of the 10-55V DC power source.

That's OK, Michael, John and I just like admiring our fine
ASCII drawings. Aren't they nice? Actually, an attractive
way to do this is to use a commercial chip with a built-in
bias generator for the power n-channel switching MOSFET.
Something like Infineon's BTS660P, for $6.69 at DigiKey.

. ___ BTS660P
. 10-58 dc >============|___|======o===(O)
. | |
. D |
. TTL >--+---- G _|_ LOAD
. | S /_\
. 10k | |
. | | |
. GND >--+-------' |
. 50V RTN >=======================o===(O)

You can switch the BTS660 on and off with a 100-volt logic-
level MOSFET in a TO-92 package, such as a Zetex ZVNL110A.
Can you tell us about this 55 volts of yours? The BTS660
has a maximum rating of 58 volts, and it probably can go
a bit beyond that, but one should know exactly what might
be encountered, in case a more complex design is in order.

The 55V source is a 12 cell lithium polymer battery pack. Thus 55V is
a bit high, but I like to be careful.

I have a memory of looking at that Infineon part and Digi-Key telling
me that it is no longer being made. But I could be remembering a
different Infineon part, I'm not entirely sure. That conversation was
at least a month ago.

Anyways, I would ideally like a less integrated solution, not because
of cost, but because I like to turn every design into a learning
opportunity, as I still have a lot of learning to do! Connecting black
boxes is too easy.

By the way Winfield - are you still in Cambridge? Your e-mail address
looks to be at the Rowland Institute. I run by there approximately
every other morning (at around 5AM... probably before you make it
in!).

Thanks!

-Michael

 

[Michael]

---
If the BTS660P's control pin isn't pulled down to 50V RTN how will
it know to turn on?

That is, shouldn't your drawing (BTW, I admire the doubly dashed
lines for the heavy power. Very nice!) look like:

. ___ BTS660P
. 10-58 dc >============|___|======o===(O)
. | |
. D |
. TTL >--+-----G _|_ LOAD
. | S /_\
. 10k | |
. | | |
. 50V RTN >======o=======o========o===(O)

Nice chip, but with worst case Ton of 400µs, Toff of 110µs, slew
rate ON of 2.2V/µs and slew rate OFF of 2.6V/µs it won't meet the
OP's stated need for a slew rate of 15V/µs.

Hi John - I don't have a desired slew rate. I just want the switch to
turn on/off without exploding. Sorry for not being clear about this
earlier.

-Michael

 

[John Fields]

Hi John - I don't have a desired slew rate. I just want the switch to
turn on/off without exploding. Sorry for not being clear about this
earlier.

 

[Michael]

Once a second, maybe? In reality it'll probably be more like once a
minute.

-Michael

 

[John Fields]

Hi John - looking at the simulation of your circuit, I am not sure if
your circuit is fast enough. If you look at the power dissipated by
the FET ((V(n004)-V(n007))*I(R4)), it peaks at about 420W when
switching. Wouldn't that kill the poor bastard? Additionally, I think
it'd be even worse in my application, as the switched load is very
inductive (about 80% of the load is motors). It looks like it takes
about half a microsecond, so perhaps is that short enough that the FET
would survive? How can one decide if this is safe or not? Is there a
rule of thumb that I don't know?

 

[John Fields]

Once a second, maybe? In reality it'll probably be more like once a
minute.

 

[Jamie]

---
Low duty cycle. Excellent!

My lovely wife has just put dinner on the table, so 30 until
tomorrow.

what's one the menu ? A Hannibal Lecter specially ?

Enjoy

 

[Winfield]

By the way Winfield - are you still in Cambridge? Your e-mail
address looks to be at the Rowland Institute. I run by there
approximately every other morning (at around 5AM... probably
before you make it in!).

I sometimes leave for work before the rush hour, and get
in between 6:20 and 7:30. The problem with this is that
I don't generally leave for home until 6:30 or 7pm, even
if I got in early, and so getting in at 6 to 7 makes a
12-hour day, longer than I really should work at one
stretch on a frequent basis. So leaving after the rush
hour is my preferred approach.

 

[Winfield]

---
If the BTS660P's control pin isn't pulled down to 50V RTN how will
it know to turn on?

That is, shouldn't your drawing (BTW, I admire the doubly dashed
lines for the heavy power. Very nice!)

You can use it in the future. I also recommend
' and , -- where appropriate, instead of + everywhere,
to close boxes, etc.

look like:

. ___ BTS660P
. 10-58 dc >============|___|======o===(O)
. | |
. D |
. TTL >--+-----G _|_ LOAD
. | S /_\
. 10k | |
. | | |
. 50V RTN >======o=======o========o===(O)

Nice chip, but with worst case Ton of 400µs, Toff of 110µs,
slew rate ON of 2.2V/µs and slew rate OFF of 2.6V/µs it won't
meet the OP's stated need for a slew rate of 15V/µs.

well, I've been assuming some kind of quasi-close
relationship between GND and RTN, but the reason
for explicitly showing them separated is to avoid
any logic-level and logic-threshold problems with
ground drops, ground bounce, etc., with 30A loads.

. ___ BTS660P
. 10-58 dc >============|___|======o===(O)
. | |
. D |
. TTL >--+---- G _|_ LOAD
. | S /_\
. 10k | |
. | | |
. GND >--+-------+--/\/\--+
. |
. 50V RTN >=======================o===(O)

It might be wise to add a ground-connecting resistor,
as shown above. 100 ohms or less... Or more.

 

[Winfield]

Hi John - I don't have a desired slew rate. I just want the switch
to turn on/off without exploding. Sorry for not being clear about
this earlier.

Well, the scene is not as bad as you might think. In order
to get low Rds(on) in the MOSFET, it's necessary to make a
very large-die part. This means that it has considerable
thermal mass to absorb heating during slow switching, and
also that it has a large die-attachment area, which rapidly
drains away heat into the package frame, and the heatsink.
To understand and analyze this scene, we rely on transient
thermal curves in the MOSFET's datasheet.

The Infineon switches, and parts of this class made by other
manufacturers, have rather slow switching speeds -- but they
are designed to handle the power dissipation resulting from
the maximum rated current while switching.

Are you saying you'd like to see a full discrete-switching
design, all spelled out? Ahem!

May I ask, what are the maximum ON and OFF time durations?

 

[John Larkin]

So something in the few-hundred microsecond or less time frame might
be safe, maybe less depending on the fets, and how many fets.

A floating power supply would be helpful, so the driver doesn't have
to swing so far.

You could do something like this:

http://img105.imageshack.us/my.php?image=fetdriverwc6.jpg

After that, you can also add a serious driver if it's needed.

John

Hey, run that from a 9-volt battery!

John

 

[Robert Latest]

Hi - I'm going to be using some N-FETs to turn on/off (from the high
side) a large (10-30A), grounded load that is being powered by a ~55V
power source.
[...]

So - do you think this is the right direction for me to be going in?
Or can anybody suggest a better method?

There are many cheap, ready-made integrated solutions to exactly this
problem. It gets a bit more complicated if the high-side switch is "on" for
extended periods of time, but if it reliably goes "off" every now and then
there's no problem.

Since you're worried about the FET overheating I guess you have frequent
switching in mind.

robert

 

[Robert Latest]

Once a second, maybe? In reality it'll probably be more like once a
minute.

Then you might want to look into the HIP4081 and friends.

robert

 

[Michael]

Hi - I'm going to be using some N-FETs to turn on/off (from the high
side) a large (10-30A), grounded load that is being powered by a ~55V
power source.
[...]

So - do you think this is the right direction for me to be going in?
Or can anybody suggest a better method?

There are many cheap, ready-made integrated solutions to exactly this
problem. It gets a bit more complicated if the high-side switch is "on" for
extended periods of time, but if it reliably goes "off" every now and then
there's no problem.

Since you're worried about the FET overheating I guess you have frequent
switching in mind.

robert

Nope, not frequent switching! The application is that there are a
couple power sources, and each power source will be switched with a
FET. A logic circuit chooses which power source to use and turns on
that FET. The circuit will use that FET for probably 10-30 minutes and
then once it is drained it'll switch to a difference source. That was
one of the main issues I ran into when initially looking for a driver
for the FET. Now I'm alot more interested in seeing if it's possible
to roll my own driver.

-Michael

 

[Michael]

I sometimes leave for work before the rush hour, and get
in between 6:20 and 7:30. The problem with this is that
I don't generally leave for home until 6:30 or 7pm, even
if I got in early, and so getting in at 6 to 7 makes a
12-hour day, longer than I really should work at one
stretch on a frequent basis. So leaving after the rush
hour is my preferred approach.

Ah, I make the reverse commute (to the outskirts of Boston), which is
bad, but probably not as bad. I generally need to leave home at around
6:45 AM or earlier and leave work at 4:30 PM or earlier.

But since I like to run in the mornings it means I have to get up at
ungodly early hours (~4:00 AM). Fun stuff. Gotta love Boston traffic!

-Michael

 

[John Fields]

what's one the menu ? A Hannibal Lecter specially ?

 

[John Fields]

You can use it in the future. I also recommend
' and , -- where appropriate, instead of + everywhere,
to close boxes, etc.

---
Yes, I can see that!

However, I prefer the look of the '+', so I think I'll stick with
that.

 

[Winfield]

Hi - I'm going to be using some N-FETs to turn on/off (from the high
side) a large (10-30A), grounded load that is being powered by a ~55V
power source. [...]

There are many cheap, ready-made integrated solutions to exactly this
problem. It gets a bit more complicated if the high-side switch is "on" for
extended periods of time, but if it reliably goes "off" every now and then
there's no problem.

Since you're worried about the FET overheating I guess you have frequent
switching in mind.

Nope, not frequent switching! The application is that there are a
couple power sources, and each power source will be switched with a
FET. A logic circuit chooses which power source to use and turns on
that FET. The circuit will use that FET for probably 10-30 minutes and
then once it is drained it'll switch to a difference source. That was
one of the main issues I ran into when initially looking for a driver
for the FET. Now I'm alot more interested in seeing if it's possible
to roll my own driver.

John posted a hand drawing of a simple scheme using
two optoisolators (John likes ordinary optoisolators).
Here's an ASCII drawing of John's drawing:

floating FET driver, JL Dec07
,---+----------,
| | | in
_____ | | A |/
---| +|--' | -> | |
| | _|_ |\v |--'
---|____-|--, --- C | |<-,
| | +---+----||--+
dc-dc | | | | |
converter | | B |/ / Rs |
| | -> | \ 1M |
| | |\v / |
| | | | |
'---+----------+---+--------+---- out


+5 --+-/\/\-+--------+
| _|_ A _|_ B
| \_/ -> \_/ ->
TTL | | |
-----|>o-----+---|>o--' dual
inverters optoisolator

A 160% CTR optoisolator like Fairchild's FOD617D can
provide 7.5mA (min) to charge or discharge a MOSFET's
gate, if the opto-s LEDs are driven by 5mA. A large-
die MOSFET like an IRF1407, rated at 75V and 92A,
http://www.irf.com/product-info/datasheets/data/irf1407pbf.pdf
with Rds(on) = 0.0078 ohms (min), has Ciss = 5600pF of
gate capacitance. In particular, it takes Qgd = 54nC
of charge to move the gate through the "Miller" region.
That's the potentially-dangerous period during which
the MOSFET is neither fully ON nor OFF, as it swings
the drain voltage from one extreme to the other. The
Miller capacitance is Crss = 190pF at 25V, which has
to be charged by our 7.5mA of gate current.

Using t = Q/i we get 54nC/7.5mA = 7.2us. During this
time the MOSFET can be dissipating what, Michael, up
to 30A * 55V = 1650 watts? Not to worry, this is a
big MOSFET, and it can handle massive heat for a short
time by absorbing it in its thermal mass. Looking at
fig 11, Effective Transient Thermal Impedance plots,
we read ZthJC = 0.008 C/W for 10us. This means that
we'll see only 1650 watts * 0.008 = 13C of junction-
temperature rise during this "dangerous" transition.

So that issue, which you were worrying about, is fine.

Clearly one painful little problem in John's design is
the dc-dc converter. But, since you're only turning
the MOSFET on once every now and then, if we change
the standby pulldown resistor Rs to 10M, it's clear
that only a modest amount of power is required to turn
on the MOSFET, enough to charge the gate capacitance,
and run the 10M resistor thereafter. Well, capacitor
C can be a 0.1uF, etc., and provide that gate charge,
if we can find a way to charge C. My suggestion is to
use another kind of opto-isolator, an extra-ordinary
type, made by IR. John suggested a 9-volt battery,
which isn't a bad idea, although a bit brute-force.
Also, that's not a semiconductor, and is frowned upon.

Basically IR's cool part is an opto-battery, a series
stack of photodiodes, powered by an LED. The PVI1050
http://www.irf.com/product-info/datasheets/data/pvin.pdf
gives us 5uA into 10 volts (if we power it with 10mA),
which we can use to charge capacitor C. Rs takes 1uA.
We calculate t = CV/i = 0.1 * 10/4 = 0.25s to charge
our 100nF cap. Very nice, that issue is fine too.

So, here's our new circuit.

floating MOSFET driver, JL, WH Dec07

,-----+----------,
_____|_ | | in, 75V max
+5 | + | | A |/ (o)
---| | | -> | |
| PVI | _|_ |\v |--'
,--| 1050 | --- C | |<-,
| | | | 0.1uF +---+----||--+
'--| | | | | | IRF1407
| | | B |/ / Rs |
,--| | | -> | \ 10M |
| |_____-_| | |\v / |
240 | | | | |
| '-----+----------+---+--------+---(o) out
gnd
470
+5 --+-/\/\-+--------+
| _|_ A _|_ B
| \_/ -> \_/ ->
TTL | | |
-----|>o-----+---|>o--' FOD617D (2)
inverters optoisolator

Now, Michael, that doesn't look too painful, does it?

 

[Winfield]

A 160% CTR optoisolator like Fairchild's FOD617D
can provide 7.5mA (min) ...

Looking at the response-time specs of the FOD617D,
http://www.fairchildsemi.com/ds/FO/FOD617A.pdf
I do see one possible issue, namely the rail-rail
current through transistors A and B, as one turns
off and the other turns on. This can last several
us, and partially discharge C. If C = 0.1uF, and
we allow a 2-volt drop, then we can handle a 20mA
shoot-through current for 0.1*2/0.02 = 10us. OK,
that isn't too bad, but let's increase C to 0.22uF
to give us more safety margin. That's easier than
providing 20us turn-on delays, etc., in the TTL-
logic cmos-inverter network, which is a commonly-
employed solution to this issue.

Here's the modified new circuit.

floating MOSFET driver, JL, WH Dec07

,-----+----------,
_____|_ | | in, 75V max
+5 | + | | A |/ (o)
---| | | -> | |
| PVI | _|_ |\v |--'
,--| 1050 | --- C | |<-,
| | | | 0.22uF +---+----||--+
'--| | | | | | IRF1407
| | | B |/ / Rs |
,--| | | -> | \ 10M |
| |_____-_| | |\v / |
240 | | | | |
| '-----+----------+---+--------+---(o) out
gnd
470
+5 --+-/\/\-+--------+
| _|_ A _|_ B
| \_/ -> \_/ ->
TTL | | |
-----|>o-----+---|>o--' FOD617D (2)
inverters optoisolator

Yep, that doesn't look too painful.