Ideas to properly synchronize transistors

[zphaze]

Hi,

Since there are many clever people here with much more experience than I, I wanted to get some counseling on how to do this.
I don't want to talk too much about my application (I don't want to explain too much why I'm going for this or that, since this was already carefully thought).
I have to get very short LED flashes pulsed at high speed for specific reasons, so the load is a set of high quality HB LEDs.

I have this schematic, which I'm trying to achieve :


I would like to synchronize Q1 and Q2 to have those state transitions:

ON -> OFF
1. Open Q1
2. Wait 10ns (ideally)
3. Close Q2

OFF -> ON
1. Open Q2
2. Wait 10 ns (ideally)
3. Close Q1

I want to switch the capacitor while it's charged to 36V, so this makes it a bit more challenging as well, because I have to drive the Q1 NMOS gate with a MOSFET driver with e.g. 45V so that when Vload=36V, Vgs = 45-36=9V. Since the MOSFET body diode could conduct when the capacitor is charged at 36V and Vload=0, I have to put the NMOS with the source at Vload node. I don't really have a better solution than a high voltage MOSFET driver IC like the LTC4440 for example, but I do have a 45V source in the circuit so this would work. I also wonder if there would not be a better way to implement Q1 than only one NMOS.

Q2 could very well be a NMOS (same kind as Q1 for example) :


The thing I'm most unsure about is the synchronization of Q1 and Q2. Basically Q1 and Q2 must never conduct at the same time. I wonder if anybody have an idea on how to achieve this synchronization. I *will* have an FPGA on this project, so I *guess* (actually, if you have suggestions for FPGAs I could take them) it could be done with my FPGA, but still, I wonder if I could not make this possible only with ICs/discrete components/logic gates in order to use a single clock! I'm trying to switch as fast as around 200ns, so this is my target.

Are there solutions on the market to synchronize two MOSFETs properly and fast enough? From my requirements, it seems I have to make the opening time faster than the closing time. Indeed, if Q1 and Q2 are the same, and Q1 and Q2 open faster than they close, then my delay requirements would be met. If you guys have ideas on implementing this, or suggestions, feel free to let me know!

Thanks!

[magic]

What is even the point of the capacitor?

Anyway, it seems the capacitor will hold a constant voltage so maybe use a PMOS instead?
You can't simply apply 45V to an NMOS gate as shown because gate insulation will break down when Q2 turns on and Vload goes to zero.

[David Hess]

There are MOSFET drivers intended to control synchronous rectifiers which will switch a pair of MOSFETs.

But if your power source is a current source, then it does not matter if both transistors conduct simultaneously.

[magic]

I suppose OP wants to prevent Q2 discharging the capacitor.

[zenerbjt]

Yes a sync fet driver seems the most springs-to-mind choice here.

Alternatively, put in a driver to drive Q1 as you wish....then install logic gates such that Q2 can never be on when Q1 is on...but yes, you will have to allow for the transistion period of the fets. An XOR gate for example, picks out the "two inputs the same" scenario......because it gives zero output when inputs are same.

There are actually a number of pwm dimming led drivers which do what you want exactly......i think they are by linear.com and ti.com (ie a current source which gets shorted when the leds are off....which is handy because the current is "all-there-and-ready" when you want the leds back on again.)

For the Q1 fet drive, you could just have an isolated PSU connected up to it...and if probz with the internal diode, then just use back-to-back fets instead. To get the gate drive signal up to Q1, use a digital isolator....or one of those bootstrap fet drive chips.
Alternatively you can use a gate drive transformer but beware the leakage L in the trxformer........of course, you can always make the GDT port the signal and also power a high side fet driver, so you avoid the leakage L...but thats another thing.

[mzzj]

Hi,
(I don't want to explain too much why I'm going for this or that, since this was already carefully thought).

I have my doubts about that but..

In your schematic you can swap the Q1 and the capacitor locations. Same functionality but easier to drive.

[magic]

#MeToo

In your schematic you can swap the Q1 and the capacitor locations. Same functionality but easier to drive.

Then it has to be P-channel once again because it will be called to block -36V from the cap during OFF time.
But it's better than my idea, probably.

[laseralex]

"very short LED flashes pulsed at high speed"

What do you consider "very short"?  Do you have a target to hit?

LIDAR uses very short pulses (10s of Amps with pulse width less than nanosecond) so some of those designs might be relevant.  TI and EPC have some great appnotes.

Your string of LEDs will have some series inductance - possibly too much to allow you to meet your target pulse width.  You might need to run multiple parallel drivers to hit your target.