Question About MOSFET Gate Driver IC

[spec]

Hi VNE,

Spec,

Thanks for your explanation in post #17. I think I now understand why the 7912 would be necessary, but I'm still a little fuzzy on specifically how I would use it. The data sheet for the UCC21520 only has an example schematic using NMOSFETs which I believe is what your option 3 from post #13 describes. If I swapped out the top side NMOSFET for a PMOSFET, as in option 2, I'd have to do a little thinking about what else I'd need to change, but specifically with respect to the 7912 you said:

But the 48V connected to the top side drain will be variable. I'll call it VTD. In reality, I'd want the output of the 7912 to always be 12V less than VTD, correct? If that is indeed the case, would I hook up the 7912 as follows?

input = VSSA, ground = VTD, output = VTD -12V

Similarly, if I went with option 3 from post #13, would I hook up the V- output of the wall wart to VSSA?

No problem about explanation- I learn a lot from these threads.

Rather than explain all the connections, if you would like to see circuits for any of the options I can probably post an outline schematic.

At this stage, if you could point at an approach that you would be interested in from a user/builder point of view, that would be the best way forward. Once that is decided we can thrash out all the technical aspects. And often you find that when a trial circuit is done problems are revealed and you may go for another design approach as a result.

Does anyone have any thoughts on changing direction to use the VOM1271 in lieu of the other driver ICs previously discussed?

Tomizett's suggestion is quite innovative and the VOM1271 is certainly a useful chip, but I am afraid that the VOM1271 just does not have enough gate drive ability, at only a few micro amps, for the size of MOSFET that will be required for your application.

Yes, you are switching infrequently, but it is is still important that the MOSFETs turn on and off smartly or they may exceed their safe operating area while they are turning on and off. There could be other problems too, with the massive gate charge and possibly leakage currents.

Sorry.

spec

 

[vne147]

it is still important that the MOSFETs turn on and off smartly or they may exceed their safe operating area while they are turning on and off. There could be other problems too, with the massive gate charge and possibly leakage currents.

Could the possible safe operating area risk be mitigated if I ensured that the output of the main power supply (the 0 - 48V supply) was never on when I turned on the MOSFETs? The whole sequence is software controlled. I could for instance send the command to turn on the NMOSFETs, wait 5 seconds (or longer), and then turn on the power supply which would be up to 25A @ 48V.

EDIT: I did some rough math, for the VOM1271. At an IF of 10 mA, the open circuit output voltage and short circuit current per the data sheet are 8.4V and 15μA, respectively. That works out to an ESR of 560 kΩ. Assuming a gate capacitance of about 20 nF, the time constant works out to be around 13 ms. I could also increase the IF up to 30 mA which would result in a 47μA output current.

Is my line of thinking incorrect, or am I neglecting to consider something important?

 

[spec]

Could the possible safe operating area risk be mitigated if I ensured that the output of the main power supply (the 0 - 48V supply) was never on when I turned on the MOSFETs? The whole sequence is software controlled. I could for instance send the command to turn on the NMOSFETs, wait 5 seconds (or longer), and then turn on the power supply which would be up to 25A @ 48V.

Well that is a neat idea.

But the output impedance of the VOM1271 would still be to high in my opinion.

But don't let me put you off having a go.

spec

 

[vne147]

Well that is a neat idea.

But the output impedance of the VOM1271 would still be to high in my opinion.

But don't let me put you off having a go.

spec

I appreciate all your input and advice up to this point, but I think I am going to give it a go. The simplicity of this solution just makes it too attractive to ignore. It'll be pretty easy and cheap to test, and while it's not the "ideal" solution for the reasons you pointed out it may be more than enough for my purposes.

It's good to know that I'll have the other options you outlined to fall back on if your concerns about the VOM1271 prove problematic. If nothing else, I learned a lot about different H-bridge designs.

Thanks!

 

[alec_t]

If you are going for NFETs on the high-side, here's a simple charge-pump boost supply for the gate drivers, using the PWM output of your MCU as an input :

 

[vne147]

If you are going for NFETs on the high-side, here's a simple charge-pump boost supply for the gate drivers, using the PWM output of your MCU as an input :
View attachment 103968

Thanks Alec. That circuit looks like it will be very handy. If I went with the UCC21520 driver IC for example, would I simply connect the "Boost" output in your circuit to VDDA (pin 16) on the IC and omit the CBOOT cap as shown in figure 34 on page 25?

 

[spec]

I appreciate all your input and advice up to this point, but I think I am going to give it a go. The simplicity of this solution just makes it too attractive to ignore. It'll be pretty easy and cheap to test, and while it's not the "ideal" solution for the reasons you pointed out it may be more than enough for my purposes.

It's good to know that I'll have the other options you outlined to fall back on if your concerns about the VOM1271 prove problematic. If nothing else, I learned a lot about different H-bridge designs.

Thanks!

No probs- I will be interested to hear how the VOM1271 circuit goes.

spec

 

[tomizett]

You could use a h-bridge gate driver such as the **broken link removed**.

Ah yes, I'd overlooked that. I've seen the same thing done in the L6204, but not on a driver for external MOSFETs before. Looks like a useful IC.

In fact if it took 5 or even 10 seconds for the MOSFET to turn on, it would matter one bit

Well... it may not be that simple. Remember that all the time the MOSFET is "switching" it's in the linear region and dissapating power. All the MOSFET-based switching circuits we're used to seeing only work because the transistor spends very little time in this state; mostly is is either carrying current but seeing very little voltage, or seeing voltage but carrying no current - either way the power dissapated is comparatively low.
With a few hundred watts to switch, as in this application you'd need to get the switching done quite fast - before too much heat built up in the die of the FET (the SOA curves on the datasheet should give you an idea of how much you can dissipate for how long).
Turning off with the VOM1271 should be no problem, as it can just short the FET gate-to-source. A back-of-the-envelope calculation suggests that the VOM1271 should be able to turn on an IRF2807 in about 10ms... that's very slow for this kind of circuit and, looking at the datasheet, you would probably have to switch at 10 times that speed in order to stay within the device's SOA.

I hadn't really intended to recommend these parts as a superior way of solving the problem, it just struck me as the only avenue that hadn't already been covered. I think I may have just proved that they're not the solution (or at least, not a simple solution).


<EDIT>
Sorry... Massively cross-posted here! I'd replied to the last post on the first page - not noticed there was a second (why does it give you as reply box anywhere other than at the end of a thread?)
Anyway - spec's already covered my main point. But, yes - give it a go, it will be interesting to see how it works. Certainly being able to switch at zero-voltage would help mitigate the turn-on time problem.
Off the top of my head I can't see any reason why you shouldn't be able to parallel several of these driver ICs to get a bit more current (except that they seem moderately expensive)?

One last warning is to remember that if your MOSFET fails when you are testing, then it will fail as a short circuit - so make sure you've got suitable precautions in place so that you don't spoil anything expensive!
</EDIT>

cheers, Tom.

 

[Pommie]

Just noticed the reply above and wanted to point out that (as far as I'm aware) the MC33883 can only be used with N fets. Also, a separate 12V supply for the chip will allow it to work down to (I think but not sure) 0V.

Mike.

 

[spec]

Hi Pommie,

Just noticed the reply above and wanted to point out that (as far as I'm aware) the MC33883 can only be used with N fets.

I may be wrong, but as the MC33883 has separate input pins for the high side and low side, NMOSFETs, can be used in the low side and PMOSFETs can be used in the high side. But you need to drive the two inputs in anti-phase instead of in phase as would be the case with all NMOSFETs.

Also, a separate 12V supply for the chip will allow it to work down to (I think but not sure) 0V.

The high side supply in the MC3383 can be configured to automatically changes to continuous if the high side MOSFETs are not switching (and thus operating the high side power supply diode and capacitor).

spec

 

[alec_t]

would I simply connect the "Boost" output in your circuit to VDDA (pin 16) on the IC and omit the CBOOT cap as shown in figure 34 on page 25?

That should work.

 

[vne147]

OK, so here's my game plan. For the testing, I'm planning on using the CSD18535KCS NMOSFET from Texas Instruments. It's a through hole component so it'll be easier to breadboard. Because of the comparatively lower performance of the PMOSFET choices, at this time I'm only going to pursue options that use only NMOSFETs.

I plan to test the following configurations:

1. Use the VOM1271 to drive the MOSFETs. I'll make sure that I command the power supply to "ramp up" so I'm always inside of the SOA for the above NMOSFET. I'm also considering adding some sort of soft start circuit in between the output of the main power supply and the H-bridge.
   
2. Use the MC33883 to drive the NMOSFETS. The IC will have a constant 12V supply and I'll connect the negative terminal of the boost cap as shown in my attached figure. Notice the differences between my figure and figure 1 on the first page of the datasheet.
   
3. Use 2 UCC21520s and Alec's charge pump circuit.

Right now I plan on testing numbers 1 and 2, and only go on to 3 if necessary. If option 2 works without a hitch, I think I'll go with that as it'll be cheaper and take up less board area.

As always, any input is welcome.

I'll report back with the results. It may be several weeks though.