LT3757 gripe

Maybe I'm missing something. What does this circuit have to do with the LT3757 switching controller?

Den mandag den 25. april 2016 kl. 21.39.29 UTC+2 skrev Tim Wescott:

missed a pin or something? Datasheet claims it can switch several nf in 20ns

-Lasse

yes, but it doesn't say it can drive a DC load like a 100 Ohms to ground.

In fact the block diagram on Fig 1 shows current limiter on the gate driver G2.

So it looks like you get a slug of current from the cap on pin 8 and then the current limiter kicks in.

Since this device has an active pulldown., why do you think you need the 100 Ohms to ground?

Mark

Goes between the gate drive pin on the LT3757 and the FET gate.

What's the FET (or gate capacitance) Did you try it w/o the speed up network?

George H.

Double, triple, quadruple-checked connections. I'm not sure I believe the data sheet any more.

Why are you using a JFET? Isn't that going to conduct with positive gate voltages? Or is that the only symbol you could use with ascii art and you are using a MOSFET?

you are right, and 7.2V typ / 40mA typ = 180R

-Lasse

Because I'm boosting to 25V, and in the past I've seen problems where the G-D capacitance of a FET would hold the gate up when there was a large voltage swing on the drain. This, in turn, would cause a slow turn-off and high switching losses.

In extreme cases, if I recall the discussion at the time correctly, the internal gate node of the chip could be at elevated voltage even when the gate pin was held tight to ground -- so a network that'll pull it a bit negative can be a Good Thing.

The network, when it is doing it's job (and not being driven by an LT3757, apparently) trades off a reduction in the DC gate voltage for a much lower (lower than ground, even) voltage at the gate pin on turn-off.

I'm getting less and less enamored with the LT3757 every day, at least for this application.

what is you input voltage?, seems like you can bypass the regulator if it is less than 17.5V by connecting it to intvcc

-Lasse

Sounds like it doesn't appreciate the 110 ohm load to ground.

What fet are you using? The specs on the driver look reasonable, 20 ns into 3.3 nF.

I'm driving INTVCC with a 12V regulated rail. And that looks just fine.

It's a MOSFET -- I figured that was implied by the job.

Total gate charge is 40nC at 12V for the FET I'm trying now.

Part of my frustration is that it works differently in simulation than it does in real life -- in simulation it's vastly better than in real life, with a much faster gate voltage transition. The speed-up network was, in part, an attempt to make the damned thing work more better.

In real life it works better without the network -- but I think that's because of the oddball behavior of the gate voltage.

Is this on a PC board, or is it a breadboard? Grounding and bypassing are really going to matter.

I've been happy with the LTC switcher parts, although I haven't used this one.

This worked great:

HV/28S840_A19.pdf

It's a PC board, although I won't disclaim any possibility of layout goofs.

I would have said exactly the same as you about LTC parts before trying this one. But I'm not sure if it's the part or me.

Yup. I disconnect the ground connection to the 100 ohm resistor and it gets much happier -- but still only 75% efficient, and I want more.

The Si7116 and DMT6007 -- as soon as I realized that the swimulation isn't accurately reflecting reality I got a bag of different FETs from DigiKey with the right footprint and started a program to swap them all in. I was on the second one when I noticed the gate drive oddity.

The actual part just seems to have a much wimpier gate drive than either the data sheet or the simulation would indicate, oh joy.

Could the problem be somewhere else, like inductor saturating or something? What's getting hot? How does the gate waveform look?

I've never seen that network applied to MOSFETs before.

Is this related to your converter from a previous thread?

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: