Controlled switching of FETs

is

h,

=A0That's

Interesting... thnaks. In my case the control is all on the one board. So performance is not improved by the 2nd ground wire, (maybe even worse?). But the monitor signals would get an offset. More of an annoyance.

If you had a constant load you might be able to 'tune' out the lead inductance with a series RC in parallel. But it sounds like you have it all in hand.

George H.

I was hesitant--not having done it--but all the cases I could dream up were okay, and I couldn't figure out any reason not to. So, I did, and it makes nice ramps where ringing impulses once reigned.

Too slow and you'll pop the FETs, but 3uS is plenty safe. 10uS is safe, usually.

-- Cheers, James Arthur

=A0That's

DS?

=A0That's

Hmm sure I could/(should) have run differential monitor lines.

George H.

oops, I meant drain. as IN D. sorry..

Jamie

Thanks. I noticed those just now when I was searching DigiKey.

For one set of loads I need ~20 milliohms, max. (I try to stay under

Some people have the weirdest ideas about grounding.

John

=A0That's

Oh. Yes, that's what I meant above by "adding ferrites." Those were to the +24v main, not the individual loads. It's pretty similar in effect though.

That helped, but still not nearly enough. There was a further concern, too, that end-users installing the upgraded units would not install the ferrites.

Oh what a web we weave, when first we practice our grounds to cleave.

-- Cheers, James Arthur

this

Cool! MOSFET versions of the LM395!

Cheers

Phil Hobbs

this

Interesting! But your driver has to be able to sink clamping current. I do similar things in chip designs (clamping like that in a relay driver ~1995), but there I have the advantage of a separate supply to manage the protection. ...Jim Thompson

rote:

ould

You mean "source" clamping current, right?

A series resistor covers that. That's also the overload feedback--the device clamps its gate terminal to its source, and you monitor the voltage. Or current, I s'pose.

-- Cheers, James Arthur

this

Both directions, actually. You have to handle basically I=(VTH+OverDrive)/(100 Ohms) ...Jim Thompson

should

t.

The gate terminal's hi-z in normal operation, and you just have to drive the usual input capacitance. That drive peaks at V(drive)/ (Rseries), then settles to zero.

On overload, the FET clamps its own gate terminal to its source terminal (usually GND, in a sane world). That's when you've got to source current continuously.

I think we're probably saying the same thing.

-- Cheers, James Arthur

Ummm, twisting the supply and ground wires increases distributed C in the sqrt(L/C) resulting transmission line, lowering the impedance.

hat's

Local bypassing : Please view in a fixed-width font such as Courier.

. . . . Ii Io . ---> --->

. . 24V >-------LLL-------+------------ . ^ ^ | ----> | . | | | | | - switched . C | | I | | | . Vi Vo =3D=3D=3D | f | | | load . | | | - . | | | ----------------+------------+ . Ii .

Believe it or not, winding the choke with Litz wire will yield better performance too.

At these currents, the C of the cable, tens of pF per foot, doesn't matter. The advantage of twisting is reducing L.

John

this

should

You have to source when the FET is trying to be off and you are commanding on. And you have to sink when the FET is clamping (on when you're saying off). I saw a +/-20mA spec in that first device. ...Jim Thompson

The act of twisting does not reduce L. It makes the wires into a controlled impedance, and reduces EMI with opposing fields.

It does reduce L. For a given current, L is proportional to the volume integral of B**2. By twisting, you make the field alternate rapidly in space, so it falls off more rapidly with distance. (Laplace's equation and all that.) The big win is running the two next to each other, of course, but twisting helps a bit more.

Cheers

Phil Hobbs