Notches in ground planes for multi-power multi-channel board

data

it

I recently did a 3.3 to 2.5 volt "regulator" to power VccAux of a Xilinx FPGA. I used a zener diode. In series. Backwards.

John

Nice... but why not just a regular silicon diode? Better regulation or something?

I do like the trick of using LEDs as a combination means of setting up biases as well as indcating power/activity/whatever.

Or two Schottkys. But I guess the zener costs less.

Be careful. Suddenly someone releases another kind of red (Oh, it looks much nicer ...) and your biases are all off the rocker.

The voltage drop came out right, 0.8 exactly. Plus, the melf package we have in stock exactly bridged the pads of the LDO that didn't work.

Yeah, an led can make a useful bias voltage, and free light.

John

What did the LDO do? The usual, such as oscillating?

biases

It dropped out! Actually, I used an LM1117, which has a dropout of around 1 volt. My bad.

The diode seems remarkably stable over a wide range of cap esr's.

John

John Larkin a écrit :

biases

Then try a TD or a GUNN :-)

message

biases

I used to love TD's. Too bad nobody makes them any more; GPD used to, but I don't think they still do. The fabrication process was insane, but it's something that an amateur could potentially do.

John

message

biases

Happened to all the boutique diodes. AFAIR you had found a source for lower cost SRDs a while ago. But the thing is, if Digikey doesn't carry something chances are it'll go lalaland soon.

There's been a trend in I/C design going on, actually, for quite some time... driven by you parts consumers...

You want to buy a circuit function that requires NO ADDITIONAL PARTS or PINS.

Therefore your LDO can't have the proper feed-forward compensation to adequately retain stability with capacitive loads... the manufacturer fudges by specifying ESR.

So it's your own fault ;-)

...Jim Thompson

Nice and quiet, too, if you can spend some current on it, especially if you use a series string to make higher voltages instead of multiplying it up.

Cheers,

Phil Hobbs

Capacitance multipliers work dramatically better than VRs for cleaning up SMPS crap.

Cheers,

Phil Hobbs

data

it

Hey gotta keep the economy going.

You simply must give us the libretto now.

Yeah but you's guys, the chip designers, typically don't ask us guys before releasing a chip to the public. I would not mind at all if an RC or whatever has to be connected if that makes an LDO unconditionally stable. After all we are all used to that from PWM chips for switchers.

Before doing the architecture have your people call my people ;-)

In the army our drill sergeant told two guys to stop singing. One of them is yours truly.

Takes an extra pin.

My people don't decide the architecture. You should know that well. MARKETING decides ;-)

...Jim Thompson

So when do we get the libretto?

Like in singing a requiem? No way my vocal cords would even get up there.