A couple of questions re: LM3478/88

Suggest you buy the eval board, substitute your parts, and make some measurements. I have just spent a couple of days with the LM5574 and it behaved nothing like their simulation!

Yep. That's reassuring :-) Ok, that way I could solve the point 2. But for the 1, I would have to test a statistically representative number of chips!

Thank you.

Watch out that you don't fall into the RHP zero trap (continuous mode, >50% duty cycle). Read page 10 again if you think that might be happening.

I was never very fond of WebBench. It always came back with "cannot be done" or whatever which in all those cases was, to say it politely, baloney. Just build it. When I used the LM3478 for the first time it fired up right away and did what I wanted it to do (in that case a pretty extreme Vin/Vout ration of more than 10). For me the only gripe with it and why I didn't use it often is it's highish price tag.

That worst case corresponds to: V_sense = 135mV (minimum) and V_sl =

I don't think this is the reason (although the Bode plot on the Webench goes totally crazy with those increased L...), but I have also tried with *decreased* L (and the V_in at the low limit) - it produces the same phenomenon - the L current grows up to the short-circuit current limit, it looks like the chip totally ignores the normal current limit (the Bode plot is ok).

I have V_in min=3.1V, max=6V, v_out=30V, I_out=0.35A, and I could not find a cheaper part.

Indeed, it looks like I'll have to build a prototype to solve the point 2), but what about the 1)?

Looking at the functional block diagram (page 9), I've got an idea: what if the output is in undervoltage condition (that the webench plots suggest)? The output of EA might then swamp the input from the slope compensation, and the PWM comparator would not get triggered at all; the switch reset would happen on the "short-circuit" condition only. But what is V_sense actually, then?

Absolutamente. IMHO WebBench isn't particularly useful, at least not for me. Calculate it by hand and build one. This is no thwack in National's direction, I have just never found these fancy "automatic design" scripts and gizmos to be very useful, regardless of the manufacturer.

As to 1): What is V_sl? And where does that formula and the factor 0.9 come from?

Where do you see V_sense? This chip has two major controls: FB which is typically used to set the output voltage (although I've used it for all kinds of other things). I_sen which limits the max current through the inductor.

So, as long as the voltage hasn't reached the target where FB would be at 1.26V this chip will run "full bore". Same if you design it so it could never get there (wrong inductor value etc.). IOW it will keep "scooping" the max amount of energy on every cycle, going to the current limit on the inductor every single time.

By V_sl I mean the "Internal Compensation Ramp Voltage", specified on page 4 of the LM3478 datasheet.

The formula is the formula for V_CS, it comes from the section "PROGRAMMING THE OUTPUT VOLTAGE AND OUTPUT CURRENT" on page 15 of the same DS.

The factor 0.9 is the maximum duty cycle (90%) in my design, corresponding to V_in = 3.1.

By V_sense I mean the "Current Sense Threshold Voltage" specified on the page 3 of the LM3478 datasheet. I understand that it is related to the I_sen in some way...

The question is: which current limit? Such that multiplied by R_sn it gets to V_sc "Short-Circuit Current Limit Sense Voltage"? Or to something lower and still controllable?

AFAIU that related to the compensation ramp to avoid sub-harmonic pumping. Are you going into continuous mode with this?

AFAIR it is for loop compensation.

That has always bothered me with this device. I do not allow inductors to get too close to saturation and also don't want to oversize the core for cost reasons. It is better to measure the actual output current and throttle back via the FB input. That is what I did. Detecting a short via the current sense pin is IMHO quite scary so I didn't do it.

Yes, the fixture is supposed to be in continuous mode, except for (very) light loads (the load can vary from 0.01 to 0.35A).

-- Regards,

Andy

Ok, that's a whole 'nother ballgame. I stayed clear of any CCM. You'll have to contend with RHP zero loop compensation and all that nasty stuff. If that goes wrong National had (probably still has) a forum on their site. Their app engineers respond quite quickly there. Or you could contact Fred Embuscado at National, he knows this stuff very well.

Ok. Thank you!