LM2676S-ADJ and ripple

[Guest]

Hi all, not sure if this is the right place however it is for my PIC project, so here we go.

I have a power supply problem, I am powering a GM862 with a LM2676 but it fails to power up correctly, I am pretty sure this is due to the ripple (logged at >700mv) from the regulator can anyone look at my design and advise how I can improve it to reduce this ripple so it works reliably across 8v-40v and max 2.5A

Input voltage = 8v-38v
Output Voltage = 3.8v
IOut Max = 2.5A

The schematic is here. http://www.xlsc.co.uk/

Thansk in advance

[mkent]

You should read the NS data sheet carefully. Your boost cap is 10x the recommended value. Your feedback circuit layout is important; keep it well away from the inductor. good luck. Mick

[Guest]

I'm new to switch mode regs I normally use linear regs, it's like a black art to me

Thanks for the pointer.

[rwyoung]

Just a little nitpick but inductors have units of Henries, not Farads. So I think you ment 22uH not 22uF for your inductor.

I also question your choice of diode. Up front I will say I didn't research it much but is it fast enough and of sufficient peak pulse current rating?

Likewise your choice of inductor, make sure it isn't going to saturate on you. In technical terms, your output will be crap.

And like "mkent" said, your boost cap is a bit on the large side, probably playing havoc with the switch waveforms.

Sorry, couldn't help it, just the old lab-instructor in me...
_________________
Rob Young
The Screw-Up Fairy may just visit you but he has crashed on my couch for the last month!

[PCM programmer]

In addition to what the others have said:

It's clear that you're using National's Webench page to design your
switching power supply. I've used it before.
http://www.national.com/appinfo/power/webench/

The problem is that you missed the fine print in the design
that they calculated for you. Your schematic shows only
one 100uf output capacitor. In fact, you must have two.
The chart of component values created by Webench
says this, for Cout:

Numcaps = 2
100.0000 uF,
0.1000 Ohms

That doesn't mean two caps totaling 100uf. It means two caps
of 100uf each, totaling 200uf.

You can't just use any capacitor, either. You must use
a special Low ESR capacitor from their list of recommended
parts. Click on the button for "Select Alternate Part" to see
the list. Also, when you click on that button, you'll a lot
of important specs for the capacitors, such as ripple current
and voltage ratings.

I also don't see any input capacitors on your schematic.
The Webench design calls for 3 caps, as follows:

NumCaps=3
47.00000 uF,
0.3200 Ohms

I've used this Webench software before and it works perfectly
well, but you've got to do what they say. You can look at their
list of recommended parts and try to find a cheaper one that's
not in the list (maybe because of a different manufacturer)
but that meets the specs. Also, sometimes you'll find a capacitor
in the list that is quite a bit cheaper than the others. You've
found a "sweet spot". The problem is that everyone else is
also seeing this cap. If you're doing this design for a company,
you better buy up that cap quickly, or it's going to be gone.
I've seen this happen with Digikey.

[Guest]

[PCM programmer]

[Humberto]

A time ago I had been playing with the LM2676. The promised spec that it will keep
+5V @2.5A output with an input of 8 to 40 it is true in part. For testing we used
the typical application circuit found in the datasheet and we found that the device hold an
output value of +5V but the ripple was innacaptable for our application. To get an output
with a ripple < 60mV we need to adjust carefully the inductor and the output cap value
but this was valid only for a delta V input <6V ie: (22 to 28V)
The output current for all this test was @400mA.

I would like to know any comments regarding the resulting performance obtained by yours.

Humberto

[Eugeneo]

I've used 3 x 2678 supplying triple output volatages with an input of 10-15 VDC. I had to use switching reg. since my voltages were 2.8,3.6,5.0 (RGB). A linear design would have ended up requiring a fan AND a bigger transformer.

Since the output current varied anywhere from 100 to 5000ma per channel, I had really bad ripple/noise that caused cpu resets (software resets when severe spi collisions were detected that couldn't be cleared). I've noticed that using the higher rated esr caps didn't really help during times of high drain. Anyways I solved the problem by putting 8 surface mount caps in parallel on the output side of each channel. The ESR is now much less than spec., but this along with some wire shielding virtually elmineted any collisions.

I was impressed by the regulator. I could short circuit the supply for several minutes @ 1ohm and the current was regulated at almost exactly 5 amps. None of the components were too hot to touch and this was wiith minimal heat sinking. Of course this can't be compared to a linear regulator with regards to heat but still, it's pretty impressive for a to-220-7.

[treitmey]

I also us the LM2676. Great web based design tools. Mine works great.

My parts person had a problem tho trying to find the lead-free version
available. It seems no one to puting there lead-free stuff out for sale yet.

How many days till july-1?