UCC1806 skips pulses at light load

I have a UCC1806 that skips pulses at less than 20% of max load. Typically they go down to 10% of max load. This one has an unusually large output inductor which should extend the min current lower. When it skips a pulse pair the output inductor goes discontinuous and causes spikes and high ripple on the 5 V output.

Here are the parameters of the suppply.

1) push pull topology typical hookup with no current slope compensation. 2) 44(ct):24(ct) transformer 3) Input 30V : output 5V @ 4A max 4) Output inductor: 100uH, low esr electrolytics at 400uF on the output 5) current sense transformer 1:75 with 37 ohm burden resistor and a 499 ohm with 100pF spike filter before the isense pin. 6) Ilimit pin is set to 5.6V though some 200K range voltage divider from a 10V reference (forget the exact values) and is decoupled with 2uF. 7) SD pin is tied low. 8) Sync and from a crystal osciliator and the Rt is tied to vref and Ct is ground (external clock operation). 9) switching Freq is 100KHz

Is there any indication from the above that could cause a problem. @800mA output current the current ramp is about .25V and clean. Then all of a suddend evey other pulse pair stops occurng reliably. raise the output curent to 1 amp and above and the supply acts normally.

any ideas on where to look?.

Reply to
Mook Johnson
Loading thread data ...

Layout, decoupling and wherever switching noise can be generated - the output rectifier/freewheeling diodes are where I would look first by gluing an RC network across them.

In my experience there *will* be a "needle" on one of the inputs to the PWM logic. Or the ground pin!

Reply to
Frithiof Andreas Jensen

check out the current sense pin. At low load, the threshold of the current sense comparator is low (by definition) so a spike on the leading edge can (and often does) trip the comparator, terminating the pulse as soon as it begins. Some controllers include 100 - 150ns of Leading Edge Blanking to avoid this issue.

Its also why there is usually an RC filter. you have 1/2k and 100pF, so about 50ns or so, which is probably too low. you will need to probe this node carefully (coax probing technique - see Linear Technology AN47). As FAJ mentioned, layout is pretty important. The cap needs to be right at the controller. ditto for timing cap.

Also check the oscillator cap waveform, a spike can get in here and seriously annoy the x806. You ought to have a nice solid 0V plane under the x806 too.

BTW the current sense spike mostly comes from the gatedrive (some too from the xfmr primary & reflected secondary capacitance). When using a current sense resistor, you can eliminate it with an npn/pnp complementary emitter follower; tie the pnp collecter to to the switch return (Source), feed the npn collector from the smps IC supply via a small resistor (eg 10R), and place a 100nF cap between the npn and pnp collectors. draw the circuit, and trace the current paths - you will see that the spike now wont flow thru the current sense resistor (but the

100nF cap current does).

As you have a CT, this is even simpler (assuming the CT is tied to the Sources of both primary switches) - connect the x806 0V to the switch side of the CT rather than the 0V side; now, as above, the gate drive current wont flow thru the CT.

HTH

Cheers Terry

Reply to
Terry Given

First time I've seen this but it makes perfect since (parden the pun).

Thanks for the tip!!!

Reply to
Mook Johnson

Did some research and found that the output inductor was made with ferroxcube 4C65 material and an ungapped 2616 pot core with 21 turns. Dont ask where they get it because I cant find a datasheet on ferroxcube.com. I had to piece together the info from various datasheets.

I did the H calculation and found that this core was well into saturation BUT the residual permiability was enough to provide 10uH of inductance.

Talk about a serious swinging choke. 100uH at low loads and 10uH at full load. Seemed odd to me. I never used one like this before so I rolled a torroidal MPP core that will not saturate, but haven't had a chance to try it yet.

Keep the good tips coming.

thanks.

Reply to
Mook Johnson

Well spotted.

swinging chokes usually help with light load regulation, as ripple decreases with increasing L, thereby decreasing the load below which inductor current goes discontinuous.

Do a "splat test" on your choke, by charging up a big cap (I use 35V and several mF, as it comes out of a bench supply), then shorting the choke across the cap and meauring I. An ideal choke gives a straight line, dI/dT = Vdc/L (until the energy stored in the choke, 0.5LI^2, is not small compared with that stored in the cap (0.5CV^2).

If you know the L you should have, at the current it needs to operate (you do) then you can calculate the cap/voltage combo that works (choose Ecap >> Echoke, 10x is enough, 100x is fabulous). I use low voltage and lots of cap cos its nice and easy, but I've also done with with mF at

600Vdc (yeeha!) for some monster chokes

what you see in a real choke is a nice straight line, until the core begins to enter saturation, when the slope increases (usually very sharply). The straight line this turns into is the leakage inductance.

when you use a material where mu varies strongly with H (eg iron powder, steel), you dont get a very sharp knee, but with gapped ferrite the knee is quite sharp indeed.

this test gives both L(I) and Isat. a digital scope makes it trivial, but if you only have an analogue scope, make a little LF osc with a 555 (eg 10Hz), use a monster FET(s) and a current sense resistor(s) (ware the peak pulse power)

Cheers Terry

Reply to
Terry Given

Hopefully, once you have sorted out your inductor things might improve. However...

Look at the block diagram of the IC. Peak current limiting is set by clamping the output of the voltage error amplifier using the internal PNP transistor which is referenced to the CURLIM pin. Max output of the VEA is

4.5V so setting that pin to 5.6V doesn't really achieve much.

CURLIM is shorted to ground during undervoltage lockout with an internal 'switch'. There is a possibilty that you are going to have reliability problems if that switch is going to have to repeatedly discharge your 2U2 decoupling capacitor. You shouldn't really need one.

Unless you take some action to actively reset your current transformer then there is a very good chance that it is going to saturate. You will loose the current sense signal and things will either go apeshit or blow up! Thou shalt not use current transformers at greater than 50% duty cycle.......

Your 'steady state' switch duty cycle might be at 30% but, if you are using one current transformer to sense current for both switches, either in the supply line or source return, then it will be operating 60%. Under transient or fault conditions it will go up to 95% and more.

DNA

Reply to
Genome

Thanks for the tip. I'll keep pecking away.

This is an old design and they finally hooked a sensitive to it and the 5V ripple was unacceptable. when someone (me) finally looked i found these problems with the PS. Talked with the floor guys and they said this is been like this for the last 10 years or so and no one had a problem with it.

ahhh sustaining.......

Reply to
Mook Johnson

Obviously not broken then. Stop trying to fix it immediately!!!!! That's why the company has a returns and repairs department.

DNA

Reply to
Genome

Don't jump to 'wrong' answers without checking.

formatting link
formatting link

I won't claim it's a good design but it might be a workable one.

Note that 4C65 is described as a low permeability ferrite. ui is quoted as

125, for power ferrites the value is in the order of 2000 and above. It doesn't specifically say it's suited to power inductors but that does not mean you might not use it. Work out the minimum number of turns......

N = LIpk/BpkAe

N = 100u.4A/300mT.80mm^2

N = 17

So..... 21 is not too far wrong........

One of the things that isn't in the above information is an effective permeability or Al value for a 4C65 core set, gapped or ungapped, but you can take a guess based on the initial material permeability of 125

Al = uo.uiAe/Le

Al ~ 300nH

Which makes the inductance approximately 130uH.

It's not 100% brilliant but it's no toooo far from being a workable solution. Bear in mind that two halves of an ungapped core will not achieve ui. The effective ungapped permeability, ue, will always be less. 3F3 is ui=2000 and an ungapped core set has ue=1500 so things have already got 25% better.

What are the chances that the purchasing department found out they couldn't buy 4C65 so someone decided to use 3F3 ungapped and did not bother checking things?

DNA

Reply to
Genome

They are trying to use this supply for a more sensitive A2D measurement and the ripple when this thing goes discontinuous is in the 250mV range on the

5V rail. way too high. Instead of putting a bandaid like a LC output filter, I thought I'd fix the problem.
Reply to
Mook Johnson

Quick fix? Put a ballast load across the output, to stop it going discontinuous. Forget trying to redesign a one-off problem - life's too short.

Reply to
budgie

Simple thing in newsgroups is......Not my problem then.

DNA

Reply to
Genome

Didn't mean to come off as harsh. With your tips I did get the supply to operate according to design.

much appreciated.

Reply to
mook Johnson

MONSTER, it's only the middle of January. I've dibbed me dibbs and now I get to use lots of swear words.

DNA

Reply to
Genome

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.