technogeek

Hi guys,

As you probably remember, I'm trying to build a mains powered switcher power supply... Well, I got it to work... sorta.

I made the values for the components for a 60V @0.7A LED bank... Now these should work for a 16V @o.7A bank, right? Here's what happens...

Powers on, and there's a high pitched whine. 100khz shouldn't be heard!! Anyways, the LED's light, which is a good sign. I check the current through them, and it shows 0.3A.

WORSE, when I touch the "negative" lead (the one coming off the inductor), the LED's brighten and the current rises to 0.4A. Meanwhile the pitch of the hum changes in frequency...

What did i do wrong? This can't be normal! Granted, the inductor is larger than it should be for just 6 LEDs, but I didn't think that would be a problem? And what's with the touch sensitivity!?!

In need of some switcher expertise..........................

kchriste

Since you are still with us, let's see a schematic of your circuit with an arrow to the point that you are touching.

__________________
--- The days of the digital watch are numbered. ---

technogeek

What do you mean "still with us"? Did you expect me to electrocute myself, or "get my answer and run"?

Attached is the schematic... Ignore L1 looking like a cap... C1 is 68uF and R2 is 0.324 ohms... the funny X things are my fuse holder pads

Also, if I just touch the insulation and it does the same thing.

Attached Images

switcher.jpg (50.0 KB, 49 views)

kchriste

Well, for starters you need a capacitor across the load (Between the two Xs pointed by the yellow arrow). The ripple on the LEDs would be huge without it and this explains the low average current you are measuring.
Search for "inductor" in eagle or "L-USL3*" to get the proper symbol for your coil when using the ADD part command.

__________________
--- The days of the digital watch are numbered. ---

justDIY

and to hold-up 700 ma, you'll need quite a large capacitor ... I'd recommend several smaller value caps in parallel

__________________
If you don't have a planet, what good are gold bars?

want to contact me directly? gmail gordonthree
check out my project website: http://projects.dimension-x.net
Favorite numbers:
09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0

technogeek

Not according to the app note I have here-

http://www.supertex.com/pdf/app_notes/AN-H48.pdf

current ripple is determined by the inductor

kchriste

Or shock and run!

I see.... I guess that is the more cost effective way of doing it. But without a cap, you won't be able to measure the LED current with a normal volt/amp meter. You'll need a resistor in series with the LED anode and a scope with a floating gnd on the probe; the "gnd" of the probe would have to goto the DC input side of the switcher (C1+) to avoid upsetting your switcher, so be carefull!
I suspect that the reason why your circuit behaves strangely, when you touch it in that special spot, is because of the high voltage & frequency at that point is being coupled into the air and into your circuit via your body. I wonder what your niebours TV looks like when you do this.
As for it oscillating at a lower frequency, make sure that the supply (pins 6&7) are properly bypassed to ground with a cap with a low esr. It is also possible that you are hearing the 120Hz rectified line voltage ripple modulating the 100Khz switching frequency....

__________________
--- The days of the digital watch are numbered. ---

kchriste

A quote from the data sheet:

__________________
--- The days of the digital watch are numbered. ---

technogeek

DOH!

I forgot it! I had it spec'd out, but forgot to order the part and put it on the board! Thanks for the reminder! (big time!)

Vin is definately >2*Vout

I think you probably hit the nail on the head with the small low esr cap though.

I guess my #1 question now though, is why is it not putting the full 0.7A into my LED's? The cap can't possibly be the reason?

kchriste

That'll do it!
No problem!

__________________
--- The days of the digital watch are numbered. ---

Oznog

Damn. I wrote an answer and I guess I hit the wrong button and it's not here.

That design will fail to operate on a fixed freq when the inductor is large for the freq. You MUST use an RC network with a time constant of over 300nS (but not too much over) on the current sense wire. Otherwise, say you have a large inductor for low ripple and insensitivity to input voltage. You shoot for 1 amp current. The inductor only decays to 990mA at the end of the cycle. The 9910 turns on the gate, say the gate charges with a brief 50mA current pulse- that goes through the gate and the current sense resistor. 9910 senses 1040mA and shuts off instantly, skipping the cycle. Happens next cycle too, even though the inductor's decayed to 980mA. Like 6 cycles later the inductor's at 940mA, gate charge makes a brief 990mA reading, not enough to trip the 9910, so the 9910 gets past that pulse then has a nice long charging cycle for once to put the inductor current back up to 1 amp. Effectively it may be operating at a fraction of the intended freq due to cycle-skipping.

The 9910 has a fixed on-time configuration if the Vout/Vin is more than 0.5 which will be stable.

The output does NOT need a cap.

Isn't that inductor huge for what you want to do? Are you even sure it's not saturating? Larger inductors usually come with lower saturation current and higher resistance unless you go with a physically larger device.

__________________
I thought what I'd do was I'd pretend I was one of those deaf-mutes.

technogeek

Thanks for your reply.

Yes it's huge, and no I'm not sure it's not saturating. It just happened to be one I have until I place an order for the right parts.

I'll give the RC network a shot. Your explanation sounds like what's happening. I'm curious why this wasn't mentioned in the app notes? For a test with this "low output" switcher, should I try dropping the frequency to 50khz? That would bring the 1.8mH inductor closer to what the design calls for.

I'll keep everyone posted.

technogeek

Well I did some experiments...

adding a small 33uF cap on the output makes the whine stop completely, and the output current increases to 0.52A. But one of the dangers with using a cap on the output, is what happens if it goes open circuit? All the current gets dumped into the cap and it explodes...

I'm crossing my fingers an RC network on the current sense resistor will do the same. It's also possible the trace length between the negative side of the bridge and the fet is too long, adding extra resistance to the tiny 0.324 ohms it is, reducing the output current..... but I used 0.060" trace width, and from the bridge to the fet is about 1/2".... Although the 68uF filtering cap is on the other side of the board... (would that do much?)

Oznog

The need for an RC network and the 300nS time constant came STRAIGHT from SuperTex tech support. I have no idea why they didn't document it. I've brought it up like 4 or 5 times here now.

Don't use an output cap. This type of buck converter is completey stable without it. With it, there's a bit of a complex response. And it's pointless anyways. The cap would have to handle like an amp of ripple current and have a very low ESR to try to filter out the small dv/dt on an LED as inductor current varies (because there's little variation in voltage even when LED current drops say 20%). Many small caps would not even have the ESR or ripple rating to do anything, and basically you have no need for a cap here.

I believe SuperTex has a some design info on the required inductor size to achieve an acceptable ripple rating.

__________________
I thought what I'd do was I'd pretend I was one of those deaf-mutes.

technogeek

Thanks, roger that..

I'll install a 1k and 300pf cap on the current sense line and see what happens...