UCC gatedrivers in MSOP8 package / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

UCC gatedrivers in MSOP8 package

1 2 next

Move Thread

LAN_403

ragnar

Sat Sept 23 2006, 05:41AM

Registered Member #63 Joined: Thu Feb 09 2006, 06:18AM
Location:
Posts: 1425

hi all,
I recently had the opportunity to sample some MSOP8. Why, you ask? The three packages in which UCCX732Xs are available, and their respective allowable power dissipation figures are:

P/DIP8 - 350mW
D/SOIC8 - 650mW
DGN/MSOP8 - 3000mW (powerpad)

The MSOP PowerPAD-8 (DGN) becomes very attractive if it can be properly heatsunk. That's my issue.

I've tried milling out a single line between tracks to form a board, but the carriage of my engraver deflects slightly as it applies downwards pressure on the tool and nosecone. The result? Not-quite-properly-enclosed tracks, with diagonal gouges wherever the tool goes up and down. There were shorts between the 'tracks' which had to be removed with a knife, then particulates blasted out with 12V.

Next, I tried bending the pins up so they wouldn't contact the coper, and soldering the chip straight down to the 3oz PCB. This was satisfactory, but very clumsy and left the pins very fragile.

My third attempt involved soldering a chip down to a thin strip of 3oz copper-clad such that the pins extend over the edge, then adding some heatsink 'wings' either side. This proved very quick, kind to the chip... but still a royal pain in the arse.

I'm looking for suggestions on what you guys do with MSOP8 w/ powerpads. I have no problem soldering the chips down, and I haven't killed one yet. The termination of the legs and time needed to mount the chip is frustrating me.

To aid any discussion, I wonder:

- Am I going overboard with the size of the heatsink?

- Should I even solder the PowerPAD on the bottom at all, and try white thermal grease instead?

- Does the only solution seem to be to get boards made up commercially? I would have thought my large area of 3oz copper is thermally superior to thermal vias etc.

ragnar

Sat Sept 23 2006, 06:31AM

Registered Member #63 Joined: Thu Feb 09 2006, 06:18AM
Location:
Posts: 1425

"For my next trick..."
(kaboom)

By not attaching the heatsink 'wings' I save a lot of thermal cycling on the chip. It was also relatively straightforward to add another strip of cladding, the copper segmented to allow a tag on the left for IN and a tag on the right for OUT.

I have detached the ENBL pin as it only gets in the way when I'm trying to solder IN. =-P

It's worth noting that at high frequencies, the UCCX7322 (noninverting) is FAR superior to UCCX7321 permutations; the duty cycle is much closer to 50% and the chip stays cooler indicating there's less of a cross-conduction issue inside.

Now I'll see how he handles driving a load... And then I may as well compare him to a DIP8 chip!

Wee.

By the way, the first pic below is a mockup of how the MSOP8 would have mounted on the milled board if the tracks were a little more reliable.

pwn4ge

PS, I didn't intend for the chip to look like it was being crucified...

Dr. Slack

Sat Sept 23 2006, 07:29AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

How about superglue it to copperclad in 'dead fly' orientation, taking the connections out on wires. Then you can do all sorts of experiments to find the least inconvenient way to heatsink. Try a bar of copper across with heatsink goo, or solder a staple of thick copper foil to the chip and to the board, or solder several staples of thickish tinned copper wire.

Sulaiman

Sat Sept 23 2006, 08:20AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3140

First - I have no experience of using powerpad devices
but lack of experience never stops me

Why not follow the manufacturers reccomendation

or drill a 2.4mm dia (1/10") hole through the board and solder a solid rod of copper to the powerpad and a heatsink on the back of the board?
Be the first to find a commonly available copper rod type thing.

I think I would try to use the IC upside-down and solder copper foil 'wings' on top?
or find a copper coin, file/sand a flat on one side, solder it edgewise to the powerpad and epoxy at/around the IC for strength.

Overall I suspect that if you're having pcb problems then I would use a TC4421/TC4422 instead, very easy to heatsink.

Avalanche

Sat Sept 23 2006, 09:05AM

Registered Member #103 Joined: Thu Feb 09 2006, 08:16PM
Location: Derby, UK
Posts: 845

The thermal vias are pretty good performers, we use them on boards at work. You just have to make sure that they lead to a very large area of copper on the other side of the board! On one of our boards, the 5v regulator uses vias to the back of the board, where it has about 120 square cm's of copper to play with. I couldn't kill the regulator, even when I overloaded the 5v line and left it running all day. So I would probably be tempted to try the vias if you're having some commercial boards made up.

Marko

Sat Sept 23 2006, 09:49AM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

I always wanted MSOP's but wuld always give up deciding i'm not good enough with SMD-pcb etching skills :(

If you ask me, I would simply use a piece of cc board, etch pins etc, and drill a big hole under the chip (big enough for solder to pass trough). Large amount of solder would conduct heat ot other side.

You wouldn't need to use multiple pieces of board that way =)

You could also heatsink the other side if you wished.

Then I would just solder it and dip the solder from other side, using it as gnd and upper side as Vcc.

I would also make sure that mosfet (or IGBT) isn't but few mm away from the UCC, otherwise inductance would be a problemat frequencies and powers I'm guessing you want to run at! (If you use resonant drive, I guess you don't need such monster cooling, just use lower voltage and higher Q!)

You can see some of waveforms I got with low-inductance drive in my class E SSTC thread

I like curcified UCC thingie, but am somehow afraid it won't be able to dissipate heat well enough.. =)

Avalanche

Sat Sept 23 2006, 10:54AM

Registered Member #103 Joined: Thu Feb 09 2006, 08:16PM
Location: Derby, UK
Posts: 845

I'm not sure how well solder would conduct heat, but that might work. I suppose an actual copper heatsink directly to the underside of the chip would beat anything, but drilling the hole and soldering it on would be the tricky bit, because the 'big' hole can only be about 2mm absolute maximum.

I had an idea though, why not solder tiny brass machine screws to the bottom of the chip first, to form a mounting point for a heatsink just like RF power transistors use. Then the head of the screw can pass through an accurately drilled hole in the board, the chip sits flush on the top of the board, then you bolt your heatsink onto the protruding screw underneath. It'd be fiddly soldering the screw on, but once you do that the rest should be easy (plus you can lay it out in a normal PCB editor).

In case I haven't explained it very well, here's a pic from an RF amp I constructed a few years ago, where the 'bolt' passes through the board and the heatsink screws to it. The only problem would be finding a bolt with a deep enough head to pass right through the PCB.

Marko

Sat Sept 23 2006, 12:24PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

2mm hole is more than enough to put the solder trough, and it's thermal conductivity will be very little restraining compared to conductivity of die/package itself.

After thinking a bit it seems that simplest approach would be best approach anyway - just soldering the chip 'factory' on the former side of board, and heatsinking the side. Just with simple hot air fan it would be as easy as first technique.

You anyway can't connect both sides of board together and use a low inductance design with Vcc/GND planes so it is practically the same thing.
(bp's second board looks somewhat like such.)

Bolt would be tricky to fit on the chip but it may be worth for little more thermal conductivity. You would have to care that head is smaller than UCC (maybe saw it off) and take care that IC doesn't hover over the board being unsolderable..

ragnar

Sat Sept 23 2006, 01:12PM

Registered Member #63 Joined: Thu Feb 09 2006, 06:18AM
Location:
Posts: 1425

The "crucified" UCC seems to work quite well. There may well be a hopeless lack of dissipative surface area and thus wasted thermal potential, but it seems to be behaving much better than the DIP8s.

Interesting to note at 13.56MHz, the output from the UCC27321DGN looks like a little spike, compressed to maybe 15% fo the whole cycle, whilst from the UCC27322DGN I get a semi-acceptable squarewave with bouncy edges, whose positive swing occupies closer to 55% of the cycle.

The big issue now is synchronisation between devices; even using a 74HC14 inverter/buffer with tweakable RC delays to each UCC, I cannot guarantee synchronism; twiddling the trigger on my oscilloscope makes the second UCCs output appear to shift out of phase, and I have no way of telling when my scope is telling the truth.

I don't think real people try to fudge fullbridges at this frequency, so I'll move to using just a single chip and perhaps multiple primaries in parallel.

Marko

Sat Sept 23 2006, 01:56PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

With hard driving at high frequencies (2..4Mhz) keeping low inductance in gate circuit is a must, and it makes a really big difference!

I got some nice looking waves with 1ohm gate resistor, and it could be even faster but I am sure I would overheat the UCC that way (IRFP450).

It already gets over 50C after I let it to run for longer period. Heatsinking can't help much there since package conducts heat poorly, and that's why those powerpad's are nice..

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob