#67 Audient iD22 part 2: the repair

 And now, getting back to the issue affecting this unit... As pointed out in the ad, the LT3439 apparently used to have a piddly little heatsink glued on top of it, but it got detached. Now, there's a chicken-and-egg type mystery here - did the heatsink glue fail because the chip overheated in the first place? Or did the chip get fried because the heatsink fell off? Regardless, investigation showed that one of the two switching transistors inside the chip seems to be stuck in a conducting state, despite the transistor itself not being shorted, though. But since it handles all the power for the analog circuitry, it's no shocker (no pun intended) that only the digital ins & outs work. 

For the sake of "future-proofing" though, this will need a more substantial heatsink (well, as substantial as can fit within the confines of the enclosure), not to mention a much more reliable mounting method. I wonder why those eight solder-filled plated through-holes are placed around the chip, as part of the ground plane... Can you imagine any sane reason?..

Either way, for the above-mentioned reasons, the through-hole components around the chip get moved to the underside of the board (plenty of space), in order to allow as big a heatsink as practical, on the top side, where the power chip resides.

First thing's first, though - removing the dud chip. Being particularly lazy and not wanting to walk all the way to my hot-air station, i figured i'd bend all the pins up and off the board, then see if i can melt the solder under the chip through the solder-filled via right under it. Except that... ehrm... Well, i never got that far.

If that chip's underside metal pad was ever, EVER, soldered to the mainboard, i'm the Pope. And yet, they tried to fix this with the band-aid comprised of the glued-on heatsink. You have GOT to be kidding me... *sigh* This is really embarrassing... This, from a company purported to have made studio mixing consoles and whatnot, and in a unit that isn't exactly the cheapest around (quite far from it, considering the feature set). And which happened to also be their flagship interface at launch, no less.

There's the chip, with the stock(?) heatsink. And with a USB plug, for scale. Nice going, guys, top-notch engineering...

Since "ain't no kill like overkill", let's do this right: clear the via of the factory lead-free solder, chuck in some flux, slap on pleeeeeenty of solder in there, and whack a sizeable heatsink on top. No glue this time - since those mounting-hole vias are available, why not strap it on with a piece of copper wire? And add a dab of Arctic Cooling MX-4, to maximize heat transfer to the heatsink...

Time for some power-testing, then. Aaaaaand... no change at all. Still getting 0.9V or so on the three "small" rails and around 5V on what should be the phantom power rail. All the diodes check out fine, and there seem to be no shorts to ground from either of the rails, on the secondary. And considering the low voltages involved, failed windings in the transformer are virtually impossible.

Drastic measures then - since it's an easy thing to try, i pulled out all four of the 33uH filter inductors on the secondaries of this power supply, in order to see what sort of voltages come up, with no load whatsoever. Turns out the +/- rails come up to +/-17V or so, the third rail goes up to about 11V, and the raw phantom power jumps up to 60V. Those are all ballpark values of what one might expect in such a context. But which one's dragging things down?.. Time to put them back one by one (well, maybe apart from the pair on the symmetrical rails).

First inductor back in was the one for the "5V" rail, which brought the raw voltage down to 6.7V. The analog rails came down to +/-16.3V, and the raw phantom rail to 52V or so. Next, the pair for the opamp rails. That brought the "5V" rail to about 5.4V, and the analog rails to +/-13.6V, while the phantom rail came further down, to 50.5V-ish. Heatsink on the LT3439's getting moderately warm, but not warm enough to burn my finger, even on prolonged contact. Resoldered the phantom inductor as well and... umm... it... seems to work fine now? Seeing 590mA draw at 12.2V from my bench supply.

Now, granted, among the above-mentioned things, i also replaced the little electrolytic next to the LT3439, sort of "just in case", but even after that, output voltages were still in the dumps. I don't get it... But hey, phantom power rail's at 47.6V (and sloooowly creeping upwards slightly), converter supply rail is a solid 5.24V, while the analog rails sit at +13.7/-13.6V. The copper wire holding down the heatsink is getting quite warm-ish, but i'm quite positive the whole setup will be loads more durable, and run considerably... well, if not "cooler", than at the very least "less baking-hot" than the stock setup (with a pathetic U-channel shaped heatsink about half the size of the ground plane outlined by the eight solder-filled holes around the chip). 

Would i be mean or unfair, if i called this a marginal design - at the very least, from a thermal standpoint? It just makes me shake my head... I'm almost surprised i haven't seen more of these dead and put up for sale. And that's not a cheap PWM converter either, even in a 2500pc reel, they're still about $4 a piece. 

Ah well, what can ya do? Maybe not buying one, or selling it off before it kicks the bucket, i suppose. The interface works again (and should stay that way, at least for a while), but that's one extra reason not to keep it around for long.