Andrew ZonenbergNov 12

Samples are here and Andrew's Back Room IC FA Lab has a new project: a large lot of HMCAD1520s from the ThunderScope project.

My lab assistant helped me catalog them and we got 30 in inventory before having to pause for dinner and bath time, which should be more than enough to get started. She had lots of fun picking up the parts, putting them in labeled tubes, and putting the tubes into the correct location on the sample tray.

For somebody who loves stickers as much as most 4-year-olds I was surprised she didn't want to label the tubes. But she had trouble getting them correctly aligned and not overhanging the lid and when I corrected her she got frustrated and asked me to do it.

But hey we're not a fancy professional FA lab here, it's a basement shop. If a preschooler shows up looking to help, we'll put her to work. It might be the only applicant we get.

Anyway, my understanding is that all of these parts were rejects from in-house testing on a larger lot of ADCs procured through sketchy supply channels. Multiple failure modes were observed but all generally manifested as high current draw or shorts on supply rails.

So let's start collecting some data on these samples.

Show threadAndrew ZonenbergNov 12

Before anyone asks: the sky gods smiled upon me (by Pacific Northwest standards).

My camera stayed rain-free for long enough to take this photo!

More clouds are moving in and it's supposed to start pouring in a few hours, so that's the extent of my aurora viewing tonight. Back to the lab.

Show threadAndrew ZonenbergNov 12

So, given the failure mode is suspected internal shorts, I'm going to definitely want to measure resistance from every pin of a few samples to ground.

I may cut that down to only measuring power pins on subsequent parts but I want a baseline to start.

But before I do that, let's get high-res package photos of a known-good part and one of the suspect parts and see how they compare.

Show threadAndrew ZonenbergNov 12

First, a known-good part from Digikey.

The top has the part number, a horizontal line, and the lot/traceability code KO68 which does not appear to be a standard JEDEC date code.

The bottom has a streaky mark on the thermal pad whose origin is not obvious, plus witness marks from factory test on each contact.

The thermal pad has a rectangular grid of 2x6 marks, while each of the outer lands has a single mark.

Show threadAndrew ZonenbergNov 12

Closeup of the probe scrub marks.

Apologies for no scale bars, pyuscope is derping today and is throwing Python exceptions when I turn on the scale bar and I don't have time to debug them. When I do the long-form report I'll add them in post to the images I include in the report, but for live-tooting the analysis it'll slow me down.

Show threadAndrew ZonenbergNov 12

Next, on to the failed parts. This is the first one I pulled out of the bag, designated 6A1 (position A1 of tray 6).

It looks *significantly* different and I am confident beyond reasonable doubt that they were not made on the same production line.

This is certainly a red flag, but if these are NOS and have been sitting around for a long time it's entirely possible there could have been changes to packaging vendors in that time. So it's not *proof* they're fakes (although combined with the fact that the parts don't work, it's certainly strong evidence).

The horizontal line on the package between the part number and lot code is missing. The font looks different. There's a slight space between HAD and 1520 on the legit part that's not on the sketchy one. The pin 1 marker is a different size.

On the bottom, the curvature of the pin 1 mark on the thermal pad is slightly different. The surface quality generally look... dirty? but this could be a sign of a NOS part that's been stored for a long time in poor conditions too. The size and shape of the solder lands are slightly different. The radius of the non-pin-1 corners of the thermal pad is larger on the good part than the bad.

Probe witness marks are visible but they look very different, for example the ground pad has a 5x5 grid of marks rather than 2x6.

Show threadAndrew ZonenbergNov 12

There are *multiple* sets of probe witness marks on the QFN pads. There's a single-point scrub looking pretty similar to that found on the known-good part, overlaid with multiple impacts from an equilateral triangle shaped pin that appears to have pressed straight into the part with minimal lateral scrubbing motion.

The 5x5 grid of witness marks on the thermal pad is also made from this triangular impactor.

Show threadchoofa
@azonenberg have you covered your photography setup before? what are you using for your camera and lenses?
Nov 12 at 7:41amWeb
Show threadAndrew ZonenbergNov 12

@choofa Low mag: Leica M80 stereo zoom microscope with 1x plan achromat objective, Leica Flexacam i5 camera. This provides 7.5 - 60x magnification range with continuous zoom, although it has a snap-in detent to lock to calibrated magnification stops that I engage when taking photos.

This uses the Leica Enersight software, which puts a red scale bar and timestamp in the bottom right corner of the image.

Show threadAndrew ZonenbergNov 12
@choofa Seen here
Show threadAndrew ZonenbergNov 12

@choofa For high mag work (most of the stuff in this thread): Labsmore LIP-X1 CNC imaging platform.

This is a custom system consisting of a 3-axis CNC motion system with a Mitutoyo VMU machine-vision microscope mounted to it.

I'm using a BigEye 10000KPA C-mount industrial camera and have a range of objectives on it, all Mitutoyo plan apo: 1x/0.025, 5x/0.14, 20x/0.42, 100x/0.90.

I also own a 10x/0.28 objective for it, but due to lack of a fifth position on the turret it's not currently installed. Usually I can do everything I need with the 5 or 20 so it doesn't get taken out often.

The X1 uses pyuscope software which puts the black bar along the bottom edge with scale bar in the left corner (although this stopped working after a recent update and I need to figure out why)

Show threadAndrew ZonenbergNov 12
@choofa X1 system on aftermarket vibration damping table from minusK
Show threadAndrew ZonenbergNov 12

@choofa For video work and macro/long shots I have a Sony A7R II with a SEL90M28G macro lens for closeups and a SEL2470ZB 24-70mm zoom lens for general purpose use.

I also occasionally use my phone (CAT S62 Pro for older stuff, now replaced by a Pixel 9 Pro) for quick stuff.

Show threadchoofaNov 12
@azonenberg much obliged. thanks!
Show threadchoofaNov 12
@azonenberg weird. last night I only saw the images. checking now, the text posts came through. thanks for being so thorough <3