Name that Ware December 2009

The ware for December 2009 is shown below.

This ware is a guest ware submitted by tmbinc. It’s a fragment of a larger, no longer functional ware, and it’s seen some rough times, as you can tell by the crushed bond wires. Despite its dilapidated state, it’s still hardware pr0n for me — a piece of hardware I’d love to have but could never afford — made by an organization that I hold in high regard.

18 Responses to “Name that Ware December 2009”

  1. thomas says:

    it’s some component of the wii, right?
    i’m guessing the 38b7d2c3f3e4028d8d56a427bcbea290 (md5)

  2. nathan fain says:

    Thinking of Thomas’s comment I took a look at the ADXL330 accelerometer which is used in the wii controller. The ADXL330 has mems and asic sandwiched together in a package. The ADXL435 however has mems and asic beside each other. A from Chipworks shows a similar 9 points for bonding on t teardown he edge between the two on the ADXL435. I would guess that the picture could be an ADXL330? Would it be pure fantasy to think that throwing the wii controller around would be the reasoning behind the mangled bond wires, or? I assume my guess is way off. In any case, here is the chipworks teardown:
    http://www.memagazine.org/backissues/membersonly/aug06/features/tearing/tearing.html

    • nathan fain says:

      oh snap, yeah I forgot that bunnie said that this was apart of or something which he would have liked to have had but was expensive. So definitely not the ADXL330 or 345. oh well.

  3. nrp says:

    Hmm. It’s made by an organization, meaning presumably a government or international organization, rather than a company. The IC is tiny, but the exposed surface has visible detail. That the surface is exposed implies that it is a sensor of some kind, and that this is a Name that Ware implies that the product as a whole depends on this IC. Being unaffordable, but still identifiable by one chip implies there are few in existence and that its not a consumer product. That you want it implies there is a cool use for it.

    Perhaps a cesium clock chip from a NIST atomic clock?

    I can’t think of many scenarios where one would get destroyed though.

  4. Brian says:

    bunnie wrote:
    >
    > a piece of hardware I’d love to have but could never afford
    > — made by an organization that I hold in high regard.
    >
    Does a piece of high-end HP or Tek test equipment hit the
    mark on that count?

    My guess would be that this is a fragment of the ceramic
    hybrid substrate out of an active probe; the die pictured
    would be a high speed buffer amplifier of some sort in the
    signal chain.

    The ceramic substrate is very skinny; assuming 6 mil pitch
    on the IC bond pads puts the whole thing at maybe 180 mils wide,
    which would make sense for a handheld probe.

    { Alternate explanation for the skinnyness is that this might
    be an amplifier substrate that is epoxied into a narrow machined
    channel of a more complex housing. }

    Notes on hybrid layout:

    – DC power rails enter top & bottom from the right

    – chip backside is tied to the top power rail -> probably a silicon die

    { Silicon backside is usually tied to V+, or floated on the substrate
    if there’s an on-die charge pump }

    – wire bondable caps at top & bottom sit on GND, bypass V+ V- rails
    ( around 40 x 40 mil -> maybe 100-200 pf or so )

    – top and bottom of IC is dedicated to multiple power & ground bond pads

    – paralleled bond wires are usually used for either lower inductance or
    DC current; since there are only two bond wires feeding the chip caps
    from the DC rails, this suggests high speed was the reason

    – there appears to be one main output{input?} line exiting{entering}
    substrate right, consistent with the guess of an amplifier

    – The other patch of metallization above this output is probably GND
    as well, the output stages often parallel the output bond wire with
    ground bond wires down to the substrate

    – no sign of DC blocking caps in the signal path on the visible portion
    of the substrate

    – assuming one output, four GND bonds, there are one or two pads left
    over on the IC right side, maybe bias or gain control

    – IC left side looks like there once were five pads bonded; assuming
    one input and two ground pads, this leaves a couple extras –
    offset or control ? , or maybe +/- input pads

    Brian

  5. The comments above make sense to me. The power looks like it comes in on the top and bottom traces on the right, then to the capacitors, which have lots of bond wires to the die. There are only two bond wires from the power traces to the capacitors, so the multiple ones to the die should be for low inductance. Pockmarks in the other traces to the right make it look as if the power rails have bond wires to the right side of the die (top and bottom lands) for even more parallel connections. I’m guessing the single one between them is the output. With all this engineering to have very low AC impedance to the output, it’s gonna be something that has to slew a lot of current in a hurry. If this were 20 years ago, I’d guess a deflection plate driver in a high performance distributed deflection oscilloscope (Tektronix 7104 or the like). But bunnie said he could never afford it, and those things can be had cheaply enough today. So I’m thinking it’s an optical network transceiver, the kind that plugs into a very high end switch or router. Since the modulation schemes used in these multi-GHz links are designed to avoid long strings of ones or zeros, those little capacitors would allow it to output a very nice logic signal at an insane speed.

    • Brian says:

      John wrote:
      >
      > So I’m thinking it’s an optical network transceiver,
      > the kind that plugs into a very high end switch or router
      >
      When trying to think of things that needed both :
      – high bandwidth DC(?) coupled amplifier paths
      – wicked skinny substrates

      I thought about optical equipment, but here’s why I went
      with ‘scope signal path’ instead:

      – If this is something from an optical electrical signal path,
      the thick film ceramic substrate [1] would suggest it’s running
      at 2.5 Gb or below – I’d expect thin film for 10 Gb or better.

      ( And where’s the fun in a measly 2.5 Gb ??? )

      – There’s only one obvious signal path, I’d expect both Tx and Rx
      paths across the width of the substrate for one of the pluggable
      module formats.

      – The driver and limiter amps I’ve seen in optical signal paths
      were often single supply (NECL/PECL), not dual supply like this.

      >
      >Since the modulation schemes used in these multi-GHz links are
      >designed to avoid long strings of ones or zeros, those little
      >capacitors would allow it to output a very nice logic signal
      >at an insane speed.
      >
      IIRC the low frequency cutoff used for SONET signal paths was
      somewhere down in the kHz range.

      The next rank of bypass caps might be lurking just off screen;
      if I were laying this out, there’d be larger valued chip caps
      hanging off the supply rails to the right.

      ————-

      >
      > If this were 20 years ago, I’d guess a deflection plate driver
      > in a high performance distributed deflection oscilloscope
      > (Tektronix 7104 or the like)
      >
      I think the Tek scopes near 7104 vintage used Tek’s square
      ceramic packages with the elastomer clamp mounting scheme.

      If you search Google Books for “7104 bond wire”, there’s a nice
      picture of the later 11A72 amplifier substrate on page 137 of:
      The Art and Science of Analog Circuit Design, Jim Williams (ed)

      ————-

      My guess that this is part of a DC coupled active probe signal
      path has some shaky assumptions as well:

      – DC coupled amplifier assumption is based in part on the lack of
      any visible series caps in signal path of the limited view shown

      – most active probe schematics I’ve seen use discretes in the front
      end and main signal path, with some sort of off-to-the-side
      DC correction path involving an op-amp

      Brian

      [1] Thick film ceramic substrate

      Both the appearance of the gold metallization ( surface roughness,
      uneven edges ) and the absence of any readily visible vias suggests
      that this is a thick film part, i.e. silk screened conductors.

      There appears to be at least a couple of inner layers, judging
      from the out of focus gold exposed within the broken edge.

      The difficulty in controlling the conductor thickness/width with
      the silkscreening process usually means you don’t see this sort of
      thick film used for designs past a few GHz ( or at least not when
      I last did chip and wire hybrid designs 10-15 years ago )

      • I like your logic here. It’s odd that the power feed at the top isn’t connected directly to the substrate pad, but via the capacitor. Perhaps it’s important the the substrate is decoupled just that little bit better. Maybe it’s one of those current probes that uses a bridge type circuit and a bucking coil to cancel out (and measure at the same time) the current flowing past a Hall effect sensor.

        I also agree that there’s probably some more capacitance somewhere, the tiny ones by the chip are just to deal with the immediate high frequency effects.

        I realize now that the top and bottom pads on the right aren’t power, the bottom one is ground, and it’s unclear where the top one goes (assuming it goes to the apparently-unconnected patch next to it). So I’m thinking the center pad on the right is the input (instead of the output), and the output(s) are on the left. Looks like it might be differential to me. It could still be optical, but you’re right, it’s not high-GHz construction. Which amuses me, as I was doing some optical stuff to 3GHz back in the mid-1980s.

        I was intrigued by Bunnie’s comment it’s made by an organization that he holds in high regard. Tektronix might have been once, but these days, they’re run by bean counters, not engineers. Most of the rest of the T&M outfits seem to be headed the same way.

        • Brian says:

          John wrote:
          >
          > It’s odd that the power feed at the top isn’t connected
          > directly to the substrate pad, but via the capacitor.
          >
          Yes, it looks like the substrate was laid out so that small gap
          could be bonded across to tie the die backside to the supply rail;
          but instead they went up to the chip cap for the bias.

          >
          > I realize now that the top and bottom pads on the right aren’t power,
          > the bottom one is ground, and it’s unclear where the top one goes
          > (assuming it goes to the apparently-unconnected patch next to it)>
          >
          I was thinking that the part ran off dual +/- supply rails,
          and some of those not-visibly-connected patches are grounds,
          tied with not-easily-visible thick film vias down to a ground
          layer below.

          Here’s an annotated version of the hybrid photo with the
          metallization labeled and a guess at the missing bond wires:

          http://sites.google.com/site/fpgastuff/misc/ntw_december_2009

          Note that I punted on the supply names ( V1, V2, GND ) in the
          above markup, as I think I swapped the polarity of V+ and V- in
          my first post about die backside bias:
          >
          > – chip backside is tied to the top power rail -> probably a silicon die
          >
          > { Silicon backside is usually tied to V+, or floated on the substrate
          > if there’s an on-die charge pump }
          >

          Most of the silicon parts I used were negative supply ECL, and had
          the die backside tied to the most negative ( VEE = -5.2 V ) supply
          rail, _not_ to the most positive (GND) rail.

          { When typing that first post, I seemed to recall that one of the
          few parts I used running off a +5V rail had the die backside tied
          to +5V; either my memory is tricking me, or perhaps the part I
          was thinking of used a funky process with an n-type substrate. }

          Brian

  6. Marty McFly says:

    Is it from the drive unit of a flux capacitor?

  7. Florian says:

    Iam guess its from an lambda sensor (oxygen sensor) which was part of an high priced car. bmw or mercedes

    • Burlap says:

      A lambda sensor is AFAIK a fuel cell that runs at high temperature. Processing may well be in the ECU and not near the sensor, as the signal seems to be easily measurable (0-1 V DC for example).

  8. Alberto says:

    Some kind of damaged aerospace gear, possibly a fragment from Columbia disaster.

    • Burlap says:

      Then it would be burned and probably not realeased to the general public. I’d guess it is part of a device has been sent (sadly) through a shredder.

      Or it was soldered into a tin that was opened with pliers.

  9. St.Jimmy says:

    Something from super high-ed radio equipment, in the multi-Ghz range.

  10. Thanks very much for sharing.
    There is no doubt that the Christian louboutin is the best high heels in today’s fashion world.