We Don’t Need No Stinkin’ Packages!

DIP, SOIC, BGA, MLF or QFP?  None, so it seems.

This morning I received an email from Texas Instruments. Normally, these things go right into the spambox but this one was a bit unique. You can now buy some of TI’s IC’s without any packaging. Yup, just trays full of silicon squares. From TI’s point of view miniaturization has reached a point where that extra 0.1″ of PCB space is now too valuable to give to a piece of worthless plastic, and bonding micro-small wires to a silicon die is a feat that any manufacturer can preform with great accuracy, reliability and speed.

Whether this is a new paradigm in manufacturing or a premature April fool’s joke, if this process catches on smartphones just went from being almost unrepairable to 100% unrepairable, and ipod nanos might just start playing back 1080p video. It’s awesome and scary at the same time.

Now, are they crazy, or just ahead of their time? Tell us what you think.

91 thoughts on “We Don’t Need No Stinkin’ Packages!

  1. All the little black dots we find in our happy meal toys and remote controls? Exact same thing. Potted in epoxy. They’ve been the mainstay of cheap electronics for a long time.

    1. I think the joke today here is this is an ad from the 1960s? I have some old electronics form the 70s with non-blobbed CoB circuits. You can clearly see the exposed bonding wires. Even an alarm clock where the LEDs are bare dies.

      1. I have a Motorola IC data book from the 1970‘s which offers tubes of passivated naked chips as one supply option. No clean room required, only a micro ultrasonic welder.

    1. The reason there are no sockets for bare silicon dies, is because the contacts’ (what we call: “pads”) locations do not adhere to any sort of standard. The pad locations are chosen to meet packaging and application requirements and vary from product to product.

      Moreover, these pads are not more than a few square microns in area… so even if there was a standard die pad arrangement and a socket was made for it, that socket would be prohibitively expensive (extreme tolerances required) compared to the cost of assembling that die into an actual package.

      I hope this answers your question sufficiently.

      1. Now, there is something called “chip scale packaging” which offers use of standard Ball Grid Array arrangements upon dies, in lieu of standard epoxy resin packaging. These chip scale package products can be socketed but this is a different concept than what your question alluded to.

    1. I was thinking that. Glop-top chips have been used in cheaper equipment for decades. Bond the bare chip to the PCB (I think, with glue), bond the pads, pour a bit of resin on, “Bob”‘s yer uncle.

      Actually, the glop-tops I mention, do they have bonding wires? Or do the pads on the chip just match up with the pads on the PCB? Cos that would save on gold wires, and save on the effort (and the expensive machines) of doing it.

      If this is an April Fool’s joke, it’s a bit of an obscure one, because it’s actually possible, and AFAIK common in industry. I…. don’t get it.

  2. The benefit of this is that it simplifies layout a lot, so you save space on the board, not only on the package itself. You don’t have to route this stupid reset pin around this stupid mosi pin: you just pull the wires where you need them.

    A bonding machine is not too expensive, but a cleanroom probably is :)

    1. Actually, clean rooms aren’t that hard or expensive to do. They just require strict attention to detail.

      There is an instructable that details building a clean room in Afghanistan for commercial mushroom spawning. They used some wood lath, sheet plastic, and duct(k) tape. Maybe a HEPA filter. If you need constant particulate cleaning, build a home-made wet scrubber with a fountain pump and a shower head.

      I always enjoyed working in the clean room. No dust or allergens.

      1. Something tells me he’s not talking about a “US-grade industrial clean room” here. He’s talking about an “it’s good enough to grow mushrooms in Afghanistan” clean room.

        That said, for the home hobbyist a duct-tape and vinyl solution may be totally viable.

    2. The class 10 cleanroom I work in cost about 12 million USD, but I honestly feel that I could duplicate it for a few thousand, including filtering equipment. It’s just a glorified steel box tube frame with vinyl sheeting stretched to form walls, with air inlet filters in the roof and air exit vents near the floor. Not all that different from a screen patio.

      There are some similar pictures of one on this pharmaceutical company’s site:

      http://www.teregen.com/pharmaceutical-compounding.html

  3. TI is “late to the party” for about 16 years becaise chip on board had been a viable mounting technology for over 16 years. Just look at the back of any small LCD display – the black blob is a bare die mounted directly to the PCB.

    1. It was but you can’t leave the secured space where you can’t find a single particle of dust to just sell them to some clients. Eventually, you buy them, and they stick the silicons to your board to an extra cost. I see it as the only valid scenario.

  4. I don’t know, this may be fake, but TI DOES sell bare dies. Look at digi-key p/n MSP430F2619SKGD1-ND or look at IC->Embedded – microcontrollers and in the filters select packaging = die. This will also turn up some maxim chips

  5. I don’t know, this may be fake, but TI DOES sell bare dies. Look at digi-key P/N MSP430F2619SKGD1-ND. Or look in the micro controllers section and filter by package = die

  6. This calls for the design of DIY-thermesonic wirebonders. +/- 10 um accuracy requirement. And lots of 1 mil gold wire ;-).

    Not to mention bare die prepping techniques, to clean the bond pads after handling.

  7. This is not even remotely new. In high reliability markets it is the norm. They’re called known good die (KGD). In treacherous, high temperature or vibration, environments the package is more likely to fail than the actual die.

    1. Also MicroSMD, in TI-speak. Put the bond pads evenly spaced across the die, put solderballs on them, and stick it on the board. Still takes up more space than this would because of the size of the solderball pads (and how they mess up layout because of their spacing demands) but it’s reliable and pretty easy to put down. TI’s currently at 0.5mm pitch between solderballs. They’re silly small.

  8. “100% unrepairable” is a bit of a stretch. Hydrofluoric acid or high-energy beam silicon layer removal processes coupled with ultrasonic or laser wire bonding, and simple metal deposition/etching can repair quite a few types of chips. Now, I understand that not everyone has a chip fab plant in their basement, but most anyone can get HF acid, and a microscope. You can also remove those annoying epoxy blobs or most any encapsulation with fuming nitric acid. I will warn you though that while science is fun it can be dangerous, and you should at least get a fume hood and gloves.

    Happy Hacking!

  9. I guess I will have to blow the dust off of my Soffa and Kulicke Ball Bonder and set up shop. Hybrids have been used for at least the last 30 years. I worked with this stuff when I got out of school.

    Can’t wait to see the DIYs for class 4 and class 3 clean rooms!

    1. iOvenMit…

      I Rotfl at that lolz…

      April Fools joke? Maybe, but even if its not, COB is not very common for lager, more complicated electronics. Like in a calculator or at the back of a LCD pannel. Those are dedicated to a function. Doing as this post suggest, smartphones will loose yet another few millimeters in thickness thanks to the removal of the packaging.

      Also, this might be exelent for high temp chips such as CPU’s and the like. Removing the packaging, the thermal couple between dye and heatsink should improve, even if only a few degrees….

  10. I LIIIIIIIIKE PLLC’s! They are especially good for the whole changeable modules thing.

    This is probably where things are headed anyway. Without the whole wide-blob-ness of the COB BLOBS. Although a DIY wirebonder might not actually be impossible. The movement of the head is controlled with a big long lever which has only a small throw when it reaches the x-y slides. And IIRC the wire can also be silver or copper. We often had copper wire when I worked in the Berkeley Nanolab.

    The DIY electronics movement is probably completely doomed either way.

  11. I think the significance is that TI is making them more widely available. They’ll support smaller quantity sales and perhaps a wider cross-section of designers. Because of the fragility of bare die, manufacturers have traditionally been pretty careful about who they would sell them to (or under what conditions).

    Actually I don’t see that it would be that great of a stretch for a very capable hobbyist or small manufacturer to design something using bare die. Here’s to seeing a whole new class of devices over the next few years.

  12. Well, DPAK packaging of power MOSFETS from IR is essentially an upside-down die in a metal cavity; the cavity gets soldered onto the PCB while the die contacts are made via solder bumps, so it’s kind-of similar. DPAKs are done this way not because of size but because of much smaller thermal resistance.

  13. The april fools joke comments are pretty annoying. They have had this up on their website on the front banner for a while and it popped in my RSS folder sometime at least a week ago.

    1. DIY transistors were available in the 1900s, for the vacuum tube style ones. Silicon ones are not nearly as hard to make as most people think, it takes me a few hours to make a couple thousand nanoscale ones. I’ve made a few at home though, they are NOT power efficient, but there is no rule against a macro-scale transistor, they can by made by ordering 2 types of silicon of your choice (p/n) then just heat the ends of the 3 pieces to fuse them together in the order you want. I’ll consider writing a tutorial.

      Best wishes and good luck!

    1. I don’t think you have to worry about that. DIY electronics are definitely changing, but the field isn’t dying. Modifying commercial products is definitely becoming more difficult. But as far as building your own projects, you’ll be able to purchase all the components you can get today for a while longer. Think about how long things like the 555 timer have been around for – 40 years.

  14. It may indeed be a jest fit for the day, but many have pointed out the reality of it. For me, first there were tubes with point to point bare wire to terminal strips hanging from bolts in a “hot” chassis. Then solid state came along and began the time of pcb’s with co-workers swearing each time they ruined a trace on a board. We switched to transistors and chips came right on their heels with .1″ spacing of which I remember co-workers lamenting that “this stuff is too tiny to work on anymore”. Replacement parts changed from components to boards and technicians with the “Right Stuff” started falling by the way as “four-layer boards” drained the last of their will to try component replacement, and as each component was a whole block diagram they knew they weren’t doing the magic like they used to, so they just swapped the board. The few of us that remained jumped to engineering.

    This may or may not have been a joke for this one day gentlemen, but it certainly is the “Writing on the Wall.”

    The days of “spit in the back and whomp it one!” are gone. I hope you can find a way to keep your hobby alive, but you should stop and think a moment because in 10 yrs what will there be? And 20?

    My suggestions? Learn to write code, or learn to start a fire with two sticks.

    1. I also saw electronics move from vacuum tube to transistors then ICs then complete functions on a chip and I disgree with you. I work as a consultant today for a major electronics manufacturer and they routinely troubleshoot and repair down to the component level. But you are correct in one aspect, field service has been at board level swap for a long time. But this is more for quality and reliability than any other reason. A major electronics manufacturer can setup a refurb line that tests, repairs and upgrades failed PC boards to new or better than new before sending them out to be installed elsewhere.

    1. THIS. Come on people, understand and practice the difference. I do work in chip fabrication research, and a few people that I talk to still call it “silicone.” It drives me absolutely crazy.

      end of rant.

  15. The real problem is the trend to packaging that is not amenable to low volume production. This is especially the case for one offs.

    Obviously the next wave of Hack a Day features will be DIY wire bonders.

  16. Oh yes, I certainly agree…. And I applaud your attempt to whail against the machine. Think you can stop time marching on? Take what I wrote and apply it to some newbie starting today. “At first it was SMT with 8 layer boards, and the old farts were complaining….. then it was… etc.

    Writing on the wall…… writing on the wall…..

  17. Sigh. Even if this was just a jest… the writing was on the wall….

    You want to change it? THIS is the place to change it. We need video microscopes, we need reasonable board mfg options IN THE HOME, we need micro soldering equipment. I’d be HAPPY to do the bonding of micro-wires IF I COULD. We have to work these out, the mfg’s will not as they are geared for mass production and anything they make will be huge sums of money. WE are willing to spend the time, rofl, right…. we all sit back and wait for ONE to do the work and we copy it. WE NEED A HACKERSPACE. We need some communications because it is unlikely the talents needed are all in the same city. Open Source.

    … or go get two sticks and head for the woods with a tent. It’s a good and pleasing alternative.

  18. Apparently this was distributed prior to April first. As pointed out this construction method is nothing new. The first sentence of the file may reveal what’s new “Texas Instruments Incorporated has expanded package options with the additional availability of bare die”

    I might be wrong, I haven’t read all five pages looking for clues, but I wouldn’t declare April fools. While interesting this isn’t some that’s going to be used in the typical home shop anytime soon

  19. happy april fools, or from my perspective:

    you can pry my dip packages from my cold dead hands )(*^%*^%$&^%)P^(*)*(&%*(&$^(&%

    and yes i WILL go on using older parts to do modern things, as long as i can do it with a 15$ – 75$ soldering iron.

    so STFU in the 70’s for inventing surface-mount and blobs and STFU for still using them and anything any less build(purchase/recycle)/hack(recycle)/repair(recycle) friendly

    1. Why the fear of SMD?
      All you need to solder them is a Toast-R-Oven from the 70’s/80’s and all you need to unsolder them is a hot air gun (or that $15 soldering iron, some flux and some magnet wire) and a pair of tweezers. SMD saves a ton of money!

    2. Surface-mount ICs are waaaay easier to solder and hack with a cheap soldering iron than a DIP package. You just want to use SOICs, rather than any finer pitch than that.

      DIP packages can be a nightmare to solder if they’re not thermal-reliefed properly. Not to mention the fact that the inductance from a DIP package will basically destroy you at moderate frequencies.

      I have no idea why people build boards with through-hole resistors, caps, and DIP chips anymore. A monkey could solder 1206s and SOICs. They’re ridiculously easy.

  20. I have seen this kind of “glop top” used in some cases for ROM chips or MCUs or other things where the manufacturer wants to stop someone being able to read out the code/reverse engineer the chip for anti-piracy reasons (especially in some cases on old video game carts)

  21. Prototyping is not going away from electronics industry, so neither will small-scale production (ever) go away. Never fear: if nothing else, the dynamics of industry itself will keep our hobby/small business going. Either some sort of human-handleable packages will always be there, or demand for the tools for manipulating chip scale components will rise, which will, after a while bring us affordable table-top chip-scale rework equipment.

  22. Presumably this is what the American NSA do on a regular basis? They’re reputed to have a silicon fab in their “secret city” for making jolly small, clever, and sometimes “one-off” things. OK, the US taxpayer is “glad” to donate untold gazillions of dollars for this work, and I don’t suppose any details will be on their web site, but this kind of thing has a habit of leaking out because of the links they have with industry (think drunken loud-mouthed salesmen at industry conferences).

    Additionally, they’ve been under the financial cosh for nearly two decades now, so getting rid of the overheads of a fab could look very attractive.

    Perhaps an intelligent review of this area would yeild something? Perhaps aspects the necessary technology and processes are already in the public domain, but thinly disguised?

    Check out what the UK journalist Duncan Campbell was able to learn about the UK spook’s nefarious deeds from published, non-confidential, sources back in the ’70s.

    Google for: duncan campbell zircon abc trial

    1. This is orthogonal to that. The NSA probably have custom ASIC, but they may still put them in a traditional package for convenience rather than bond them directly to a circuit board. It’s not as if they run a volume production line that is super sensitive to unit cost.

  23. Just curios, and please, no flamers… Why does one need a clean room for this? I dont have clean rooms when I solder 80pin TQFP, very tiny spacing. So if I dont have a clean room, and a wire is soldered, and there is dust next to it, so what??? whats the difference in dust and a glob of epoxy??

    1. >Just curios, and please, no flamers… Why does one need a clean room for this?

      The die is unprotected, so any “dust” that is reactive (and that means any dust from pencils, erasers, plastics…) could be a problem to the super-thin metal and semiconductor layers. And electrostatic discharge is a huge issue. And even inert dust sitting on a wire-bond pad will be a problem.

  24. Aren’t we just talking “hybrids” (chips that get connected and assembled together inside a shared package—they still often look like a “chip” package from the outside) here? I worked on hybrids in the early ’80s that went in the MX (later known as “Peacekeeper”) missile. And a lot of the high speed convertors were hybrids when I worked for TRW LSI Products in the very early ’80s—to the point that their high speed converters were advertised as “monolithic” (not hybrid) explicitly.

  25. There are ways to solder these “on the cheap”, one of which is lasers.
    Mixing solder with nanosized carbon (iirc 50um) to make it absorb the laser light, makes things a lot easier and there is the possibility of rebonding wires if something goes wrong.

  26. Wow, are you guys stuck in the 80’s? IC manufacturers have been selling bare die’s for a very long time. De-pot a control module from a 90’s cadillac… You will see what I mean.

    Packaging an IC adds significant cost. If a customer can buy the bare die and bond it out themselves, they can realize some considerable savings!

    Anyway, this is nothing new at all. I am kind of shocked that you guys have never heard of this… Kind of says something about your knowledge level when it comes to electronics, and it’s rather lame that you are now causing a bunch of readers (who don’t know any better) to think that this is some kind of new technology or method. Shame.

Leave a Reply to JriddlesCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.