Stories
Slash Boxes
Comments

SoylentNews is people

posted by janrinok on Tuesday November 15 2022, @03:09AM   Printer-friendly

https://www.righto.com/2022/11/the-unusual-bootstrap-drivers-inside.html

The 8086 microprocessor is one of the most important chips ever created; it started the x86 architecture that still dominates desktop and server computing today. I've been reverse-engineering its circuitry by studying its silicon die. One of the most unusual circuits I found is a "bootstrap driver", a way to boost internal signals to improve performance.

This circuit consists of just three NMOS transistors, amplifying an input signal to produce an output signal, but it doesn't resemble typical NMOS logic circuits and puzzled me for a long time. Eventually, I stumbled across an explanation: the "bootstrap driver" uses the transistor's capacitance to boost its voltage. It produces control pulses with higher current and higher voltage than otherwise possible, increasing performance. In this blog post, I'll attempt to explain how the tricky bootstrap driver circuit works.


Original Submission

This discussion was created by janrinok (52) for logged-in users only, but now has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: 3, Informative) by maxwell demon on Tuesday November 15 2022, @05:47AM (9 children)

    by maxwell demon (1608) on Tuesday November 15 2022, @05:47AM (#1279816) Journal

    An interesting read. And a reminder that simplified mental models, though helpful in most situations, can also limit you if you are not able to think beyond them. Sometimes the details you normally abstract away are the key to achieve things that otherwise seem impossible.

    --
    The Tao of math: The numbers you can count are not the real numbers.
    • (Score: 4, Interesting) by kazzie on Tuesday November 15 2022, @06:15AM (8 children)

      by kazzie (5309) Subscriber Badge on Tuesday November 15 2022, @06:15AM (#1279821)

      These are the kind of reads that I devour with keen interest, then start thinking "can I bring this into my students' teaching, or is it going to over-complicate things"?

      Last year I had the opportunity to teach a module on electronic principles and semiconductor theory right before teaching digital logic to the same group. It was the first time I'd been able to bridge the two together properly. In my education, a while back, digital logic had been treated separately with abstract gates, and I took me a while to get my head around the idea of transistors.

      I only looked at (simplified) methods of building logic gates out of transistors for the first session or two, but then let things be abstracted away into logic gates, flip-flops, etc. But it did pay off when we came to looking at TTL vs CMOS (still sat in many curriculums, though hardly anyone uses TTL now), as they were able to relate the power drain and switching issues to the transistors they'd looked at previously. Without having looked at the underlying structure of the transistors, it's just a list of properties assigned to one label or another.

      I've decided this article is going to be included in my teaching next time, even if only as a 'further reading' link at the end of a session. It's too good to be overlooked, and hopefully some of my students will get similarly enthused by reading it.

      • (Score: 4, Informative) by janrinok on Tuesday November 15 2022, @06:52AM (2 children)

        by janrinok (52) Subscriber Badge on Tuesday November 15 2022, @06:52AM (#1279826) Journal

        Thanks for your own post and the from one from Maxwell Demon that you are replying to.

        We welcome feedback - but we seem to get far more criticising stories than telling us what the community find interesting. Your comments have been noted.

        The difficulty is find similar stories to regularly put on the front page. If anyone knows of a particular site, or better still an RSS feed, that has more of the same then I would welcome a link to it. Owl (15206) seems to be our resident contributor of this and similar stories.

        • (Score: 3, Interesting) by Rich on Tuesday November 15 2022, @02:13PM

          by Rich (945) on Tuesday November 15 2022, @02:13PM (#1279858) Journal

          I've praised Ken Shirriff's content before. Not only does he have these immaculate die shots, he also completely breaks them down when describing the function. Probably the most hardcore nerd content on the entire internet.

          But now I wonder if these "dynamic designs" of P/NMOS days all hinge on this kind of voltage boosting, or if there are other tricks that require a minimum clock rate. I was under the impression that constructs like DRAM cells were mostly responsible for this until I read TFA. It only came with CMOS that CPUs became "static".

        • (Score: 2) by owl on Tuesday November 15 2022, @04:04PM

          by owl (15206) on Tuesday November 15 2022, @04:04PM (#1279870)

          Owl (15206) seems to be our resident contributor of this and similar stories.

          Glad to hear that at least some of the Soylentutil's enjoy these kinds of stories.

          Ken's page (which the story links to) itself links to his other stories. Most are quite detailed.

      • (Score: 1, Insightful) by Anonymous Coward on Tuesday November 15 2022, @07:53PM (4 children)

        by Anonymous Coward on Tuesday November 15 2022, @07:53PM (#1279901)
        I think TTL is still widely used by electronics hobbyists, as it's very cheap, durable (far more so than CMOS chips which can be easily damaged by ESD), and still easy to interface to popular microcontrollers like the ATTiny85 and Arduinos. I still have a ton of 74LSxx series chips and have used them on occasion to build all kinds of projects.
        • (Score: 2) by kazzie on Wednesday November 16 2022, @07:06AM (3 children)

          by kazzie (5309) Subscriber Badge on Wednesday November 16 2022, @07:06AM (#1279975)

          Agreed, it's still in use in the hobbyist scene, but my impression is that it's largely dead in industrial sectors. The range of LS/ALS chips in production is dwindling, and it's now getting harder to find 5V-compatible CPLDs, FPGAs, and other bits and bobs.

          • (Score: 2) by Rich on Wednesday November 16 2022, @01:40PM (2 children)

            by Rich (945) on Wednesday November 16 2022, @01:40PM (#1280019) Journal

            Who uses actual LS (and friends) anymore? HC uses a fraction of the power, has nicer output voltages, and if it has to be fast, there's AHC. All the CMOS ICs have ESD protection diodes, I've never had one of those die from ESD. Don't handle them with rubber pincers that you clean on your cat's fur, and keep them away from similar ESD sources (like old synthetic carpets that get you shocked when touching door handles...), and you're good. The ESD danger is real with single MOSFET transistors where there are no protection diodes for the gate, but with your usual IC don't worry.

            • (Score: 2) by kazzie on Wednesday November 16 2022, @06:36PM (1 child)

              by kazzie (5309) Subscriber Badge on Wednesday November 16 2022, @06:36PM (#1280068)

              That's just the point I was making, further up. CMOS wins out in almost all scenarios, but TTL is still taught alongside it in education.

              One area that TTL is a lot more forgiving is in breadboarding circuits our of logic chips. The fact that floating inputs float upward (due to internal pullups) means that debugging students' creations is a lot simpler, especially when they get on to building state machines. A loose wire or floating input on a CMOS gate can make for subtle errors or outright borked behaviour, which can be awkward to trace. At least with TTL the floating input will tend to be high at all times, and a consistent fault is easier to locate.

              In the broader sense, there's also the disconnect between the use of discrete logic gates in DIP packages for teaching, and the modern reality of programming FPGAs and CPLDs in software. I think that learning the bare bones with discrete packages is an important stepping stone to getting your head around digital logic systems, but I wonder whether educational and hobbyist orders will be enough to keep LS / TTL chips in production in the long term.

              • (Score: 3, Informative) by Rich on Wednesday November 16 2022, @07:22PM

                by Rich (945) on Wednesday November 16 2022, @07:22PM (#1280074) Journal

                A loose wire or floating input on a CMOS gate can make for subtle errors or outright borked behaviour, which can be awkward to trace.

                Haha. That'll teach 'em! I had a 4099 (DIP) enable input floating and it worked for 10 minutes with the device cold and then stopped. :)

                Regarding the CPLD vs. fixed function issue, I don't think the 74xx and 40xx families will go away, at least not in CMOS form. They are too useful as glue everywhere. What I think is a little strange is that the entry level CPLDs are near-extinct (ATF22V10 sold out, and Lattice trying to get rid of the GAL22V10...) because they are really valuable in the "long tail" of the market, industry specialty applications from one-offs to maybe a few hundred units.

                The DIP vs SMD issue is something else. I think this hinges only on extended military contracts. If you look at the CD4099 datasheet from Texas, it's pretty obvious that they just keep historic scans - but everything is qualified with military temperatures. Otherwise the "brand" manufacturers drop low-margin parts or outsource them, like NXP 1N4148 are really from Tak Cheong these days. I think Texas and Microchip (not counting Rochester) are about the only shops still selling DIP across a reasonable spectrum. However, if DIP would go away, it's not too catastrophic, because it's still possible to put some small package on a carrier for tinkering. Except for everyone's favourite in electronic music, the LM13700, where they discontinued DIP and only SOIC remains, which is too fat for a simple carrier. Super annoying, but fortunately I have a stash of mostly compatible NE5517 DIPs and "Xinluda" (ever heard of them???) seem to come to the rescue (LCSC have XD13700 at really convincing prices in stock today).

  • (Score: 1) by pTamok on Tuesday November 15 2022, @07:37AM

    by pTamok (3042) on Tuesday November 15 2022, @07:37AM (#1279832)

    Add my vote to it being an interesting read. Thank you.

(1)