Zeptobars

You've probably seen https://zeptobars.com, but if not, there are lots of nice chip images there.

I love posts like these. For more of this stuff I recommend checking out ZeptoBars. https://zeptobars.com/en/ Lots of die shots and I’m always blown away by the beauty of some chips when designers fully know it is unlikely their art will see the light of day.

If anyone is curious what a real vs fake FTDI chip looks like under the hood (de-capped chip) this is a great analysis and some beautiful pictures. http://zeptobars.ru/en/read/FTDI-FT232RL-real-vs-fake-supereal Exerpt: "What's the economic reason of making software fake of well-known chip instead of making new one under your own name? This way they don't need to buy USB VID, sign drivers in Microsoft, no expenses on advertisement. This fake chip will be used right away in numerous mass-manufactured products. New chip will require designing new products (or revisions) - so sales ramp up will happen only 2-3 years later. Die manufacturing cost is roughly the same for both dies (~10-15 cents)."

These images would be a fantastic resource on Wikimedia Commons (for use in Wikipedia articles). The licensing (Creative Commons Attribution 3.0 Unported License) is compatible as per [1] and [2]. I've got things started with the 74HC595 die photo at [3] and [4]. [1] http://zeptobars.ru/en/contacts [2] https://commons.wikimedia.org/wiki/Commons:Licensing#Well-known_licens… [3] http://commons.wikimedia.org/wiki/File:74HC595-HD.jpg [4] https://en.wikipedia.org/wiki/7400_series

And for those who just love seeing silicon dies up and close, you might want to check out these amazing pictures to really put things into perspective: http://zeptobars.ru/en/read/open-microchip-asic-what-inside-II-msp430-… Really cool to see the different designs used in various components, or even compared to other CPUs/MCUs. The pictures above were taken by dissolving off the inline package casing with sulfuric acid. No 8085, but there's a picture of the Z80A available, and you can make out the similar areas to the 8085 (register file, control logic, ALU, etc... the quality is insane, the Z80A picture is 6200x6000 px). There's also some earlier photos at http://zeptobars.ru/en/read/how-to-open-microchip-asic-what-inside if you want some more high-res silicon imagery.

Chips made on modern processes are dominated by capital costs. Old chips are made on old foundries, which are fully depreciated and therefore have no capital costs. If you made a 90nm foundry today, then it probably couldn't sell those microcontrollers for less than $100 each, just like a modern CPU. >why can't existing chip companies, some startup or GM themselves spend $10's of millions building a fab that can unblock $100's of millions of product, and alleviate the shortage? They can't do it fast enough. Standing up a new foundry takes years under the best circumstances, and today you couldn't do it all, since all the tooling is sold out and deeply backordered. If GM could snap their fingers today and magic a cleanroom into existence, they'd still be waiting a hell of a long time to put tools in it. Secondly on the price question, microcontrollers just have way fewer gates than a desktop CPU. A dozen registers, a thousand bytes of RAM, a few kilobytes of flash. This makes them physically smaller, which lets you put more on a wafer, and makes the unit price cheaper. The total die area of the ATmega8 is just 7.9 square millimeters... at the 500nm node! https://zeptobars.com/en/read/atmel-atmega8 This gets you 8,100 dies from a single 300mm wafer. (If there are any 300mm foundries at 500nm, which there probably aren't) Thirdly, what makes you think anyone could build a 90nm foundry at all? Take a look at a list of foundries:…

Most of the commentators in this topic are probably EE guys, may be I can find my old classmates here, hi guys:-) Joke aside, AD is probably one of the most profitable companies with unprecedented monopoly on analog components and circuits business. Last week somebody mentioned about Visa and Mastercard duopoly in HN, but with this Maxim acquisition, AD is now the Visa and Mastercard combined while TI is the American Express. To see how profitable AD is please check this article on the AD's 3rd generation transceiver chip and if you are not reading it, basically it's more profitable than drug [1]! I am using every single versions of this AD's transceiver chip for my work, now it is in the latest 6th gen. Suffice to say if you're building 5G transceiver you will need AD's 5th gen transceiver. It's probably not an exxaggeration to say that ITU radio/wireless cellular standards' bandwidth actually following the AD's transceiver chips generations since 5G system can be built by the 5th gen chip but not the 4th gen chip (it's for 4G). The closest competitors are transceivers from LT (already acquired by AD) and the Lime Microsystems from Cambridge, UK. [1]https://zeptobars.com/en/read/AD9361-SDR-Analog-Devices-DAC-ADC-65nm

https://zeptobars.com/en/read/wch-ch32v003-risc-v-riscv-microcontrolle… This page estimates 90nm for the CH32V003, and I found another post very roughly estimating 130nm. And a pi pico isn't all that fancy either at 40nm. And should I be very worried about counterfeit microcontrollers? It seems like a lot of effort, and like it would probably still work.

With exceptions, was reading recently an interesting die analysis/estimate of the costs and margins of manufacturing of the AD9361 chip (a 65nm, digital radio transceiver, introduced 12 years ago and still selling at retail today for $300/$400): https://zeptobars.com/en/read/AD9361-SDR-Analog-Devices-DAC-ADC-65nm Relevant quotes (and the current retail price if anything is higher now then when the article was written): “ Retail price of AD9361 at distributes is 275$, volume price from manufacturer is 175$. That is quite an impressive added value! For 1,68$ of manufacturing cost we are getting 173,32$ of added value! Even Intel with their x86 or drug cartels could NOT do business like that.” Of course, the actual margin needs to take into account NRE and other costs (and the above link does get into that) but, in this case, the manufacturing is a tiny sliver of the costs.

The first link in the article is to a .10 RISCV part. https://zeptobars.com/en/read/Nyquest-Technology-NY8A051H-8051-smalles… The RPI2040 is under a buck in bulk (the chip). The Pine Ox64, The Milk-V duo and the RPI packages are all interesting products (at a sub 10 bucks each). The MILK-V duo S is very appealing (form factor wise). If they ever mange to get POE/HDMI in version out it will be a killer product (even more so if they get all that into a sub 20 buck price point).

I had some mosfet gate drivers (IR2112) that were really marked as IR2113, a higher tier part. The only real way to tell what you've got is to remove most of the epoxy, then boil the rest in nitric/sulfuric acid. After much mess and fume, you'll be left with the raw die and can inspect the markings. I've done this a couple of times, it's quite cool. In my experience, most "fakes" are crappier parts marked up to be better, or chips that have failed QA. Rarely do you get a whole clone, although for some very high volume or simple, old parts, it's a bit more common. If you like this, check out [zeptobars](https://zeptobars.com/en/), they post high quality dieshots, and sometimes fakes. IR2113/IR2112 die: https://www.youtube.com/watch?v=e6DfBuPwAAA

There are some nice images of the ESP32 die here https://zeptobars.com/en/read/Espressif-ESP32-Wi-Fi-Bluetooth-2.4Ghz-I… and yes does not look like a multi die chip.

There's no FW on it as a programable microcontroller would be far too expensive, it's just glue logic die under a black resin blob, optimized to the absolute bare minimum number of transistors, which you can easily reverse engineer as the die shots are now public[1], but it doesn't help you since you can't reprogram it. [1] https://zeptobars.com/en/read/Casio-F-91W-OKI-quartz-watch

I guess you'll enjoy this one too: https://zeptobars.com/en/read/AD9361-SDR-Analog-Devices-DAC-ADC-65nm

Some time ago I posted exactly about that : https://zeptobars.com/en/read/how-to-open-microchip-asic-what-inside My latest microscope is this one, but Casio photo was made by temporary microscope I assembled from Thorlabs & Olympus parts: https://3.14.by/en/read/olympus-bx60-fluorescence-metallurgical-micros…

This same site has an article on laser decapsulation not too long ago: https://zeptobars.com/en/read/laser-decapsulation-fail-not-so-fast It appears to not work so well, but the video looked very cool.

That's a good public example, here's a post for anyone that missed it: https://zeptobars.com/en/read/FTDI-FT232RL-real-vs-fake-supereal

I suspect the minimum die size is dominated by all the pads required. Edit: maybe not, I found a die shot: https://zeptobars.com/en/read/FTDI-FT232RL-real-vs-fake-supereal - even on the genuine 2005 die there's lots of empty space. Could probably squash that by 50% with only moderate effort without even a process shrink. Now I'm wondering what volume FTDI see. Can't be too high or it would be a more competitive market.

Ironically, decades later this clone of the 8080 was itself reverse-engineered by someone in Ukraine: https://zeptobars.com/en/read/KR580VM80A-intel-i8080-verilog-reverse-e…

> This is clearly a smaller chip than the ESP32 Well.. only if it's made on a more advanced technology. See, the RAM on an ESP32 actually takes near 85% of the total die, see https://zeptobars.com/en/read/Espressif-ESP32-Wi-Fi-Bluetooth-2.4Ghz-I… The remaining 15% are for the radio, the peripherals and the cores. As there are 2 cores, each possibly takes less than 1% of the die.

I try not to use them, but they seem fine for basic SPI/UART/etc. Someone did an analysis of the dies, and apparently they compare favorably in some ways: https://zeptobars.com/en/read/GD32F103CBT6-mcm-serial-flash-Giga-Devic… I haven't tried anything complicated with the peripherals though, and the thing is, STM32F103 chips are so cheap already. The design is over 10 years old, and ST is so good at making them that most "64KB" chips actually have 128KB of Flash. The GD32V chips are very cool, though; they keep the STM32-alike peripherals, but they use a faster RISC-V CPU core (rv32imac).

Datasheets are here: http://dl.sipeed.com/LONGAN/Nano/DOC/ As from my experience with GD's STM32 clones, they did not have a real flash, but copied all your code from SPI-like internal flash to main SRAM on startup. I wonder if they did pull the same stunt with GD32V's EDIT: what is inside stm32 clone: https://zeptobars.com/en/read/GD32F103CBT6-mcm-serial-flash-Giga-Devic… EDIT2: original page now returns 404, it was this board: http://dl.sipeed.com/LONGAN/Nano/Spec/Sipeed%20longan%20nano%20Datashe… for $4.90 (https://www.seeedstudio.com/catalogsearch/result/?cat=&q=GD32VF103…

Chip photography site zeptobars.com has some photographs of modern transistors, and while some have interesting geometry, I didn't see anything that I'd consider an integrated circuit, i.e. something that would be drawn as more than a single transistor symbol on a schematic. https://zeptobars.com/en/read/NXP-PMST3904-NPN-BJT https://zeptobars.com/en/read/NXP-MMBT3904-npn-bjt https://zeptobars.com/en/read/NXP-2N7002-TrenchMOS-N-Channel-MOSFET There are some "discrete" transistors with built-in ESD protection, e.g. NCV8440A, but those are unusual, and under normal operating conditions they work just like a normal transistor so I'd still count them as discrete components.

There are things (anti-things?) you want and there are the things the market provides. https://www.digikey.com/product-detail/en/analog-devices-inc/AD9361BBC… The AD9361 is exactly that all purpose programmable radio. http://www.analog.com/en/products/rf-microwave/integrated-transceivers… https://zeptobars.com/en/read/AD9361-SDR-Analog-Devices-DAC-ADC-65nm

I recognize that chip at the bottom of the article! That's a Hitachi HA17555, which was on Zeptobars just a few months ago[0]. [0]: https://zeptobars.com/en/read/Hitachi-HA17555-555-timer

@joshvm looks like we ran out of thread depth.. That's super cool. I wanted to show you this too if you hadn't seen it already: http://zeptobars.ru/en/read/KR580VM80A-intel-i8080-verilog-reverse-eng…

Zeptobars has some nice shots of silicon dies. As a random example http://zeptobars.ru/en/read/Nordic-NRF24L01P-SI24R1-real-fake-copy I estimate that those images are taken with equivalent of about 60000 DPI. Even with that sort of resolution you can't really resolve the details in the smaller features. In comparison typical macro lenses have 1:1 magnification, so e.g. D800 DSLR with 36MP sensor would have theoretical maximum of roughly 5000 DPI, and I would expect the reality to be quite a bit harsher than that.

Yeah, but looking at the physical differences between the chips it's probably not something that difficult to ensure. The counterfeit chips are built in a totally different fashion, with different size and materials, on a different process tech. http://zeptobars.ru/en/read/FTDI-FT232RL-real-vs-fake-supereal

Elsewhere Hackaday links to a Russian microelectronics company's blog[1], where they decapped real and fake FTDI chips and found that: "It seems that in this case Chinese designers implemented protocol-compatible "fake" chip, using mask-programmable microcontroller. This way they only needed to redo 1 mask - this is much cheaper than full mask set, and explains a lot of redundant pads on the die. Fake chip was working kinda fine until FTDI released drivers update, which were able to detect fake chips via USB and send only 0's in this case. It was impossible to foresee any possible further driver checks without full schematic recovery and these hidden tricks saved FTDI profits." [1]: http://zeptobars.ru/en/read/FTDI-FT232RL-real-vs-fake-supereal

In you want more: http://zeptobars.ru/en/ The nice thing is they decap old/common devices.