http://zeptobars.com/en/ Microelectronics. Die-shots. Artificial intelligence. Lasers. en-us Tue, 10 Jun 2006 04:00:00 GMT Mon, 30 Dec 24 17:34:01 +0300 webmaster@zeptobars.com 120 10 http://zeptobars.com/en/read/ASM3242-PCI-Express-USB3.2-20Gbps-host-controller-PCIE Die size 3719x3937um. Thanks for the ASMedia chip collection to cyrozap!


]]> Sun, 01 Dec 24 13:52:20 +0300 http://zeptobars.com/en/read/Siliconix-DG190BP-dual-SPDT-JFET-analog-switch-chiplet-Vishay
Main die size 1113x2027 µm.


]]> Sun, 17 Nov 24 06:26:37 +0300 http://zeptobars.com/en/read/Casio-F-91W-quartz-watch-homage-china Casio F-91W, but fortunately does not misuse Casio trademark. Chip is 4 times smaller in Chinese version, it looks simpler, purely digital.

PCB is optimized for cost to an extreme : everything that can be integrated is integrated. There is not even a decoupling capacitor which is probably ok for digital circuits with low slew rate.

Die size 1384x1245 µm.




]]> Wed, 30 Oct 24 17:19:39 +0300 http://zeptobars.com/en/read/Segnetics-74F651F-bus-transceiver

]]> Sun, 20 Oct 24 11:08:09 +0300 http://zeptobars.com/en/read/AD561JD-10-bit-laser-trimmed-DAC


Datasheet has top metal photo and high-level architecture:

]]> Tue, 01 Oct 24 13:26:39 +0300 http://zeptobars.com/en/read/RCA-CA3033-high-current-opamp

Schematics below:


]]> Mon, 02 Sep 24 20:04:34 +0300 http://zeptobars.com/en/read/RCA-CD40163BD-CMOS-Synchronous-Programmable-Counter

]]> Sun, 01 Sep 24 08:19:03 +0300 http://zeptobars.com/en/read/Kwazar-K140UD1B-Soviet-opamp-Ukraine-Kyiv K140UD2 which had much higher open-loop gain.

Die size 1225x1228 µm.


]]> Thu, 01 Aug 24 08:36:10 +0300 http://zeptobars.com/en/read/tt04-tinytapeout-silicon-inside-gds-sky130 TinyTapeout project (TT04) I got to design a microchip with a bunch of oscillators, had it working in my hands and inspected under microscope. Now we'll be able to verify if manufactured chip matches GDS design files and is there any secret sauce!

Feel free to inspect source design using KLayout, SKY130 layer definitions, and finally full original GDS design. You can hide metals and DECAP cells in KLayout to make it easier to look around and compare to photos.

You can also watch video interview with decapsulation process here.

Die size 5256x3662 µm.

TT04 with metal removed, side illumination for looks:


On top metal everything is covered by power distribution and fill, so not much details below are visible:


On the right side there are few silicon doodles:








To see under top metal we need to chemically etch glass & metal layers using hydrofluoric acid (HF). Here we can see Caravel management block with RISC-V CPU at the bottom, system controller in the middle and multiple user designs covering lower half of the chip:


At higher magnification we see that all metals and polysilicon are gone, we likely only see STI (Shallow trench isolation) structures. This is IO cell:


DFF SRAM of the Caravel's RISC-V CPU. DFF memory is used to reduce tapeout risks, but much less area-efficient vs conventional SRAM array:


This is central TinyTapeout controller:


And finally, this is my design:


Chip covered by immersion oil during photography:

]]> Thu, 11 Jul 24 19:26:52 +0300 http://zeptobars.com/en/read/Nyquest-Technology-NY8A051H-8051-smallest-microcontroller 10-cent microcontroller - Nyquest Technology NY8A051H pushes price expectations much, much further down: Its high-volume pricing is just 1.5 cents, and it is currently the cheapest microcontroller in the world* (cheaper than Padauk, the holder of the previous record). I had to order a whopping 20pcs as a minimum order, and it was still under $1. Please note that despite its name - it is not 8051-compatible.

For these 1.5 cents, you don't get much: 1k words of OTP memory and 48 bytes of SRAM, but this might be just enough for many simple applications. And if you need even less and would like to save another fraction of a cent, Nyquest offers a range of simpler 4-bit MCUs.

Die size 497x418 µm. This is where 20pcs minimum order came in handy : the first 2 dies were lost as I was not expecting the die to be this small. A single 300mm wafer could fit 337,000 NY8A051H microcontrollers. It is truly an economic marvel how one is able to manufacture, test, package, and sell microcontrollers at this price point. It is barely larger than a simple BJT.

* in stock at LCSC.


]]> Mon, 01 Jul 24 05:16:54 +0300