ZeptoBars - RSS feed http://zeptobars.com/en/ Microelectronics. Die-shots. Artificial intelligence. Lasers. en-us Tue, 10 Jun 2006 04:00:00 GMT Sun, 28 May 17 04:11:51 +0300 webmaster@zeptobars.com 120 10 <![CDATA[1107PV2 - 20MHz 8bit Soviet flash ADC : weekend die-shot]]> http://zeptobars.com/en/read/1107PV2-20MHz-8bit-flash-ADC-soviet


Closer look at comparators:

]]>
Sun, 30 Apr 17 12:24:26 +0300
<![CDATA[1107PV3A - 100MHz 6bit Soviet ADC : weekend die-shot]]> http://zeptobars.com/en/read/1107PV3A-100MHz-flash-ADC-soviet Die size 3959x2755 µm.


]]>
Mon, 24 Apr 17 01:42:36 +0300
<![CDATA[LR433A - Chinese 433MHz SAW resonator : weekend die-shot]]> http://zeptobars.com/en/read/LR433A-SAW-resonator-433MHz-chinese
It was packaged this dirty (it looks more like photoresist processing defect). Die size 3681x769 µm.


Comb pitch 3.2µm (line+space). Another specimen had cute defect: only miracle left 2 terminals not shorted:

]]>
Mon, 17 Apr 17 09:03:42 +0300
<![CDATA[Siemens CFY77-10 - 20GHz AlGaAs / InGaAs HEMT : weekend die-shot]]> http://zeptobars.com/en/read/Siemens-CFY77-10-20GHz-AlGaAs-InGaAs-HEMT-transistor

]]>
Tue, 14 Mar 17 01:21:16 +0300
<![CDATA[A look inside Russian 28nm MIPS CPU - Baikal-T1]]> http://zeptobars.com/en/read/baikal-t1-mips-Imagination-Technologies-P5600-Warrior Baikal-T1 - Russian implementation of dual core Imagination Technologies MIPS P5600 32 r5 with on-board 10Gb Ethernet. Baikal was the first to implement P5600 core in silicon.

CPUs itself (had to go through 4 pieces to get a good die):


BGA-substrate and heat-spreader are similar to other modern CPUs. It's flip-chip BGA:


Die itself has contacts over all it's area, majority of these are VCC/GND nets. This is required not because of high power consumption (it's actually not that high, ≤5W), but rather to reduce inductance of VCC/GND nets. Again, majority of modern CPU's/SoCs use the same approach:


Closer look at Ethernet controller (likely 10GbE KR/KX4):


After etching metal layers - we can see auto-synthesized sea of standard cells (Multiclet CPU had similar standard cell "waves" despite significantly larger 180nm technology), lots of generated memory/register file instances (typically supplied by foundry), few supposedly monitoring blocks scattered across the area (tiny vertical blue-white rectangles). Significant part (~25%) of die area is not used by transistors, and just filled with dummy cells.


Taking a closer look: This is what I suppose a monitoring block (could measure temperature/inverter loop oscillation speed for example). Around the block - sea of dummy cells:


Columns of standard cells in highest optical resolution. 1px=28.5nm here (area on the photo has 28µm width - that's just ~0.03mm). Optical resolution is limited by diffraction to ~200nm which makes image to look quite soft. Standard cells look similar to 180nm for example - same columns of transistors back-to-back : [P N] [N P] [P N]... (i.e. neighbor columns of standard cells are mirrored). Columns with P-type transistors are slightly wider:


One of the smallest generated memory blocks. SRAM area uses only small fraction of it (the rest goes for row/column decoders, sense amplifiers and IO logic). Again, around it we can see sea of dummy cells (one cannot just leave it empty - this would cause uneven height which will ruin lithography of metal layers):


Finally - photo of Baikal CPU in immersion oil, right after last microscope shoot:


In my view Baikal-T1 is a large step forward for Russian civilian microelectronics. It uses modern commercial core, it is designed using industry-standard flow and manufactured using high-volume commercial technologies. Finally it uses open source compiler and OS - which makes it much easier to support and work with comparing to infamous Elbrus. Using non-standard solutions in situations where they could have been avoided was the curse of Russian microelectronics, until Baikal.

That's it for now. If you like articles like this one - you can now support us via Patreon or other means.]]>
Tue, 28 Feb 17 21:39:56 +0300
<![CDATA[Hitachi HA17555 : weekend die-shot]]> http://zeptobars.com/en/read/Hitachi-HA17555-555-timer Die size 747x762 µm.



For some reason different die sample had significantly different color:

]]>
Sun, 19 Feb 17 06:33:37 +0300
<![CDATA[National LM394CH - super matched BJT : weekend die-shot]]> http://zeptobars.com/en/read/National-LM394CH-super-matched-bjt
Die size 1449x1347 µm.



Chip is soldered to ceramic substrate holder:
]]>
Thu, 16 Feb 17 23:37:06 +0300
<![CDATA[Espressif ESP32 - integrated Wi-Fi transceiver: weekend die-shot]]> http://zeptobars.com/en/read/Espressif-ESP32-Wi-Fi-Bluetooth-2.4Ghz-ISM Espressif ESP32 sets the new level for widely available WiFi/Bluetooth transceivers. Now with 2x LX108 CPU core and 520kB RAM. Surprisingly - includes integrated hall sensor. Manufactured on TSMC 40nm ULP technology. Huge upgrade over their previous highly successful ESP8266.

Hopefully next time we'll see their take on 5.8GHz ISM band.

Die size 2960x2850 µm (~double the area of ESP8266).



Closer look at RF part - some parts resemble ESP8266:


Thanks for the chips and modules to Espressif .]]>
Sun, 05 Feb 17 03:04:45 +0300
<![CDATA[Analog Devices AD584 - precision voltage reference : weekend die-shot]]> http://zeptobars.com/en/read/AD584-reference-analog-devices Analog Devices AD584 is a voltage reference with 4 outputs : 2.5, 5, 7.5 and 10V. Tempco is laser trimmed to <15ppm/°C and voltage error to ~0.1%.


Die size 2236x1570 µm.

One can refer to a die photo from AD datasheet showing a bit older design of the same chip:

]]>
Fri, 27 Jan 17 00:01:04 +0300
<![CDATA[Vishay TSOP4838 - IR receiver module : weekend die-shot]]> http://zeptobars.com/en/read/Vishai-TSOP4838-IR-receiver-Remote-Control Vishay TSOP4838 - decodes IR commands sent with 38kHz modulation. This modulation (and narrow-band filter on receiver module) is required to eliminate ambient light sources which could flicker somewhere at 50-100Hz or 20-30kHz (bad CFL/LEDs). Black (IR transparent) plastic also helps with background noise.


Die size 590x594 µm.

Photodiode is on a separate die:

Die size 1471x1471 µm]]>
Mon, 16 Jan 17 17:37:37 +0300