Very interesting. Since finding the Padauk microcontrollers I've had a strange urge to use them in trashy applications and also see how far the software can be crushed to make them do great things.
Can you try to erase it by xray? Because flash should be erasable by ionizing radiation, and UV is blocked by the package. It probably not worth to erase, it is cheaper to replace.
@@adamrak7560 most EPROMs nowadays are EEPROMs, meaning ELECTRICALLY erasable programmable read-only memory. There is a variant of this chip that you can reflash, the PFS154C
The power transistor on the top is telling me, it is not any charge based memory (nor Eprom, nor eeprom, nor flash; those need higher voltage at nearly no current). The high current device suggests it is some high current zapping memory (programming a cell actually means blowing out a diode or something dimilar), so truly physically an OTP (once blown, the component just can not be "unblown"). These use to be way more compact arrays than any of the charge based cell (contain a very small Zener, so two minimal contacts, plus a selection vertical PNP - one minimal contact for the emitter, the base is the column select NWell stripe), so all is about the size of a minimal MOS in the given process. Flash needs a capacitor, one selection and one tunneling MOS, so about 3 or 4 minimal MOS size per bit cell. Beside the size, the other advantage is the state does not depend on any floating charge, so in principle no data retention problems. The only drawback is the impossibility to reprogram and the high currents needed for programming.
A bit about the cost of these, since I was curious: - They're using Magnachip Semiconductor and Powerchip Semiconductor Manufacturing to manufacture these chips. - The oldest logic node that they provide is 180 nm, which uses a 200 mm wafer. - If you tune the scribe spacing, edge loss, and defect density, they're producing 75000 chips per wafer. - The slightly conservative 180 nm wafer cost is $700. - Which means that each one of these costs $0.01 to produce. - Take away the retailer margin, packaging, and maybe shipping, and Padauk's easily got a 50%+ margin on these.
This is blasphemy, for someone in the 1950s who spent $5000 on a giant processor with capabilities that was less than this dirt-cheap, energy-efficient, ultra-portable, disposable MCU.
Good grief... I thought the smallest die that was practical was 1mm in any dimension. The mechanical means of cutting and handling these chips is just as impressive as the lithography! Hats off to you, sir. Teardowns are always fascinating, even for a mere 3 cent part. 8)
The funny thing about this is that there were already 5 cent name brand, reflashable chips since the early 2000s in lowend Microchip PICs and Atmel Tiny4. But just through the economics of distributors it was hard to actually buy them for less than ~15 cents. This 3 cent micro is mostly a marvel of efficient trading rather than efficient process engineering! How can they turn a profit on such low margins?
You wrote a typo in the video title, it's Padauk not Paduak. Thanks very much for the excellent video. I am fascinated by these chips and built a programmer so that I can play around with the open source tools that have been built around these inexpensive microcontrollers.
welcome back , good to hear your voice again. It would have been nice to have matched the pinout with the die to see if any of the special functions on the pins were visible. Also to have compared the 5 pin version to see if the same auto layout was used or if a fresh layout was run. thanks !
Reminds me of the memes posted on old computer repair videos where you had to debug TTL "I'm a redstone engineering prodigy don't worry lads I got this"
Hi my friend. Take also a look at the WCH CH551, CH552, CH553 and CH554 amd decap them. I made a video on how to prgramm them and i wonder if in all 4 of them is the same but just software locked. They are super cheap 8051 micros with usb support and the ch554 even has usb host. And they have an integrated usb bootloader. They are useable with just 2 caps and thats it.
I really love these silicon analysis of yours. I don't know if you're aware of the long gone 3Dfx video cards. I have a dead Voodoo Banshee that i would be willing to sent you if you're willing to shows us it's die under the microscope :)
Here is a good study on wafer cost: anysilicon.com/major-pure-play-foundries-revenue-per-wafer-2017-2018/ 8" (200 mm wafer). $625 ~53,000 dies per wafer 1.18 cents per part......
I doubt they are using only 4", I would rather guess for at least 8". The simple reason is the cost: 8" could be about 50% more expensive wafer than the 4", but contains 4x more of the final chips. So at the end offers at least 3x lower silicon cost per chip.
this is extremely interresting, I would like to know what you can use them for / aka, can someone direkt me to a chanel where I can learn about such micro controlers. I would like to lern to use them, I have just no understanding what good they are. ^^"""
Dave from the EEV Blog took a look at it. Dunno if someome actually did domething useful with it though. Coding has to be done in assembler, which is not exactly popular.
@@oilybrakes Not entirely true, in Dave's videos he demonstrates a C compiler that came with the programmer, though it was for something called "mini-C" which was made up by the manufacturer and not ANSI at all. So it's C like and not actually C, but importantly it's not assembly,
rockets4kids A 6502 fits into about 4500 transistors, add some more for about 128 bytes of static ram flip-flops, I/Os. Maybe that chip has around 6000. This is no NPX LPC810 of course which has a 30MHz 32-bit arm cortex-m0+ , 1K static ram, 4K flash, 4 timers, 4 pwm, 4 uarts, 6 GPIO etc., but that part is already around $1.30. All the same, I'm disturbed to see foreign economies develop their own solutions, things like these can come back to bite us.
Lots of things. Replacement of simple circuitry (think replacing a 555 and a few logic ICs), in simple tech (battery chargers, drills, PWM controllers, dimmers)...
Whatever you can code into 1024 bytes of memory and use 6 pins for. My RGB controllers use less than 1K of assembly code. This chip could be used for IR transmitters and receivers, timers and controlling things like hand dryers and flashlights.
But only if you know how to write code for them. I don't. but i could easily build a circuit using a 555 (or even multiple 555s) +logic to do a certain task. No expensive programmer or debugging lots of lines of code needed. You design a schematic and PCB layout, then you build it and it just works. anyone that looks at can see how it works by just reverse engineering the PCB.
A lot can be done with it, assuming the tool sets exist. I think SDCC May support it. ATTINY is obviously easier but for the price point, yeah wow. Cheaper than STM8
@@BLRSharpLight Taiwan no longer claims Mainland China. You need to read up a bit. This changed some years ago. Taiwan is in the process of removing the word ”China “ from its passport, national airline and many other Government bodies. I live here and am well versed in the CURRENT situation. You are quoting history. There is also a strong movement to change Taiwans name away from ROC and make it officially Taiwan.
@@Mikere5 no no no the DPP do not claim mainland China, the KMT in the other hand they considered themselves the legitimate rulers of all China, from the Pakistan-Indian border to the South China Sea.
Plenty of videos on photolithography on youtube, but a very quick, inaccurate primer: Coat silicon wafer in photosensitive polymer, place stencil over ir, expose to ultraviolet light, develop photosensitive polymer, clear the unexposed parts. Now the areas of the silicon that were under those parts are bare, and the rest of the die is still covered in the polymer. Submerge wafer in acid to eat away part of the silicon in those bare areas. Wash. Repeat. Repeat. Repeat...
Also, the size of the features shown in this video is many, many thosands of times bigger than modern 7-10-14nm processes. This is decades old technology, current processes are toying with sizes where the individual transistor gates are a few dozens of atoms wide.
@@kaboom555 WOW!!! Thank you for the explanation!!! How in gods name do they create a stencil with details THAT small and fine!?! Also, how does the acid get into those tiny gaps to eat away the silicon when you have surface tension to deal with?? How does the acid wick into gaps THAT thin? Normally surface tension would prohibit that. Thats unreal!!
@@StreuB1 that's where the inaccuracies in my explanation start... It used to be an actual hand drawn stencil, which was then projected onto the wafer with a lens that made the image smaller, the opposite of binoculars. Nowadays it's made, I believe, with laser light and interference patterns. As for the etching process itself and surface tension, no idea...
Yes, and what can it do? I can emagine that two of the connectors are for the power, and two connectors are for communication, and that all interaction goes with typing in Morse-code.
The company is from Taiwan not mainland China, but yes China have the capability of making chips for almost 40 years? SMIC was founded in 2000 and they started with a 250nm process node, smaller process node than this MCU who is probably made in 350nm process node.