Difficulties with the purchase of sdtam 16Mx32 512Mbit.

[AVI-crak]

Posted on July 02, 2018 at 12:53

At the moment, almost all the memory chips of such a configuration are discontinued. They were not popular at the time of active sales, and after the end of support for manufacturers - leftovers remained at an inadequate price ($ 25 per chip). At the same time, new memory chips, with much higher capacity and speed of operation - have a significantly smaller price tag.

Prompt a way out:

1) Use microcontrollers from other manufacturers - those that have a lower cost of a microcontroller + memory bundle.2) Use microcontrollers from other manufacturers - those that have sdram memory and microcontroller in one package.3) Continue to use ST chips - with an obviously overpriced finished product, and humbly wait for bankruptcy.

Question number two:

What is the difficulty for ST company - to manufacture microcontrollers with sdram memory in one package.Or as an option, a compact module: microcontroller + sdram + quartz (hour + 25MHz). Not a fee for development - but a module for embedding in industrial devices.#sdtam-16mx32-512mbit

[Tesla DeLorean]

Posted on July 02, 2018 at 14:17

Perhaps design using SDRAM parts that people like Micron have provided 10-year availability guarantees on, rather than catch-of-the-day parts. Your choices are likely to be more constrained, and less competitively priced.

Pretty sure the SDRAM process isn't quite the same as micro-controller targeted one.

Putting multiple/stacked die also has a whole bunch of headaches and availability/re-validation issues. I know ATMEL had huge issues back in the day with the AT91 devices and the made BOTH the CPU and FLASH memories involved.

As new versions of the parts are released you will have to re-validate your own design using those parts, make software changes to accommodate or recognize the evolution. Or just embrace the need to spin the board slightly on 12-24 month basis to get to the most aggressive part pricing, using the parts going into the mass-market products (phones, tablet, PC) which is driving the cost basis downward due to volume.

The Broadcom/RPi chips had a stacked BGA, but long term availability of that likely to be transient. Most Chinese manufacturers are in constant motion they build the current design and move on, availability beyond 18 months is irrelevant.

People make module boards, but there is a premium for those too.

https://www.ghielectronics.com/products/ucm

 

Pushing your design choices onto third parties to solve isn't the route to success here. They typically have entirely different business models and agendas, and a 10-year guarantee might not survive a merger, bankruptcy or natural disaster.

[waclawek.jan]

Posted on July 02, 2018 at 15:01

What is the difficulty for ST company - to manufacture microcontrollers with sdram memory in one package.

Not an unreasonable request, given the huge interest in the high-end 'F7s and 'H7s (I am not at this end and don't quite understand that interest, but I don't understand this world as such either). ST used to make stacked die/memory, the uPSD3xxx used to consist of the mcu and memory die stacked on each other, at least the higher-end ones. 10 years++, could the know-how got lost?

Btw. at least ISSI appears to actively manufacture 16Mx32 SDRAMs and the price appears to be a bit less than what you said, e.g.

https://www.mouser.sk/ProductDetail/ISSI/IS42S32160F-7TL?utm_term=IS42S32160F&qs=N2Tevpb%252bvoaxGx4J5%2fLNHw==

 . BGA tends to cost less given the high number of pins. 2pcs 16Mx16 cost significantly (half order of magnitude) less and are available from several sources, couldn't that be an option, too?

JW

[AVI-crak]

Posted on July 02, 2018 at 15:13

The options are few IS42S32160 x1, AS4C16M16 x2.

IS42S32160 is discontinued, but the factory can start the converter for an inadequate price tag.AS4C16M16 is in active production - but such chips need two, and this is an increase in area.

For IS42S32160, I see a price of $ 21 for one chip.

[Tesla DeLorean]

Posted on July 02, 2018 at 20:50

>>10 years++, could the know-how got lost?

However some practicalities and limits may have been learned. Insanity is repeating things and expecting different results.

As product life cycles shrink, validation issues expand, and issues with sourcing also expand with each additional die/part pulled into a package. Testing becomes an issue. It is not that things can't be done, but rather the cost trade-offs. You can get small parts now, wafer scale, you can get 8-layer PCB, make flex-PCB, you can mount die and wirebond on your own substrate if you want.

You want it to have no incremental or exponential cost, oops!

The responsibility for making niche products for everyone's must-have requirements is not the role of the mass-market chip vendor.

Processes have diverged, and we are already seeing MB of FLASH on die, and large SRAM. Stuff that was stack-die is pulled on-die. Frequency, temperature and packaging create issues.

At least one RF product I'm aware of has issues of noise from a stacked QSPI flash die. Current flows and proximity become issues. Different temperature and thermal expansion coefficients can become issues, causing stresses and reducing life span.

So the know how is not lost, but rather they know better.