NAND Flash upgrade from 4GB to 8GB

flash

We used to use H27UBG8T2BTR in all our OLinuXino and SOM boards with ¬†NAND flash. Hynix stopped 4GB NAND manufacturing back in 2014, fortunately there were huge stock supplies in Asia and we kept buying this part until recently, now this stock is gone and these ICs are either not in stock either some refurbished pulls from old boards with unknown quality, so once we couldn’t find these at official Hynix distributors we decided to move to another ICs.

Our first attempt was to use the bigger version H27UCG8T2ETR-BC but after losing months trying to make uboot working with these we gave up. The problem? There is ZERO information about this flash and zillion of combinations of block sizes and parameters, which we have no time to test all.

Fortunately Micron’s MT29F64G08CBABAWP is supported by Allwinner uboot and Livesuit and relatively easy to find.

I write relatively as memory prices are CRAZY last months, both RAM and Flash, here is the price trend from Hong Kong IC memory stock exchange:

memory

With these rising prices all official distributors and manufacturers slow down their sales waiting for bigger prices, if you ask officially Micron for quote they can’t give you price and say – allocation, minimum delivery time 20 weeks, good luck ūüôā

What we found is that all Android images work well with Micron 8GB part, the Linux images we had to work with 4GB will not work with the 8GB parts as the Allwinner Android uboot (which is used if you want to boot from NAND) has bugs and 1 out of 10 times will fail to boot if the memory is bigger than 4GB (32 bit arithmetic bug? nobody knows as this is binary blob) and mainline uboot has no NAND flash support, so if you want to boot Linux from NAND Flash the solution is to make small 16MB FAT16 partition for the Allwinner uboot to boot then you can use second partition with EXT4 format and there will be no problems.

To make things more complex we use eMMC Flash for some of our boards (A20-OLinuXino-LIME2 and A64-OLinuXino) which are in industrial temperature grade.

To separate eMMC from NAND Flash we add letter ‘e’ or ‘n’ in the board names.

A20-OLinuXino-LIME2-e4GB means LIME2 with 4GB eMMC Flash memory

A20-OLinuXino-LIME2-n8GB means LIME2 with 8GB NAND Flash memory

eMMC memory works like SD card, it’s faster than NAND and has wider operating temperature range.

We upload end of this week new Linux images which will work on both old 4GB and new 8GB boards.

A64-OLinuXino Open Source Hardware board with 64-bit Cortex-A53 processor is in released

A64-OLinuXino-

A64-OLinuXino-22

A64-OLinuXino OSHW board is now released. Current revision is Rev.C.

Features are:

  • A64 Cortex-A53 64-bit SoC from Allwinner
  • AXP803 PMU with Lipo charger and step-up
  • 1 or 2GB or DDR3L @672 Mhz
  • 0 / 4 or 16GB of industrial grade eMMC
  • SPI Flash in SO8 package with hardware WP (not assembled)
  • USB-OTG and USB-HOST
  • HSIC connector¬†(not assembled)
  • Gigabit Ethernet
  • BLE/WiFi module
  • HDMI and MIPI display connectors
  • microSD card
  • Debug console serial connector
  • Audio In and Out
  • LCD display connector
  • GPIO 40 pin connector (not assembled)
  • UEXT connector (not assembled)
  • 5V power jack
  • Dimensions: 90×62.5 mm

For the moment we have three models:

  • 1G0G with 1GB RAM, no Flash, no WiFi/BLE
  • 1G4GW with 1GB RAM, 4GB eMMC and WiFi/BLE
  • 2G16G-IND with 2GB RAM, 16GB eMMC with industrial grade components -40+85C

The optional connectors and SPI Flash etc may be assembled upon request for small fee.

A20-OLINUXINO-MICRO now available also in Industrial temperature grade -45+85C

A20-OLINUXINO-MICRO-EMMC-3

We are selling for some time already A20-OLinuXino-MICRO Rev.J where few things were improved:

  1. We changed the LAN PHY from Realtek to Microchip as latter is more reliable supplier for both commercial and industrial temperature components, we searched desperately Realtek PHY in industrial temperature grade but without success.
  2. We extended the input working voltage from 6-16VDC to 8-24VDC
  3. We changed the NAND Flash to eMMC (but old NAND style flash is still possible to assembly)

The Ethernet PHY change requires new patches on the Uboot and Linux images which are already uploaded.

Ground Penetration Radar with LIME2 the ultimate tool for all treasure hunters

Screenshot from 2017-04-04 11-45-57

Ground penetration radar is cool device which allows you to see what is below earth surface. The principle is same as the radars which operates at air. They send pulses and sense the reflections of this pulses back. When the surface is not consistent you get reflections which allow you to “see” cavities of metals under the surface, some GPRs can scan as deep as hundred meters below the surface. This is ultimate tool for all treasure hunters, but also very useful if you want to see water and other fluids underground.

We spotted interesting post at our forum for GPR made by company from Plovdiv with LIME2.

The device looks very nice in the custom box they made:

7-small

The GPR has antenna and WIFi connection so you can connect and visualize the results with any device:

10-small

 

FOSDEM and TERES I update

fosdem-1

FOSDEM was great place to present the TERES I, as there were the right people who share open source values.

We got lot of people passing through our table and got tons of feedback and suggestions. Thanks guys!

One of the most asked questions was: “can you make it with more RAM memory?” ¬†The answer is yes, we can and we even have the memory chips which easily allow TERES I to be with 2GB RAM, but these are quite expensive than the mainstream memories as they are just two 4Gb memory chips put in one package with two Chip Select signals which A64 supports. As these memories are not so used their prices is more than double of the normal DDR memory price and this is why we though initially that this will be expensive option to consider. If we put 8Gb DDR memories on TERES I instead of 4Gb and make the RAM memory 2GB this will affect the price with EUR 15. Perhaps we should just add this as option when we assembly the main boards and people can decide what to buy 1GB or 2GB RAM version.

The 4GB mainboard eMMC flash is also quite humble for storage. Micron who manufacture this eMMC has also 16GB and 32GB versions, but they are very slow to deal with, just to receive quotation from their sale reps here needs weeks and even if you order they will tell you something like 20-30 weeks delivery time. For the moment this is no go. We use these 4GB eMMC in our other boards this is why we decided to use here too. The 4GB eMMC is in stock and we can produce right now. The micro SD card connector kinda solves the storage issue as you can use up to 128GB micro SD cards, but they are not so fast as the eMMC on board. Anyway if we could make deal with Micron later we could offer mainboards with more on board eMMC.

Another interesting suggestion was to add small SPI flash on the mother board. Initially as hardware guys we decided to not do this when we designed the main board. our though was: if you have eMMC 4GB on board which you can read with 100MB/s and write with 10MB/s why would you need slow serial SPI flash with small capacity like 4MB on board? We got explanation at FOSDEM why some people would love to have it. The SPI Flash could be used to have boot code from which A64 will always boot first. The SPI Flash could be hardware write protected, i.e. you will be always sure that your processor boots trusted software and no one could overwrite it except if he has physical access to your laptop and open the plastic and put the SPI to read/write mode again. The mainboard eMMC could be overwritten from user space, so considered as compromised media for secure boot by the people who care about their security.

One more good feedback was that A64 has OTP bits, one of which once set A64 will refuse to boot unsigned code, and no one knows how to sign code for A64 yet, so this effectively bricks your device and all existing A64 boards and tablets and laptops suffer from this problem. One could now write malicious software which to set this OTP bit and brick the A64 devices on the market. Fortunately there is A64 pin which enables and diables the OTP writting at hardware level, so we are going to disable OTP write by hardware.

Another question was “can I have better display with higher¬†resolution?” Sure you can! TERES1 laptop has eDP interface which all laptop LCDs share, so any LCD with eDP interface will work, and of course will require proper Linux setting. Right now if you do not like the current 1366×768 resolution you can search for other 11.6″ with eDP connector and spend almost as the cost of the TERES I for LCD with IPS and 2K or 3K resolution, but franky we do not see the point of fancy graphics as TERES I would never be laptop for graphics developers anyway.

Other ¬†people were asking: “can I have 16GB of RAM, SSD disk” No, unfortunately this is not possible. A64 is humble low cost processor, it has no SATA, so SSDs are not possible to connect except via USB which spoils the SATA speed. And IIRC 4GB RAM is the max A64 can address, but needs total mainboard rerouting. At this stage TERES1 would never be developer laptop which you can use to build Linux kernels, or do fancy graphics etc. We see it more like hacker tool. We made it lightweight, our target is to run on battery for long time, so you can get it while you travel. We work on FPGA internal board which you will be able to program with FOSS and to add Oscilloscope and Logic Analyzer capability at later stage while keeping all other hardware intact so it will be add-on board. We think of TERES1 as to become portable lab for hackers, to may program Arduino boards, sniff protocols at hardware level, capture analogue signals etc. it will be still good to browse internet, edit text files, code embedded software, but do not expect it to replace your desktop.

Good suggestion was to bring out the debug UART console so developer could see kernel messages without opening the laptop and solder wires. So we decided to add multiplexer and will bring Serial UART TX, RX, GND on the audio headphone jack which will be multiplexed by software, so the developers could just plug serial cable to Audio jack and debug.

Some people suggested us to place Arduino connectors near the touchpad where you add Arduino shields, this need a bit of consideration. When you do prototype work and experiments there is always possibility to damage your hardware, even now when we use Arduino we always put USB-ISO between it and the computer we use to program. With this protection we are sure that if we short something or feed high voltage to the shield by accident we will not damage our computer. So this is great idea but needs a bit of thinking.

Some people were asking: “do you think if this will be commercial success project, your laptop is so expensive, there are Chinese laptops for $50, $60, $100?”

Frankly we do not care too much about this, our core business is development boards and if you follow our blog you see that we have enough work. We spend more than year on TERES I so far and it was fun project and we learned a lot during the development. Now if it will be liked by many people or not is not so important, the important thing is that we made first step to bring to people, who appreciate open source an platform and template which they can use and improve both harware and software wise.

This is why we selected KiCAD as our CAD for designing TERES I. What is the point to release OSHW made with Altium or Eagle which will require your community to spend thousands of EUROs if want to study or modify your files? Every time you use proprietary tool to make Open Source you just decrease your community base just to people who can afford to buy the tools you used to design.

Now TERES I gives freedom to everyone to download KiCAD, ¬†the CAD files from GitHub and people can view how it’s done, learn something new, and if you do not like something you can modify it up to your taste. You can’t do with any other laptop on the market.

Note that this is just the first step, the development and the fun will continue. Once we finish the software, add on boards we will look around for more new SOC candidates as well.

Not at least everyone was asking “when it will be available for sale?”. We build our first three prototypes 3 days before FOSDEM. While we worked more than year on the harware and we solved all issues there, the software is in quite initial state.

For the moment the only working Linux Kernel which supports all A64 features is the Allwinner Android Kernel. This Kernel is full of binary blobs, but the only one which could be used for demo. Beside the binary blobs many other things are broken, like the power management, different drivers like the LCD backlight PWM, wake up from suspend, eDP converter is not set properly and works just in 15 bit color mode etc etc. We have the hardware for 50 laptops ready (developer edition), but we do not want to ship before we take care for the software. At other hand we do not want to ship TERES I with Android or RemixOS also which are complete with binary blobs and will never be Open Source.

So let’s hope we will have good enough Linux support in couple of months and we can start the shipments, until then please wait patiently, we spend over an year and now we are close to the final ūüôā

All above suggestions requiring the hardware modifications for the debug console, OTP lock, SPI Flash will be implemented in the next run when the software is completed.

 

A20-OLinuXino-LIME2 now with PCB revision G

a20-olinuxino-lime2

A20-OLinuXino-LIME2 now is assembling on same PCB Revision G as A20-OLinuXino-LIME2-eMMC.

What are the improvement:

  • Ethernet PHY is changed to RTL8211E replacing the obsolete RTL8211CL no need for kernel patches;
  • we drop the odd shape which was necessary to fit LIME2 in the plastic box as we now have range of metal boxes;
  • the four mount holes now have grounding for better contact with chassis;

Meantime we silently work on further improvement for next revision (to be released March 2017):

  • adding SPI boot Flash;
  • replacing RTL8211E to industrial grade PHY, so the board can be produced completely in -40+85C operating temperature;

 

Pre-Christmas Crazy times – Expansion of production capacity with three new Samsung machines

christmas-balls1

Probably many of you wonder why we are so silent recent months and nothing new is posting on the blog. The reason is quite casual Рwe are overwhelmed with work!

Starting from September something odd happens. Our OLinuXino customers went in crazy mode and we start getting almost twice more orders than usual. All who have been dealing with production know that to double capacity needs time.

This is the reason for the delay with iCE40HX8K-EVB, A64-OLinuXino, TERES release – there is simple no free window on the SMT assembly machines and testers to may run them, the machines are busy with OLinuXino and SOM assembly.

The small orders are shipping on time i.e. in 1-3 working days, but all bigger OLinuXino and SOM orders for 50+ boards are now shipped with a bit of delay as we have quite backlog for these boards.

We apologize to all our established customers which do wonder why the orders which we usually shipped to them in 1 week before, now are shipped with delay of 3-4 weeks.

We assure you that we do everything humanly possible to ship all orders ASAP.

To add little more crazyness the three new Samsung machines arrived two weeks ago. We wanted to build new space for them and working more than year on this, but the lazy and ignorant Bulgarian administration still didn’t issued us permit to start building, so we have to install them in the old building where we run out of space.

20161118_102525

20161118_102818

One SM471 and  two SM482 with printers, loaders, unloaders, conveyors, packed in 20 wooden boxes were unloaded from the containers:

20161118_102435

Immediately we start to break walls, extend doors, and other funny things so the big machines may enter the building.

20161126_085319

20161126_085332

20161127_075039

We did same 1 year ago when our first Samsung came, but shortsighted decided to rebuild the walls with intention the new machines to be installed in the new building, without taking into account how long and difficult the building permit is to get in Bulgaria.

Now the three machines are installed and testing, but as you guess they will not run alone, so we will need new employees to train.

New Year – new luck we say here, I hope things will go back from crazy to normal mode in January!

EDIT:

Some words about the machines SM471 is the Samsung fastest and greatest from SM series. It has two heads with 10 nozzles each and dual rail conveyor which allow two different boards to be assembled at same time. With maximal performance 75000 csp (which you never reach on real boards). The machine can place down to 01005 components which are with size 0.4 x 0.2 mm (400 x 200 microns)!

SM482 is flexible mounter which can place both small and big components, we already have one such machine and it performs very well. The listed maximal speed is 28000 cps.

Each SM482 machine comes with full set of 8, 12, 16, 24, 32 mm feeders and tray changer which can hold up to 80 trays with components.

Both SM471 and SM482 support the new splice-less tape feeders which can be load even with component stripes without need to have initial empty cells on the tapes.

Previous Older Entries