Mercurial > whatever / changeset

Binary file content/images/numworks/flex_design.jpg has changed

Binary file content/images/numworks/flex_pcb.jpg has changed

Binary file content/images/numworks/flex_pcb_connected.jpg has changed

Binary file content/images/numworks/partial.jpg has changed

Binary file content/images/numworks/rastnest.jpg has changed

Binary file content/images/numworks/repaired.jpg has changed

Binary file content/images/numworks/success.jpg has changed

Binary file content/images/numworks/test_ips_module.jpg has changed

--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/content/numworks-lcd.rst	Wed Jul 12 21:35:29 2023 +0200
@@ -0,0 +1,186 @@
+==========================
+ Numwork N0110 calculator
+==========================
+
+:Author: David Douard
+:Category: Electronics
+:Tags: Numworks, N0110, repair, calculator
+
+
+My daughter's `Numwork <https://www.numworks.com>`_ N0110 graphing calculator
+got fried when plugged in a cheap USB PSU. Black screen, heating up killing the
+battery in a matter of a few tens of minutes.
+
+The design being originally Open Source (hardware and firmware), it makes sense
+to try to fix it instead of just buying a new one (it is not a cheap
+calculator).
+
+The schematic is pretty simple: built around a `STM32F730V8T6 <https://www.st.com/en/microcontrollers-microprocessors/stm32f730v8.html>`_, there are
+only a few components surrounding the MCU:
+
+- an LCD (280x320, 2,8", ST7789V driver) driven using the 16 bits 8080
+  interface, taking advantage of ST's `FSMC <https://www.st.com/resource/en/product_training/STM32F7_Memory_FSMC.pdf>`_ controller,
+
+- a 64MB flash (`AT25SF641 <https://www.renesas.com/us/en/products/memory-logic/non-volatile-memory/spi-nor-flash/at25sf641-64mbit-27v-minimum-spi-serial-flash-memory-dual-io-support>`_) to store apps and user data,
+
+- LiPo charger (`RT9526AGE <https://www.richtek.com/Products/Battery%20Management/Single%20Cell%20Li-Ion%20Battery%20Charger/RT9526A?sc_lang=en&specid=RT9526A>`_),
+
+- `RT9078 <https://www.richtek.com/Products/Linear%20Regulator/Single%20Output%20Linear%20Regulator/RT9078?sc_lang=en&specid=RT9078>`_ 2.8V voltage regulator to poser the MCU from the LiPo battery,
+
+- LCD backlight DC-DC converter (`RT9365GQW <https://www.richtek.com/Products/LED%20Driver/White%20LED%20Driver/RT9365?sc_lang=en>`_)
+
+- 3 LEDs plus a number of passives, keypad, reset button,
+
+- LiPo battery,
+
+- USB port surge protection diodes (`USBLC6-2SC6 <https://www.st.com/en/protections-and-emi-filters/usblc6-2.html>`_)
+
+
+The list of fried devices was pretty much every active component:
+
+- MCU,
+
+- LCD,
+
+- 2.8v regulator,
+
+- USB protection diodes.
+
+
+Due to the global chip shortage, I had to gamble on Aliexpress to find a
+STM32F730. I bought one from `there <https://fr.aliexpress.com/item/1005004924466026.html>`_. Against all odds, the delivered chip
+looks genuine and is working properly.
+
+So I replaced the MCU and the 2.8v regulator. At this point, I wasn't sure
+about the status of the LCD. I did not know the USB protection diodes was dead,
+so uploading a firmware was quite a challenge; USB based DFU was not working
+(obvioulsy), but it was not obvious to understand why; I wasn't sure if a blank
+STM32 should enumerate ok as USB device, or if bootstrping a firmware was
+required before.
+
+I also did not want to upload Numworks' latest official firmware nor using
+their WebDFU tools since they changes their licensing policy and moved to a
+closed source model, official firmwares since version 16.3 now lock the
+bootloader to prevent custom or open source firmware from being installed.
+A complete explanation of this (in French) is `provided here <https://tiplanet.org/forum/viewtopic.php?f=97&t=25561>`_
+
+So I started looking at the alternative firmware projects for the Numworks:
+`Omega <https://getomega.dev/install/latest>`_, `Phi
+<https://phi.getomega.dev/>`_, `Khi
+<https://www-fourier.univ-grenoble-alpes.fr/~parisse/nws.html>`_ and `Upsilon
+<https://lolocomotive.github.io/Upsilon-website/install>`_ (!) Plus starting
+from a blank STM32 instead of an upgrade from an existing installation added a
+bit of confusion.
+
+
+After some time I got convinced I had successfully uploaded a firmware, but
+noting showed up on the LCD: diagnostic of a dead LCD as well...
+
+The tricky part is that the LCD used in the calculator is pretty common on
+principle: standard size, usual LCD controller. The very one used in the
+calculator seems specificly made for the Numworks: specific ribbon cable and
+specific pinout. Could not find a compatible device on usual sources.
+
+To validate I did indeed have a successfully replaced the MCU and uploaded a
+working firmware, I did a nasty rastnest experiment with a ST7789V LCD module I
+had around.
+
+The result was pretty ugly
+
+.. image:: {static}/images/numworks/rastnest.jpg
+   :alt: The test setup to connect a ST7789V based LCD module to the Numworks,
+         using a breadboard to interconnect the 2 interfaces.
+   :class: image-process-large-photo
+
+but I had *some* results:
+
+.. image:: {static}/images/numworks/partial.jpg
+   :alt: The LCD module shows half a screen with some content mixed with garbage.
+   :class: image-process-large-photo
+
+
+First I was thinking the issue could be related to the speed of the
+communications between the MCU and the LCD: the signal path was far from
+optimal, with a ribbon cable, a breakout board, a bus of 10cm jump wires, a
+breadboard, another bunch of 10cl jump wires, yet another bunch of jump wires,
+another breakboard, and at last, the ribbon cable of the LCD module... So I
+spent some time hacking the Omega source code to reduce the transmission speed;
+but that did not help.
+
+To figure out what was wrong I ended up using my DSLogic logic analyzer. The
+culprit was a pair of swapped jumpers (strangely there was only one pair of
+messed connections). Since it was a pair of data lines above D7, it was
+partially working. So the messed half display.
+
+With the 2 wires back in order, result was much better...
+
+.. image:: {static}/images/numworks/success.jpg
+   :alt: The LCD module shows the Numworks main screen... but flipped
+   :class: image-process-large-photo
+
+For some reason, this display showed the image reversed. Anyway, I was
+confident enough to spend a bit more money buying a replacemement LCD module.
+
+Choosing a suitable replacement was not an easy task:
+
+- It needs to use the ST7789V controller.
+
+- It needs to be the proper size; seems easy, it's a standard 2.8" module...
+  well no, most of these modules are 66.95x47.80mm while the module in the
+  calculator is 66x50mm. Also the thickness must be about 2.1 or 2.2mm max.
+
+- It needs to have the proper ST7789V signals available, 16bits 8080 mode;
+  especially the Tearing Effect (TE) signal is used by the firmware to optimize
+  display refresh timing.
+
+- It needs to be an "12 o'clock" IPS LCD module (not a TN one) so the viewing
+  angle is adapted when used in the calculator.
+
+I found a few references on AliExpress (almost) matching all these contraints
+and bought `this one <https://fr.aliexpress.com/item/1005004628969756.html>`_.
+
+Which worked nicely:
+
+.. image:: {static}/images/numworks/test_ips_module.jpg
+   :alt: The IPS module works just fine, the image is correctly oriented.
+   :class: image-process-large-photo
+
+Obviously, the ribbon cable pinout is not the same as the original LCD module.
+So an adapter had to be made. At first I thought I could just solder a bunch of
+thin enameled wires, but this was not a very good idea; too hard to do
+properly, without damaging the ribbon cable etc. So I decided to try the new
+Flex PCB service JLCPCB (and others) now offer. Never did a Flex PCB design,
+was a nice opportunity to give it a try.
+
+I did a simple design, with one side being connected to 30 pins FCP connector
+on the Numworks side, and the other side with a 40pins FCP connecter soldered
+onto the Flex PCB.
+
+The design, printed on paper to check dimensions, looks like this:
+
+.. image:: {static}/images/numworks/flex_design.jpg
+   :alt: Paper print of the flex-PCB adapter.
+   :class: image-process-large-photo
+
+The flex-PCB arrived a few days later:
+
+.. image:: {static}/images/numworks/flex_pcb.jpg
+   :alt: Flex-PCB adapter side-to-side with the paper print.
+   :class: image-process-large-photo
+
+Was not sure how it would support hand soldering the FCP connector, but it went
+ok.
+
+.. image:: {static}/images/numworks/flex_pcb_connected.jpg
+   :alt: Flex-PCB adapter installed in the Numworks.
+   :class: image-process-large-photo
+
+It fits pretty nicely in the Numworks enclosure; the only modification I had to
+make it to trim a bit some plastic pads to fit and center le LCD module.
+
+.. image:: {static}/images/numworks/repaired.jpg
+   :alt: The Numworks fully reassembled running the Omega firmware.
+   :class: image-process-large-photo
+
+
+The Kicad design for the Flex-PCB is available `here <https://git.sr.ht/~douardda/numworks-lcd-adapter>`_.