0200-2024-05-01.md

5 Apr 2024

Previous journal:	Next journal:
0199-2024-04-04.md	0201-2024-05-11.md

Belated update (TT chips, and analog)

Since my last update I:

• Started working at Efabless, during the busy April 2024 tapeout period.
• Completed an analog submission to TT06, called tt06-grab-bag
• Started looking at a unity gain voltage buffer or other opamp solution for redoing my DACs on TT07.
• Received my TT03 chip/board, and TT03p5 (alpha).

Bringing up my RP2040 board

I think I forgot to document that I received the 5 RP2040 boards I designed (back in 0190) and they generally look good and the first one I plugged in seemed to function correctly:

When plugging in USB, it shows up on my PC as a drive:

Normally USB_BOOT would need to be pulled low to go into this mode, but it does this automatically if there isn't a valid firmware. If a valid firmware were already loaded on the onboard flash, and RUN were pulled high (actually it seems it works with it floating, despite arguments), it would start executing that.

In this mode, I can download the generic Raspberry Pi Pico MicroPython firmware UF2 file (also referenced from the official Pi Pico doco), and drag it onto the drive as it presents in Windows, and the firmware flashes to the chip. The drive should then disappear indicating that the firmware is now running.

I then downloaded Thonny and did "Install only for me".

In the bottom-right corner of Thonny, click to select the interpreter, and it recognises I have an RP2040 on COM3. Select it.

help() in the interactive window shows the RP2040 board is alive.

I've got 3 GPIO LEDs on this board:

• D3: GPIO17, active high.
• D4: GPIO15, active low.
• D5: GPIO16, active low.

D4 and D5 are illuminated, I guess because they are defaulting to being inputs, and either have weak pull-downs, or this is just the input path load.

list(map(lambda e: machine.Pin(e), [17,15,16]))
# => [
# Pin(GPIO17, mode=ALT, pull=PULL_DOWN, alt=31),
# Pin(GPIO15, mode=ALT, pull=PULL_DOWN, alt=31),
# Pin(GPIO16, mode=ALT, pull=PULL_DOWN, alt=31)]

Set all 3 of these GPIOs to 1 (should turn on D5, turn of D3/4):

for e in [15,16,17]: p = machine.Pin(e, machine.Pin.OUT); p.value(1)

Now turn D3 off, D5 on:

machine.Pin(17).value(0); machine.Pin(16).value(0)

Toggle D3 using PWM:

d3 = machine.PWM(17)
d3.freq(8) # This is as slow as it can go.
d3.duty_u16(32767) # 50% duty cycle.

Get existing system speed, then change it to 20MHz:

machine.freq()
# => 125000000 or 125MHz.
machine.freq(int(20e6)) # 20MHz seems to be the lowest.

After replugging the USB, the MicroPython firmware starts running again.

Voltage buffer experiments

Basic MOSFET current source:

Ditto:

Using ngspice raw files for rendering

I want to take an ngspice .raw file and use it to approximately render how something would look on a VGA display.

In time this will be useful when I've got cosimulation (Verilog+SPICE) working for actual output from my VGA driver design.

For now I just want to sample an analog output at every 40ns and use that to render a pixel luminosity.

I could use set filetype=ascii to generate a readable output format instead of binary. Either that or I could learn to parse the .raw file.

I really only need to capture V(out) so that should make the file more lean.

It looks like there was a branch of ngspice for writing LXT files (see also) but I don't know much about that yet.

It may be possible to process ngspice .raw files with the ltspice Python package.

Cosimulation

Using Matt's R2R DAC demo

1. Install Verilator. On my relatively clean Ubuntu 22.04 system, this gives Version: 4.038-1:

   sudo apt update
   sudo apt install verilator

   NOTE: I was using 4.228 back in July 2023 (see 0110). Even the MPW8 VM uses 4.227.

   Hmm, I've found that this Verilator version is too old, and complains about this in Matt's design:

   %Error: ../verilog/rtl/r2r_dac_control.v:44:9: Unknown verilator lint message code: 'WIDTHEXPAND', in '/*verilator lint_off WIDTHEXPAND*/'
      44 |         /*verilator lint_off WIDTHEXPAND*/ 
         |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   

2. Let's try installing Verilator from source:

   1. Uninstall: sudo apt remove verilator
   2. Prereqs: sudo apt install git make autoconf g++ flex bison help2man
   3. Clone and pick version 4.228:

      cd ~
      mkdir -p JUNK
      cd JUNK
      git clone http://git.veripool.org/git/verilator # or maybe: https://github.com/verilator/verilator
      unset VERILATOR_ROOT
      cd verilator
      git pull
      git checkout v5.024 #NOTE: v4.228 is the latest version in the 4.x branch

   4. Configure:

      autoconf
      ./configure

   5. Build:

      make -j$(nproc)

   6. Install:

      sudo make install
      #NOTE: To make this work, I had to first:
      # sudo make uninstall
      # sudo rm -rf /usr/local/share/verilator
      # sudo apt install help2man

   7. Also:

      which verilator
      # => /usr/local/bin/verilator
      verilator --version # NOTE: Might need to open a new shell to avoid a cached old copy
      # => Verilator 5.024 2024-04-05 rev v5.024

3. Clone Matt's repo: https://github.com/mattvenn/tt06-analog-r2r-dac

4. Go into sim dir

5. Run:

   ngspice vlnggen ../verilog/rtl/r2r_dac_control.v

6. Edit mixed.cir to update the .lib line with your own PDK path, i.e. change it from:

   .lib /home/matt/work/asic-workshop/shuttle-2404/pdk/sky130A/libs.tech/ngspice/sky130.lib.spice tt
   

   ...to:

   .lib /home/anton/.volare/sky130A/libs.tech/ngspice/sky130.lib.spice tt
   

Notes

• Change xschem default terminal from xterm to gnome-terminal:
  • In local xschemrc, near line 289, add: set terminal { gnome-terminal }
  • Anton's 'Ngspice interactive' line was: $terminal -fa 'Monospace' -fs 14 -e {ngspice -i "$N" -a || sh}
  • Replaced with: $terminal --zoom=1.5 -- bash -c "ngspice -i \"$N\" -a || sh"
  • NOTE: Potential conflict between local xschemrc (setting the terminal) and the modified terminal config line (which seems to be global).
• System-wide Magic config file: /usr/local/lib/magic/sys/.magicrc
• I installed the xuanli.spice extension in VSCode for SPICE (.cir) script syntax highlighting.
• http://www.h-renrew.de/h/ngspice/ngspicedemo.html
• https://www.isotel.eu/mixedsim/blender/2wheel/index.html#exporting-charts-schematics-and-data
• It looks to me like ngspice time sampling is variable depending on the precision you specify for your stimuli. For example, even if your transient analysis is set to 40ns resolution (tran 40ns 64us), it will be finer than that if a given voltage source has rise/fall timing which ramps at a smaller time scale...?

TODO

• Doco for tt06-grab-bag
• Full sim of extracted tt06-grab-bag layout
• Create a 'fork' of tt06-grab-bag and make it a generic TT project (not specific to TT06 or TT07) -- also make it a 'VGA tests' focus.
• Prep TT07 submission
• Install gnuplot and learn to use it in ngspice, e.g. see near here
• Find out how to determine max current that can be supported and realistically delivered by sky130 devices internally (e.g. NFETs).
• Try voltage buffers on R2R DAC output -- bias it: DAC currently gives [0, 1.8V], but ideally we want [0, 0.7V].
• Clean up VM(s)