We have been keeping a few things under wraps, waiting for the right moment to share the full picture. That moment is now. In the month of September there have been lot of updates which I want to share with our Mechaverse.

Over the next four posts, we are going to walk you through the sequence of events that changed our trajectory. It’s a long story, but it’s one you need to hear. Let’s get started.

Comet Status

Our revision 6 prototypes are finally now stepping out of rigorous Validation and testing, over the past two months our teams have been scrambling to get all peripherals up and working. Here is the current status of all peripherals that have been tested and their status.

Core Peripherals - SOC, RAM, PMIC, eMMC

Our core peripherals that keep the board running are working fine and we have not identified any critical issue with them over the past 2 months of testing. Memory has been tested for both 4GB and 8GB and eMMC for 64 GB and 128 GB.

Display

Our little bright display of joy has been thoroughly tested, we have kept it running on the prototypes in our office for days at stretch. We are working with our display manufacturer to address some minor issues we have identified in the display and have placed an order for 120 quantities for the next batch. Fun fact: The older Comets had 480x480 display i.e. ~200K pixels, the new one has 1240x1080 i.e. 1.3M pixels packed inside a 4" display!

WiFi and Bluetooth

The WiFi/BT module has been a source of grief for us all September throwback to Wake me up when September ends. One of the features of our WiFi/BT module is that it supports an internal and external antenna. We observed early on that the internal antenna would give us ridiculously terrible speeds and fluctuations, whereas the external antenna worked absolutely fine. Since then we have been engaging Ublox to help sort with out with multiple calls, emails and isolated tests. They finally helped us conclude that the traces and ground on the right side of the module where the internal antenna lies is the source of our problem

With an off-the-shelf flex pcb antenna, we tested speeds up to 400 Mbit/s on a WiFi 6 router. We also identified that we used 3V3 power rails, but if we switch to 1V8 we can raise these speeds further.

USBs

All USBs are mostly working, we have had some role swapping issues and few pin swapping issues here and there but largely we have been able to get all USBs up with 3.0 speeds and also PD and Power-in on the two USB Type-Cs. Here is a demo of the Comet running high speed internet over a Gigabit ethernet to Type-C. For all those who miss the RJ45, I swear I am not trying to add salt to your wounds.

Camera

Another peripheral keeping us awake at nights, we managed to pull raw image from the sensor but for the life of us we couldn't the ISP working with it. After much ado over nothing we realized that the ISP from NXP ¯\_(ツ)_/¯ for IMX219 camera is mapped to CSI1 and we are using CSI2. We are checking with NXP if they can recompile their precious closed-source ISP to use CSI2 or we change the pins to CSI in the upcoming DVT. I recently found that there was some support added in the kernel for libcamera with IMX219 on IMX8MP - there is hope!

Audio Codec

The Audio Codec is performing well, we have been able to get the onboard speaker, dual microphones and the headphone jack working. We currently have an issue where Pulse/Pipewire does not automatically switch from the speaker to headphone jack when plugged in but instead decides to play simultaneously 🤦.

We are NOT happy with the speaker we are using, by 'we' I mean me because I like me some good audio! We are changing the speaker part and adding a bigger enclosure design to try and boost the audio further. I also hate wires inside the Comet so we are moving to a push connector like the Camera and Display.

HDMI with Audio

This one I was very excited about, I somehow love the idea of the small Comet powering a large display! We got HDMI working now even with Audio working well. Here is a quick shot of Luanti (Minecraft FOSS version) running on a large TV from the Comet. The resolution somehow got capped at 800x600, but after fixing an I2C issue, we were able to run an external 1080p display in parallel with our inbuilt 1240x1080p display. Also a small video of us using it on TTY with a wireless keyboard.

PCIE: NVMe

We tested our PCIe 3.0 with NVMes, our first test was using an EVM-make SSD which didn't work and somehow heated up our SOC, when we checked its CONFIG pin we realized they are setting it to SATA mode whereas it was a PCIe ssd, we still don't understand why. But the Raspberry Pi SSDs (which are actually Samsung SSDs) worked fine, and we tested write speeds upto 630 MB/s and reads at 430 MB/s . This is an easy 5x gain compared to the 110 MB/s in the older Comet. If you notice carefully, you will see the adapter - this adapter converts 2230 M-Key to 2242 B-Key. Hopefully sometime next month we can test a discrete GPU as well, thanks Jeff (Geerling)!

Test Results (using fio): link

Pogo Pins: All 40 of them

All 40-pins on the pogo pin connector have been mapped properly. We also observed some connection issues in the center pins due to less contact between the pogo pins and the housing. To address this we decided - (i) Not to add solder paste to the contacts on the pcb for improved contact (ii) Increase compression and length of the female pogo pins (iii) Add a screw on top to ensure the pogo pin connector is pressed properly.

PCIE: LTE ⏳

We are testing with Quectel EM05 CAT4 LTE module, we have been able to get the modem detected and are able to fire AT commands but we are debugging some power consumption and SIM detection. This is one WIP item at our end, I wish I had more to share here 🙁

Power Delivery ⏳

The Comet+1 is designed to consume up to 35-40W of power from USB PD chargers, we are now working on testing these limits by adding more and more load to the device, verifying fast charging and power output to peripherals. We also got our batteries to deliver upto 5A of current, so the Comet on battery will also be able to drive a number of peripherals.

Mechanicals

Our mechanicals are also getting prepped now for R7 for better fitting, durability and thermals management. We are also running thermal and drop-test simulations to improve the design further. You can now see two parts merging in to form the main housing of the Comet. I love this wing/curve design of the new Comet.

We are also testing the extension area with N50 magnets for better grip compared to the commodity N35 we used last time.

Other Peripherals

Other long-list of peripherals have been tested and are working fine -

• Micro SD Card ✅
• RTC (via battery) ✅
• 4-Channel 16-Bit ADC ✅
• 6-Axis Gyroscope ✅
• Ambient Light Sensor ✅
• Ambient Light Sensor ✅
• Secure Element ✅
• Haptic Motor ✅

DVT: Design Validation Test Prototype

In the next few weeks, we will be incorporating all the issues identified in the EVT and release the DVT for manufacturing that will arrive by early November. The DVT will be manufactured in stages, with an eventual total of ~100, these units will also be distributed to the pilots, more on that in the coming updates.

That's it for now! See you tomorrow with the next update in the series.