NVIDIA's press release states that "Jetson TX1 is the first embedded computer designed to process deep neural networks -- computer software that can learn to recognize objects or interpret information." The 3.4x2inch module includes a Tegra X1 ARM Cortex-A57 processor with 256-core NVIDIA Maxwell graphics, 4GB of LPDDR4 memory, 16GB of eMMC storage, 802.11ac WiFi and Bluetooth, and Gigabit Ethernet support.
AnandTech Article: http://www.anandtech.com/show/9779/nvidia-announces-jetson-tx1-tegra-x1-module-development-kit
The Jetson TX1 Development Kit will be available for preorder starting Nov. 12 for $599 in the United States. The kit includes the Jetson TX1 module, a carrier board (pictured below), and a 5MP camera. The stand-alone module will be available in early 2016 (for $299 in bulk).
The Jetson TK1 (not TX1) was released in 2014 to encourage the development of products based on the Tegra K1 processor. However, according to AnandTech, developers were using the Jetson TK1 outright as a production board, choosing to focus on peripheral and software development instead of system hardware development. With the new TX1, all of the I/O connectivity is provided on a carrier board, enabling rapid development on the credit-card sized TX1 module. After development is finished, the TX1 module can be directly deployed in products, such as drones.
NVIDIA used a drone application to promote the Jetson TX1
Comments
@Dustin ... just a santa whish! Could you get TX1 developer sets in the hand of Randy, Tridge ...
Consumer grade USB 3.0 plug - is a bad idea. It's clumsy, and has unnecessary requirement of using the stock usb3 cable, which is, in airborne case - is a loss of precious weight just for the cabling, plus it's not reliable at all.
I would personally prefer 10 pin PicoBlade or 1.25mm pitched header(the same type as found on the majority of the PC motherboards).
The same goes for the HDMI as well - stock connectors are really unnecessary - 1.25mm pitched interface is far more usable - if needed the hobbyist can crimp the cable to the required length and it is far more usable.
For users who are awkward with the normal connector types - it is easy to design a breakout board, which is serving the necessity(if needed) to route out the connections to consumer grade connectors
Here's a diagram indicating the main connector pin orientation. Note that this is the top-down view, so the connector you are seeing on the back.
The lead style depth is -5.0
If there is space for extra USB, I bet most would find it useful. You will want to consult our reference designs for high-speed signals like USB3 and CSI, and the soon-to-be-published OEM Carrier Design Guide.
Nice board, juergen. But should we also discuss, if we need some additional break out lines (maybe as solder points)? This could be done on the area between SD and Hdmi out for e.g. I would like to see the 4 PINs for an additional USB 2 Port (for 3G), the pins for uart and maybe a I2C and/or SPI port ...
best .cj
@ Patrick
With the USB3 connector I don't have any information on the bandwidth capacity of picoblade or other connectors, but given that PC motherboards use common headers to route out USB3 ports into a PC case, I'd say the connectors don't play a significant role in attenuating USB performance. It would be worth investigating though before locking down the connector type.
Sometimes I think old fashioned hardware developers don't get that new applications need to be coded not just manufactured now a days. Locking up the coding part "for profit" means that it limits the sale of hardware. So an expensive design suite simply tries to recover some software/tooling costs for the manufacturer, but once the developer has software for a product the hardware manufacturer can't profit from the software suite anymore (or not much).This limits hardware sales as well and is the reason why many companies are providing development toolchains with their hardware for 'free' and factoring toolchain costs into the hardware instead, in the hope of increasing hardware sales.
No-one wants to fork out money to "try" a maybe useless or non-functional idea, so expensive design suites just don't support non-conventional ideas at all. Many of the new ideas are therefore happening in the open source space, provided that the hardware is priced so that would be developers can afford the risk to give it a go. I'd say most DIYers with adequate skills or willingness and capacity to learn are comfortable trying out something under $500, but bulk at anything more than that for hardware if they still have to sink a whole load of hours to get the code to work as well. The TX fits the bill that way if the carrier board is priced competitively.
I've had a look at the Zedboard it's a bit pricey I think. There's also the Matchstix (which is even more expensive) and Ettus B200 mini (which is a bit better) that have the right dimensions and performance for UAV SDR. Of course cost is relative, and because the actual cost is hidden in the software development, through thousands of hours of coding etc to get hardware to work as desired, many projects don't even get started due to lack of budget, or marketability uncertainty to recover costs. Still boils down to money...so low hardware entry cost and "free" software suites are key in getting "free" code work (and new ideas) done by the open source community. Linux, Android and 3DR are a classic case.
Regards JB
I agree with you 100% concerning usage of FPGA , and especially the new SOC like Xilinx Zynq or Altera Cyclone V. The main issue dealinf with
This technology is the steep learning curve of VHDL or Verilog and the smalle developpers community. I wish that Xilinx will release a free version of Vivado design suite INCLUDING the HLS , High Level Sunthesis , that allows to works on these directly with OpenCv.
Did you had a chace to look at what Avnet juste released: http://zedboard.org/product/picozed-sdr-som
Concerning the design of the board, are picoblade have the capacity to transport high bandwidth signals like USB3?
@Jurgen
RE: Samtec connector
The Carrier Spec and drawings call out these P/N's from Samtec:
These REF P/N's don't return google results, however here's the SEAF catalouge. I'll get clarification from the hardware team regarding the P/N scheme, what the "A" lead style depth is (-5, -6, -6.5), and what corners of the connector pins A1 and H50 correspond to.
RE: Toshiba HDMI to CSI bridge
Here is an active Devtalk topic about using the bridge, with interest from others in the community as well. It seems it will work with some driver tweaks.
RE: Carrier integrated heat-spreader
If you didn't see it already, you might find the Thermal Design Guide useful. Although JTX1 module TDP is rated up to 15W (during games or GFXBench), typical usage is measured at 6-8W. For example, during DNN image classification with CUDA. I do a lot of DC power measurements here, intended for battery pack.
@ Jurgen
For me USB3 would be ok on a Picoblade header if that's easier to accommodate on the board. Typically consumer plugs/sockets aren't the best way to wire up a UAV, especially if weight cable dimensions are an issue on a small frame. It's also a hassle to source specific manufactured cables with the right USB plugs etc (like straight right/left angle etc). Sometimes it's easier to run one side of the cable into a header plug, which at least makes customizing the cable length easier. Cables could then be sold extra and be terminated by the user into the header plug as required.
Vibration is another consideration, however I can't see this type of board going onto a badly balanced vibrating airframe, but "rough" landings might result in some issues, especially if the leverage on plugs/cables rip the USB3 sockets off the PCB. The best way to design UAV gear for crashes is to make components soft and to "easily disassemble" on impact. That way typically the energy is dispersed better, provided one maintains enough rigidity so that it doesn't come apart in normal operations! It's a it of a balancing act. ;-)
If possible it might make sense to make a modular type PCB where for example some components like HDMI inputs and processing can't be left off leaving just USB3's etc. But that needs to be factored into the BOM/PnP cost to see if it's at all worth making different runs of the board.
Sorry I can't help you atm with your other queries.
-
@ Patrick
I think SDR will become critical in making any type of robotics more resilient, adaptive and self aware. In my opinion I can't see why we should continue making "custom hardware" if we can predominately, within reason of physical properties (like antennas etc), make the hardware we require by customizing the software. The point is that with single chip tuners like those from Limeware and Analog Devices there's just no real advantage in customized RF hardware apart from making money from components no one else can source! I see SDR transceivers coming down to the $100-200 mark within the next 2 years, which put them in competition of conventional RF telemetry. If we're already heading that way, we should make sure they are accommodated in the companion board(s) to make them future proof.
With SDR's it's common to have the data crunching done off board by PC/server etc, with FPGA's handling the signaling. In a UAV there's an opportunity to use the aircraft as an airborne relay or "untethered flying antenna", simply by streaming whatever data the airborne SDR is tuned to receive to the ground for further processing. With the TX1 onboard some of this could be done onboard as well, but if imaging recognition/avoidance processing is being used, then it's better to only conduit the RF to ground as a digital stream for processing there.
I see one of the great advantages in SDR as the "multi band self-healing mesh network". This will give it the ability to operate as a network node to connect other participants, without being too constricted by what RF protocol is being used, from Bluetooth to 4G. For example deploying a few UAV's could result in re-establishing a base mobile phone network after a natural disaster, at least between victims and first responders. In fact it could become a form of "free internet" for them that don't yet have internet access aka Mesh Potato.
I think that there is as much, or maybe even more potential by giving flight to SDR as there is by making imaging airborne.
Regards JB
@JB, Dustin and Patrick
Thank you for the great feedback.
My carrier board is designed that the TX1 plugs in on top. So the heat spreader is exposed on top. So an extra heat sink could be added. I guess the whole unit could also be mounted upside down, so the heat spreader attaches to the frame.
USB3: thank you for this feedback. I have now reduced the number of HDMI inputs (via CSI-2) from 3 to 2. The third CSI-2 bus I will put on a 22 pin FPC connector with 0.5mm pitch. I will use the pinout of the RPi-CM carrier board. Now there is space for two micro USB 3.0 connectors. Is the micro version ok? Is it vibration resistent enough? Or should I put the signals on a 10 pin PicoBlade header?
I am planning to use the Toshiba TC358743XBG on my carrier board to convert HDMI to CSI-2 (with 4 lanes). Does anybody know more specifics about the status of driver support for this chip by the TX1 (and also TK1)? Some of my customers use the B100 module (with the same chip) and my adapter board to the Jetson TK1. But they have issues with the driver.
B100 HDMI to CSI-2 adapter module
Stacking height: I have not seen a real TX1 yet. The data sheet specifies a Samtec SEAF-50-05-S-08-02-A-K on the TX1. "A" is a variable and specifies the height of this female connector. "A" may be -5.0, -6.0 or -6.5. Does anybody know, which version is placed on the TX1 modules?
Where is pin A1? The data sheet is not 100% clear, where A1 is located. My best guess is on the top left. Can somebody confirm this?
Regards Jurgen