We have bought an m5 compute stick to see if we can use it as the engine in an on-set device for tv- and filmproductions.
We're running CentOS, and in the software part of the process, everything was working just fine.
The setup is an DekTec USB3 capture card which allocates bandwith enough for 10-bit 444 1080p RGB video + metadata capture. This unit was attached either directly to the USB3 port on the stick, or to the HUB on the power adapter.
Additionlly we have aZTE LTE modem. This was mostly mounted on the hub.
In this stage, soft-and hardware seemed to work just fine.
Then we need to make this an on-camera/on-set tool...
First attempt was to put a passive USB-c to USB-3 HUB on the USB-C port.
Attach the LTE modem to the passive port and keep the capture card in the USB 3.
The Intel power-adaptor did not manage to let the compu stick even start with this configuration.
Which tells me that the power on USB might be a tiny bit weak.
We then replaced the Intel 2,2 A adapter with a 3A USB-C wall adapter.
In this configuration, the LTE modem and the stick starts, but not the capture device., and with a wide variety of active USB3 hubs, the stick does not work when sitting on a HUB.
In this process we have seen the USB ports being unavailable, being clocked down to USB2, turning itself off and on etc etc etc.
We have confirmed this behaviour with ither USB devices than those we need to use, including mouse and keyboard.
From what I see and read, this sounds like the behaviour of a faulty stickOur plan now is to power each of the USB devices from an external source, while keeping the data connections directly to the stick.
But I am more than curious about people having made a reliable setup with the computer stick + external USB hardware, after seeing this and reading the posts I have read around the subject.
The stick is such a perfect platform (seemingly) for a wide range of embedded devices, but that is only true, if it is a reliable computing platform in the first place, and the USB behaviour we see on the stick, seems to be able to very random.
Any ideas or additional documentation appreciated.
Gunleik_Groven, Thank you very much for joining the Intel® Compute Stick communities.
In order to try to find if there is any documentation about the USB ports on the stick, I will do further research on this matter and as soon as I get any updates I will post all the information on this thread.
In the mean time, on the link below you will find the TPS (Technical Product Specification) for the stick, in there you might find additional details about the USB functionality:
Any further questions, please let me know.
Yes, we read that before even buying it.
There is one particular interesting behavior. We have tried a number of powered external USB3 hubs, including hubs that are stated to be isochronous.
While those seem to work on other computers, they don't seem to work for USB3 on the stick. The exception being the hub built into the power. Which kind of tells us that this unit has a few tricks not too common in USB hubs. Is there some "smart proprietary" electronics in that hub?
Also the stick seems to be very fragile for changes in the effect drawn. The LTE modem has spikes (within USB3 specs), and that seems to be enough to make the compute-stick go into re-boot or shutting off other USB ports.
As nothing gives us the result we need from the capture-card when mounted on a hub, we will power the USB devices directly from our own power distributor for the next (and possibly last) set of experiments, and we have also ordered a new stick to verify that this is not due to some malfunction on the stick we have been developing on.
We could imagine to get a larger number of these, if we can make them run stably over some time It is kind of "almost perfect" as a platform.
The only thing we miss is a core m7 version. And of course, stability.
Thanks for looking into this.
There isn't really anything proprietary or any kind of special design around the USB port, we try to test and validate most common devices but we cannot validate every single device out on the market but if there is business justification I can submit a request for testing the DekTec* USB3 (is it this one DTU-351? and the ZTE LTE modem, I would need a direct URL to look at specifications for this one). I cannot promise anything at this point but if you send me a better description of the project: company name, location, amount of units that you are planning to purchase, system details and timeline; I will be more than glad to submit that internally and evaluate the possibility of testing these devices.
On the other hand, I checked at http://www.dektec.com/Products/Usb/DTU-351/ http://www.dektec.com/Products/Usb/DTU-351/ (is this the capture card that you are testing?) if yes, I see that this unit is bus-powered, so no power supply is required. Could this be the problem?
Yes, that's the capture card and it is bus-powered, but we plan to test to feed it from a different source than the Stick (through the USB cable for now, but from an external source, as it doesn't work to put it on a HUB)
The ZTE USB2 modem is also bus-powered. This works on a HUB, but we woulld need to put that HUB on the USB-C side and unless THAT hub is powered (which most are not) separately from the distribution on the power from "your" adapter, it still doesn't work."
We have gauged the LTE modem and see that it "spikesin power draw depending on activity, and that those spikes can shut down the stick.
As to specifics on plans and volume, I would prefer to do that on an offlist mail correspondence.
gunleik (at) quine (dot) no
We are involved in this, though:
And this is related.
I have been checking on this regards and there shouldn't be anything special about the USB port on the Compute Stick but still this device is supposed to be very low power this may affect the overall performance when USB connected to devices that need additional power.
Wouldn't be a good idea to test a more robust solution for your needs? Perhaps an Intel NUC?
We used NUC (previous gen with mini PCIe ) as a prototype for both this and a different more complex appliance, but it is too big, simply for the uses we intend for the stick.
The compute stick has the right form and enough computational powers for a "mini-box".
Though an m7 version would have been better
For the more complex use, the fact that all ports are built in on the NUC, while we need all contacts to be custom and robust in an on-set environment, adds even more to the size on the NUC.
Also the NUC doesn't offer us the IO flexibility we need for the big one. (more USB channels (both 2 and 3), multiple serial ports, GPIO with interrupt.
We miss the NUC, but it didn't work. It was too "in the middle" and when the mini-PCIe was killed off for M2 and that wasn't compensated with more USB3 and 2 channels, it became unusable for us. Unfortunately.
M2 doesn't fully replace mini PCIe... yet
Just a quick update on the USB behaviour.
We are now on Windows 10, but we still see inconsistencies in USB3 connectivity.
It seems to set all devices to USB2 randomly, like it did under Linux.
Robust USB3 behaviour is critical to make the capture-card work at all.
We are on a fresh stick with a fresh windows install