POE VR Basestations

L

LuxTerra

Limp Gawd
Joined
Jul 3, 2017
Messages
220
I haven't jumped into the VR pool yet, but I was thinking about wiring up some category cable/jacks for the future. For now, none of the basestations are POE enabled, but I'd hope we get there. In the mean time, has anyone attempted this setup with something like this: Tycon Systems POE-SPLT-4812G?

http://tycononline.com/POE-SPLT-4812G-Active-Splitter-8023afat-POE-to-12VDC-25W-Output_p_68.html

Of course it's a bit pricy for what it is, but the idea of clean install has its appeal. Plus, then it would be powered from my network closet switch/UPS.

Thoughts?
 
If you're talking about the tracking sensors, none of the commercially available ones use ethernet at all. The Vive's only require power, and the Oculus Rift's sensors are powered via USB, and require active USB extension cables.

Furthermore, the bandwidth required by the Rift's sensors is greater than that provided by 1Gbit ethernet, meaning they'd have to either include a full set of 10Gbit ethernet gear (cables, switch and a PCI-E NIC for the host PC), or make the user buy all that separately, which would be a non-starter for most consumers.

I really like the idea of POE for the sensors, though. It'd be great to see that in a future version once the cost of 10Gbit comes down, or they come up with a lower bandwidth tracking scheme.
 
RazorWind said:
If you're talking about the tracking sensors, none of the commercially available ones use ethernet at all. The Vive's only require power, and the Oculus Rift's sensors are powered via USB, and require active USB extension cables.

Furthermore, the bandwidth required by the Rift's sensors is greater than that provided by 1Gbit ethernet, meaning they'd have to either include a full set of 10Gbit ethernet gear (cables, switch and a PCI-E NIC for the host PC), or make the user buy all that separately, which would be a non-starter for most consumers.

I really like the idea of POE for the sensors, though. It'd be great to see that in a future version once the cost of 10Gbit comes down, or they come up with a lower bandwidth tracking scheme.
Click to expand...

I was talking about the tracking sensors. Even if it was power only for the wireless options like the Vive. I'm highly skeptical of the 1Gbps bandwidth required per tracking sensor. That seems absurd and begs the question how Vive can do it on *significantly* less bandwidth (i.e. wireless). Forget what the WiFi wireless channel is marketed as transmitting, divide by 10 and you're in the ballpark of what really happens (shared half-duplex TDMA channel, etc.). And a 10x divisor is if you have half way decent equipment...I'm running 3x Ubiquiti AC-HDs.

Where did the 1Gbps number for Rift come from? Are you sure they're not just using USB 3.0 because USB 2 is must slower and USB is a relatively terrible protocol for what they're doing, but hey, everyone has the ports?

All that aside, my cabling I have to run is Commscope GigaSpeed X10D Cat6A...so ya, it would handle even absurd tracker bandwidths requirements. Over the short channel I'd have, 40Gbps is within reason if you have the switch support it. Anyways, POE isn't compatible with 10GBase-T. However, I do agree if it's really above 1Gbps per sensor that kills it for most users.

Appreciate the discussion.
 
You can basically sum up the difference between the Rift and Vive as mainly, the Vive uses IR sensors on the headset and controllers to sense the IR lights the lighthouses are sending out, while the Rift the headset and sensors are transmitting an IR signal pattern that is being read by 1080p USB webcams to calculate positional info. Assuming something like h264 or equivalent as the video compression, that is 5-8Mbps @ 1080p30 or 10-16Mbps for a 1080p60. It's possible that the reason they say a 1Gbps is they are not using video compression. If my calculations are right, the rift camera are 8bit gray scale, so a 1080p60 stream is ~1Gbps bandwidth.
 
LurkerLito said:
You can basically sum up the difference between the Rift and Vive as mainly, the Vive uses IR sensors on the headset and controllers to sense the IR lights the lighthouses are sending out, while the Rift the headset and sensors are transmitting an IR signal pattern that is being read by 1080p USB webcams to calculate positional info. Assuming something like h264 or equivalent as the video compression, that is 5-8Mbps @ 1080p30 or 10-16Mbps for a 1080p60. It's possible that the reason they say a 1Gbps is they are not using video compression. If my calculations are right, the rift camera are 8bit gray scale, so a 1080p60 stream is ~1Gbps bandwidth.
Click to expand...

Thank you for the clarification. Uncompressed video would in fact do it. I was curious so I did a bit more Googling and found Oculus's USB recommendations. The bandwidth isn't nearly that high, assuming I'm reading it correctly.

https://www.oculus.com/blog/oculus-roomscale-balancing-bandwidth-on-usb/

Oculus recommends no more than two USB devices per controller regardless of 3.0/2.0 spec. Their reasoning is that practically speaking, it's a shared channel arrangement (like the WiFi example I gave earlier). Also, the fact that 2.0 can support two devices and so can 3.0, suggests that it's not a bandwidth issues, but rather a protocol issue. USB isn't a great choice for this application, but everyone has them.

Oculus also claims that you'll be using 80% of USB 2.0 bandwidth with two sensors. They also correctly state that the practical bandwidth of USB 2.0 is closer to 300Mbps, not the theoretical peak of 480Mbps. 80% of 300Mbps is 240Mbps, but that's for two sensors. Thus, a rough estimate of the required bandwidth per sensor is 120Mbps and well below GbE (also, remember full-duplex). USB 2.0 is half-duplex, but we can safely assume that the data is mostly flowing in one direction. (https://en.wikipedia.org/wiki/USB)

That makes a ceiling mounted jack for POE even more appealing. Mount the sensor, plug one short cable in, and go play. :)

Edit: I bet it wouldn't take much given the relative protocol efficiencies and a bit of optimizing and you could get it to work on 100Mbps Ethernet. You'd still need GbE to the host PC in order to aggregate multiple sensors, but it's not far off.
 
LuxTerra said:
I was talking about the tracking sensors. Even if it was power only for the wireless options like the Vive. I'm highly skeptical of the 1Gbps bandwidth required per tracking sensor. That seems absurd and begs the question how Vive can do it on *significantly* less bandwidth (i.e. wireless). Forget what the WiFi wireless channel is marketed as transmitting, divide by 10 and you're in the ballpark of what really happens (shared half-duplex TDMA channel, etc.). And a 10x divisor is if you have half way decent equipment...I'm running 3x Ubiquiti AC-HDs.

Where did the 1Gbps number for Rift come from? Are you sure they're not just using USB 3.0 because USB 2 is must slower and USB is a relatively terrible protocol for what they're doing, but hey, everyone has the ports?

All that aside, my cabling I have to run is Commscope GigaSpeed X10D Cat6A...so ya, it would handle even absurd tracker bandwidths requirements. Over the short channel I'd have, 40Gbps is within reason if you have the switch support it. Anyways, POE isn't compatible with 10GBase-T. However, I do agree if it's really above 1Gbps per sensor that kills it for most users.

Appreciate the discussion.
Click to expand...

1Gbps is the maximum theoretical bandwidth of consumer grade wired ethernet. This being a consumer product they're hoping to sell to the masses, it's not really reasonable to expect their customers to have or be willing to buy anything more skookum than that.

LuxTerra said:
Thank you for the clarification. Uncompressed video would in fact do it. I was curious so I did a bit more Googling and found Oculus's USB recommendations. The bandwidth isn't nearly that high, assuming I'm reading it correctly.

https://www.oculus.com/blog/oculus-roomscale-balancing-bandwidth-on-usb/

Oculus recommends no more than two USB devices per controller regardless of 3.0/2.0 spec. Their reasoning is that practically speaking, it's a shared channel arrangement (like the WiFi example I gave earlier). Also, the fact that 2.0 can support two devices and so can 3.0, suggests that it's not a bandwidth issues, but rather a protocol issue. USB isn't a great choice for this application, but everyone has them.

Oculus also claims that you'll be using 80% of USB 2.0 bandwidth with two sensors. They also correctly state that the practical bandwidth of USB 2.0 is closer to 300Mbps, not the theoretical peak of 480Mbps. 80% of 300Mbps is 240Mbps, but that's for two sensors. Thus, a rough estimate of the required bandwidth per sensor is 120Mbps and well below GbE (also, remember full-duplex). USB 2.0 is half-duplex, but we can safely assume that the data is mostly flowing in one direction. (https://en.wikipedia.org/wiki/USB)

That makes a ceiling mounted jack for POE even more appealing. Mount the sensor, plug one short cable in, and go play. :)

Edit: I bet it wouldn't take much given the relative protocol efficiencies and a bit of optimizing and you could get it to work on 100Mbps Ethernet. You'd still need GbE to the host PC in order to aggregate multiple sensors, but it's not far off.
Click to expand...

I don't think you're right, at least not totally. At lease some of the sensors, it appears, need to have a high bandwidth connection to the host PC, hence the recommendation that at least two be connected via usb 3. You may be right that part of issue is the way the protocol works in that USB just doesn't share the channel efficiently enough, but raw bandwidth does appear to be a factor, and my own experience with my Rift would seem to confirm this. It can track headset OK if you plug all the sensors in via USB2, but once you start wanting to use the handy-twiddlers, it gets real flaky unless you plug at least one of the sensors into a USB 3 port.

I would argue that, if POE is something future system designs want, the best way to go about it might be to do some of the interpretation of the images on the sensors themselves, and then only transmit vector data for the position of the headset's IR emitters back to the PC for use in the tracking process. Obviously, this would make the sensors more expensive, but that might be OK if it meant that less expensive host computers could keep up. Better still might be to just adopt the Vive approach, which seems to get around this problem entirely.

On the subject of POE specifically, I can see that being a really attractive option for more permanent installations, and maybe there could be a high end version of the systems that uses it, but I don't see making that a requirement for a consumer product any time soon, since there are so many extra parts that are needed for it. You'd need a switch. You'd need cables. The host PC may need to have an extra NIC in some situations. On the other hand, everyone has USB, and that works OK as-is.
 
Are you going to put in some of the USB over Cat6 extenders as well? The high speed ones are still crazy expensive.

If I were doing construction, I'd probably just make sure I had pull tape to all my boxes and then once my plans get closer to fruition pull the cable and place the necessary equipment. Right now you'd spend money on something that doesn't really work with either HMD... You'd be better served with active USB cables than any POE network cables; and that's only if you end up with a Rift. (Which IMO is crushing it on the available software end of things...)
 
Just an FYI, I have both my Rift sensors hooked into USB 2.0 and they work great. No problems.
 
RazorWind said:
1Gbps is the maximum theoretical bandwidth of consumer grade wired ethernet. This being a consumer product they're hoping to sell to the masses, it's not really reasonable to expect their customers to have or be willing to buy anything more skookum than that.
Click to expand...
Already agreed too.



RazorWind said:
I don't think you're right, at least not totally. At lease some of the sensors, it appears, need to have a high bandwidth connection to the host PC, hence the recommendation that at least two be connected via usb 3. You may be right that part of issue is the way the protocol works in that USB just doesn't share the channel efficiently enough, but raw bandwidth does appear to be a factor, and my own experience with my Rift would seem to confirm this. It can track headset OK if you plug all the sensors in via USB2, but once you start wanting to use the handy-twiddlers, it gets real flaky unless you plug at least one of the sensors into a USB 3 port.
Click to expand...

It seems that sensors plug into the USB 2 ports do run in reduced bandwidth model. What's interesting about it is that the measured reduced bandwidth does NOT match Oculus's claims. Actual measurements are 500Mbps on USB 3 and 50Mbps on USB 2. There might be some oddities with how it's measured.


Here as well: https://imgur.com/a/I3JLI


RazorWind said:
I would argue that, if POE is something future system designs want, the best way to go about it might be to do some of the interpretation of the images on the sensors themselves, and then only transmit vector data for the position of the headset's IR emitters back to the PC for use in the tracking process. Obviously, this would make the sensors more expensive, but that might be OK if it meant that less expensive host computers could keep up. Better still might be to just adopt the Vive approach, which seems to get around this problem entirely.
Click to expand...
Agree, particularly about the Vive comment. The Rift seems like a brute force solution and that never scales well. Perhaps, Rift can't wait for the TB3 version. With IP POE security cams starting to do motion detection and other process it may happen.

RazorWind said:
On the subject of POE specifically, I can see that being a really attractive option for more permanent installations, and maybe there could be a high end version of the systems that uses it, but I don't see making that a requirement for a consumer product any time soon, since there are so many extra parts that are needed for it. You'd need a switch. You'd need cables. The host PC may need to have an extra NIC in some situations. On the other hand, everyone has USB, and that works OK as-is.
Click to expand...
Ya, LCD. You could do it today with the Vive and a POE to 12V converter like I listed in the OP. However, that's only slightly cleaner than running the power bricks around to wall outlets.
 
RazorWind said:
...Furthermore, the bandwidth required by the Rift's sensors is greater than that provided by 1Gbit ethernet, meaning they'd have to either include a full set of 10Gbit ethernet gear (cables, switch and a PCI-E NIC for the host PC), or make the user buy all that separately, which would be a non-starter for most consumers...
Click to expand...
Incorrect.

USB 3 spec is 60MB/s, 1Gbe is 125MB/s. These are the spec'ed theoretical limits but I commonly transfer files at about 110MB on Ethernet. I could see a power injection over Ethernet coupled with USB to Ethernet potentially working....but it'd be a hell of a lot more expensive than using powered USB extensions.
 
USMCGrunt said:
Incorrect.

USB 3 spec is 60MB/s, 1Gbe is 125MB/s. These are the spec'ed theoretical limits but I commonly transfer files at about 110MB on Ethernet. I could see a power injection over Ethernet coupled with USB to Ethernet potentially working....but it'd be a hell of a lot more expensive than using powered USB extensions.
Click to expand...
Um, no. USB3 's theoretical throughput is 5Gbps, or 640MBps. USB2's theoretical throughput is 480Mbps, or ~60MBps.

https://en.wikipedia.org/wiki/USB_3.0


https://en.wikipedia.org/wiki/USB#2.0HS

The wikipedia articles cite their sources for this, one of which is the original USB spec document.
 
LuxTerra said:
I haven't jumped into the VR pool yet, but I was thinking about wiring up some category cable/jacks for the future. For now, none of the basestations are POE enabled, but I'd hope we get there. In the mean time, has anyone attempted this setup with something like this: Tycon Systems POE-SPLT-4812G?

http://tycononline.com/POE-SPLT-4812G-Active-Splitter-8023afat-POE-to-12VDC-25W-Output_p_68.html

Of course it's a bit pricy for what it is, but the idea of clean install has its appeal. Plus, then it would be powered from my network closet switch/UPS.

Thoughts?
Click to expand...

Total waste of time for VR. Rift is the last successful consumer headset whose trackers need to be connected to the PC. People do not want to string tracker wires back to the PC.

Now that tracker-less VR is a thing (Windows VR) that's what most people will buy. Enthusiasts will be ok with Vive style trackers but those only need power. Even that's a pain in the ass so I'm guessing we'll see rechargeable lighthouses in the not too distant future too.
 
BecauseScience said:
Total waste of time for VR. Rift is the last successful consumer headset whose trackers need to be connected to the PC. People do not want to string tracker wires back to the PC.

Now that tracker-less VR is a thing (Windows VR) that's what most people will buy. Enthusiasts will be ok with Vive style trackers but those only need power. Even that's a pain in the ass so I'm guessing we'll see rechargeable lighthouses in the not too distant future too.
Click to expand...
Rechargeable lighthouses is an interesting idea. Although, constantly putting them up and down could get old.
 
BecauseScience said:
Total waste of time for VR. Rift is the last successful consumer headset whose trackers need to be connected to the PC. People do not want to string tracker wires back to the PC.

Now that tracker-less VR is a thing (Windows VR) that's what most people will buy. Enthusiasts will be ok with Vive style trackers but those only need power. Even that's a pain in the ass so I'm guessing we'll see rechargeable lighthouses in the not too distant future too.
Click to expand...
It depends on how much information is going to be transferred. If each eye was 4K 90fps that's a lot to transmit?
 
You must log in or register to reply here.
Back
Top