Behind the scenes: live-streaming the Shimano SuperCrit with on-board cameras
Last Sunday, CyclingTips live-streamed the elite men’s race at the St. Kilda Cycling Club Shimano SuperCrit including live footage from on-board cameras. As far as we’re aware, it’s the first time on-board vision has been used for a live race broadcast anywhere in the world.
It wasn’t perfect by any stretch, and we weren’t expecting it to be. Rather it was a chance to test the technology and show that it is possible. After all, you have to start somewhere.
We asked one of the brains behind the operation, Tom Reynolds from CharterMason Giant Racing, to give us the low-down on the system they used, what worked, what didn’t and what the future might hold.
Cameras on bikes. It’s one of those inexplicably complex and tedious issues that cycling must deal with. We’ve all seen the first on-bike videos released as post production efforts, but I wondered: “Where are the live cameras?”
And while I’m sure that the-powers-that-be have more than just Astana, shortened Grand Tours and sock height to concern themselves with, we thought the best way forward on this issue was just to forge ahead.
Co-conspirator on this bikecam project Andy Richardson and I cut our teeth in motorsport. Andy was chief engine engineer for a V8 Supercar team and then moved into production vehicles. Today he is the chief engine engineer on the most powerful Australian car ever produced. He’s good with his hands, can solder, design and print 3D parts, make carbon fibre bits ‘n’ bobs, program devices and troubleshoot in a pressure environment like no one I know.
While he’s been making cars go fast, I’ve been a team media manager or series media manager in Superbikes, V8 Supercars and Australian Rally. Each of these series rely on on-board cameras to tell the story to a massive degree. So when I became involved in the National Road Series in 2012 I was aghast that even a simple on-board camera was banned at all levels.
Having decided that an on-bike camera streaming live was a good idea in theory, we realised early on that hacking existing technology would be the go. As avid drone enthusiasts — I made last year’s SuperCrit video, while Andy recently built a six-rotor drone on his day off — we wondered whether it would be possible to hack existing drone tech into a workable solution.
Of course it turned out to be far more complex and difficult than originally thought. A drone-based two axis gimbal was overkill (and too heavy!) as bikes don’t pitch front to rear — they just go left to right, whether tipping into a corner or when the rider is standing to sprint.
Andy not only hacked a standard gimbal down to a single axis, he then rewrote the software to make it work. The initial results were super-promising. The biggest issue we faced (and still face) is that the camera is effectively on the end of a tiny springboard and mild vibrations are amplified. As usual we “have a theory” on that and will be testing an improved setup shortly.
The setup we ended up with was:
Mobius Camera (Beta version!) now to a GoPro4 -> Analogue out -> Transmitter -> over-the-air back to base station -> analogue out into TV capture USB device -> UStream streaming software -> **HOPE**.
After some testing at the St. Kilda Cycling Club White Street circuit (testing transmitter range) and out at Sandown (thanks to Carnegie-Caulfield Cycling Club), it was apparent that the tiniest glitch could result in the system entirely falling over.
There were too many links in the (above illustrated) chain. The camera inexplicably stopping transmission, camera cabling breaking, gimbal failing (it came loose during one test!), transmitter and battery failing (we had to go bigger for the gimbal and it still fell over after one hour) …
With “disastrous embarrassment” as a motivator, we decided three days before the SuperCrit to build a second/fall-back, basic-as-can-be rig with a GoPro4, no gimbal and a transmitter (see the video at the end of the post for a video comparison of the two setups).
In our discussions with CT, I was super pessimistic as there were just so many things that had gone wrong and our experience in motorsport told us that it’s a pretty risky trying to test while racing. SuperCrit was upon us and while we’d done a fair bit of testing, we also knew that some things just had to go wrong.
Murphy’s Law being the bastard that it is, the things that did go wrong were things we had not even considered or had never failed previously. First the video capture device decided that it was not going to accept an audio input. Then the sun came out and overheated the laptop, causing it to hang and then crash.
Umbrellas were provided, the computer restarted and then the audio worked. Our video switcher also failed. We ended up manually switched plugs – hence the lag in shots. This was now so primitive that perhaps watching it on a 4:3 CRT TV was the go…?!
The things that went right were pleasing: the gimbal rig worked for 70 minutes. This was not the full race distance, but we can improve on that and it was around what we expected. The gimbal-less GoPro4 and transmitter worked right to the end. We think there may be 100 minutes of transmission possible from that rig; a terrific outcome.
The Telstra 4G we feared would slow to a crawl was utterly brilliant and we could have upped the quality were we not worrying about everything else. Our pre-race speedtest indicated 28mb/sec upload. While it was great at White Street, South Melbourne, a change of location will inevitably bring challenges.
After the SuperCrit — and over a beer — we started planning version 2.0. The GoPro 4 footage is just outstanding, so while we could go to a tiny 3CCD system, the weight saved will be offset by a loss of quality and no on-board recording facility.
For transmission we have two options, and these are budget dependent. A full MotoGP-spec system is possible, but at a cost of at least $10k. There best be a compelling business case or an angel investor if we’re going to get that up and running!
For domestic Australian cycling (especially crit racing) we already have a new system in the pipeline with multiple digital receivers and gear which auto-switches between signals as the bikes come into and out of range. This could offer much improved full-lap coverage.
We could do this using our new, more stable gimbal system which now has super-stiff custom-made carbon brackets and is 30% smaller than the SuperCrit prototype. This system would be amazing at Bay Crits, the Nationals and even at club racing. We could live stream a feed to a big screen on site or once again to the internet.
We’re even keen on building a seat-based camera with remote pan. This is yet another hack from drone technology. Hell, we could even build a camera that you could view via goggles and then pan left or right by turning your head!
So what about WorldTour or other professional stage racing? As it stands, we believe it would be possible to build a live on-board rig … with some pretty big ifs. We see it like this:
1. We’d need a system not too far different to what we currently have, but with a bespoke smaller camera system with on-board recording and a slightly better transmitter system. Costly, but it would work.
2. Footage from the cameras would be beamed to a helicopter above or perhaps to a motorcycle and then on to the chopper.
3. Due to battery issues, a software system to “wake” the camera up and then send it to sleep is important. You could perhaps get 90 minutes of live broadcast at best with existing smaller batteries. “Wake-up” software would mean you could “check in” with a rider and then sleep the system so you are good to go for the finish.
4. The bike build might have to be unique to reduce ballast, run custom wiring and aerial locations. Weight would be down to 500gms in total.
5. On mountain stages it might even be possible to get the rider to change bikes. It was done in the Tour de France in the past.
As we see it you can either wait for the future to come to you, or you can just get out there with a soldering iron, some heat shrink and a hobby website and get amongst it. Stand by.