The Internet Of Football

While football in the United States means something totally different from what it means in the rest of the world, fans everywhere take it pretty seriously. This Sunday is the peak of U.S. football frenzy, the Super Bowl, and it is surprisingly high-tech. The NFL has invested in a lot of technology and today’s football stats are nothing like those of the last century thanks to some very modern devices.

It is kind of interesting since, at the core, the sport doesn’t really need a lot of high tech. A pigskin ball, some handkerchiefs, and a field marked off with some lime and a yardstick will suffice. However, we’ve seen a long arc of technology in scoreboards, cameras — like instant replay — and in the evolution of protective gear. But the last few years have seen the rise of data collection. It’s being driven by RFID tags in the player’s shoulder pads.

These aren’t the RFID chips in your credit card. These are long-range devices and in the right stadium, a computer can track not only the player’s position, but also his speed, acceleration, and a host of other statistics.

A Stripe of a Different Color

The company behind the technology is Zebra, and they make RFID solutions for many different industries. At first, this might seem a little gimmicky. Do you really need to know that the halfback is running at a certain speed? Does the color commentator really need to tell us that the quarterback’s average acceleration is down 30% today for some reason?  Maybe not, but then again baseball fans have obsessed about statistics for years. However, there are other consumers of this data. Coaches, scouts, and trainers can all use the horde of data to assess players and possibly plan improvements.

It is all part of an NFL program known as NFL NextGen Stats. If you want to know which team has the fastest runners, for example, you can find out on that site. We can now reveal that running back Matt Breida was the fastest ball carrier in the 2019 season with a peak speed of 22.3 miles per hour. Who knew statistics were so exciting?

Of course, all this data is subject to number crunching, so the site can tell you how one quarterback performs compared to the average of all quarterbacks and there are a number of fancy colorful graphs. There are even YouTube videos based completely on these harvested stats.

Cracking Open the Hardware

The RFID transponder is about the size of a coin and reports position and accelerometer readings at 25 Hz. Every compatible stadium has 20 receivers that ship the incoming data to Zebra’s San Jose, California control room in a little more than 100 milliseconds. The company claims they report data to broadcasters in under 500 milliseconds. Material on the vendor’s web site indicates the measurements of position are good to within six inches.

We understandably couldn’t find a lot of technical details on the Zebra or NFL websites. But we got lucky when we went to find some images. The photos above are screenshots from a Zebra promotional video which shows tag itself is clear enough that you can read the FCC ID!

Taking Bob Baddeley’s advice about FCC filings we managed to literally get a look inside of the tag itself. Above you can see the block diagram from the filing that indicates a PIC microcontroller and a CR1616 3V battery for power. A charge pump gives it 3.3V and the RF generation is at 6.55 GHz.

The filing even includes internal photos of the PCB. At 6.5 GHz everything is an antenna or a transmission line, so you can see there is a very distinctive PCB feature near U8 and Q1 that form part of the circuitry. If the big coil at L1 looks scary at 6 GHz, don’t be alarmed. That has to be part of the “magnetic interface” from the block diagram. That coil isn’t carrying anything that high in frequency.

For practice sessions, the tag also can send data in real-time via Bluetooth. This joins a host of other bio instrumentation coaches are starting to use during practice sessions like heart rate monitors. The NFL doesn’t allow this during games, though. But a coach might see that a player is, for example, dehydrated by monitoring stats gathered like this.

DIY

We started thinking about how hard this would be to do for your own backyard athletes. Long-range RFID gear is available, although it can be costly and only makes sense when harvesting data from numerous tags. However, a small wearable package that records data for later is workable. We wondered if some of the better fitness trackers could work in a pinch?

You can hack existing trackers or you can opt for open source. There are even some hacks for the cheap knock offs if you don’t want to lose the name brand tracker to a particularly nasty sack.

Living in the Future

Given that many of us remember when radio was a big purchase and a personal computer was unthinkable, it is funny to see radio and computers so commonplace and so tiny. It’s a story that repeats itself. Putting a computer in a toy, or even a car, would have been outlandish science fiction until relatively recently. The shoulder pad sensors are

So if you are watching the big game and you hear some oddball stat, you’ll know where it came from. Or, just hang out on the NFL stats site and find out the ratio of weight to acceleration for each player or something like that.

35 thoughts on “The Internet Of Football

  1. Curious if there are g-force sensors integrated inside? Would probably be a good idea to put them into their helmets. Could get real time and longitudinal data correlations between being continually hit versus specific data regarding concussion and other injury data.

    1. There is an accelerometer in the block diagram so, yes there are g-force sensors in there. You can get really small and lower power six axis accelerometers (X, Y, Z, pitch, roll, and yaw).

    2. Helmet sensors already exist and the data is garbage. Think of it this way – the helmet is designed to reduce the impact to your brain, so measuring the acceleration of the helmet is not going to tell you much unless you know the transfer function between the helmet and brain. Since that’s going to vary wildly based on the type of hit, the helmet data doesn’t really tell you anything.

      I worked at a company that made impact monitoring mouthguards because the teeth are firmly fixed to the skull in which the brain is riding, so the transfer function is a lot simpler. This is generally acknowledged to be the best, most wearable way to get brain impact data. Even still, there’s just not enough data yet to know what types of hits are most likely to cause injury.

    3. I believe they’re used from college level up to pull athletes off the pitch if they may have suffered a concussion level impact (as subsequent impacts after a concussion significantly increase risk).
      Accuracy between helmet and mouthguard-based sensors may vary, but they should be calibrated using sensors in an instrumented dummy, so should be fairly accurate – at least, a lot better than the coach’s visual judgement.

      1. The tough part was during the competitive bidding phase, everyone was developing while the customer was watching over our shoulders, while the game was going on. After that was done, and we secured the contract, we needed to go to games to refine things and be ready in case anything happened. We had all access passes and were able to even go out on the field to find tags that had fallen off (rare).

  2. With enough of this kind of technology you can have the devices calling balls and strikes, safe or out on the basepads in baseball, off-sides, goals, icing, etc. in hockey. Detecting illegal blocks in football or tripping in hockey will be tough but I’m sure they’re on it. Since this stuff is so cheap, everyone will be able to play team sports with quality impartial refs. Imagine an umpire who actually understands the infield fly rule. And everyone gets meticulous stats for everything downloaded to their phones, to boot.

      1. Most routes are determined pre-snap according to the defense. That’s why you get hard counts by the offense, it tricks the defense into revealing what coverage they are in. Most bad passing plays are simply down to the qback and the receiver having different “reads” of the defense, and don’t “agree” on where to be.

  3. A few years back, one of my first big projects was building a time keeping machine for motocross. I didn’t want to track the bikes, but since they are super bulky and riders might even jump through the finish line, so i was wondering how I can detect iff a motorcycle is crossing the finish line and if so, which one. It turned out not to be as easy. Infrared was out, because there is a possibility of two motorcycles crossing the line besides eachother. Then I looked into UHF RFID, which gets used in some professional solutions, but this was too expensive for me at this time.
    My hacky solution was using attiny85’s with nrf24l01 modules which were constantly publishing their ID on the lowest possible transmit power. They don’t have a received signal strength indicator (RSSI), but one register which indicates if the received signal strength is above some threshold (say one-bit-RSSI). With a directional antenna on the receiver this worked quite nice.
    Now why am I writing this? Because this article is a similar use-case and I might take another take on this time keeping again. The problem I had on this hacky-solution was that the battery life was utter garbage. Any ideas how I could possibly solve that problem (relatively cheap)?
    Also: I believe the really good solutions use an inductive circuit across the finish line, but i am not quite sure how to start on that one.

    1. Horizontal IR doesn’t work, but what about vertical? Mount emitters on the bikes pointed up, and a bunch of receivers on a crossbar above the finish line, with shrouds around them so their field of view is limited?

  4. “These aren’t the RFID chips in your credit card. These are long-range devices and in the right stadium, a computer can track not only the player’s position, but also his speed, acceleration, and a host of other statistics.”

    Number and severity of blows to the head?

    1. The NFL knows that the real problem is players leading with the crown of the helmet – allowing them to project the force associated with their body mass through the hit. All you really need is a one-dimensional accelerometer mounted on the crown that causes an LED to flash for 30 seconds when a certain limit is exceeded. It could even be done with little more than an analog output accelerometer, an LM324, a 555, a 2N2222 and a handful of passives. If the LED flashes, the player is ejected. Fixing the cause is better than measuring the symptoms.

      1. Better still, no helmets!

        (Only half kidding. The reduction of an infrequent, obviously painful event into a chronic punishing has changed the nature of the damage, but debatably increased the quantity.)

        Also bare-knuckles boxing. Not so good for the bloody faces / broken fists, but a hell of a lot easier on the grey matter. Pick your poison.

  5. I didn’t follow all the links, so apologies in advance if this question’s been answered. Do the teams get access to these data? I’d think they’d be interested in their team and individual player performances. If they do get them, when do they get them? Is it real-time or near real-time, or post game only?

    1. What’s the reference for those 100ms/500ms latency numbers? Started looking around on the system out of curiosity and haven’t been able to see those figures anywhere. I’m probably missing out on an interesting piece of writing somewhere…

      1. Honestly, I do not recall. However, Zebra’s web site has plenty of info if you look for MotionWorks which is the name of their system. In particular “Dart” is the 6GHZ tags.

    1. And some are readable and some are read/writeable. Still really a battery powered radio transmitter. It’s like calling an ankle bracelet, a numbers station, and a cellphone RFIDs. Still sounds like taking the train to Buzzword Town to me.

      1. It’s active RFID, if it was passive, they’d have to literally pave the football stadium floor with RFID readers (think the things you walk thru going into a music store) – passive RFID has a range measured in inches in most cases, active RFID has a range that can be measured in the hundreds of yards. Common uses of active RFID tags are to locate/inventory large expensive equipment, trucks, cargo containers. Some are as big as hardcover books, with batteries that last months.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.