Teradek Bond: Review

Continuing my research into the range of devices I call “cellular multiplexers” or “CellMuxes,” I was delighted to spend some time with Richard Payne of Holdan, the UK distributor of the Teradek Bond, and who I was put in touch with by dealer CVP.

Undeniably, with its modem bays full, the Bond is one of the most eye-catching devices in the market. The Bond looks a little like it could walk to an event on its own (as you will see in the accompanying image, it looks like a spider on its back!) So as far as being a bit of kit that can cause a stir in the thrusting backroom chats of the broadcast professionals, the Bond most certainly has the “cool” factor.

Frankly, I wanted one from the moment I first saw a photo. And having reviewed one (and yet not managed to yet convince the supply chain that it is in thier interests to give me a free one to keep) I am still hankering....

So what’s so good?

Well first and foremost, the cigarette-pack size is striking. The Bond is approximately 1/4 the size of any of its direct competition. It’s important to highlight something here: There are variations on the CellMux theme, but essentially there are three key components I am tracking that offer the user the same workflow. The CellMux workflow in essence allows a live video transport to be established between the camera and the playout center/distribution head-end with the connection being established from the feild on an ad-hoc basis and using immediately available wireless links including Wi-Fi, 3G, and soon 4G. In effect the workflow is the same as a satellite outside broadcast uplink, with a typical presentation being SDI in at the event, and SDI out at the playout.

Teradek Bond

At the moment, the key playing feild that these devices compete in is link aggregation of 3G networks, where a single mobile/cellular data service may not be able to offer or sustain the bandwidths between 1Mbps and 2Mbps required for broadcast (or near-broadcast) quality video. These devices aggregate several such services together and use them in parallel to create a single, higher availabilty and higher capacity connection.

Cellular Multiplexers: A Background

While they are all nuanced at a micro level, the key macro level components in a CellMux are as follows:

Video and Audio Compression/Encoding

The central component involves capturing the incoming video; encoding to a suitable video format for transmission given the bandwidth: adapting to the variability of the bandwidth; forward error correction to manage the unpredictability of the availability of the link; and suitability for decompression at the remote end (playout/headend) for onward use in a broadcast environment. Latency is often a key performance indicator here, where, to deal with the link layer variance, often a higher latency has to be added than in a typical video encoding environment to handle the adverse network conditions.

Link Aggregation

This is the core process of slicing up the video/audio stream in such a way that it can be logically transmitted over several routes. Most engineers are familiar with the notion that, in a packet network, each packet could take a different route from origin to destination, and multiple streams of data can be sent from the origin to different destinations sharing the same origination point using a process called “multiplexing.” In a “link aggregation protocol,” however, the origins may vary for a single given stream of data.

The analogy is that a book is posted to you one page at a time, with some pages being sent from one depot, and others being sent by another depot, but all arriving at your door in a sensible order for you to be able to reconstruct the story in a timely and logical fashion. In a cellular link aggregator each depot would be akin to each network modem. This is a complex process to manage, and as such all the link aggregators typically have a counterpart “demultiplexer” at the receiving end, which combines the data (and requests any lost bits or handles the losses gracefully) before handing the stream to the next stage in the workflow.

Antenna

The final key factor in the system is the radio link itself. Some link aggregation and CellMux devices simply use 3G dongles, and these have their own cellular radio systems. Others have a built-in antenna with specific characteristics, and require only the SIM cards for each cellular service to be added.

Cellular Multiplexer vs. Link Aggregator

So let me now clarify the distinction between my term cellular multiplexer and the underlying term of link aggregator: Essentially a link aggregator manages the radio link and the bonding of these cellular channels together to provide the user with a single internet connection. It goes no further. As far as the operator is concerned, the effect is to have a high bandwidth internet connection. A cellular multiplexer combines video and audio encoding with link aggregation in one device, changing the presentation to the operator from an IP conenction to an AV connection such as SDI.

Unlike link aggregators, CellMuxes tightly combine the video and audio encoding process with feedback on the network conditions so that, for example, if the network availability is reduced the video/audio encoding bitrate can be throttled back. While you can indeed plug a typical video encoder into a link aggregator, if you set a bandwidth of (for example) 2Mbps because this is available to you at the start of the event,you would in all likelyhood lack any ability to automatically drop to 750Kbps when the network availaiblity drops. Certainly you would need somethink between the encoder and the remote link to maintain some intelligent feedback to ensure throttling occured.

A CellMux would offer you this adaptability by design.

Teradek Bond vs. LiveU

So why all this detail in a review of the Teradek Bond? Well it’s important to understand that the Bond itself is strictly a 3G link aggregator only, and yet since it only works with its counterpart, the Teradek Cube, it is assumed you have a Cube in order to get any value from the Bond. As such, the two combined become a CellMux.

The Bond uses external 3G modems, and the effect of plugging these in is what gives the cool “spiders legs” appearance. There are some limitations here, particularly when compared with the competing LiveU product range. If you use such 3G modems in the context of an emergency news feed, you may find that the modem's range limits the reach to the local/nearest phone network cells. That’s not a problem if you are the only operator wanting access to data services on those cells, but in reality at an emergency you are typically contending with many other users of that mobile data service. LiveU has a patented high-gain antenna system which, in this particular situation, sets it head and shoulders above its competition, since the high gain allows their unit to reach far over the local cell masts to those a considerable distance away, which in turn massively increases the network availability - so in mission critical situations this particular feature undoubtedly provides an edge.

However it must also be pointed out that this LiveU edge comes at a significant premium. If mission criticality to this extent is not key to your decision, then it is almost undeniable that the Bond offers much better value for money than any other CellMux system on the market.

For the comparible cost of a LiveU system I could buy 4 or 5 Teradek Bond/Cube systems. LiveU also comes with a monthly service fee and includes the cellular fees, where the Teradek requires you to arrange your own cellular contracts. In fact you could probably buy another Teradek system every two or three months instead of spending on the LiveU service.

That said, it is important to emphasise that no matter how many Teradeks you have, if you can’t get the signal out because the antenna can’t reach the cellular services to link, then it has failed its primary purpose.

This makes the choice between the two very much a business decision, and is excellent for both players in a growing and competitive market.

Putting the Teradek Bond through the Paces

So on to quality. Obviously, once the systems are connected we want to know that they are usable for video. As I have mentioned the Bond is, itself, only a link aggregator, but has some added ability to communicate with its counterpart encoder the Cube, so I want to be careful not to confuse this article by producing a review of what is actually the Cube. That said, they are so closely coupled that it would also be impossible to not comment on quality.

I set Richard Payne of Holdan a pretty tough challenge. In my home, I typically only receive GPRS signal strength. I don’t even manage EDGE, let alone 3G. The local cell tower is only about half a mile away, but it only has one cell on it, and there is a hill just breaking the line of sight. Oh, and my neighbor’s house sits in that line too.

I could see that caused a trace of concern as Richard got set up. I should mention that the setup is amazingly quick: The Bond and Cube screw in a stack onto the top of the camera, and take up no more space than two cigarette packs would. With the modems in and the SDI into the devices, the power taken from the camera, and a small USB cable linking the Bond and Cube, the whole setup took less than three minutes, with the link live in about an additional three mins after Richard set up a Cube receiver on my home broadband and connected its SDI to a monitor. Somewhere in Amazon’s cloud was the Teradek Sputnik software running as the demuxer and presenting the consolidated stream in a way that the receiver could connect and haul it back, over IP to my front room.

The latency was typically 5 seconds, which is great for nearly all applications, though it would complicate two-way interviews over the link in live-to-air bulletin scenarios. This latency is common in CellMuxes as a neccesity of dealing with such volatile network conditions, and while it can be “tuned down”—probably to about two seconds—this tuning will increase dropout, stutter, and error-based artifacting.

What impressed me the most was that, despite the fact the modems were clearly struggling (indicated by the modem LED link quality indicators on the Bond), the picture quality was very good—even though the target bandwidth as set to 2Mbps. My visual take on it is that it the output was comparable to a 750Kbps stream. We pointed the camera at some football shown on my TV, and the codec in the Cube flattened the grass out completely, but cleanly, allowing the bandwidth to be used for the player motion.

At 2Mbps I was very impressed that we got a picture at all (given my local cellualr data issues), but it was noticable that when we stood near or moved the camera at all, or used our own mobiles or Wi-Fi laptops anywhere near the camera, the radio interference gave the Bond a real headache, with the signal breaking up and artifacts rendering the quality unusable. Richard then dropped the target bitrate to 1.5Mbps (note the bitrate in this mode was a target, and the Bond and Cube negotiate the actual bitrate adaptively and continuously). This dramatically improved the stability of the signal, and without any particular drop in the quality of the image. With the caveat that I am someone who is pragmatic about quality vs. smoothness, I was very happy with the overall output.

So what are my conclusions?

Frankly I think the Teradek Bond/Cube CellMux will divide the market with LiveU. For cost-no-object and SLA-is-key applications, LiveU has a safe lead.While the TVUpack and the AVIWEST IBIS DMNG are close competition for LiveU in price, the LiveU patents relating to its high-gain antenna should protect its position well.

However, the much more affordable Teradek Bond /Cube combination, with no fixed on-going commitment, will certainly find its place in many production companies' equipment bags and in nearly all curcumstances will do the job. Its affordability means it can be “in the bag just in case,” which, in my mind, is a perfect place for technology geared to ad-hoc occasional use. The total cost for the Bond, Cube 150 Encoder, Cube 300 Decoder, and MPEG TS license is $6,580.