With the announcement this week that Optus has decided to shutter the operations of Unwired, I thought this would be a fitting time to produce a blog post as a tribute to the service.
For those who don’t know, Unwired was an Internet Service Provider in Sydney and Melbourne which offered wireless internet. Their service was somewhat unique at the time, they pushed a nomadic wireless service which could be used at any place within the coverage area you desire, however they did take your address so that they could use the data to plan network capacity accordingly. Their network was also unique and “cutting edge” at the time, being an IP only WiMAX (or pre-WiMAX) service. They didn’t have the capacity issues that 3G services had, and they didn’t have the outrageous data costs that the other services of the time had, although they were expensive compared to fixed services.
At the time, the service was intensely marketed to those who move house a lot or maybe were renting. We definitely used the service as we were renting, the main convenience being the lack of re-connection fees and the lag time between connection and moving. That was a great convenience. Unfortunately, for Unwired, they never really had many customers at all, and their reputation was relatively poor for numerous reasons.
The company was founded when they happened to win a microwave allocation in the 3.4Ghz band and decided to establish a multimedia service. The ISP was then bought out by Seven Network (yes, the TV network), which had hoped to leverage the network to provide multimedia services to people on the go, but it never really worked out. They did some work to try and reduce the amount of money they were bleeding at the time, but they weren’t really successful against the relentless data discounting that fellow mobile telco companies were doing. Then it was bought out by Optus, probably for the lucrative spectrum licenses they had, which leads us to where we are today – being shuttered and closed down. Interestingly, many of the spectrum licenses were held under shell companies called AKAL and BKAL, to try and hide their existence from competitors. The original term of these licenses was for 10 years, and in other states, these licenses were held by Austar (if memory serves me right).
There were many problems though. The service itself was very lacklustre in speed and reliability. Being microwave in nature, it was touted as being working even with no line of sight, but that was rarely the case. It was affected by weather and other multipath issues. In fact, these modems are HATED by many satellite enthusiasts because they cause severe interference to LNBs operating in the C-band, especially those with extended frequency coverage, and as their signals were much stronger than those from satellites, even spillover was enough to desensitize a satellite TV system into failure. Filters for this interference were extremely expensive and difficult to come by as well. They claimed to use microwave backhaul, which didn’t help latencies. They seemed to have made conscious decisions to limit the amount of base stations – they had substantial care to try and maximize coverage but unfortunately, due to their stinginess, it wasn’t really satisfactory. There were many blackspots and areas where there were competing signals. Ultimately, contention was the main killer, causing massive packet loss and speed losses to the point that VoIP rarely worked adequately and file uploads were almost always doomed. Their promise was only 50% of the advertised speeds for 60% of the time. If anyone was promising that, I can’t say that it’s really a solid deal.
They did offer services with no excess fees, offering throttling when you hit the data limit rather than cutting you off directly. There was also the option to pay for a data allowance top-up to restore full speed, and try to milk the customer for a bit more money. This strategy was one that was later recommended for even 3G internet when it was clear “unlimited” plans were unfeasible. They eventually even introduced a prepaid service to try and get some more occasional customers. Then they did some work in weeding out unprofitable customers (like me) by issuing Acceptable Usage Policy violations for high levels of uploads etc.
So it was a sub-par connection, for a moderate price, that didn’t work at high speeds (They offered 1Mbit/s maximum from memory), with patchy coverage and ultimately, the network collapsed under the poor opinions.
So why does it deserve a blog post? It was special. Yeah, that’s right. The system itself was based on the Navini Ripwave pre-WiMAX system – this itself was a unique selling point. WiMAX was the standard which was being ratified at the time that was supposed to be the next 4G standard offering metropolitan area networks. At the time if it was built, if you were smart, you would probably invest in this thinking that you’d be the next big thing. (In fact, they knew that bet was up and started another company, which offers 4G using the TDMA version of LTE called Vividwireless, which is also being shuttered by Optus.) The standard itself took a long time to be standardized, and Navini was a big player in the negotiations that eventually had a fall-out. Smart people were buying into the Navini system under the expectation that it would become the WiMAX system of the future and it would be directly upgradable to WiMAX compliance. Unfortunately there were issues with standardizing frequency bands for WiMAX use and MIMO training techniques which held up the standard from memory. The interesting thing was that this system involved the use of smart beamforming, and MIMO, which was reflected in the high upfront cost of the modems. This was apparently also marketed as a security thing – i.e. other people won’t hear your data.
This system did have something which other networks didn’t – they specifically made this network “fairer” than the others. The network operated using time-slotted access, and each modem had a fixed number of timeslots to use with the base station. Based on data loading, some random-access slots could be allocated to modems on demand thus alleviating congestion while allowing a minimum QoS to each modem. Further to this, (in a strange way), the modems were also capable of “relinquishing” their allocated slots if they had no data to transmit. Yeah, I was a sucker for believing in this – in fact, this was the cause of many peak-time modem-roams to adjacent “lesser loaded” stations. The closest station to me is ~ 1.8km away in Guildford. Next closest is 3.2km to Condell Park. Unfortunately, even with the Guildford base, it was very marginal. When it connected to Condell Park, it was as good as not having a connection at all. (Unwired had a habit of naming base stations using a numbering system, and people had driven around using their modems to pin-point the modem base stations using the distance readouts, and also using the ACMA database. The base stations themselves look like a grey multiple-slotted toaster (due to the multiple antennas for MIMO), and normally are clustered in threes or sixes.)
The modems themselves were Navini Networks Ripwave MXs (and then later, Cisco branded ones) looked like this. They were often referred to as “rabbits” due to their ‘ear’ antenna, but technically they were known as webbits. This particular version is the Ethernet with LCD edition, there was an earlier variant with three LED indicators for signal quality instead. The power supply was a 6v 3A supply, earlier there were lesser powered adaptors which caused problems with malfunction of some modems and were exchanged. The badge on the side that said “Unwired” unfortunately fell off.
These modems connect via Ethernet, although variants involving USB were available. Unfortunately, no matter which variant you choose, you still need to use the external power adaptor with it. For those who want to be a bit more mobile, external battery packs were offered which clipped into the holes at the bottom, being an integrated package. They weren’t cheap though.
Other than that, you can use the PCMCIA version of the modem, but that had poorer coverage on a network with already poor coverage. And that modem would cost you somewhat more. The network itself was stated as a nomadic wireless, rather than mobile wireless, so you can’t use it while you’re in a moving car or train as the connection would drop out. The beamforming might have something to do with that one. Interestingly, there was no IP portability between base stations which made a dreadful experience. When the modems would roam from station to station, all the IP sessions would break and that would make you curse and swear.
Moving into a new house means moving around the rabbit for optimal reception. In order to make this happen, there is a set of diagnostic utilities called Navini Diagnostics which allows you to see the signal levels and signal quality as well as a front panel LCD which you can use to guide you. Yes, we had to turn these things around, sit them high and low. It was a lesson in frustration.
The positioning of the device was extremely important to getting a good signal as the device has three antennas – a patch panel on each side of the body and an “ear” that was more omnidirectional but had less gain. Should the modem be swapping antennas a lot, the connection would go flakey. So much for antenna diversity. How do we know what the ‘rabbit’ is doing? Well, we’ll get into that later, but first, lets admire the insides.
Prying the modem open was not an easy task – it’s very difficult, and unfortunately, my ‘rabbit’ suffered a broken LCD as a result. But prying it open lets us admire the insides and just how this thing worked.
Cracking it open required a lot of persuasion from a flat head screwdriver. Minor damage to casing. Inside, we find a lot of metal – in fact, it’s a heatsink that encases the PCB from both sides, secured together with screws. Using foam-rubber tape, the patch panel antennas are attached to the sides of the heatsink, and have to be peeled back to get the heatsink off (irreversibly damaging the foam rubber tape itself).
There is an interesting opening at the bottom of the heatsink set. I’ve been told that this is for use for calibration of the modem itself, as frequency drift over temperature can be important to the modem, so they probably test it and produce correction data in the ROM. More on that later. Of interest is that the modem used a 10BaseT connection, only 10 megabits by ethernet, and half duplex too. So it could never compete with modern connection.
So once we unscrew the many screws and peel back the antennas, we are rewarded with the PCB:
This picture shows us the powerhouse the modem is. Three TI DSPs and or OMAP CPUs, it’s more than most appliances have. I guess it’s a sign of the fact that these modems weren’t yet popular enough yet to have fully integrated ASICs and SOCs made, and the use of CPUs and DSPs would have provided flexibility through firmware updates to change modem operating parameters.
Not much on the front half of the top side, there’s a crystal I can recognize but that’s about it. Time to flip it over …
On the back, there’s SMSC MAC for LAN (I think), an SST flash memory chip and the rest of it is likely to be for the RF side of things. The whole combination puts out a LOT of heat, typical operation sees temperatures about 60-70 degrees!
Ultimately, I reassembled the modem and it still worked, although it was out of service as we had terminated our contract with Unwired, having migrated to dial-up and then ADSL2+.
So back to the original question, how do we know what the ‘rabbit’ is thinking? Well, there was a tool called “Navini Diagnostics” which provided much diagnostics data. It is not compatible with Windows 7 (initial install screens on Windows 7, but the program doesn’t work), so I had to dig out an old XP machine to get it working. The communication for the diagnostic data was performed by UDP packets sent periodically from 169.254.254.250.
The install page, I love the slogan.
We also need to run an old version of Java to get it to work properly. Hmm. Anyway, this is what the main screen looks like, some basic signal quality and strength trend data is given, but otherwise, it’s mostly spartan.
Interestingly, these modems also featured over the air updating for firmware updates. Through the lifetime of the modems, I had only seen it update once – from 4.5.0(UD) to 4.5.2(JD). The modems themselves featured two roms, so that should an update fail due to the modem powering off during update, then the modem will not be bricked. One of the things about the update was an apparent increase in signal strength into good-regions despite the signal not actually improving. So while it might show good, the service wasn’t really any better. Way to trick the end user. Worse still, should a firmware update actually occur when someone was using the modem, the modem will reboot itself thus breaking all the connections.
The real power comes with the special key-combination of CRTL+SHIFT+F10, which unlocks the hidden options. Yum!
The signal trend option can be used to graph various options. Nice! You can see that the modem is capable of 25dBm transmit power (316mW) and was topping out from time to time, suggesting that the base station is about as far as you can reliably connect. The signal level is a paltry -90dBm to -100dBm.
One of the impressive things I found was the ability to see the signal constellation, so you can judge what the signal quality is like:
Aside from that, there’s also a long list of information available on the modem itself – signal to noise ratios, BERs, temperatures, reboot reasons, distance to base station. Click on any of the thumbnails to enlarge:
You can also get all the specific data from the modem – my modem’s data dump:
******************************************************** Equipment ID :FFBD63FE Network Id :345 Frequency Band :3.5 Interface Type :LCD Ethernet Active Software Version :4.5.2 built at 08:36 AM on 02/14/2007 Standby Software Version :4.5.2 built at 08:36 AM on 02/14/2007 Hardware Version :91-03632-19-revC ******************************************************** Boot Loader Software Version :4.0 built at 02:45 PM on 05/28/2004 Band Width Limiter :4 Manufacture Date :06172006 Band Begin Frequency :3475000000 Band End Frequency :3600000000 Band Search Step :500000 Serial Number :0004 6abd 63fe Machine Address :0004 6abd 63fe Transmit Flatness :0000 0000 0000 0000 0000 Hardware Feature Mask :5 CPLD Version :05-A Calibration Temp Reference :0x32 Factory Reserved :0017 02f0 f9e7 0304 05ff 04fb 0000 0000 0000 0000 0000 0000 0000 0000 0000 Calibration Data 0046 ffd4 0011 0012 0013 0014 0015 0016 0017 0018 0019 001a 001c 001f 0021 0024 0026 0029 002b 002e 0032 0035 0039 003c 0040 0043 0048 004c 0051 0056 005a 005f 0064 0069 006d 0072 0077 0037 003b 003f 0043 0047 004b 0050 0055 0059 005e 0063 0068 006c 0070 0075 007a 007e 0080 0083 0086 0088 008a 008d 0090 0093 0096 0098 009b 009e 00a1 00a6 00aa 00ae 00b1 00b4 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0001 00f6 00f6 0046 ff93 0011 0016 001b 0020 0025 0029 002b 002d 002f 0031 0034 0036 0038 003a 003c 0040 0043 0046 0049 004c 004e 0051 0054 0057 005a 005e 0060 0062 0064 0066 006b 006e 0071 0074 0077 007b 007e 0081 0084 0087 0088 008a 008c 008e 0090 0091 0094 0097 009a 009d 009f 00a3 00a7 00ab 00af 00b1 00b3 00b5 00b7 00b9 00bc 00be 00c0 00c2 00c4 00c8 00ca 00cc 00ce 00d0 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 Number of History Frequency :4 History Frequency List :3170 2975 3375 3175 Number of Colocated Frequency :1 Colocated Frequency List :3175 Number of Neighbor Frequency :0 Factory data version :2 Tx flatness step size :8 Tx flatness :2053 770 256 0 1 1030 2057 2570 Rx flatness step size :8 Rx flatness :513 0 -1 0 0 257 514 514 Tx temperature compensation step size :10 Tx temperature compensation :25 -258 2 1032 3341 3341 3341 3341 -258 2 Rx temperature compensation step size :10 Tx temperature compensation :0 0 0 0 0 0 0 0 0 0 Reserved :4d 4d 0 0 ********************************************************
You can also do a location-test log which logs all the signal strength data for later analysis, and if you attach a GPS, you can also perform a drive-test log which has the details of position and signal strengths as well. A very powerful and necessary tool to get the most from your connection.
So there we are, pretty much the ins and outs of Unwired and their modems. Both literally and figuratively. As you can see, I had high hopes for it back then as the technology was wonderful. Now this will just remain another one of those technologies which has since expired and become irrelevant. It wasn’t the only one though – I will cover another similar technology in one of my future posts.
[Aside: Hi Jeff!]