Product survey: Checking and verifying Bluetooth traffic

The final extent of Bluetooth-enabled products is something that at present we can only imagine. But for certain, we’ll see lots of existing products — such as PCs, printers, and portable office equipment that are presently anchored by connecting cables — given welcome release. Equally important will be new lines of communication with PDAs, digital cameras, and portable phones. Think about domestic equipment and someone, somewhere, will be puzzling out how your TV, fridge, and hi-fi can also benefit from Bluetooth. Further still, Bluetooth is sure to attract educational, entertainment, and toy businesses. Will we see Bluetooth-enabled fluffy toys? You can put money on it.

With this diversity of product development, it’s a great relief that the whole Bluetooth business is very tightly regulated. This regulation constrains the way not only how you’ll design, but also test, your Bluetooth-enabled product.

A Bluetooth wireless interface subdivides into three layers of design: radio layer, baseband protocol layer, and the application layer. You’ll buy an approved Bluetooth radio chipset or module to manage standard RF and baseband signal handling, so most of your work will be designing and testing the application-specific features of the wireless interface. Again, regulation helps. The Bluetooth SIG classifies all products and this, in turn, describes what implementation, or product profile, from the overall Bluetooth specification relates to your product. A LAN access profile, for example, describes services and connection states for an application layer that includes: initialisation of LAN access service; shutdown of LAN access service; establish LAN connection; lost LAN connection; disconnect LAN connection.

As regards testing, certainly you’ll need something to check out the firmware that implements your product’s Bluetooth profiles and that means some level of protocol analysis. Our survey (below) covers several dedicated Bluetooth protocol analysers with broadly similar performance. The Bluetooth community calls these test units “sniffers” because they can passively sense and translate data via RF from any Bluetooth product in range. In this mode, and linked to your PC via USB, they capture, decode, and display data packets, commands, and events at baseband and higher protocol levels.

Some sniffers also have a piconet mode where they can actively participate as a master or slave. In a piconet, Tektronix’s BPA100 can initiate limited modes of operation including sending your product intentional error signals to partially stress test it.

CATC’s Merlin protocol analyser remains a sniffer-only product but, now, includes Merlin’s Wand hardware to add piconet capability. Unlike Tek’s self-contained BPA100, Merlin’s Wand is an additional piece of hardware with its own USB link. It’s significant though, that only the Wand hardware joins a piconet. The Merlin box itself therefore still sniffs all piconet traffic, even that from the Wand. If you want to expand a piconet, you can add additional Wands (1k Euros). A recent firmware upgrade to version 1.6 gives Merlin profile filtering.

If your application requires a top-end protocol tester, then Rohde & Schwarz’s PTW60 is becoming the industry standard for full protocol and profile compliance testing. This unit may be an overkill and too pricey for some developers. Basic cost is around 60k Euros, although this can rocket to over 200k Euros depending upon how many software library test cases you purchase. As a piconet master, the unit can initiate all prescribed Bluetooth test cases and even allows you to develop your own test scenarios — both features well beyond the scope of sniffers.

You can study a more detailed comparison of all Bluetooth protocol analysers at www.palowireless.com/bluearticles/bluetoothanalyzercompare1.asp.

Circumventing RF
In successfully sniffing around your product, you can argue that protocol analysers also conveniently confirm the RF and baseband layer operation of your design. They do but protocol analysers give you no quantitative results about these layers and only tell you that the radio works.

You might well believe, in principle, that you don’t need to measure the radio Bluetooth interface in the same way you don’t expect to measure bi-directional driver capability when you design-in a GPIB chip, for example. Your argument might hold if it weren’t for more Bluetooth regulation, namely, compliance testing.

All products, before you can market them as “Bluetooth-enabled”, must pass a formal bluetooth compliance test carried out by an approved independent test house (see the Box “Conquering compliance” below). As well as exhaustively checking protocols, compliance testing also looks closely at the RF and baseband parameters of a wireless interface (Ref 2). Just as with the EMC directive, CE-marking and so on, you should approach these tests as unavoidable, rather expensive formalities that your product should pass the first time. To build that assurance, you’ll want to carry out at least some level of in-house pre-compliance testing. This, in turn, means you’ll need a degree of RF measurement capability and expertise. Our survey on page 17 also includes dedicated Bluetooth test sets that, although primarily intended for production test, can also help you to do this work.

RF is the single big-unknown as far as many Bluetooth designers and test engineers are concerned. Most of the work in designing and testing a wireless interface into a host product concerns software. If you’re a software specialist, and most Bluetooth designers and testers will be, then feel at least a small level of foreboding. What you’re designing with is a 2.4-GHz, frequency-hopping, 720-kbps, 10-metre range radio. The 2.4 GHz gives a 125-mm/5-inch wavelength. So, a turn of your head can change the whole standing wave pattern in the multiplicity of transmissions, Bluetooth and alien, in which your product must reliably operate. Such a movement may shift the precise location of your product from an RF anti-node to a node. Frequency hopping should counter such problems but the data rate may suffer as the system strains to overcome collisions or initiates retransmissions. In short, in replacing cables with RF, you’ve exchanged Ohm’s law for an inverse square law and introduced an overwhelming number of variables where few existed before. What’s more, you’re now bordering the territory of an endangered species — the RF designer. Let’s face it, you’re an RF virgin and you’d better bone up, at least, on basics.

Reference designs offer refuge
You can some take some comfort from knowing that Bluetooth radio chip or module vendors (Ref 3) provide reference radio designs — already compliance tested — for you to adopt in your host product. The idea is that meticulously replicating the reference will greatly diminish your RF worries. In practice, it appears that few designers adopt reference designs in their entirety. Although it seems like asking for trouble to deviate from a design with certified GHz credentials, a number of reasons for making the move arise. Physical dimensions or layout limitations in host products are the main reasons, but power supply voltages, regulation, and power capacity are others. In theory, if you deviate from an already-certified reference design, then you should have the new design re-qualified (although nothing exists to police that eventuality). The critical parameters affected by changing the physical design will be output power and receiver sensitivity. These factors, in turn, directly affect the radio’s range. If you get to pre-compliance and your product won’t meet its 10-metre range, then your chip vendor will be less supportive than if you had slavishly adopted the reference. Even so, chip/module vendors have a vested interest in the success of your product. While Bluetooth remains in its infancy, and while your RF knowledge is thin, building a close relationship with your radio vendor is a wise move. Nonetheless, all this further endorses the view that you can’t design and test Bluetooth products without some RF measurement capability.

Dedicated RF test sets
The two test sets in our survey target Bluetooth production test but you can also usefully employ them for some in-house RF testing prior to compliance, or pre-compliance. Agilent’s E1852A test set is a blind box that operates from an MS-Windows display on PC via a parallel port. Anritsu’s MT8850A comes as a half-rack benchtop box that you can operate either locally or on a PC via RS-232C or an IEEE-488 interface. Both test sets come with a support software suite. Both sets can operate as piconet masters with the ability to force your product into its test modes. In this mode, the test sets can conduct a wide range of transmitter and receiver RF measurements, such as peak and average power, sensitivity, frequency accuracy and drift, and BER. Both sets enable flexible frequency hopping, from preset hopping routines to disabling hopping altogether and selecting a single channel frequency.

A significant feature of Anritsu’s MT8850A is the ability to generate “dirty transmitter” signals, which enable you to investigate how well your product performs under stressful conditions. Bluetooth is designed primarily as a robust communication system and is certainly not state of the art. By this means, the Bluetooth SIG aims to accommodate the widest range of applications in an unlimited number of environments. For example, the Bluetooth specification allows receiver channel frequency to have up to a 75-kHz error after a frequency hop and at the commencement of a new data packet. Furthermore, the specification allows channel frequency to drift by another 40 kHz during the time taken to transmit one packet. A preset dirty transmitter function allows you to assess your product’s receiver sensitivity under these adverse conditions.

Unplug-fests are like beer-fests
Ignoring for a moment how to test Bluetooth in production, can you now safely assume your development test trail has terminated? Well, not quite. Regulation has one last test trick up its sleeve.

Testing a product’s profiles in isolation is one step, having it pass compliance testing is another, but neither or both those steps prove that your product will actually work in the wild. In theory, and with those two test steps behind you, it certainly should. Now, as further proof, come a hotchpotch of tests, which take place at what the Bluetooth community calls unplug-fests, or what you probably know better as “interoperability” trials. Bluetooth SIG declares that participation in unplug-fests is not mandatory but highly recommended.

Unplug-fests, organised at intervals by the Bluetooth SIG, should be fun events where you release your product into a roomful of other equally unsuspecting products to see what happens. Will your product link up to the piconet? Will it identify what other products are on the net? Will it communicate with the products it’s supposed to and disregard the others? These are just a few of the imponderables that unplug-fests allow you to investigate.

One drawback of unplug-fests is that they occur infrequently and, if your product misbehaves at one of these events, you have very limited time and probably no chance to take corrective action. On returning to base, you may have a quick fix and then want retest interoperability straightaway. As unplug-fests take place at roughly three-month intervals, that sort of delay only increases time-to-market pressures. One company, Wireless Futures (Milton Keynes, UK, www.thebluelabs.com), aims to relieve this pressure by providing permanent unplug-fest facilities. The company is now in the process of establishing a pool of Bluetooth reference devices with various product profiles in order to replicate a wide range of unplug-fest conditions. In addition, the company offers lab facilities for on-the-spot modifications. If you find it’s too tough designing-in Bluetooth, then this company also provides a Bluetooth stack development service.

Other Bluetooth test avenues
While our survey considers only test equipment dedicated to Bluetooth designs, other Bluetooth test options exist. Manufacturers of common test equipment, such as spectrum analysers, signal generators, and DSOs offer “personality” firmware options that optimise those instruments for selective Bluetooth measurements.

These instruments satisfy the needs mainly of the designers of radio chipsets and modules. The instruments make discrete physical-level measurements and have no ability to operate in a piconet.
  In addition, top-end radio test sets that cover a wide range of communications standards now also provide Bluetooth options. Test houses use these test sets for compliance testing. High-speed versions of them target high-throughput test in production.

Ten-second production testing
While regulation broadly dictates the way you test Bluetooth products in design and development, in production you have a free hand. The guiding principle here will be how little test can you get away with and still have confidence that your product’s Bluetooth interface works. Industry forecasters reckon that, at maturity and on average, Bluetooth interfaces will have to cost no more than 5 Euros. Test cost fits somewhere inside that figure and clearly there’s not much room. Overall product cost and volume, of course, also factor into the equation.

In practice, what will happen is that manufacturers will devote somewhere between zero and ten seconds maximum for testing a Bluetooth interface. The simplest way to inspire confidence in a good level of Bluetooth operation is by a selection of RF and baseband tests. There’s hardly a need to exhaustively test Bluetooth product profiles because they’ll be safely cast into firmware and, anyway, will have been thoroughly tested in development. However, a simple functional test, such as a Bluetooth printer actually printing, initiated via the Bluetooth interface also usefully adds confidence.

While speed of test is critical, and high volume justifies it, production engineers will tend to favour top-end radio test sets, such as Rohde & Schwarz’s CMU200, to achieve throughput. These test sets are not dedicated Bluetooth instruments but rather universal test sets optimised for Bluetooth. These may be just the same test sets that manufacturers already use for mobile phone test. These test sets are also high cost (50k Euros) so some producers may consider Anritsu’s MT8850A test set. At much lower cost, and as a dedicated Bluetooth tester, its measurement speed challenges top-end units. For example, the test set can measure transmitter power at 100 readings/s, which means for example you can run a go/no-go test on 100 frequency hops in less than one second.

In the end, what you test in production will depend upon where the product goes. On the one hand, it’s unlikely that Mums will complain if the fluffy toy little Joey had for Christmas fails at two metres. On the other hand, up-market PDA users with “gadget-man mentality” will most certainly kick up if your product performs intermittently above 9.5 metres.

For many companies, especially those of high repute with established product lines, adding Bluetooth puts a lot at stake. It’ll be no good saying it worked fine with cables. Bluetooth is marvellously ambitious and should be a credit to our industry. It’s a bold step forward with great quality-of-life potential. It’s also a small step into the unknown.

References
1. Bluetooth Special Interest Group (SIG) web site, www.bluetooth.com.
2. Robinson, Angus, “On Yours Marks for Testing Bluetooth”, Test & Measurement Europe, June-July 2000, pages 17-24.
3. Cambridge Silicon Radio, www.cambridgesiliconradio.com, Bluetooth chipsets. IBM, www.ibm.com, Bluetooth PCMCIA cards.

Brian Kerridge has been writing on design and test in electronics for ten years with both EDN and TME. He has 25 years’ industry experience of design, engineering management, and marketing in T&M, control electronics, and military radar. He is a Chartered Engineer and MIEE.

A BQTF is an accredited organisation with the skills and equipment to test a product against the Bluetooth specification. Some BQTFs may not offer qualification for every single aspect of the Bluetooth standard. Nonetheless, these BQTFs will still be busy because many product profiles will be limited to a few specific applications. And, anyway, some of the Bluetooth specification is optional. A BQTF performs measurements on your behalf only for the combination of radio, protocol, and profiles for your product. Based on the results, your BQTF prepares a test report that forms part of a compliance folder that it submits to a BQB.

The BQB reviews all submissions to ensure that all the correct tests have been passed. After that, your product is formally listed as Bluetooth qualified and you can go ahead and apply a Bluetooth sticker, and then sell it as a Bluetooth-enabled product.