Troubleshooting WLAN Radio Designs: Part 1
By Richard Abrahams, Intersil , CommsDesign.com
Aug 8, 2002 (4:38 AM)
URL: commsdesign.com
The proliferation of wireless LAN (WLAN) technology has become a catch 22 for today's PC and consumer electronics developers. On one hand, WLAN systems provide a new feature set or business opportunity. On the other, these companies are now being forced into the radio arena, which maybe outside their area of expertise.
Like any radio architecture, 802.11b WLAN systems are complex, requiring a keen understanding of RF design, signal processing, and more. Fortunately, manufactures have made the development of WLAN systems easier by offering complete chipset and reference designs. But, even with these complete solutions in hand, many equipment designers still need to make tweaks to these reference designs in order to optimize them for their particular system architectures.
Here's where problems set in. As designers make tweaks, they can quickly run into some tough RF problems, antenna matching headaches, spurious signals, and more. In this two-part series, we'll detail some of the key problems that designers will encounter when implementing WLAN technology in a system design. In part 1, we'll look at troubleshooting coplanar waveguide and transmit output power problems. In part 2, which will appear online next week, we'll look at troubleshooting antenna matching, RF/IF spurious signal, and receiver sensitivity problems.
Most Common Problems
Table 1 describes general symptoms and possible problems which may be causing the specific symptom. We'll look at how to diagnose each of these problems in more detail in both Part 1 and Part 2 of this article. During our discussions, the device under test (CUT) will be n 802.11b radio.
<http://img.cmpnet.com/commsdesign/csd/2002/aug02/intersil-table1.gif>
Most of the problems highlighted in Table 1 require design engineers to use RF testing techniques. The non-50-ohn coplanar waveguide problem, however, requires a different approach. Thus, before looking at the other problems in detail, let's explore the coplanar waveguide issue.
Non-50-ohm coplaner waveguide PC traces are probably the most common cause of poor range in a WLAN radio. A poor impedance match can cause both low transmitter power and poor receiver sensitivity.
In a typical WLAN reference design, RF signals are distributed throughout the radio using 50-ohm coplanar waveguide with ground printed-circuit traces. Coplaner waveguides are usually favored since they offer both low loss and a relatively high degree of shielding.
The impedance of a coplaner waveguide printed-circuit trace strongly depends on the dielectric constant of the insulating material and its thickness. Unfortunately, unless special measures are employed, generic FR4 PCB material has a relatively poor control over both parameters, possibly resulting in substantial deviation from 50 ohms.
Most major PCB manufacturers have developed methodology to provide production printed circuits having impedances of 50 +/-5 ohms. Adherence to this standard will ensure proper propagation of the RF signals throughout the PC board.
It is important that both the PC vendor and the company purchasing the PC board institute quality control procedures to verify the board impedance on an ongoing basis. Fortunately, there are a large number of designs being produced by many companies using the ubiquitous, inexpensive FR4 material. Many PCB vendors have therefore agreed to provide controlled impedance material.
It is also good practice to incorporate a test microstrip transmission line approximately six inches long in the tooling area of each PC array panel. Because of possible impedance variations over a large full PC panel, experience dictates that it is indeed necessary to test each smaller-sized array.
The impedance of the smaller-sized array may be easily verified in production using a time domain reflectometry (TDR) measurement. Polar Instruments is one company providing semiautomatic TDR impedance measurement equipment.1
The layout of a typical coplaner waveguide structure is shown in Figure 1 . In this diagram, the critical dimensions are the thickness of the FR4 insulating material, its dielectric constant, and the width of the RF signal trace. These dimensions have the most pronounced effect on the impedance.
<http://img.cmpnet.com/commsdesign/csd/2002/aug02/intersil-fig1.gif>
FIGURE 1: Diagram of a typical coplaner waveguide.
To a lesser degree, the spacing between the RF signal trace and the top ground traces also influences the impedance of the structure. Note that the top- and bottom-ground planes are connected by an array of ground vias. These ensure a low impedance ground connection between the layers and prevent a higher order propagation mode and improve isolation. It should also be noted that the total width of the top and bottom grounds are not critical as long as they are wide compared to the width of the center trace.
Diagnostic Testing Equipment
Clearly, problems like a non-50-ohm coplanar waveguide can provide huge headaches for system designers. Therefore, it's extremely important for design engineers to setup a proper test bench for tracking down these problems. The following is a recommended set of RF test equipment that can aid when diagnosing problems:
1. RF spectrum analyzer, Agilent model 8593A or equivalent.
2. RF watt meter, Agilent model E4419B with E4412A power hed.
3. Vector network analyzer (VNA), Agilent Model 8753D with calibration kit.
4. Semi-rigid coaxial cable terminated in an SMA connector (several required) [colloquially called a "pipe"].
5. Differential probe, Agilent model 1141A with 1142A probe power supply.
6. Digital storage oscilloscope (DSO), Tektronix TDS3054 with P6247 differential probe.
7. Host computer loaded with s test utilities software package.
8. Flexible high quality coax cables terminated in male SMA connectors (2 required). Source: W.L. Gore or Huber-Shuner.
9. Miscellaneous SMA adapters.
10. Coaxial blocking capacitor, Agilent model 11742A or equivalent, (2 required).
11. SMA 50-ohm dummy load.
12. Murata MXGM76RD1000 RF test cable (some WLAN's may utilize a different RF in/out test connector).
13. PCMCIA extender card, mechanically modified for 3-V keying.
14. Magnetic pickup loop: Short length of semi-rigid coax with one end terminated in an SMA connector and center conductor of the other end bent into a 3/16 inch loop and then soldered to the outer jacket. Insulate the whole assembly with shrink sleeving.
Troubleshooting Low Transmitter Power
The amount of output power pumped out from the transmitter has a clear impact in the range and performance of a WLAN system. Therefore, when designers are experiencing poor range, this is one of the first areas that systems designers should evaluate.
There are four key ways for designers to diagnose transmit output power problems. The first is measuring the RF power output delivered to the antenna.
To measure the output power delivered to the antenna, designers must first establish a reference RF power output at the antenna. To do this, connect the RF output jack to the spectrum analyzer using a flexible coax cable. Once this is done, follow the steps below.
1. Carefully insert the DUT into the host computer.
2. Bring up the specific test software providing a continuous transmitter function. Select a channel in the middle of the band (e.g. CH6). Set the spectrum analyzer to a center frequency of 2437 MHz, a span of 66 MHz, and an amplitude of +10 dBm. Leave all other settings at their default values.
3. Ensure that the automatic level control (ALC) in the transmitter is set.
4. Begin a continuous transmission and note that a spectrum similar to that in Figure 2 is present. The bounceback in signal level appearing on either side of the main spectrum, termed "regrowth", is caused by nonlinearities in the transmitter.
5. For a compliant IEEE 802.11b transmitter, the first regrowth should be at least -30 dBc and the second regrowth better than -50 dBc.
6. Turn off the transmitter. Disconnect the coax cable from the spectrum analyzer and connect it to the RF wattmeter. Once again, begin a continuous transmission. Let the transmitter warm up for approximately 30 seconds then note the wattmeter reading. This is the reference power delivered to the antenna.
7. Consult the reference design specifications and verify that the correct power output has been obtained.
It's important to note that there are two different front-end configurations used on typical WLAN reference designs, normal and enhanced (see Figure 3) . The enhanced front end yields slightly higher power output due to reduced insertion loss at a tradeoff of somewhat increased cost.
<http://img.cmpnet.com/commsdesign/csd/2002/aug02/intersil-fig2.gif>
Figure 2: Typical CCK channel 6 transmitter spectrum.
<http://img.cmpnet.com/commsdesign/csd/2002/aug02/intersil-fig3.gif>
FIGURE 3: Block diagram of a normal and enhanced transmitter front-end design.
Measuring Output Power at the PA
Once designers have checked power output at the antenna, they should then evaluate the power being delivered from the power amplifier (PA). Note: when making this measurement, we'll assume that we have confirmed that the coplaner waveguides on our PC board are close to 50 ohms.
The output of the PA typically contains a matching network to ensure an optimum match to 50 ohms. To measure output power, therefore, designers must carefully remove this network and rebuild it such that the network is isolated from the output trace on the PC assy, thus forming a "teepee" in mid air.
Solder the center conductor of a coax pipe with ultra short leads to this junction and also carefully ground the jacket of the pipe to the ground plane near the PA output. It is also prudent to add an additional mechanical support for the pipe by soldering the shield to a second ground point of the PC board.
Carefully dress the pipe so that the radio assembly may still be inserted into the host computer. Connect the free end of the pipe to the spectrum analyzer using one of the flexible SMA cables and, if necessary, the coaxial blocking capacitor. Next, follow the procedure of steps 1 to 7 above.
The difference in level between that measured at the antenna and at the output of the power amplifier is the insertion loss in the PA/antenna path. The typical loss for designs with a normal front end is approximately 4 dB whereas the loss for those with enhanced front ends is 2 dB.
Check Front-End Component Losses
If the PA output appears normal while insertion loss between the PA and Antenna remains high, designers should then check the losses in the transmit/receive (Tx/Rx) switch, front-end filter, and the diversity switch.
The exact details of the measurements depend on whether the design uses the normal or enhanced front end. The example below details measurement of the Tx/Rx switch loss in a radio with a normal front end. This same measurement, however, is easily adaptable for measurement of the other front-end components. A vector network analyzer (VNA) may also be used as an alternate to the procedure detailed, however the detailed procedure below is usually more convenient to implement.
When measuring loss in a Tx/Rx switch, designers should:
1. Disconnect any matching network between the PA output and its coplaner waveguide. Solder a short piece of pipe directly to the input of the coplaner waveguide at the output of the PA right where the matching network normally connects.
2. Similarly, unsolder the bandpass filter and connect a second pipe across its input terminals.
3. Place one SMA coax cable on the signal generator and another on the RF watt meter.
4. Set the signal generator frequency to 2437 MHz (i.e. the middle of the ISM band) and its RF output level to 0 dBm). Be sure to disable any modulation or sweeping of the generator.
5. Join the free ends of the coax cables together using an SMA double female adapter and the coaxial blocking capacitor.
6. Note the wattmeter reading. In general this reading will be somewhat different from the level set on the generator due to calibration errors and cable losses.
7. Disconnect the cables previously connected in step 5 above and connect the cable from the generator, through the blocking capacitor, to the pipe connected to C56. The blocking capacitor is required so the DC bias on the input pin of the Tx/Rx switch is not disturbed. Connect the free end of the cable from the watt meter to the pipe on the input pads of the bandpass filter.
8. Apply power to the DUT and place the card in continuous transmit mode. Note the difference in RF wattmeter between this reading and the reference reading noted in step 6. This is the insertion loss of the Tx/Rx switch.
Some Other Transmit Power Tests
If the insertion losses of the front-end components appear normal and the PA chip does not appear to be faulty, the next step would be to check the drive to the PA. To do this, designers must:
1. Unsolder the input coupling capacitor to the PA.
2. Solder one end of a 7-pf 0402 size capacitor to the pad of the input coupling capacitor away from the PA. Note that a 7-pf capacitor is self resonant at approximately 2.45 GHz so it appears like a direct short circuit at this frequency.
3. Connect a short piece of pipe to the free end of the 7-pf capacitor and carefully ground its jacket to the ground plane right near the pad.
4. Connect a short, low loss SMA coax cable from the pipe to the RF watt meter.
5. Apply power to the DUT.
6. Note the wattmeter reading. This is the drive to the PA. Note that if the transmitter contains an ALC circuit, this drive may be heavily compressed due to the loop now being open.
Some designs using the enhanced front-end use a bandpass filter (or a pair of bandpass filters in cascade) between the up/downconverter IC and the PA. The insertion loss of these filters may be checked using the procedure highlighted in outlines above in check front-end component losses.
On to Part 2
That wraps up part 1 in our series on troubleshooting WLAN radio designs. In Part 2, which will appear online next week, we'll explore antenna matching and receiver sensitivity issues.
References
1. Polar Instruments (UK), www.polarinstruments.com
2. Abrahams, Richard L. "Measurement of WLAN Receiver Sensitivity", Technical Bulletin, TB382, ( intersil.com
About the Author
Richard Abrahams is a senior principal engineer in the Wireless Applications Group of Intersil Corporation. He received the BSEE and MSEE from Rensselaer Polytechnic Institute. Richard can be reached at rabraham@intersil.com. |
