Parallel Tests Promote GSM Phone Production

  The idea behind parallel testing is to use a single test station but to run tests on one or a small number of products concurrently. GSM mobiles lend themselves to this form of test because their design means that different sections of the same product operate somewhat independently. For example, with a suitably fast tester, you can overlap tests on the transmitter and receiver of the same mobile. This process can halve the total test time compared with present methods of running all transmitter tests and then following up with all the receiver tests.

  The particular tests that you can run in parallel include audio, display, keypad, and the GSM functional tests. The GSM tests verify RF conformance with ETSI specifications and take up most of the testing time. In the case of dual-band GSM mobiles, parallel testing is particularly advantageous because you can consider these products as two single-band mobiles on a common RF interface. Without parallel testing on dual-band mobiles you would need almost twice the time that it takes to run a regular GSM test.

Speeding Up Receiver Test
In order to implement parallel GSM tests, you should start by taking a close look at the different test phases. Compare the time taken to measure receiver sensitivity and transmitter quality — the major aspects of functional test — and you find that the receiver test takes two to three times longer. So, in order to gain optimal advantage in using parallel testing, first you have to speed up receiver test.

  For GSM mobiles, you determine the sensitivity of a receiver by its bit error ratio (BER) at low receive signal levels. The mobile sends back the decoded data it receives, and the test instrument compares these data to the original data that were transmitted. The tester uses the number of differing bits to calculate the BER. To improve speech performance, digital receivers use a series of signal processes to recover errors in transmission. In most current GSM receivers, a convolutional or Viterbi decoder processes the data bits before the mobile can send them back. However, only uncoded bits are usable for receiver sensitivity measurement and, thus, you can use only every fifth bit (class II bits) to determine this sensitivity.

  To enhance the efficiency of this type of measurement, ETSI introduced what is called the C-loop in phase 2 of GSM. This loop retrieves the data before the decoder (see Figure 1). In this way, a tester can use all the bits of a burst to determine BER, which speeds up receiver sensitivity testing by as much as five times and gives the method its name: Fast BER. Using Fast BER, the measurements of receiver sensitivity and transmitter quality now take about the same time. Therefore, making these two measurements in parallel could almost halve testing time. The next question is: Can you determine transmitter quality at the same time as measuring receiver sensitivity?

Treat Tx Tests With Caution
In a GSM functional test, your test instrument determines transmitter quality by measuring burst parameters and modulation parameters. The burst parameters are peak power and power/time template, and burst length and timing advance. These values are independent of the transmitted data, so they can be measured in parallel with the receiver test. However, test procedures generally evaluate the parameters at a number of power levels (often three to five levels). The test instrument signals to the mobile the change of transmitted power by a command sequence, and so any measurement of receiver sensitivity has to be interrupted momentarily.

  Such procedures raise the issue of whether the receiver test influences the measurement of transmitter modulation parameters. As the data appear in the modulation contents — and, of course, the BER measurement uses the data — so this data pattern can influence the modulation parameters. This data pattern has little effect on the measurement of phase error, but it does affect frequency error because an imbalance of "1" and "0" bits can force a frequency shift of as much as 67 kHz. Therefore, firstly, you should select random data with a uniform distribution, which in turn means that you learn more from the BER and phase error measurement. Secondly, the tester should correct the frequency error measurement to compensate for any imbalance in the data. Once you have done that then you have the basis for measuring receiver sensitivity and transmitter quality in parallel.

DSP Enables Parallel Test
You must not compromise measurement accuracy when you implement parallel testing because only precise measurement allows optimum alignment and dependable results. Your tester requires a large amount of computing power to test the receiver and transmitter of a mobile phone in parallel, and only the arrival of fast digital signal processors (DSPs) have made this technique possible. Testers such as Wavetek’s Model 4400 feature three high-speed DSPs for the purpose, and task distribution requires careful design.

  The 4400’s first DSP takes care of the signalling by arranging the data contents of the bursts to be sent and decoding the bursts that are received. This "signalling" DSP also computes BER. A second "measurement" DSP operates independently to determine transmitter quality. A third DSP in the RF module controls modulation of the GSM bursts, frequency, and output level.

Testing Multiple Mobiles
Running parallel tests inside your test instrument is just one version of parallel testing. Another version is the functional testing of several mobile phones at one test station simultaneously. In this version, there are two basic parallel methods you can use: one is called asynchronous; the other is synchronous. In the asynchronous method, mobiles leave the test station sequentially. The synchronous method tests all mobiles as batches, which arrive and leave the test station together.

Asynchronous Parallel Test
Asynchronous parallel testing reduces test time, particularly on existing production lines. Overall tests generally involve audio, keypad, and display tests in addition to the functional GSM test. Simultaneous testing of the receiver and transmitter by the GSM test equipment shortens this step to the overall duration of the audio, keypad, and display test. Therefore, now you can even implement these tests in parallel with the GSM functional test.

  If you add a second fixture and a switching matrix to the test station, then you can test two mobile phones simultaneously (see Figure 2). This idea itself is not new, but instruments such as Wavetek’s 4400 now automatically perform the entire GSM functional test using a test program within the instrument.

  Formerly, testers required hundreds of GPIB control commands and results had to be exchanged on the bus with a controller. Today, one command is enough to start the complete test and one request is enough to retrieve all the results in a block. The concept is easy to implement in the software of the test station controller. You can program the test instrument internally for the GSM test, leaving the controller to handle the audio, keypad, and display test.

  This method of asynchronous parallel testing is not only limited to GSM functional tests, but it can also integrate the phone alignment operation. A test station with two RF instruments is capable of handling four mobile phones simultaneously. The set-up can process several steps including: firmware download; RX and TX alignment; keypad, display, and audio test; and GSM functional test.

Synchronous Parallel Test
Unlike asynchronous parallel testing, the synchronous method executes the same step simultaneously on all mobiles under test. Figure 3 shows, for example, that batches of four mobiles can arrive, be tested, and leave the station together.

  In synchronous parallel testing, you have to take precautions against interference between the mobiles. A microphone test on one mobile must not influence the same test on the mobile next to it, for example. This situation is relatively easy to avoid in audio tests by using shielding.

  The situation is quite different in the GSM functional test. In this case, you have low levels of sensitivity measurement on the one hand and high transmit levels on the other. The difference in the two levels can be up to 140 dB.

  Even the best mechanical and electrical isolation construction of the fixture is of limited help here. However, GSM, as a time division multiple access (TDMA) system, offers a solution through the timing offset of a mobile’s transmit and receive operation. If all mobile phones on test receive at the same time, there are no transmitters to disturb reception. Equally, if all mobiles transmit simultaneously and at about the same level, the proportion of RF power is likewise balanced and interference is negligible.

  For all mobile phones under test to transmit and receive synchronously, the test equipment itself has to be synchronised. Test instruments can offer a frame sync output and input for this purpose. All the test instruments — one per mobile-under-test — operate in synchronism like a network of base stations, as do all the mobile phones under test.

  Again, the test instrument’s ability to run an internal program reduces GPIB communication to a minimum. A start command, a group event trigger, and a result request are all the instructions a system controller needs to issue. In this way, you can easily test a group of mobiles by testing the transmitter and receiver sections of each mobile in parallel.

Günther Klenner is a product manager with Wavetek Wandel & Goltermann Wireless Division, Germany. He has a masters in engineering and business administration, and has more than 20 years experience in the RF business.