Veri Dönüştürücüleri İle Tasarlanan Seğirme Ölçüm Yöntemleri Ve Uygulamaları

thumbnail.default.placeholder
Tarih
Yazarlar
Çetin, Emre
Süreli Yayın başlığı
Süreli Yayın ISSN
Cilt Başlığı
Yayınevi
Fen Bilimleri Enstitüsü
Institute of Science and Technology
Özet
Bu çalışmada veri dönüştürücüleri ile tasarlanan seğirme ölçüm yöntemlerinden ve bunların uygulamalarından bahsedilmiştir. Öncelikle seğirmenin tanımı yapılarak değişik seğirme türlerinden bahsedilmiş ve seğirme ölçümü ile ilgili literatür araştırması yapılmıştır. Burada geçmişteki araştırmacıların yaptığı bazı seğirme ölçüm çalışmalarından bahsedilmiştir. Daha sonra seğirmenin bir sistemde işaret gürültü oranına etkisi incelenmiştir. Benzer şekilde seğirme ile faz gürültüsü arasındaki ilişki de detaylandırılarak faz gürültüleri eğrilerinden seğirme hesabının nasıl yapılacağı anlatılmıştır. Bu şekilde cihazların katalog verilerinden seğirme hesapları yapılmış ve elde edilen sonuçlar daha sonra geliştirilen yöntemlerin ölçüm sonumlarını analiz ederken korelasyon amacıyla kullanılmıştır. Çalışmanın devamında doğrudan seğirme ölçümü için kullanılabilecek cihazlardan bahsedilmiş ve laboraturvarda bulunan seğirme ölçüm cihazları ile seğirmesi ölçülmek istenen işaret kaynaklarının seğirme değerleri elde edilmiştir. Yine bu değerler geliştirilen yöntemlerin ölçüm sonuçlarını analiz ederken korelasyon amacıyla kullanılmıştır. Çalışmanın devamında bir sistemdeki toplam seğirmeyi ölçmek için kullanılabilecek çift FFT ve çift Histogram yöntemleri tanıtılmıştır. Bu iki yöntem daha önceden Texas Instruments firması çalışanları tarafından patentleri alınan yöntemlerdir. Bu yöntemler yeni nesil 16 bit Pipeline (Boru Hattı) ADC’lere (Analog Sayısal Dönüştürücü) uygulanmıştır. Her bir yöntem için ayrıntılı prensip şemalar ve denklemeler verilmiş, oluşturulan algoritmalar ve MATLAB ortamında yazılan kodlar ayrıntılarıyla açıklanmıştır. Bunun devamında bahsedilen iki yöntem yardımıyla Rohde&Schwarz SMA100A ve Agilent 33250A işaret üreteçlerini içeren düzenekler kurularak her düzeneğin seğirme ölçümleri yapılmıştır. Ölçüm sonuçları daha önceden yapılan doğrudan ölçüm sonuçları ve katalog verileri ile karşılaştırılarak yöntemlerin başarımları test edilmiştir. Bunlara ek olarak ADC’lerin iç seğirmelerini ölçmek için kullanılabilecek farksal örnekleme metodundan bahsedilmiş ve AD9268 16 bit Pipeline ADC’li test düzeneğinde belirtilen yöntem uygulanmıştır. Son olarak da bir veri dönüştürücü sisteminde giriş ve saat işaretlerindeki seğirmeleri birbirlerinden bağımsız olarak ölçmeye dayanan yöntemden bahsedilmiştir. Bu yöntemdeki algoritma ayrıntılı bir şekilde anlatılarak belirli hata parametrelerinin nasıl ortadan kaldırıldığı anlatılmış ve önerilen yöntem 10 farklı deney düzeneğine uygulanarak seğirme ölçümleri yapılmıştır. Veri dönüştürücü sisteminde giriş işaretindeki seğirmenin saat işaretinden çok fazla olduğu durumda saat seğirmesinin ölçümünde tutarsız sonuçlar elde edildiği görülmüştür. Bu durumu düzeltmek için tasarlanan yöntemde geliştirme yapılmış ve gerçeğe daha yakın sonuçlar elde edilmiştir. Son olarak da geliştirilen yöntemleri kullanarak tasarlanabilecek bir seğirme ölçüm cihazından bahsedilmiştir. Böylece çalışmanın devamı niteliğine olacak yeni bir projenin tanıtımı yapılmıştır.
Jitter measurement methods using data converters and their applications are explained in this study. Firstly, jitter is defined and different types of jitter including timing jitter, periodic jitter and cycle-to-cycle jitter are explained. Also a literature search about jitter measurement methods describing some of the previous works to measure system jitter is done in the study. Afterwards, phase noise is explained and its relation with jitter is visuliazed. A MATLAB code is written to compute jitter over a frequency band from a phase noise plot. The code uses piece-wise linear approximation methods to compute jitter from a phase noise plot. Related equations are described and written code is used to calculate jitter of signal generators used in the experiments by their phase noise plots. After that relation between system jitter and signal to noise ratio is revealed. Two experimental setups are used, one consisting Rohde&Schwarz SMA100A very low phase noise signal generator and other consisting Agilent 33250A high phase noise signal generator. It is seen that the setup consisting Agilent 33250A has signal to noise ratio 20dB less than the one consisting Rohde&Schwarz SMA100A. Also known direct jitter measurement methods are explained in the study. These methods involve taking histogram of zero crossings of the DUT (Device Under Test) signal using a fast sampling scope to measure jitter, using a spectrum analyzer to downconvert the DUT signal to a lower frequency band to measure jitter from its phase noise data, or using a signal source analyzer to measure jitter directly from phase noise graph of a DUT signal. Before introducing the proposed jitter measurement methods, two signals generators used in the experiments, Rohde&Schwarz SMA100A and Agilent 33250A, are tested with direct jitter measurement equipments. ITU VLSI laboratory has Rohde&Schwarz FSU26 Spectrum Analyzer, Agilent E5052B signal source analyzer and Agilent DSA-X 93204 fast sampling oscilloscope at the time these study has been carried out. So these two signal jenerators’ jitter is measured directly with fast sampling oscilloscope, spectrum analyzer and signal source analyzer. Also signal jenerators’ phase noise plots derived from device datasheets are used to calculate their jitter from previously written MATLAB code. All these measurements are used to make correlations with proposed jitter measurement methods. Experimental results show that fast sampling scope Agilent DSA-X 93204 is unable to compute jitter accurately for signals which have a moderate slew rate (for example a 80MHz sine wave). It is seen that jitter results of fast sampling scope improves as signal slew rate increases, so it has a slew-rate dependent jitter measurenet floor. Experimental results show that Rohde&Schwarz FSU26 spectrum analyzer can not maintain jitter measurement below 1ps RMS (Root Mean Square). That is because internal phase noise of FSU26 (mostly mixers and other internal circuits) limits the jitter measurement floor of the device. Also spectrum analyzers may spoil the original parameters of the DUT signal while downconverting the signal so this also may affect the jitter results of of the DUT signal and may cause the measured value to deviate from original specifications. Experimental results show that Agilent E5052B signal source analyzer is quite capable of measuring jitter directly. Rohde&Schwarz SMA100A jitter is measured 124 fs RMS by signal source analyzer at 100Hz to 20MHz frequency band, and this result is convenient with the device specifications. So Agilent E5052B is used as a good reference when testing proposed jitter measurement methods. Firstly, overall jitter measurement methods are introduced. There are dual FFT method and dual Histogram method. Dual FFT method is based on SNR (Signal to Noise Ratio) changes at different frequncies and amplitudes. These changes are related with system jitter and jitter can be found from two different SNR values derived at different amplitudes (or different frequencies). Dual FFT method is similar to how spectrum analyzers measure the jitter, it examines all the frequency band and computes the jitter from SNR values at different amplitudes. Dual FFT method equations and develeped algorithm are described in the study. After that, overall jitter measurements are done using dual FFT method. This idea is a TI (Texas Instruments) patent [8] and it is applied to new 16 bit pipeline ADCs (Analog to Digital Converter) by this study. Another overall jitter measurement method is dual Histogram method. Dual Histogram method makes a DC noise pattern or a low frequency sine wave to alias to the Nyquist band by changing input or clock frequency of an ADC. If DUT signal is the inpu signal, the method changes the clock signal or vice-versa. After that the method takes the histogram of the maximum slew-rate points (zero crossings) and minimum slew-rate points (peak values) of the aliased waveform. Minimum slew-rate points have zero slope so a jitter error here does not create a voltage error. So the difference between the variance of maximum slew-rate points and the variance of minimum slew-rate points gives the total variance that jitter causes. By this total variance and by computing slew-rate, overall jitter can be found. Method equations and develeped algorithm is described in the study. Dual Histogram method is similar to how fast sampling scopes measure the jitter. The method does not examine the whole spectrum, instead it examines the zero crossings and peak values. Thus some deterministic components can not be measured with dual Histogram method. This idea is a TI (Texas Instruments) patent [15] and it is applied to new 16 bit pipeline ADCs (Analog to Digital Converter) by this study. All ADCs have internal aperture delay and thus aperture jitter. To measure ADC internal jitter, time diffferential sampling method is introduced. One signal generator is used as both input and clock of the DUT ADC. Input and clock transmission lines are adjusted so that clock samples the input at maximum and minimum slew-rate points. After that point dual Histogram method can be used to compute internal ADC jitter. Experimental setups show that time differential sampling method measures AD9268 internal jitter 120fs RMS. AD9268 datasheet specifies the aperture jitter 70 fs RMS but there is no information about how this value is obtained (by measurement or by a simulation). But it is clear that 120fs RMS aperture jitter measurement is quite good and time differential sampling method seems succesful in this case. After doing overall and internal jitter measurements, proposed methods are compared with direct measurement methods described before. It is seen that for Rohde&Schwarz SMA100A signal generator dual FFT and dual Histogram methods have similar results with Agilent E5052B signal source analyzer. Fast sampling scope Agilent DSA-X 93204 and FSU26 spectrum analyzer failed to meet device specifications. It is seen that for Agilent 33250A signal generator, dual Histogram method gives similar results with fast sampling scope Agilent DSA-X 93204 and Agilent E5052B signal source analyzer. On the other hand dual FFT method’s results are similar with FSU26 Spectrum Analyzer. This is because dual FFT method examines the whole frequnecy band like spectrum analyzer. Dual Histogram method examines zero crossings and peak values and takes their histogram to compute jitter like fast sampling scopes. Lastly, a new method to measure input and clock jitter independent from each other is introduced in this study. This method uses controllable delay lines at two ADC inputs to compute jitter. Input or clock source frequency (of which is not DUT signal) is adjusted so that a low frequency sine wave is aliased to the Nyquist band. For input jitter measurements, delay line is adjusted to have propogation delay of one input period. Common input signal is splitted to both ADCs so ADCs sample inputs at the same time. As common input is splitted to two ADCs by different transmission lines, gain and offset errors will be present at ADC outputs. So, offset and gain errors are corrected and by point to point subtraction of the outputs of the ADCs , clock jitter effects cancel. After the subtraction of two ADC outputs, it is seen that there is also a phase error between two ADC outputs. As a result the difference of both ADC outputs become a sinewave instead of a noise pattern. To overcome this issue, FFT (Fast Foruier Transform) of the difference of both ADCs is taken and fundemental components are made zero. After that inverse FFT of the difference of both ADCs is taken and a noise pattern is derived from this transformation. Then histogram of minimum and maximum slew-rate points are taken and by computing variances and slew-rates, DUT jitter is computed. The same method is applied to measure clock jitter. For clock jitter measurements, delay line is adjusted to have propogation delay of one clock period. 10 measurements with different setups are done with new proposed method. At the case where clock jitter is our concern and input source has high jitter and clock solurce has low jitter, the proposed method fails. This is because at the interval clock source deviates (DUT signal), input source deviates more than clock and input jitter effects are seen at the ADC outputs. So measured jitter values become wrong at this case. To overcome this issue, clock and input frequencies are made equal and similar setup used in input jitter measurement is prepared. With this improvement, both ADCs sample their inputs at the same time, eliminating the problem faced before. After that, for clock jitter measurements, both ADC outputs are subtracted from each other. After that maximum and minimum slew-rate points are specified and their histogram is taken. By computing variances and slew-rates, the jitter is computed. At the final stage, a jitter measurement device that can be developed from the proposed methods is introduced as a block diagram. This may be the continuation of the proposed study at the future. The main drawback here will be the DUT bandwidth, which will be ADCs input bandwidth in this case (approximately 650MHz). Improvements in the ADC design will also improve the methods’ accuracy and bandwidth in the near future.
Açıklama
Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 2012
Thesis (M.Sc.) -- İstanbul Technical University, Institute of Science and Technology, 2012
Anahtar kelimeler
Seğirme, Ölçme, Elektronik Ölçme, Ölçme seti, Ölçme-değerlendirme, Bilgisayar Destekli Ölçme, Ölçme sistemleri, İstatistik, Analog Sayısal Çeviricileri, Jitter, Measurement, Measurement set, Electronic measurement, Assesment-evaluation, Computer-aided measurement, Measurement systems, Statistics, Analog digital converters
Alıntı