SECTION 5 UNDERSAMPLING APPLICATIONS Fundamentals of Undersampling Increasing ADC SfdR and enob using External SHas Use of Dither Signals to Increase ADC Dynamic Range Effect of ADC Linearity and Resolution on SFDR and Noise in Digital Spectral Analysis Applications Future Trends in Undersampling adcs
1 SECTION 5 UNDERSAMPLING APPLICATIONS Fundamentals of Undersampling Increasing ADC SFDR and ENOB using External SHAs Use of Dither Signals to Increase ADC Dynamic Range Effect of ADC Linearity and Resolution on SFDR and Noise in Digital Spectral Analysis Applications Future Trends in Undersampling ADCs
sectiOn 5 UNDERSAMPLING APPLICATIONS Walt Kester An exciting new application for wideband, low distortion ADCs is called undersampling, harmonic sampling, bandpass sampling, or Super- Nyquist Sampling To understand these applications, it is necessary to review the basics of the sampling process The concept of discrete time and amplitude sampling of an analog signal is shown in Figure 5. 1. The continuous analog data must be sampled at discrete intervals, ts, which must be carefully chosen to insure an accurate representation of the original analog signal. It is clear that the more samples taken(faster sampling rates), the more accurate the digital representation, but if fewer samples are taken(lower sampling rates), a point is reached where critical information about the signal is Nyquist's Criteria given in Figure 5.2. Most textbooks state the Nyquist theorem 9 actually lost. This leads us to the statement of Shannons Information Theorem ar along the following lines: A signal must be sampled at a rate greater than twice its maximum frequency in order to ensure unambiguous data. The general assumption is that the signal has frequency components from de to some upper value, fa. The Nyquist Criteria thus requires sampling at a rate fs> 2fa in order to avoid overlapping aliased components For signals which do not extend to de, however, the minimum required sampling rate is a function of the bandwidth of the signal as well as its position in the frequency spectrum SAMPLING AN ANALOG SIGNAL *antiRat Figure 5.1
2 SECTION 5 UNDERSAMPLING APPLICATIONS Walt Kester An exciting new application for wideband, low distortion ADCs is called undersampling, harmonic sampling, bandpass sampling, or Super-Nyquist Sampling. To understand these applications, it is necessary to review the basics of the sampling process. The concept of discrete time and amplitude sampling of an analog signal is shown in Figure 5.1. The continuous analog data must be sampled at discrete intervals, ts , which must be carefully chosen to insure an accurate representation of the original analog signal. It is clear that the more samples taken (faster sampling rates), the more accurate the digital representation, but if fewer samples are taken (lower sampling rates), a point is reached where critical information about the signal is actually lost. This leads us to the statement of Shannon's Information Theorem and Nyquist's Criteria given in Figure 5.2. Most textbooks state the Nyquist theorem along the following lines: A signal must be sampled at a rate greater than twice its maximum frequency in order to ensure unambiguous data. The general assumption is that the signal has frequency components from dc to some upper value, fa. The Nyquist Criteria thus requires sampling at a rate fs > 2fa in order to avoid overlapping aliased components. For signals which do not extend to dc, however, the minimum required sampling rate is a function of the bandwidth of the signal as well as its position in the frequency spectrum. SAMPLING AN ANALOG SIGNAL Figure 5.1
SHANNONS INFORMATION THEOREM AND NYQUISTS CRITERIA Shannon: An Analog Signal with a Bandwidth of fa Must be Sampled at a Rate of f 2fa in Order to Avoid the Loss of Information The signal bandwidth may extend from DC to fa(Baseband Sampling)or from f, to f2, where fa=f2-f,(Undersampling, Bandpass Sampling, Harmonic Sampling, Super-Nyquist) Nyquist: If f <2fa, then a Phenomena Called Aliasing Will Occur. Aliasing is used to advantage in undersampling applications Figure 5.2 In order to understand the implications of aliasing in both the time and frequency domain, first consider the four cases of a time domain representation of a sampled sinewave signal shown in Figure 5.3. In Case 1, it is clear that an adequate number of samples have been taken to preserve the information about the sinewave In Case 2 of the figure, only four samples per cycle are taken; still an adequate number to preserve the information Case 3 represents the ambiguous limiting condition where fs=2fa. If the relationship between the sampling points and the sinewave were such that the sinewave was being sampled at precisely the zero crossings (rather than at the peaks, as shown in the illustration), then all information regarding the sinewave would be lost. Case 4 of Figure 5.3 represents the situation where fs<2fa, and th information obtained from the samples indicates a sinewave having a frequency which is lower than fs/2, i.e. the out-of-band signal is aliased into the Nyquist bandwidth between dc and f/2. As the sampling rate is further decreased, and the analog input frequency fa approaches the sampling frequency fs, the aliased signal approaches dc in the frequency spectrum
3 SHANNON’S INFORMATION THEOREM AND NYQUIST’S CRITERIA Shannon: An Analog Signal with a Bandwidth of fa Must be Sampled at a Rate of fs>2fa in Order to Avoid the Loss of Information. The signal bandwidth may extend from DC to fa (Baseband Sampling) or from f1 to f2 , where fa = f2 - f1 (Undersampling, Bandpass Sampling, Harmonic Sampling, Super-Nyquist) Nyquist: If fs<2fa , then a Phenomena Called Aliasing Will Occur. Aliasing is used to advantage in undersampling applications. Figure 5.2 In order to understand the implications of aliasing in both the time and frequency domain, first consider the four cases of a time domain representation of a sampled sinewave signal shown in Figure 5.3. In Case 1, it is clear that an adequate number of samples have been taken to preserve the information about the sinewave. In Case 2 of the figure, only four samples per cycle are taken; still an adequate number to preserve the information. Case 3 represents the ambiguous limiting condition where fs=2fa. If the relationship between the sampling points and the sinewave were such that the sinewave was being sampled at precisely the zero crossings (rather than at the peaks, as shown in the illustration), then all information regarding the sinewave would be lost. Case 4 of Figure 5.3 represents the situation where fs<2fa, and the information obtained from the samples indicates a sinewave having a frequency which is lower than fs /2, i.e. the out-of -band signal is aliased into the Nyquist bandwidth between dc and fs /2. As the sampling rate is further decreased, and the analog input frequency fa approaches the sampling frequency fs , the aliased signal approaches dc in the frequency spectrum
TIME DOMAIN EFFECTS OF ALIASING CASE 1: fs =8 fa CASE 2: fs =4 fa CASE 3: fs=2fa CASE 4: fs =1.3 fa Figure 5.3 The corresponding frequency domain representation of the above scenario is shown in Figure 5.4. Note that sampling the analog signal fa at a sampling rate fs actually produces two alias frequency components, one at fs+fa, and the other at fs-fa. The upper alias, fs+fa, seldom presents a problem, since it lies outside the Nyquist bandwidth. It is the lower alias component, fs-fa, which causes problems when the input signal exceeds the Nyquist bandwidth, f$/2
4 TIME DOMAIN EFFECTS OF ALIASING Figure 5.3 The corresponding frequency domain representation of the above scenario is shown in Figure 5.4. Note that sampling the analog signal fa at a sampling rate fs actually produces two alias frequency components, one at fs+fa, and the other at fs–fa. The upper alias, fs+fa, seldom presents a problem, since it lies outside the Nyquist bandwidth. It is the lower alias component, fs–fa, which causes problems when the input signal exceeds the Nyquist bandwidth, fs /2
FREQUENCY DOMAIN EFFECTS OF ALIASING ASED BANDWIDTH CASE t ' a',. REPEATS 2 CASE 3 a CASE 4 Figure 5.4 From Figure 5. 4, we make the extremely important observation that regardless of where the analog signal being sampled happens to lie in the frequency spectrum(as long as it does not lie on multiples of fs/2), the effects of sampling will cause either the actual signal or an aliased component to fall within the Nyquist bandwidth between dc and fs/2. Therefore, any signals which fall outside the bandwidth of interest, whether they be spurious tones or random noise, must be adequately filtered before sampling. If unfiltered, the sampling process will alias them back within the Nyquist bandwidth where they can corrupt the wanted signals Methods exist which use aliasing to our advantage in signal processing applications Figure 5.5 shows four cases where a signal having a 1MHz bandwidth is located in different portions of the frequency spectrum. The sampling frequency must be chosen such that there is no overlapping of the aliased components. In general, the sampling frequency must be at least twice the signal bandwidth, and the sampled signal must not cross an integer multiple of f/2
5 FREQUENCY DOMAIN EFFECTS OF ALIASING Figure 5.4 From Figure 5.4, we make the extremely important observation that regardless of where the analog signal being sampled happens to lie in the frequency spectrum (as long as it does not lie on multiples of fs/2), the effects of sampling will cause either the actual signal or an aliased component to fall within the Nyquist bandwidth between dc and fs/2. Therefore, any signals which fall outside the bandwidth of interest, whether they be spurious tones or random noise, must be adequately filtered before sampling. If unfiltered, the sampling process will alias them back within the Nyquist bandwidth where they can corrupt the wanted signals. Methods exist which use aliasing to our advantage in signal processing applications. Figure 5.5 shows four cases where a signal having a 1MHz bandwidth is located in different portions of the frequency spectrum. The sampling frequency must be chosen such that there is no overlapping of the aliased components. In general, the sampling frequency must be at least twice the signal bandwidth, and the sampled signal must not cross an integer multiple of fs /2
MINIMUM SAMPLING RATE REQUIRED FOR NON OVERLAPPING ALIASING OF A 1 MHZ BANDWIDTH SIGNAL Figure 5.5 In the first case, the signal occupies a band from dc to 1MHz, and therefore must be sampled at greater than 2MSPS. The second case shows a 1MHz signal which occupies the band from 0.5 to 1.5MHz. Notice that this signal must be sampled at a minimum of 3MSPS in order to avoid overlapping aliased components. In the third case, the signal occupies the band from 1 to 2MHz, and the minimum required sampling rate for no overlapping aliased components drops back to 2MSPS. The last case shows a signal which occupies the band from 1.5 to 2.5MHz. This signal must be sampled at a minimum of 2. 5MsPS to avoid overlapping aliased components This analysis can be generalized as shown in Figure 5.6. The actual minimum required sampling rate is a function of the ratio of the highest frequency component fMAX, to the total signal bandwidth, B Notice for large ratios of fmAX to the bandwidth, B, the minimum required sampling frequency approaches 2B
6 MINIMUM SAMPLING RATE REQUIRED FOR NONOVERLAPPING ALIASING OF A 1MHz BANDWIDTH SIGNAL Figure 5.5 In the first case, the signal occupies a band from dc to 1MHz, and therefore must be sampled at greater than 2MSPS. The second case shows a 1MHz signal which occupies the band from 0.5 to 1.5MHz. Notice that this signal must be sampled at a minimum of 3MSPS in order to avoid overlapping aliased components. In the third case, the signal occupies the band from 1 to 2MHz, and the minimum required sampling rate for no overlapping aliased components drops back to 2MSPS. The last case shows a signal which occupies the band from 1.5 to 2.5MHz. This signal must be sampled at a minimum of 2.5MSPS to avoid overlapping aliased components. This analysis can be generalized as shown in Figure 5.6. The actual minimum required sampling rate is a function of the ratio of the highest frequency component, fMAX, to the total signal bandwidth, B. Notice for large ratios of fMAX to the bandwidth, B, the minimum required sampling frequency approaches 2B
MINIMUM REQUIRED SAMPLING RATE AS A FUNCTION OF THE RATIO OF THE HIGHEST FREQUENCY COMPONENT TO THE TOTAL SIGNAL BANDWIDTH B= SIGNAL BANDWIDTH MAXIMUM SIGNAL FREQUENCY 's.MINIMUM REQUIRED SAMPLING RATE Figure 5.6 Let us consider the case of a signal which occupies a bandwidth of 1MHz and lies between 6 and 7MHz as shown in Figure 5.7. Shannon s Information Theorem states that the signal (bandwidth= 1MHz) must be sampled at least at 2MSPS in order retain all the information(avoid overlapping aliased components). Assuming that the adC sampling rate, fs, is 2MSPS, additional sampling frequencies are generated at all integer multiples of f: 4MHz, 6MHz, 8MHz, etc. The actual signal between 6 and 7MHz is aliased around each of these sampling frequency harmonics, fs, 2fs, 3fs 4fs,., hence the term harmonic sampling. Notice that any one of the aliased components is an accurate representation of the original signal (the frequency inversion which occurs for one- half of the aliased components can be removed in software). In particular, the component lying in the baseband region between de and IMHz is the one calculated using a Fast Fourier Transform, and is also an accurat representation of the original signal, assuming no ADC conversion errors. The FFT output tells us all the characteristics of the signal except for its original position in the frequency spectrum, which was apriori knowledge
7 MINIMUM REQUIRED SAMPLING RATE AS A FUNCTION OF THE RATIO OF THE HIGHEST FREQUENCY COMPONENT TO THE TOTAL SIGNAL BANDWIDTH Figure 5.6 Let us consider the case of a signal which occupies a bandwidth of 1MHz and lies between 6 and 7MHz as shown in Figure 5.7. Shannon's Information Theorem states that the signal (bandwidth = 1MHz) must be sampled at least at 2MSPS in order to retain all the information (avoid overlapping aliased components). Assuming that the ADC sampling rate, fs , is 2MSPS, additional sampling frequencies are generated at all integer multiples of fs : 4MHz, 6MHz, 8MHz, etc. The actual signal between 6 and 7MHz is aliased around each of these sampling frequency harmonics, fs , 2fs , 3fs , 4fs , ...., hence the term harmonic sampling. Notice that any one of the aliased components is an accurate representation of the original signal (the frequency inversion which occurs for one-half of the aliased components can be removed in software). In particular, the component lying in the baseband region between dc and 1MHz is the one calculated using a Fast Fourier Transform, and is also an accurate representation of the original signal, assuming no ADC conversion errors. The FFT output tells us all the characteristics of the signal except for its original position in the frequency spectrum, which was apriori knowledge
INTERMEDIATE FREQUENCY (IF) SIGNAL BETWEEN 6 AND 7 MHZ IS ALIASED BETWEEN DC AND 1MHz BY SAMPLING AT 2MSPS BASEBAND fs=2MSPS SIGNAL de To Figure 5.7 A popular application of undersampling is in digital receivers. A simplified block diagram of a traditional digital receiver using baseband sampling is shown in Figure 5.8. The mixer in the rf section of the receiver mixes the signal from the antenna with the RF frequency of the local oscillator. The desired information is contained in relatively small bandwidth of frequencies Delta f In actual receivers, Delta f may be as high as a few megahertz. The local oscillator frequency is chosen such that the Delta f band is centered about the IF frequency at the bandpass filter output. Popular IF frequencies are generally between 10 and 100MHz. The detector then translates the Delta f frequency band down to baseband where it is filtered and processing, but the simple diagram serves to illustrate the concept, s processed by a baseband ADC. Actual receivers can have several stages of rf and IF SIMPLIGIED DIGITAL RECEIVER USING BASEBAND SAMPLING FSECTON FFSECnON D>③ LOWPASS 花TE网 Figure 5.8 In a receiver which uses direct IF-to-digital techniques(often called undersampling harmonic, bandpass, or IF sampling), the If signal is applied directly to a wide bandwidth ADC as shown in Figure 5.9. The ADC sampling rate is chosen to be at least 2 Delta f. The process of sampling the if frequency at the proper rate cause one of the aliased components of Delta f to appear in the de to f/2 Nyquist
8 INTERMEDIATE FREQUENCY (IF) SIGNAL BETWEEN 6 AND 7 MHz IS ALIASED BETWEEN DC AND 1MHz BY SAMPLING AT 2MSPS Figure 5.7 A popular application of undersampling is in digital receivers. A simplified block diagram of a traditional digital receiver using baseband sampling is shown in Figure 5.8. The mixer in the RF section of the receiver mixes the signal from the antenna with the RF frequency of the local oscillator. The desired information is contained in relatively small bandwidth of frequencies Delta f. In actual receivers, Delta f may be as high as a few megahertz. The local oscillator frequency is chosen such that the Delta f band is centered about the IF frequency at the bandpass filter output. Popular IF frequencies are generally between 10 and 100MHz. The detector then translates the Delta f frequency band down to baseband where it is filtered and processed by a baseband ADC. Actual receivers can have several stages of RF and IF processing, but the simple diagram serves to illustrate the concepts. SIMPLIGIED DIGITAL RECEIVER USING BASEBAND SAMPLING Figure 5.8 In a receiver which uses direct IF-to-digital techniques (often called undersampling, harmonic, bandpass, or IF sampling), the IF signal is applied directly to a wide bandwidth ADC as shown in Figure 5.9. The ADC sampling rate is chosen to be at least 2 Delta f. The process of sampling the IF frequency at the proper rate causes one of the aliased components of Delta f to appear in the dc to fs /2 Nyquist
bandwidth of the ADC output. DsP techniques can now be used to process the digital baseband signal. This approach eliminates the detector and its associated noise and distortion. There is also more flexibility in the dsp because the adC sampling rate can be shifted to tune the exact position of the Af signal within the The obvious problem with this approach is that the adc must now be able to accurately digitize signals which are well outside the dc to f/2 Nyquist bandwidth which most ADCs were designed to handle Special techniques are available however, which can extend the dynamic range of ADCs to include IF frequencies IMPLIFIED DIGITAL RECIEVER USING IF SAMPLING RF SECTION BANDPASS DSP Is?2Af Figure 5.9 Let us consider a typical example, where the IF frequency is 72.5MHz, and the desired signal occupies a bandwidth of 4MHz ( B=4MHz), centered on the IF frequency(see Figure 5.10). We know from the previous discussion that the minimum sampling rate must be greater than SMHz, probably on the order of 10MHz in order to prevent dynamic range limitations due to aliasing If we place the sampling frequency at the lower band-edge of 7OMHz(72.5-2.5), we will definitely recover the aliased component of the signal in the dc to 5MHz baseband. There is however, no need to sample at this high rate, so we may choose any sampling frequency 10MHz or greater which is an integer sub-multiple of 70MHz, i.e., 70-2 35000MHz,70÷3=23.333MHz,70÷4=17.500MHz,70÷5=14.000MHz,70:6 11.667MHz, or 707= 10.000MHz. We will therefore choose the lowest possible sampling rate of 10000MHz(70-7)
9 bandwidth of the ADC output. DSP techniques can now be used to process the digital baseband signal. This approach eliminates the detector and its associated noise and distortion. There is also more flexibility in the DSP because the ADC sampling rate can be shifted to tune the exact position of the Df signal within the baseband. The obvious problem with this approach is that the ADC must now be able to accurately digitize signals which are well outside the dc to fs /2 Nyquist bandwidth which most ADCs were designed to handle. Special techniques are available, however, which can extend the dynamic range of ADCs to include IF frequencies. SIMPLIFIED DIGITAL RECIEVER USING IF SAMPLING Figure 5.9 Let us consider a typical example, where the IF frequency is 72.5MHz, and the desired signal occupies a bandwidth of 4MHz (B=4MHz), centered on the IF frequency (see Figure 5.10). We know from the previous discussion that the minimum sampling rate must be greater than 8MHz, probably on the order of 10MHz in order to prevent dynamic range limitations due to aliasing. If we place the sampling frequency at the lower band-edge of 70MHz (72.5–2.5), we will definitely recover the aliased component of the signal in the dc to 5MHz baseband. There is, however, no need to sample at this high rate, so we may choose any sampling frequency 10MHz or greater which is an integer sub-multiple of 70MHz, i.e., 70÷2 = 35.000MHz, 70÷3 = 23.333MHz, 70÷4 = 17.500MHz, 70÷5 = 14.000MHz, 70÷6 = 11.667MHz, or 70÷7 = 10.000MHz. We will therefore choose the lowest possible sampling rate of 10.000MHz (70÷7)
INTERMEDIATE FREQUENCY(IF)SIGNAL AT 72.5MHz(=2MHz)IS ALIASED BETWEEN DC AND 5MHZ BY SAMPLING AT 10MSPS 每=100M tsa 10 000 MSPS 725:2MHE dc To 5 MHz igure 5.10 There is an advantage in choosing a sampling frequency which a sub-multiple of the lower band-edge in that there is no frequency inversion in the baseband alias as would be the case for a sampling frequency equal to a sub-multiple of the upper band-edge. (Frequency inversion can be easily dealt with in the DSP software should it occur, so the issue is not very important) Undersampling applications such as the one just described generally require sampling AD Cs which have low distortion at the high input IF input frequency. For instance, in the example just discussed, the ADC sampling rate requirement is only 10MSPS, but low distortion is required (preferably 60 to 80dB SFDR at the IF frequency of 72 5MHz A large opportunity for bandpass sampling is in digital cellular radio base stations For systems which have RF frequencies at 900MHz, 70MHz is a popular first -IF frequency. For systems using an RF frequency of 1.8GHZ, first-IF frequencies between 200 and 240MHz are often used In broadband receiver applications, one adC digitizes multiple channels in the receive path. Individual channel selection and filtering is done in the digital domain. Narrow band channel characteristics such as bandwidth, passband ripple, and adjacent channel rejection can be controlled with changes to digital parameters (i.e filter coefficients). Such flexibility is not possible when narrow band analog filters are in the receive path Figure 5.11 illustrates the kind of input spectrum an ADC must digitize in a multichannel design. The spectral lines represent narrow band signal inputs from a variety of signal sources at different received power levels. Signal"C"could represent a transmitter located relatively far away from the signal sources"A"and B". However, the receiver must recover all of the signals with equal clarity. This requires that distortion from the front-end rf and IF signal processing components
1 0 INTERMEDIATE FREQUENCY (IF) SIGNAL AT 72.5MHz ( 2MHz) IS ALIASED BETWEEN DC AND 5MHz BY SAMPLING AT 10MSPS Figure 5.10 There is an advantage in choosing a sampling frequency which a sub-multiple of the lower band-edge in that there is no frequency inversion in the baseband alias as would be the case for a sampling frequency equal to a sub-multiple of the upper band-edge. (Frequency inversion can be easily dealt with in the DSP software should it occur, so the issue is not very important). Undersampling applications such as the one just described generally require sampling ADCs which have low distortion at the high input IF input frequency. For instance, in the example just discussed, the ADC sampling rate requirement is only 10MSPS, but low distortion is required (preferably 60 to 80dB SFDR) at the IF frequency of 72.5MHz. A large opportunity for bandpass sampling is in digital cellular radio base stations. For systems which have RF frequencies at 900MHz, 70MHz is a popular first-IF frequency. For systems using an RF frequency of 1.8GHZ, first-IF frequencies between 200 and 240MHz are often used. In broadband receiver applications, one ADC digitizes multiple channels in the receive path. Individual channel selection and filtering is done in the digital domain. Narrowband channel characteristics such as bandwidth, passband ripple, and adjacent channel rejection can be controlled with changes to digital parameters (i.e. filter coefficients). Such flexibility is not possible when narrowband analog filters are in the receive path. Figure 5.11 illustrates the kind of input spectrum an ADC must digitize in a multichannel design. The spectral lines represent narrowband signal inputs from a variety of signal sources at different received power levels. Signal "C" could represent a transmitter located relatively far away from the signal sources "A" and "B". However, the receiver must recover all of the signals with equal clarity. This requires that distortion from the front-end RF and IF signal processing components
