Wideband A/D Converter Front-End Design Considerations When to Use a Double Transformer Configuration By Rob Reeder [[email protected]] Ramya Ramachandran Consider the input, x(t), to the transformer. It is converted into a pair of signals, x1(t) and x2 (t). If x(t) is sinusoidal, the differential output signals, x1(t) and x2 (t), are of the form x1 (t ) = k1 sin(ω t ) x 2 (t ) = k2 sin(ω t − 180° + ϕ ) = − k2 sin(ω t + ϕ ) (1) The ADC is modeled as a symmetrical third-order transfer function: BACKGROUND Transformers are used for isolation and to convert signals from single-ended to differential. A factor often overlooked when using transformers in the front-end circuitry of high-speed A/D converters is that they are never ideal. With sinusoidal input signals, any imbalance introduced by the transformer delivers an imperfect sinusoidal wave to the input of the ADC, and results in overall data-conversion performance worse than the ADC could otherwise provide. We consider here the effects of input imbalances on ADC performance and provide examples of circuitry to achieve improved results. About Transformers The wide variety of available models from many manufacturers can make transformer selection a confusing process. The challenge is compounded by the differing approaches taken by suppliers in specifying performance; they often differ in the choice and definitions of the parameters they specify. h(t ) = a0 + a1 x (t ) + a2 x 2 (t ) + a3 x 3 (t ) (2) Then y(t ) = h( x1 (t )) − h( x 2 (t )) y(t ) = a1 [ x1 (t ) − x 2 (t ) ] + a2 x12 (t ) − x 22 (t ) + a3 x13 (t ) − x 23 (t ) (3) Ideal Case—No Imbalance When x1(t) and x 2 (t) are perfectly balanced, they have the same magnitude (k1 = k 2 = k) and are exactly 1808 out of phase ( = 08). Since x1 (t ) = k sin(ω t ) x 2 (t ) = − k sin(ω t ) (4) y(t ) = 2a1k sin(ω t ) + 2a3k 3 sin 3 (ω t ) (5) Some key parameters to consider when selecting a transformer to drive a particular ADC are insertion loss, return loss, magnitude imbalance, and phase imbalance. Insertion loss is a guide to the bandwidth capability of the transformer. Return loss, also useful, allows the user to design the termination to match the transformer’s response at a particular frequency or band of frequencies—especially important when using transformers with greater than unity turns ratios. We will focus here on magnitudeand phase imbalance, and how they affect the ADC’s performance in high-bandwidth applications. Applying the trigonometric identity for powers and gathering terms of like frequency, Theoretical Analysis Now suppose the two input signals have a magnitude imbalance, but no phase imbalance. In this case, k1 k2, and = 0. Even with a wide bandwidth rating, the coupling between the transformer’s single-ended primary and differential secondary, though linear, introduces magnitude- and phase imbalances. When applied to a converter (or other differential-input device), these imbalances worsen even-order distortion of the converted (or processed) signal. While usually negligible at low frequencies, this added distortion in high-speed converters becomes significant at roughly 100 MHz. Let us first examine how the magnitude- and phase imbalance of a differential-input signal, particularly the second-harmonic distortion, affect the performance of an ADC. x1(t) x(t) h(t) ADC x2(t) h(t) Figure 1. Simplified block diagram of the ADC front end using a transformer. Analog Dialogue 40-07, July (2006) (6) This is the familiar result for a differential circuit: even harmonics cancel for ideal signals, while odd harmonics do not. Magnitude Imbalance x1 (t ) = k1 sin(ω t ) (7) x 2 (t ) = − k2 sin(ω t ) Substituting Equation 7 in Equation 3 and again applying the trigonometric power identities, ( ) ( )) ( a 3a y (t ) = 2 k12 − k22 + a1( k1 + k2 )+ 3 k13 + k23 4 2 − y(t) XFMR a3k 3 3a3k 3 y(t ) = 2 a1k + sin(ω t ) − sin(3ω t ) 4 4 ( ( a2 2 k1 − k22 2 cos 2ω t − ( ( a3 3 k1 + k23 4 )) sin ωt )) sin 3ω t (8) We see from Equation 8 that the second harmonic in this case is proportional to the difference of the squares of the magnitude terms, k1 and k2, viz., 2nd harmonic ∝ k12 − k22 http://www.analog.com/analogdialogue (9) Phase Imbalance Assume now that the two input signals have a phase imbalance between them, with no magnitude imbalance. Then, k1 = k2, and 0. x1 (t ) = k1 sin(ω t ) (10) x 2 (t ) = − k1 sin(ω t + ϕ ) Substituting Equation 10 in Equation 3 and simplifying, 3a k 3 y(t ) = a1k1 + 3 1 ( sin ω t + sin ω t cos ϕ + cos ω t sin ϕ ) 4 a k 2 (11) − 2 1 ( cos 2ω t − cos 2ω t cos 2ϕ + sin 2ω t sin 2ϕ ) 2 The noise floor, second harmonic, and third harmonic here reflect the composite performance of the converter and front-end circuitry. The converter distortion coefficients (a2 and a3) and noise were computed using these measured results, combined with the 0.0607 dB of magnitude imbalance and 148 of phase imbalance at 170 MHz, specified for a standard 1:1 impedance ratio transformer. These coefficients are used in Equation 8 and Equation 11 to compute y(t), while the magnitude- and phase imbalances are varied in the ranges 0 V to 1 V and 0 degrees to 50 degrees, respectively (the imbalance ranges of a typical transformer in the 1-MHz-to-1000-MHz range), and observe the effect on the second harmonic. The results of the simulations are shown in Figure 4 and Figure 5. a k 3 − 3 1 ( sin 3ω t − sin 3ω t cos 3ϕ + cos 3ω t sin 3ϕ ) 4 –80 –85 From Equation 11, we see that the second-harmonic amplitude is proportional to the square of the magnitude term, k. (12) Observations A comparison of Equation 9 and Equation 12 shows that the second-harmonic amplitude is more severely affected by phase imbalance than by magnitude imbalance. For phase imbalance, the second harmonic is proportional to the square of k1, while for magnitude imbalance, the second harmonic is proportional to the difference of the squares of k1 and k2. Since k1 and k2 are approximately equal, this difference is small. As a test of the validity of these calculations, MATLAB code was written for the model described above to quantify and illustrate the impact of magnitude- and phase imbalances on harmonic distortion of a high-performance ADC with a transformer input (Appendix A). The model includes additive white Gaussian noise. DYNAMIC RANGE (dB) 2nd harmonic ∝ k12 0.1MF 0.1MF 336 +AIN AD9445 366 INPUT Z = 506 3pF 366 0.1MF ADC INTERNAL INPUT Z 336 –AIN 0.1MF 2k6 Figure 2. Front-end configuration of the AD9445 with transformer. 0 DEVICE: AD9445-125 DEVICE NO.: 051107a AVCC: 5V DVCC: 3.3V ENCODE: 125MSPS ANALOG: 45.303MHz SNR: 72.87dB SNRFS: 73.89dBFS UDSNR: 0dB NF: 28.85dB SINAD: 72.65dB FUND: –1.024dBfs IMAGE: 0dBc 2ND: –93.15dBc 3RD: –87.78dBc 4TH: –97.95dBc 5TH: –98.65dBc 6TH: –96.79dBc WOSPUR: –95.65dBc+ THD: –85.75dBc SFDR: 87.78dBc NOISE FLOOR: –116.01dBFS SAMPLES: 32768 WINDOWING: BH4 ANALOG INPUT = 170MHz –10 –20 AMPLITUDE (dBFS) –30 –40 –50 –60 –70 –80 3 –90 + –100 65 2 4 –110 –120 –130 0 5 10 15 20 25 30 35 40 FREQUENCY (MHz) 45 50 55 60 Figure 3. Typical FFT of AD9445, 125 MSPS, IF = 170 MHz. –100 –105 –110 SECOND HARMONIC –120 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 MAGNITUDE IMBALANCE (V) Figure 4. Harmonics plotted vs. magnitude imbalance only. –80 SECOND HARMONIC –85 –90 DYNAMIC RANGE (dB) 10nF ANALOG INPUT –95 –115 The coefficients, ai , used in the MATLAB model are for the AD9445, a high-performance 16-bit, 125-MSPS ADC. The AD9445, in the front-end configuration shown in Figure 2, was used to generate the FFT shown in Figure 3, from which the coefficients were derived. XFMR ADT1-1WT 1:1 Z THIRD HARMONIC –90 –95 THIRD HARMONIC –100 –105 –110 –115 –120 0 5 10 15 20 25 30 35 40 45 50 PHASE IMBALANCE (Degrees) Figure 5. Harmonics plotted vs. phase imbalance only. Figure 4 and Figure 5 show that (a) the third harmonic is relatively insensitive to both magnitude- and phase imbalance, and (b) that the second harmonic deteriorates more rapidly with phase imbalance than with magnitude imbalance. Thus, to achieve better performance from the ADC, a transformer configuration with improved phase imbalance is needed. Two feasible configurations, the first involving a double balun, and the second a double transformer, are shown in Figure 6 and Figure 7. Analog Dialogue 40-07, July (2006) BALUN 1:1 Z INPUT AD9445 using a single transformer input (Figure 10) and a double balun input (Figure 11) show that this is indeed the case; a +10-dB improvement in SFDR is seen with a 300-MHz IF signal. OUTPUT OUTPUT 0 BALUN 1:1 Z DEVICE: AD9445 DEVICE NO.:1 AVCC: 5V DVCC: 3.3V ENCODE: 125MSPS ANALOG: 32.195MHz SNR: 71.17dB SNRFS: 72.95dBFS UDSNR: 83.92dB NF: 28.89dB SINAD: 67.28dB FUND: –1.779dBfs 2ND: –70.93dBc 3RD: –75.29dBc 4TH: –102.18dBc 5TH: –96.69dBc 6TH: –94.24dBc WOSPUR: –91.85dBc+ THD: –69.55dBc SFDR: 70.93dBc NOISE FLOOR: –115.07dBFS SAMPLES: 32768 WINDOWING:BH4 –10 –20 –30 XFMR 1:1 Z INPUT XFMR 1:1 Z AMPLITUDE (dBFS) Figure 6. Double balun configuration. OUTPUT –50 –60 –70 –130 5 10 15 20 25 30 35 40 FREQUENCY (MHz) 45 50 55 60 0 DEVICE: AD9445 DEVICE NO.:1 AVCC: 5V DVCC: 3.3V ENCODE: 125MSPS ANALOG: 32.195MHz SNR: 71.71dB SNRFS: 72.13dBFS UDSNR: 84.79dB NF: 29.71dB SINAD: 67.01dB FUND: –0.415dBfs 2ND: –81.55dBc 3RD: –69.27dBc 4TH: –96.23dBc 5TH: –82.63dBc 6TH: –92.43dBc WOSPUR: –82.63dBc+ THD: –68.81dBc SFDR: 69.27dBc NOISE FLOOR: –114.24dBFS SAMPLES: 32768 WINDOWING: BH4 –10 –20 –30 SINGLE XFMR DOUBLE XFMR DOUBLE BALUN AMPLITUDE (dBFS) AMPLITUDE IMBALANCE (dB) 0 Figure 10. Single transformer input, FFT of AD9445. 125 MSPS, IF = 300 MHz. 3.5 3.0 –40 –50 –60 3 –70 –80 2 5 + –90 6 4 –100 –110 2.5 –120 2.0 –130 –140 1.5 1.0 0 100k 1M 10M 100M 1G FREQUENCY (Hz) Figure 8. Magnitude imbalance from 1 MHz to 1000 MHz. 70 60 0 5 10 15 20 25 30 35 40 FREQUENCY (MHz) 45 50 55 60 Figure 11. Double balun input, FFT of AD9445. 125 MSPS, IF = 300 MHz. 0.5 SINGLE XFMR DOUBLE XFMR DOUBLE BALUN 50 Does this mean that to achieve good performance one must couple two transformers or two baluns onto the ADC’s front end? Not necessarily. The analysis shows that it is essential to use a transformer that has very little phase imbalance. In the following examples (Figure 12 and Figure 13), two different single transformers were used to drive the AD9238 with a 170-MHz IF signal. These examples show that there is 29-dB improvement in second harmonic when the ADC is driven by a transformer that has improved phase imbalance at high frequencies. 40 1 0 ENCODE: 62MHz SAMPLES: 16384 ANALOG: 15.8784MHz –10 –20 30 –30 –40 20 AMPLITUDE (dB) PHASE IMBALANCE (Degrees) 6 4 –120 0.1µF The imbalances from these configurations were compared using a vector network analyzer on specially designed characterization boards. Figure 8 and Figure 9 compare the magnitude- and phase imbalance of these configurations with that of a single transformer. 4.0 + 5 –100 –140 4.5 3 –90 Figure 7. Double transformer configuration. 5.0 2 –80 –110 OUTPUT 0.1µF –40 10 0 100k 1M 10M 100M 1G FREQUENCY (Hz) –60 3 –70 4 –80 + 5 6 –90 SNR: 61.1dBc SNRFS: 61.67dBfs THD: –50.96dBc SINAD: 50.56dBc SFDR: 51.57dBc NOISE FLOOR: –100.8dB –100 Figure 9. Phase imbalance from 1 MHz to 1000 MHz. –110 The above figures clearly show that the double configurations have better phase imbalance at the cost of slightly degraded magnitude imbalance. Therefore, using the results of the above analysis, it appears that the double-transformer configurations can be used to achieve better performance. FFT plots of –130 Analog Dialogue 40-07, July (2006) 2 –50 FUND: –0.567dBfs 2ND: –51.02dBc 3RD: –70.71dBc 4TH: –82.33dBc 5TH: –83.9dBc 6TH: –88.86dBc –120 0 3.1 6.2 9.3 12.4 15.5 18.6 21.7 FREQUENCY (MHz) 24.8 27.9 31.0 Figure 12. Single transformer input, FFT of AD9238. 62 MSPS, IF = 170 MHz @ –0.5 dBFS, second harmonic = –51.02 dBc. 1 0 ENCODE: 62MHz SAMPLES: 16384 ANALOG: 15.8746MHz –10 –20 FUND: –0.504dBfs 2ND: –80.56dBc 3RD: –77.09dBc 4TH: –98.55dBc 5TH: –88.81dBc 6TH: –95.01dBc –30 AMPLITUDE (dB) –40 –50 –60 –70 3 –80 2 + 5 4 –90 6 –100 SNR: 62.81dBc SNRFS: 63.31dBfs THD: –75.21dBc SINAD: 62.57dBc SFDR: 77.24dBc WOSPUR: –87.21dBc NOISE FLOOR: –102.4dB –110 –120 –130 0 3.1 6.2 9.3 12.4 15.5 18.6 21.7 FREQUENCY (MHz) 24.8 27.9 CONCLUSION The phase imbalance of a transformer can worsen the secondharmonic distortion when the transformer is used as a front end for processes (such as A/D conversion, D/A conversion, and amplification) with high-IF inputs (>100 MHz). However, by employing a pair of transformers or baluns, significant improvements can be readily achieved, at the cost of an additional transformer and extra PC board space. Single-transformer designs can achieve adequate performance if the design bandwidth is small and a suitable transformer is chosen. However, they do require a limited matching of bandwidth, and they can be expensive or physically large. In either case, choosing the best transformer for any given application requires detailed knowledge of the transformer’s specifications. Phase imbalance is of particular importance for high-IF inputs (>100 MHz). Even if it is not specified in the data sheet, most transformer manufacturers will provide phaseimbalance information upon request. A network analyzer can be used to measure the transformer’s imbalances as a check, or if the information is not readily available. ACKNOWLEDGEMENTS The authors wish to thank Ravi Kummaraguntla, Andy Morgan, and Chad Shelton for their help in the theoretical analysis and for their lab support. 1.Reeder, Rob, “Transformer-Coupled Front End for Wideband A/D Converters,” Analog Dialogue, Vol. 39, No. 2, pp. 3-6, 2005. 2.Mini-Circuits Data Sheet ADT1-1WT. 3.Pulse Data Sheet CX2039L. 4.Mini-Circuits Application Note: “How Transformers Work.” 5.The Mathworks Matlab program. 6.Analog Devices Data Sheet AD9445. 7.Analog Devices Data Sheet AD9238. 8.M/A-COM Data Sheet TP101. 9.Sprague-Goodman Data Sheet GLSB4R5M102. % Matlab code to study the effect of magnitude and phase imbalance of input % signals on the output % Oct 19, 2005 clear all; close all; % Error terms that magnErrdB = 0; %in phaseErr = 50; %in sd_noise = 100e-6; can be set by the user dB degrees %std dev of noise 31.0 Figure 13. Single transformer input, FFT of AD9238. 62 MSPS, IF = 170 MHz @ –0.5 dBFS, second harmonic = –80.56 dBc. FURTHER READING APPENDIX A MATLAB code used in this experiment: % Convert dB magnErr to voltage level magnErr = 10^(magnErrdB/20); % Coefficients a0=0; %dc offset a1=0.89; a2=0.00038; a3=0.0007; %coefficients of 1st,2nd,3rd harmonics %to match AD9445 typical FFT fin = 100; %input freq - does not affect calculations t = 0:1:2047; %Input signals x1 = 0.5*sin((t/2048)*2*pi*fin); x2 = 0.5*(magnErr)*sin(((t/2048)*2*pi*fin)-pi(phaseErr*pi/180)); %Each differential signal multiplied by the transfer function y1 = a0 + a1*x1 + a2*x1.^2 + a3*x1.^3; y2 = a0 + a1*x2 + a2*x2.^2 + a3*x2.^3; %Output y = y1 - y2; noise = sd_noise*randn(1,length(y)); y = y + noise; % figure; plot(1000*t(1:80),x1(1:80),1000*t(1:8 0),x2(1:80),1000*t(1:80),y(1:80)); %Take the FFT fft_y = fft(y/1024, 2048); Pyy = 10*log10(fft_y.*conj(fft_y)); freq_axis = 0:1:1023; % figure; plot(freq_axis, Pyy(1:1024), ‘-d’); % title(‘Frequency content of the output’); % xlabel(‘Frequency (Hz)’); % axis tight; %Print fundamental and 2nd, 3rd harmonics f = Pyy(101) h2 = Pyy(201) h3 = Pyy(301) Analog Dialogue 40-07, July (2006)