Talanta 71 (2007) 1998–2002 Performance of a sound card as data acquisition system and a lock-in emulated by software in capillary electrophoresis Marcos Mandaji a , Tiago Buckup b , Rafael Rech b , Ricardo Rego Bordalo Correia b , Tarso Ledur Kist c,∗ a Centro de Biotecnologia, Universidade Federal do Rio Grande do Sul, Porto Alegre, RS, Brazil b Instituto de Fı́sica, Universidade Federal do Rio Grande do Sul, Porto Alegre, RS, Brazil c Instituto de Biociências, Universidade Federal do Rio Grande do Sul, Av. Bento Gonçalves, 9500, Caixa Postal 15093, 91501-970 Porto Alegre, RS, Brazil Received 7 June 2006; received in revised form 7 September 2006; accepted 7 September 2006 Available online 6 October 2006 Abstract The performance of fluorescence detectors in capillary electrophoresis is maximized when the excitation light intensity is modulated in time with optimal frequencies. This is especially true when photomultiplier tubes are used to detect the fluorescent light. The photomultiplier tube amplified raw output signal can in principle be captured directly by a personal computer sound card (PCSC) and processed by a lock-in emulated by software. This possibility is demonstrated in the present work and the performance of this new setup is compared with a traditional data acquisition system. The results obtained with this “PCSC and lock-in emulated by software” were of the same quality or even better compared to that obtained by conventional time integrators (Boxcars) and data acquisition boards. With PCSC the limits of detection (LOD) found for both naphthalene-2,3-dicarboxaldehyde-derivatized tyrosine and alanine were 3.3 and 3.5 fmol (injection of 5 nL of samples at 0.66 and 0.70 ␮mol/L), respectively. This is at least three times better compared to conventional systems when light emitting diodes (LEDs) are used as the excitation source in fluorescence detectors. The PCSC linear response range was also larger compared to conventional data acquisition boards. This scheme showed to be a practical and convenient alternative of data acquisition and signal processing for detection systems used in capillary electrophoresis. © 2006 Elsevier B.V. All rights reserved. Keywords: Light emitting diode; Capillary electrophoresis; PC sound card; Lock-in amplifier; Fluorescence 1. Introduction It was recently shown that light emitting diodes (LED) are inexpensive alternatives to lasers in fluorescence detection systems used in capillary electrophoresis (CE) [1,2]. With the appearance of LEDs with high optical power, small spectral width, and the option of wavelength peaks ranging from UV to the near IR, the application of these devices increased significantly in fluorescence detection systems in capillary electrophoresis. Several works demonstrated low limits of detection (LOD) for many analytes using LEDs as excitation sources in CE Abbreviations: LED, light emitting diode; NDA, naphthalene-2,3dicarboxaldehyde; PLL, phase-locked loop; PMT, photomultiplier tube; PCSC, personal computer sound card ∗ Corresponding author. Tel.: +55 51 3316 7618; fax: +55 51 3316 7003. E-mail address: tarso@adufrgs.ufrgs.br (T.L. Kist). 0039-9140/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2006.09.007 [3–10]. Su et al. [3,5] analyzed riboflavin in beer and urine and obtained a LOD of 1–20 ng mL−1 with a blue LED in the detection system. Jong and Lucy [9] obtained the LODs of 50, 20, and 3 nM for riboflavin, eosin Y, and fluorescein, respectively. Yu et al. [4] also used a blue LED to analyze pathogen cells in fish fluid and obtained a LOD of 4.2 × 104 cells/mL. Tsai et al. [7] assembled a fluorescence detector with UV LED. According to these authors a LOD of 2.0 × 10−10 M and 3.0 × 10−8 M was obtained for NDA-derivatized reserpine and dopamine, respectively. Yang et al. [8] obtained a LOD of 10 nM for FITC-labeled phenylalanine. Zhang et al. [10] also obtained a LOD of 10 nM for FITC-labeled arginine. In these works [3–9] the authors showed several optical setups and sample pre-concentration strategies to obtain low LODs. Zhang et al. [10] assembled the detector with dual modulation of the source light to enhance the signal-to-noise ratio. Several strategies are used to minimize the LODs. A few of these strategies are the use of an appropriate methodology of sample M. Mandaji et al. / Talanta 71 (2007) 1998–2002 1999 into very attractive data acquisition systems for slow varying signals (below 20 Hz), as is the case in capillary electrophoresis. In this work the setup and performance of signal processing made by a lock-in emulated by software in a personal computer is tested. In addition, the data acquisition was made by a usual sound card installed on the mother board instead of a data acquisition card. This new setup was assembled in parallel with a traditional system to compare the performances. 2. Experimental Fig. 1. Schematic diagram of a lock-in amplifier. Filtering and the phase-locked loop (PLL) were emulated by software written in Labview. pre-concentration, good optical components, a good optical setup, LEDs with high optical output power, a sensitive light sensor, a good electronic system to filter the noise, and a data acquisition system with high resolution and sampling rate. The LEDs’ light intensity is easily modulated and this in turn leads to many new possibilities and strategies of signal processing and data acquisition. Usually, the source light is modulated and the signal is processed by a lock-in amplifier or a time integrator in fluorescence detectors for CE. This technique is widely used to detect low signal levels buried in high noise. Today several lock-in amplifiers are digital and the signal processing is made by a digital signal processor (DSP) with on-board software [11–13]. The scheme of a typical lock-in amplifier is shown in Fig. 1. A lock-in amplifier extracts frequency and phase information from a reference signal and generates a pure harmonic wave. This task is made by a digital phase-locked loop (PLL) circuit. The harmonic wave is mixed with a pre-amplified and pre-filtered signal. For each frequency component in the input signal, the mixer stage generates two output components: one with the frequency equal to the difference between the frequencies of the internal reference and the signal component (ωR − ωS ) and another component equal to the sum of the two frequencies (ωR + ωS ). Since the signal and internal reference have the same frequency then its difference will be a DC signal for a specific phase. It is usually referred as the X component and is proportional to the relative phase of the internal reference (cos φ) and signal amplitude. A low-pass filter, of sub-Hz cutoff frequency, will allow it to be detected, but the sum ωR + ωS will be filtered out. If the steps of filtering and mixing are performed in quadrature, then the Y component can be extracted independently from the X component. In a lock-in amplifier the measurement and the comparison of these two components allows the amplitude and the relative phase of the signal to be determined. The resolution (bits) and the sampling rate of analog to digital converters (A/D) installed on the data acquisition cards is another important parameter to achieve low limits of detection. A data acquisition card for CE usually has an A/D with a resolution of 14–16 bits and sampling rates higher than 50 kHz. Personal computers usually have two analog doors for data acquisition, they are the two channels of the on-board sound cards. The sound cards usually have high resolution A/Ds (16 or 24 bits) and sampling rate up to 192 kHz. These characteristics turn these devices 2.1. Electrophoresis parameters The capillary was conditioned at the beginning of each run by rinsing it with NaOH 1 M for 2 min, followed by water for 1 min and running buffer (25 mM borate, pH 9.0, 15% methanol) for 2 min. The NaOH, boric acid, KCN and methanol were purchased from Merck (Darmstad, Germany), amino acids (Sigma–Aldrich) were derivatized with NDA (Sigma–Aldrich). The derivatization reaction was performed in 25 mM borate buffer pH 9.0 with KCN and NDA at concentrations 10 times higher than tyrosine and alanine. 2.2. Instrumentations The CE instrument used in this work was built in our laboratory [2,14] and was equipped with a fused-silica capillary (Polymicro Technologies, Phoenix, AZ, USA) with an i.d. of 50 ␮m, o.d. of 363 ␮m, and a total length of 47 cm (42 cm to the detector). The high-voltage power supply (Series 205B; Bertan, Hicksville, NY, USA) has an output voltage ranging from 0 to ±30 kV. An UV LED with emission maximum at 405 nm (10 mW) and a blue LED with emission maximum at 475 nm (5 mW) were used as the excitation sources for naphthalene-2,3dicarboxaldehyde (NDA) derivatized amino acids and fluorescein, respectively. Fig. 2 shows the instrumental setup of the CE instrument and of the two systems assembled in parallel used for data acquisition and signal processing. A function generator (CFG250; Tektronix, Beaverton, OR, USA) was used for both driving the LED and generating a trigger signal for the Boxcar integrator and a reference signal for the channel 1 of a personal computer sound card (PCSC). The signal from the photomultiplier tube (PMT) (Model E850-02; Hamamatsu, Bridgewater, NJ, USA) was enhanced by an adjustable pre-amplifier (13 AMP 003; Melles Griot, Boulder, CO, USA). The output of this pre-amplifier was send to two directions by a “T” like BNC connector. One branch was used to feed the time-integrator or Boxcar (model 132/164; Princeton Applied Research, Princeton, NJ, USA) called system A. In this system (system A) the signal was integrated by the Boxcar and recorded by means of a conventional data acquisition board (CIO-DAS08; Computer Boards Mansfield, MA, USA) using a home written HP VEE routine (HP VEE, Version 3.21). The second branch was used to feed the channel 2 of a PCSC (Vibra 128 PCI card, model CT4810, Chipset CT5880, Creative Labs) of a second 2000 M. Mandaji et al. / Talanta 71 (2007) 1998–2002 Fig. 2. Scheme of the experimental setup. The output of a function generator (1) is used to supply the LED, the Boxcar trigger, and the PCSC reference signal. The light of the LED (2) is collimated with lenses onto the capillary detection window (3) and the induced fluorescence light is detected by a PMT (4) after passing a spectral filter and a spatial filter (more details are shown in Ref. [2]). The PMT output signal is enhanced by the pre-amplifier (5). System A works with a time integrator (6) and a conventional data acquisition system, while the system B works with a PSCS and lock-in emulated by software. Both systems receive the same output signal from function generator (dashed line) and same output signal from pre-amplifier (solid line). PC and was called system B. In this system (system B) the signal was processed by a Labview routine that emulates a lock-in and the results were recorded on this same PC where the PCSC is also installed. With this setup the data of a single CE run could be simultaneously processed and recorded by both systems and compared. 3. Results and discussion Samples were run several times with 1, 2, 4 and 6 ␮M of derivatized tyrosine and alanine. They were injected at 1 kV for 30 s into the capillary and ran at 15 kV and 25 ◦ C for LOD determination. The results are shown in Fig. 3. Considering the LOD as the concentration that produces a signal-to-noise ratio equal to 3, the LOD in system A was 9.1 and 10.6 fmol (injection of 5 nL of samples at 1.82 and 2.12 ␮mol/L) for NDA-derivatized tyrosine and alanine, respectively, while in system B the LOD was 3.3 and 3.5 fmol (injection of 5 nL of samples at 0.66 and 0.70 ␮mol/L), respectively. Therefore, system B showed to be about three times better (Fig. 4) than system A in this regard. These results show us that emulated lock-ins associated with a sound card have a performance as good as conventional lock-ins and data acquisition systems. In general, enhancement of S/N comes from two aspects: A/D resolution and the intrinsic phase-sensitive detection mechanism [15]. The resolution of a lock-in can be estimated by the A/D card resolution. In the case of a sound card with a 16 bits A/D chip, the S/N that can be measured is at least around 48 dB. This is due to the PCSC high resolution and the ability of the emulated lock-in to make in-phase measurements (PLL). In this sense, the use of A/D cards with higher resolution and higher Fig. 3. NDA-derivatized tyrosine and alanine samples were diluted step-by-step and electrophoresed up to the LOD: (A) the results for system A and (B) for system B. Fig. 4. Signal-to-noise ratio of NDA-derivatized tyrosine (filled symbols) and alanine (empty symbols) as a function of concentration measured by system A (circles) and system B (squares). 2001 M. Mandaji et al. / Talanta 71 (2007) 1998–2002 sampling rates would allow even lower LOD to be achieved. This is especially true in situations were strong noise components are present, i.e., the A/D scale is almost full: a one-count signal in a 24 bit (one in 224 counts) A/D will be hardly detected in a 16 bit (one in 216 counts) A/D because it does not have enough resolution depth. In our case, the A/D resolution did not play a strong role because the noise was not strong. Sampling rate is another important parameter that must be optimized in order to maximize the S/N. According to the Nyquist Theorem [16], the sampling rate must be at least two times higher than the highest frequency observed in the signal. In our measurements, the signal (triggered LED) had a repetition rate of 0.5 kHz; therefore a sampling rate of at least 1.0 kHz should be used. However, sound cards can sample only in four available rates: 8, 11, 22, and 44 kHz. In our setup a sampling rate of 8 kHz was used. The S/N can be improved even more by using higher frequencies, as known from the literature [16]. These sampling rates are still low for multi-capillary systems, which limits its application to single capillary systems. PCSC showed to be a good alternative to conventional data acquisition boards. For low analyte concentrations the PCSC exhibited excellent linear responses and linear regression variances, higher than 0.9741 for peak height and peak area in electropherograms shown in Fig. 3. The relative standard deviation of the four electrophoretic runs was less than 5% for all points using peak area and peak height. This shows that the use of a PCSC as data acquisition system with a lock-in emulated by software is a good alternative to traditional data acquisition boards. Another point investigated was the range of linear response of the PCSC. For this, the UV LED was replaced by a blue LED (475 nm) and fluorescein was used as the analyte. For the test of linearity on the upper limit of the working range a solution of fluorescein was chosen to avoid solubility problems and instabilities that are frequently associated with NDA-derivatized amino acids at higher concentrations. The nonlinear behavior of both systems in the upper limit is examined in Fig. 5. The work- Fig. 5. Linear response of the peak area. Data obtained from system A was showed in (A) and from system B in (B). The averages and standard deviations of four repetitions are shown. Table 1 Linear regression variance data of different fluorescein concentration ranges Analytea (␮M) 20–150 20–200 20–250 20–300 a b Peak height Peak area System A System B System A System B 0.9728 n.a.b n.a.b n.a.b 0.9819 0.9824 0.9559 0.7623 0.9894 n.a.b n.a.b n.a.b 0.9983 0.9870 0.9694 0.8641 Analyte concentration range (fluorescein). Data not available (analog signal input overflow). ing range of the A/D of system A was from 0 up to 10 V. The output signal of this system was linear with concentration from 0 to 150 ␮mol/L of fluorescein. Note that Fig. 5A shows the concentration of the samples against peak areas and not peak heights. According to the results shown in Fig. 5A we conclude that system A has a linear working range from 0 to 150 ␮mol/L of fluorescein. The working range of the A/D of system B was from −5 up to 5 V, but in practice only the range 0–5 V was used. Moreover, the limit signal of 5 V was only reached by 300 ␮mol/L of fluorescein. As shown in Fig. 5B, the output signal of this system was linear with concentration from 0 to 200 ␮mol/L of fluorescein. From Fig. 5A and B it is clear also that system B has a wider working range in concentration compared to system A. Moreover, by using an offset of −5 V in system B, the working range can easily be increased even more, from 0 to 400 ␮mol/L. The linear regression variances of both systems in the upper limit interval are shown in Table 1. For higher concentrations the relationship is no longer linear, this saturation may be a result of many factors thus requiring a calibration curve for each device. However, by varying the detector’s gain, the amplifier gain, or both parameters simultaneously allows the samples to be analyzed in a wider range of concentrations. 4. Conclusion The PCSC with emulated lock-in showed a good performance and the ability to replace Boxcars and conventional data acquisition boards widely used in detection systems in capillary electrophoresis. This system showed to be more sensitive and exhibited a larger working range in concentration. The same performance was also observed for all analytes tested, which includes peptides, protein co-factors, carboxylic acids, and also indirect fluorescence detection of anions (data not shown). Moreover, the applicability of this setup (PCSC with emulated lock-in) is expected to work with all kind of excitation sources that can be modulated, for instance diode lasers and some lamps. Considering that the PCSC used in this work had an A/D with only 16 bits and maximum sampling rate of 44 kHz, it is possible to even improve the performance using PCSC with A/D of 24 bits or higher and with sampling rates of 96 kHz or higher. Therefore, the above tested setup and emulated lock-in showed a good cost-benefit ratio. This is also attractive for applications where low space is available, as all the data acquisition components already are present in entry level PC. 2002 M. Mandaji et al. / Talanta 71 (2007) 1998–2002 Acknowledgements The authors would like to acknowledge the financial support from Coordenação de Aperfeiçoamento de Pessoal de Nı́vel Superior (CAPES), Conselho Nacional para o Desenvolvimento Cientı́fico e Tecnológico (CNPq), and Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (FAPERGS). References [1] M.L. Chabinyc, D.T. Chiu, J.C. McDonald, A.D. Stroock, et al., Anal. Chem. 73 (2001) 4491. [2] S. Hillebrand, J.R. Schoffen, M. Mandaji, C. Termignoni, et al., Electrophoresis 23 (2002) 2445. [3] A.K. Su, Y.S. Chang, C.H. Lin, Talanta 64 (2004) 970. [4] L.J. Yu, L.L. Yuan, H.T. Feng, S.F.Y. Li, Electrophoresis 25 (2004) 3139. [5] A.K. Su, C.H. Lin, J. Chromatogr. B 785 (2003) 39. [6] F. Xu, M. Jabasini, B. Zhu, L. Ying, et al., J. Chromatogr. A 1051 (2004) 147. [7] C.H. Tsai, H.M. Huang, C.H. Lin, Electrophoresis 24 (2003) 3083. [8] B. Yang, F. Tan, Y. Guan, Talanta 65 (2005) 1303. [9] E.P. Jong, C.A. Lucy, Anal. Chim. Acta 546 (2005) 37. [10] T. Zhang, Q. Fang, S.-L. Wang, L.-F. Qin, P. Wang, Z.-Y. Wu, Z.-L. Fang, Talanta 68 (2005) 19. [11] J. Gaspara, S.F. Chen, A. Gordillo, M. Hepp, P. Ferreyra, C. Marque, Microprocess. Microsys. 28 (2004) 157. [12] M.O. Sonnaillona, F.J. Bonettob, Rev. Sci. Instrum. 76 (2005) 024703. [13] A. Restelli, R. Abbiati, A. Geraci, Rev. Sci. Instrum. 76 (2005) 093112. [14] T.B.L. Kist, C. Termignoni, H.P.H. Grieneisen, Braz. J. Med. Biol. Res. 27 (1994) 11. [15] R.B. Northrop, Analog Electronic Circuits: Analysis and Applications, Addison-Wesley, 1989, 521 pp. [16] M.J. Demler, High-Speed Analog-to-Digital Conversion, Academic Press, 1991, 182 pp.