NDT.net • Nov 2005 • Vol. 10 No.11

Design of a low noise preamplifier for ultrasonic transducer

L. Svilainis*, V. Dumbrava
Signal processing department, Kaunas University of technology,
Studentu str. 50, LT-51368 Kaunas, Lithuania, tel. +370 37 300532, E-mail.:svilnis@ktu.lt

*Corresponding Author Contact:
Email: svilnis@ktu.lt, Internet: www.ktu.lt/ultra/

Published in ISSN 1392-2114 ULTRAGARSAS, Nr.2(55) 2005


Abstract

The low noise preamplifier, used for the ultrasonic transducer, design approach is presented. Optimization is based on noise model for both electrical and the acoustical part of receiver. Every noise component has been investigated separately, using various transducers and operational amplifiers models. Noise behavior has been presented in diagrams for analysis. Every noise component has been minimized separately. For signal propagation evaluation, narrowband signal case has been considered. In has been shown that different ultrasonic transducers amplifier circuitry and amplifier type has to be chosen individually in order to improve signal-to-noise ratio.

Introduction

Development of air-coupled ultrasonics is raising demands for electronics used. Because of large difference of acoustic impedances of solid body under investigation and air, testing signal is attenuated. Attenuation can reach up to 100dB, so powerful transmitters and low noise receivers should be used. Since a conventional ultrasound equipment [1] is not suitable in this case, development of dedicated equipment is on its way. Among high power excitation techniques, users are trying to squeeze out better parameters from the transducer preamplifier [2, 3].


Fig. 1. Air-coupled ultrasonic NDE system setup

The ultrasonic transducer is excited by an electrical pulse and transmits the ultrasonic pulse into air and only then a signal is entering the medium under investigation. The ultrasonic pulse travels in the material and is emitting some energy into air. This energy is picked up by a receiving air-coupled transducer.

In this paper we concentrate on the receiving transducer and the preamplifier in order to obtain the best signal to noise performance.

Possible solutions

A few possible solutions for performance optimization are available:
  • Simplest approach would be to use the amplifier with lowest noise components specified by a manufacturer. If an operational amplifier is used, the corresponding datasheet will indicate the voltage and current noise spectral densities en and in. Minimizing those two seems to be sufficient to reach the best noise performance. As it can be seen from further investigation, this approach is not giving a unique solution – what is good for one type of ultrasonic transducer might be of little or no use for other.
  • Maximize a signal level, by increasing the amplifier input impedance, so the signal generated by the transducer is not dampened. Another way is to apply matching circuits [4, 5], so the signal cable or the transducer impedance is best matched for minimum reflections of band flatness. But this requirement usually calls for a high input impedance, which is expected to generate more noise. Also the noise level is left unaccounted here.
  • Minimize the preamplifier output noise, taking into account all possible noise sources influences [2]. However, there is a need to separate the sources of noise which can not be modified or which would degrade the signal performance.
  • Minimize the noise figure. The noise figure is expressing the signal-to-noise ratio degradation while signal is passing the amplifier electronics. Again, as can be seen from a further investigation, this method also has some shorts comings like signal source noise increase is reflected as the noise figure improvement.
  • The complex approach. Since we are speaking about the measurement, major importance is the signal-to-noise ratio (SNR), because SNR is influencing the measurement errors. Simple to say, it still requires some techniques to be developed. Namely, as with methods above, we need some expression of the signal level and how it is influenced by a circuit parameters. Also, a noise behavioural model needs to be developed. Both should be combined in order to give a single result.

Noise model

The noise model can be developed either into analytical form [4] or be modelled using some circuit oriented software like PSPICE [6]. Despite that the PSPICE modelling seems to offer the easiest and closer to reality approach, we decided to use the analytical model, since in such a case noise contributors can be analyzed separately, cutting out the sources level of which can not be modified or controlled. For this purpose we have developed the analytical noise model, incorporating both transducer and electronics units [7].


Fig. 2. Electrical equivalent of the circuit noise to be modeled

The noise spectral density of the is input resistance Rt

eT2 = 4kTRt;      (1)

where k=1.380658.10-23 [J/C] - Bolzmann constant, T is the absolute ambient temperature.


Fig. 3. Noise model circuit

The transducer noise spectral density is determined by the real part of the transducer impedance [4]

es2 = kT Zs      (2)

The feedback circuit resistors contribute the noise densities

e12 = 4kTR1 ;     e22 = 4kTR2 .       (3)

With the amplifier noise gain

we get the equation for the output noise density calculation:

Signal propagation model

To simplify the development of the signal model, we subdivide the task into two possible applications:
  • narrowband inspection, when continuous in time (CW) or close to (CW burst) signals are used (e.g. for attenuation measurement);
  • wideband inspection, when signals used occupy some definable bandwidth (e.g. pulsed signals).
The transducer signal depends on various factors, like transmitting and receiving transducer matching to air, test object coupling to air, transducers orientation, object defects interaction with signals etc. In order to take at least close approach to the factors mentioned, complexity of the model is increasing [5, 6]. There have been attempts to simplify such a solution, with quite useful results [8]. We are taking a different approach. In this paper we will concentrate on a narrowband inspection analysis, so signal is close to the operating frequency and the voltage source internal impedance is determined by Rs. In such a case the signal propagation through amplifier circuits is calculated as

where etr is the transducer signal internal voltage source. Even further, since we are interested on how amplifier circuit elements are influencing the signal level, we do not need an absolute value of the signal received. We just need to have some sort of indication how the signal level will change with circuit parameters. So, we will be investigating the signal level change, regarding the maximum signal level achievable. Since best propagation is achieved when Rt is infinity, Eq.6 becomes

where SF denotes the normalized signal.

Signal propagation analysis

In order to start analysis of signal propagation, we have investigated various air coupled transducers [2, 5, 9] in order to evaluate the range of transducer parameters. In order to restrict ourselves we have concentrated on (400- 800) kHz frequency range transducers, because of popularity of this range. Results of investigation are presented in Table 1.

Transducer Rs, [kW] Cs, [pF] Ls, [ľH] C0, [pF]
Tr 1 9,3 1758 88,7 3333
Tr 2 410 173 561 480
Tr 3 1800 26 1600 65
Table 1. Air coupled transducers parameters

By analyzing the data presented in Table 1, one can see that Rs range is within 10OW to few kW. Therefore we have chosen such a range in the signal propagation analysis. Results of application of Eq.7 for resistance range mentioned are presented in Fig. 4.


Fig. 4. Normalized signal propagation vs. input resistance Rt and transducer Rs

It can be seen that the ratio of input resistance Rt and the transducer impedance Rs should be above 10 times in order to have lowest signal attenuation (the highest SF).

Noise analysis

In order to evaluate the importance and influence of circuit parameters, we now investigate Eq.5 components. The operational amplifier internal noise voltage source en is of major interest. As mentioned above, the simplest approach would be to minimize this parameter, by choosing an appropriate amplifier. In addition to that, this member has the simplest expression. We have chosen three amplifiers as representatives of achievable limits. One (LMH6624) is presenting the lowest voltage noise, an other (OPA657) is exhibiting the lowest current noise source and the third one (CLC425) has both current and voltage noise in between the first two candidates. The parameters are presented in Table 2.

We start analysis using those parameters which do not interfere with other components of noise generation. Since the feedback circuit is not influencing the input components (the transducer and the input resistor) noise, we analyse these components first of all. In order to be able to neglect the R1 and R2 noise, their voltage spectral density has to be 1/3 (less than 5%) of that of contributed by en. Let as solve Eq.5 components for this requirement. The maximum R1 value for R1 thermal noise exclusion is calculates as

The maximum R1 value for R2 thermal noise exclusion is

The maximum R1 value for exclusion of the amplifier current noise in- generated voltage when loaded by R1 and R2 parallel connection is

It can be proved that the reasonable gain value to use in calculations is about 100. The results obtained for this gain level will hold true for all gain values in the range from 2 to 200. The calculations results, basing all the assumptions and Eq.8, 9 and 10 are presented in Table 2. The last two columns summarize the values recommended for R1 and R2. In order to suit all amplifiers, in subsequent calculations we will be using the lowest values – R1=10W, R2=1kW.

Amplifier en,
nV/Hz
in-,
fA/Hz
R1max (9)
W
R1,(10)
W
R1,(11)
W
R1,
W
R2,
KW
LMH6624 0.95 2300 10 610 140 10 1
OPA657 4.8 1.3 160 15610 1243200 160 15
CLC425 2 800 30 2710 840 30 2.7
Table 2. Amplifier parameters and corresponding limit and recommended values for R1 and R2 at G=100

The next candidate for analysis should be the component which we can not be modified, e.g. transducer. Only transducer noise spectral density component e2ntrtot at the amplifier output with the gain 100 is presented in Fig.5. In order to reduce the complexity, the input resistance Rt is taken to be close to infinity.


Fig. 5. Transducers noise spectral density at the amplifier output

The output noise can be calculated by integrating component e2ntrtot over a frequency range. One can see that noise is large at the transducer resonant frequency and is degrading by going apart. For the sake of simplicity, we take the integral over the wide frequency range 150 kHz to 2.5MHz, so there is no need for taking into account an individual transducer bandwidth. The value obtained represents the noise RMS value at the amplifier output:

In such a way we have the ability to analyse noise RMS at the amplifier output behaviour versus the input resistance Rt. The diagrams obtained are presented in Fig.6. In order to be able to judge about the noise influence, the operational amplifier voltage noise source en contributed RMS level at the output is added. From the signal propagation analysis it can be seen that the input resistance Rt is desired to be over 10 times of the transducer Rs. Therefore Rt=10 Rs values for best signal performance are indicated for all three transducers.


Fig. 6. Transducers noise RMS at amplifier output vs. Rt

One can see that using Rt which is optimal for signal propagation will cause an increase in the output noise level. For Tr1 this increase is insignificant with any amplifier, Tr2 noise will increase by about 2 times for the lowest en amplifier and will be insignificant with other two. Tr3 noise will increase about 3 times if the lowest en amplifier is used or increase 1.7 times if average en amplifier is used, but remains almost the same in the case of a relatively large en. In the same way we can calculate the output noise for the resistance Rt thermal noise component e2nRttot getting the EnRtRMS at the amplifier output (Fig .7).


Fig. 7. Rt thermal noise RMS at the amplifier output versus it's value

This time we see that it makes sense to increase the Rt value further above 10Rs, so the Rt thermal noise component can become negligibly low. This behaviour can be explained by damping of the thermal noise source voltage by a lower impedance of the transducer circuit. Taking the same integral over the frequency range we calculate the current noise source in the generated voltage contribution to the noise RMS. The Ein+RMS voltage at the amplifier output is presented in Fig. 8.


Fig. 8. Current noise source in+ generated output noise RMS versus Rt

It is evident, that the LMH6624 amplifier is not suitable for the transducer Tr3 with high Rs and Roe values. The Ein+RMS value for this combination is extending ten times above the en contributed noise level. The same comes to Tr2 and LMH6624 combination, only difference is not that large. Application of the OPA657 amplifier significantly reduces the Ein+RMS level – it is lowest part of the drawing, so is not even notable. But it should be noted here that the OPA657 exhibits quite large en (5 times that of LMH6624), so this fact should be taken into account when shifting to the OPA657 amplifier. Finally, we calculate the total amplifier output noise EntotRMS. The results are presented in Fig. 9.


Fig. 9. Total output noise RMS versus Rt

It is clearly visible that major noise contribution is by en. But for transducers with large Rs and Roe (Tr2 and Tr3) current noise at the desired values of Rt starts prevailing over en. So, basing only noise analysis it can be predicted the that transducer Tr3 should be used with OPA657 or with CLC425 – amplifiers with a lower current noise in+.

SNR analysis

Taking all noise sources into account and calculating the signal level at the output we normalize the value obtained to SNR value at Rt infinity and the lowest Rs transducer operating with the best en level amplifier LMH6624. Then the circuit with best noise performance SNR is 0dB. The SNR analysis results are presented in Fig.10.


Fig. 10. Normalised SNR change at the amplifier output versus Rt

Analysing the results obtained one can see that lowest en amplifier is suitable for a quite large range of transducer Rs values (10W to 400W – the transducers Tr1 and Tr2). But such en does not leave much room for noise reduction the in case of large transducer Roe and Rs (the transducer Tr3). Even further – bias currents of amplifiers investigated were not taken into account, because the resistance at the negative amplifier input is quite low (10W) and the positive input DC resistance is high (it is desired to have it above 10kW), the bias currents skew will cause output DC value out of limits. In literature [1] authors recommend using a low bias current amplifier in such a case (e.g.OPA657). Our solution is different – we break amplifier feedback loop, so the amplifier DC gain gets equal 1. Therefore, the output DC shift due to bias currents skew is low. Taking into account that DC coupling is not needed in ultrasonic applications, the bias shift may be neglected. So, the high bias current amplifier can be used even with a high level of biasing mismatch. As a result of the analysis we present final schematics for the preamplifier (refer to Fig.11). The preamplifier will be used with relatively low impedance transducers [5, 9].


Fig. 11. Final amplifier schematics

Some additional components have been added to schematics to keep the calculated noise performance. The capacitor C6 in series with R1 is introducing the high pass filter with the -3 dB cutoff frequency of 150 kHz. The capacitor C5 will add a low pass filter with the 2.5 MHz cutoff frequency.

Conclusions

Several approaches have been used to improve noise performance of a transducer preamplifier for relatively narrowband signal case. By analysing every noise component it has been shown that en has major contribution to a noise level. The noise current source is of next importance. The approach have been demonstrated how in- component can be reduced. Unfortunately, in+ component influence is dependant an transducer parameters, so it can not be completely taken out. For some transducer types, with high Rs or Roe this component starts prevailing over the en part. In order to have full noise performance evaluation, the SNR analysis has been used. The final design is best suited for relatively low source impedance transducers. If the amplifier for other type of transducer is needed, the circuit parameters should be adjusted in order to get the best noise performance.

Acknowledgements

We would like to thank the Dr. Darius Kybartas for his consultations and useful literature support.

References

  1. Svilainis L., Puodžiunas V. Ultrasonic NDE system: the hardware concept. Ultragarsas. Kaunas: Technologija. 1998. No.29. P.34-40.
  2. Yanez Y., Garcia – Hernandez M. J., Salazar J., Turo A., Chavez J. A. Designing amplifiers with very low output noise for high impedance piezoelectric transducers. NDT&E International. 2005. Vol.38. P. 491-496.
  3. Turo A., Salazar J., Chavez J. A., Kichou H. B., Gomez T. E., Montero de Espinoza F., Garcia – Hernandez M. J.. Ultra-low noise front–end electronics for air–coupled ultrasonic non-destructive evalution. NDT&E International. 2003. Vol.36. P. 93-100.
  4. Dumbrava V., Adomaitis K. Modeling of matching ultrasonic transducers. Elektronika ir elektrotechnika . 2002. Vol 41. P.55-59.
  5. Kažys R., Vladišauskas A., Žukauskas E. Wideband air – coupled ultrasonic transducers. Ultragarsas. Kaunas: Technologija. 2004. No.52. P.21-28.
  6. Maione E., Tortoli P., Lypacewicz G., Nowicki A., Reid J.M. PSpice modelling of ultrasound transducers: comparison of software models to experiment. IEEE transactions on ultrasonics, perroelectrics and frequency control. 1999. Vol.46. P.399-406.
  7. Dumbrava V.,Svilainis L. Noise model for ultrasonics transducer preamplifier. Elektronika ir elektrotechnika. Kaunas: Technologija. (submitted for publication).
  8. Augutis S. V., Dumcius A., Gailius D., Styra D., Jacenas S. The ultrasonic method of hardboard testing. Ultragarsas. Kaunas: Technologija. 2004. No.53. P.14-19.
  9. Žukauskas E., Cicenas V., Kažys R. Application of air – coupled ultrasonic technique for sizing of delamination type defect in multilayered materials. Ultragarsas. Kaunas: Technologija. 2005. No.54. P.7-11.

© NDT.net |Top|